University of Southern California Dissertations and Theses

Investigating brain aging and neurodegenerative diseases through omics data

(USC Thesis Other)

Investigating brain aging and neurodegenerative diseases through omics data

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Investigating Brain Aging and Neurodegenerative Diseases Through Omics Data

By

Carlos Alberto Galicia Aguirre

A Dissertation Presented to The
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOLOGY OF AGING)

August 2023

Copyright 2023 Carlos Alberto Galicia Aguirre
ii

ACKNOWLEDGMENTS
I would like to express my profound gratitude to Dr. Lisa Ellerby, my primary advisor, for her
unceasing guidance, support, patience, and insightful feedback throughout the course of my
dissertation.
To my committee members, Dr. Simon Melov, Dr. Judith Campisi, and Dr. Sean Curran, I
express my heartfelt thanks. Your knowledge, expertise, and incisive comments were of
immense value to the development of my work.
My sincere appreciation goes to my current and past colleagues in the Ellerby lab, especially Dr.
Kizito-Tshitoko Tshilenge and Dr. Sally Mak, whose friendship and collaboration have been a
source of joy, learning, and inspiration.
I also wish to thank Dr. Nicolas Martin, Dr. Serban Ciotlos and Tommy Tran, who provided
crucial support in the area of single cell genomics. Without your help, my research would have
been significantly more challenging.
My gratitude extends to Dr. Akos Gerencser and the morphology core at the Buck Institute,
whose help and resources made it possible to carry out this research.
I am also grateful to Dr. Birgit Schilling, the past and current members of her laboratory, as well
as the past and current members of the Campisi Lab, whose knowledge and assistance were
necessary to complete this work.
I would like to express appreciation to our collaborators in the laboratories of Dr. Sean Mooney,
Dr. Michelle Ehrlich, and Dr. Christian Neri.
A special appreciation is extended to Angelina Holcom, whose support and companionship
were invaluable throughout this journey.
iii

Lastly, my deepest thanks to my family, particularly my parents, for their enduring support,
unconditional love, and constant encouragement. Your faith in me has been a beacon in my life,
guiding me through every challenge and success.
This work would not have been possible without the contributions, encouragement, and support
from each one of you. For everything you have done, I am deeply and sincerely grateful.
In the following chapters, Chapter 3 is a version of the BioRχiv manuscript entitled
“Transcriptomic Characterization Reveals Disrupted Neurodevelopmental Trajectories in
Huntington's Disease and Possible Therapeutic Avenues” posted April 30, 2023 by Carlos
Galicia Aguirre, Kizito-Tshitoko Tshilenge, Elena Battistoni, Alejandro Lopez-Ramirez, Swati
Naphade, Kevin Perez, Sicheng Song, Sean D. Mooney, Simon Melov, Lisa M. Ellerby.
Chapter 4 is a version of the journal article entitled “Proteomic Analysis of Huntington's Disease
Medium Spiny Neurons Identifies Alterations in Lipid Droplets”, published in Molecular and Cell
Proteomics, March 21, 2023, by Kizito-Tshitoko Tshilenge, Carlos Galicia Aguirre, Joanna Bons,
Akos A Gerencser, Nathan Basisty, Sicheng Song, Jacob Rose, Alejandro Lopez-Ramirez,
Swati Naphade, Ashley Loureiro, Elena Battistoni, Mateus Milani, Cameron Wehrfritz, Anja
Holtz, Claudio Hetz, Sean D Mooney, Birgit Schilling, Lisa M Ellerby.
Chapter 5 is a version of the journal article entitled “Postnatal Conditional Deletion of Bcl11b in
Striatal Projection Neurons Mimics the Transcriptional Signature of Huntington's Disease”,
published in Biomedicines in September 23, 2022, by Sicheng Song, Jordi Creus Muncunill,
Carlos Galicia Aguirre, Kizito-Tshitoko Tshilenge, B Wade Hamilton, Akos A Gerencser, Houda
Benlhabib, Maria-Daniela Cirnaru, Mark Leid, Sean D Mooney, Lisa M Ellerby, Michelle E
Ehrlich.
Chapter 6 is a version of the journal article entitled “FOXO3 targets are reprogrammed as
Huntington's disease neural cells and striatal neurons face senescence with p16INK4a
iv

increase”, published in Aging Cell in November 19, 2020, by Jessica Voisin, Francesca Farina,
Swati Naphade, Morgane Fontaine, Kizito-Tshitoko Tshilenge, Carlos Galicia Aguirre, Alejandro
Lopez-Ramirez, Julia Dancourt, Aurélie Ginisty, Satish Sasidharan Nair, Kuruwitage Lakshika
Madushani, Ningzhe Zhang, François-Xavier Lejeune, Marc Verny, Judith Campisi, Lisa M
Ellerby, Christian Neri.

v

TABLE OF CONTENTS
ACKNOWLEDGMENTS ................................................................................................................................... ii
LIST OF TABLES .............................................................................................................................................. x
LIST OF FIGURES .......................................................................................................................................... xiv
ABSTRACT .................................................................................................................................................... xix
CHAPTER 1: Introduction .............................................................................................................................. 1
Aging ................................................................................................................................................ 1
Aging across multiple species .......................................................................................................... 8
Human Aging .................................................................................................................................. 10
Brain Aging ..................................................................................................................................... 11
Senescent cells and hippocampal function .................................................................................... 13
Huntington’s Disease ..................................................................................................................... 15
Omic Data to Study Brain Aging and Neurodegenerative Diseases............................................... 16
CHAPTER 2: Materials and methods ........................................................................................................... 18
Experimental procedures for chapter 3 ......................................................................................... 18
Experimental procedures for chapter 4 ......................................................................................... 21
Experimental procedures for chapter 5 ......................................................................................... 34
Experimental procedures for chapter 6 ......................................................................................... 41
Experimental procedures for chapter 7 ......................................................................................... 57
CHAPTER 3: Transcriptomic Characterization Reveals Disrupted Medium Spiny Neuron Trajectories
in Huntington's Disease and Possible Therapeutic Avenues. (C Galicia Aguirre et al., 2023) .................... 62
Abstract .......................................................................................................................................... 62
Background .................................................................................................................................... 63
Results ............................................................................................................................................ 65
vi

Differentiation and Characterization of HD72 NSCs and Developing MSNs. ................... 65
scRNAseq Reveals Dysregulation of Pathways Related to HD Pathology as MSN
Maturation Occurs. ........................................................................................................... 73
HD Alters the Neurodevelopmental Program of MSNs. ................................................... 77
Predicted Developmental HD Modifiers Based on Transcriptional Data can Reverse
HD Phenotypes. ................................................................................................................ 83
Discussion ...................................................................................................................................... 90
Supplemental Figures .................................................................................................................... 94
Supplemental Tables .................................................................................................................... 100
CHAPTER 4: Proteomic Analysis of Huntington’s Disease Medium Spiny Neurons Identifies
Alterations in Lipid Droplets (Tshilenge et al., 2023) ................................................................................ 102
Abstract ........................................................................................................................................ 102
Background .................................................................................................................................. 103
Results .......................................................................................................................................... 107
Generation and Characterization of MSNs from HD-iPSCs ............................................. 107
Proteomic Analysis of HD-MSNs with FAIMS-DDA MS on a Orbitrap Lumos system ..... 110
Proteomic Analysis of HD-MSNs with DIA-MS ................................................................ 115
Visualization of the Proteomic Analysis of HD-MSNs with FAIMS-DDA MS ................... 115
Functional Enrichment and Protein Network Analysis Reveal Molecular Hallmarks
of HD ............................................................................................................................... 117
Reactome Functional Interaction Network for Isogenic HD-MSNs Upregulated
Proteins ........................................................................................................................... 118
Clusters 0, 1, 4, 6 – Extracellular Matrix Organization, Replicative Senescence ............ 121
Clusters 2, 3 – Muscle Contraction, Netrin Signaling and Angiogenesis ........................ 121
vii

Cluster 5 – Septin Signaling Pathways in HD .................................................................. 121
Cluster 7 – Glycosaminoglycan Biosynthetic Process and Wnt Signaling ....................... 125
Clusters 8,9 – Oxidation Reduction Process and Fructose Metabolic Process ............... 125
Cluster 10 – DNA Signaling Is a Top Enriched Pathway in HD MSNs and Implicates
MCM Proteins ................................................................................................................. 125
Clusters 0 – VEGFR, Integrin and Fc R Signaling Pathways ............................................ 129
Clusters 1, 2 – Axonal Guidance through Ephrin Signaling and the Stathmin Pathway . 129
Clusters 3, 8 – Dysregulation of APOE Signaling and Lipid Metabolism in HD72-MSNs . 129
Cluster 4 – TGFβ, SMAD and TCF12 Signaling Pathways ................................................ 136
Clusters 5, 7 – Downregulation of HLA- and CNS-Related Proteins in HD-MSNs ........... 136
Cluster 11 – Transcriptional Initiation, Elongation, and Termination from the RNA
Polymerase 1 Promoter .................................................................................................. 137
Comparison of HD-MSNs Proteomic Data Set to Human and Mouse HD Proteome
and Modifier Data Sets ................................................................................................... 138
Drug Signature of HD-MSNs ............................................................................................ 138
HTT Protein Interaction Network Overlap ...................................................................... 139
Discussion .................................................................................................................................... 139
Supplemental Figures .................................................................................................................. 145
Supplemental Tables .................................................................................................................... 166
CHAPTER 5: Postnatal Conditional Deletion of Bcl11b in Striatal Projection Neurons
Mimics the Transcriptional Signature of Huntington’s Disease (Song et al., 2022) ..................... 167
Abstract ...................................................................................................................................... 167
Background .................................................................................................................................. 168
Results .......................................................................................................................................... 171
viii

Transcriptomic analysis of D9-Cre-Bcl11
btm1.1Leid
mice .................................................... 171
Bcl11b reduction results in differentially expressed genes that correlate with
pathways dysregulated in HD. ........................................................................................ 175
Transcriptional network impacted by Bcl11b deletion ................................................... 181
Disruption of BCL11B function in a human HD MSN model. .......................................... 191
Discussion .................................................................................................................................... 195
Supplemental Figures .................................................................................................................. 197
Supplemental Tables .................................................................................................................... 203
CHAPTER 6: Human Huntington's disease neural cells and striatal neurons face senescence with
p16INK4a increase while reprogramming FOXO3 targets (Voisin et al., 2020) ........................................ 204
Abstract ........................................................................................................................................ 204
Background .................................................................................................................................. 204
Results .......................................................................................................................................... 206
Ryk-ICD binds to Armadillo repeats 9-10 of ß-catenin. .................................................. 206
Human HD NSCs reprogram FOXO3 targets. .................................................................. 208
FOXO3 binding sites are enriched for co-regulator motifs. ............................................ 212
FOXO3 binding sites overlap between human C116 and mouse NSCs. ......................... 212
F3T reprogramming in human HD NSCs implicates regulators of cell senescence. ....... 213
FOXO3 represses ETS2 expression in human HD NSCs. .................................................. 214
ETS2 positively regulates p16
INK4a
expression in human HD NSCs. ................................. 215
Prepatterned HD NSCs show cellular senescence features in striatal neurons. ............ 219
FOXO3 and p16
INK4a
oppositely modulate the vulnerability of human HD NSCs. ........... 224
p16
INK4a
mRNA levels are increased in the striatum of HD knock-in mice ...................... 228
Discussion .................................................................................................................................... 229
ix

Supplemental Figures .................................................................................................................. 232
Supplemental Tables .................................................................................................................... 249
CHAPTER 7: Exploring Transcriptional Changes in the Aged Hippocampus Driven by Whole-Body
Clearance of Senescent Cells. ................................................................................................................... 251
Abstract ........................................................................................................................................ 251
Background .................................................................................................................................. 252
Results .......................................................................................................................................... 253
Using single nuclei profiling to study effects of clearance of senescent cells in the
hippocampus................................................................................................................... 253
Comparison of clearance of senescent cells via a transgenic model of p16-positive
cell ablation and 25HC. ................................................................................................... 257
Discussion .................................................................................................................................... 275

x

LIST OF TABLES
Supplemental Table 3.1. DEGs in HD72 and C116 NSCs. IPA on canonical pathways
predicted to be dysregulated in NSCs. Bioplanet enrichment analysis for HD72 NSCs. .......... 100
Supplemental Table 3.2. DEGs in HD72 and C116 developing MSNs. IPA on canonical
pathways predicted to be dysregulated in MSNs. Bioplanet enrichment analysis for HD72
developing MSNs. Comparison of DEGs found in HD72 developing MSNs and other HD
models. ................................................................................................................................... 100
Supplemental Table 3.3. Differentially spliced genes in NSCs and developing MSNs.
Enrichment analysis of shared differentially spliced genes in HD72 developing MSNs,
R6/1 striatum, and HD patient striatum.................................................................................... 100
Supplemental Table 3.4. Single cell DEGs in HD72 NSCs, early progenitors, intermediate
progenitors and mature MSNs. IPA of DEGs in early progenitors, intermediate progenitors
and mature MSNs. IPA comparison analysis for canonical pathways dysregulated in early
progenitors, intermediate progenitors and mature MSNs. ....................................................... 100
Supplemental Table 3.5. Comparison of DLX targets during striatum development with
DEGs in HD72 developing MSNs. ........................................................................................... 100
Supplemental Table 3.6. Small molecules predicted to reverse transcriptional
dysregulation in HD72-developing MSNs. Additional information on predicted small
molecules and predicted mechanisms of action. ..................................................................... 100
Supplemental Table 3.7. DEGs found after Cerulenin treatment in HD72-developing
MSNs. IPA of DEGs found after Cerulenin treatment. Signatures of other small molecules
with similarities to Cerulenin. ................................................................................................... 100
Supplemental Table 3.8. qPCR primers and probes used. ...................................................... 101
Supplemental Table S4.1. DIA window isolation scheme of the DIA MS acquisition method. .. 166
xi

Supplemental Table S4.2. Protein, peptide and peptide spectrum match identification
results obtained from the FAIMS-DDA MS data set. ............................................................... 166
Supplemental Table S4.3. Protein quantification and statistical analysis results obtained
from the FAIMS-DDA MS data set. ......................................................................................... 166
Supplemental Table S4.4. Comparison and validation of the significantly changing protein
candidates obtained from the FAIMS-DDA MS data set with the significantly changing
proteins obtained from the DIA MS data set. ........................................................................... 166
Supplemental Table S4.5. Protein quantification and statistical analysis results obtained
from the DIA MS data set. ....................................................................................................... 166
Supplemental Table S4.6. Custom background of proteins. .................................................... 166
Supplemental Table S4.7. Pathway enrichment analysis of the upregulated and
downregulated proteins from the FAIMS-DDA MS proteomic analysis. ................................... 166
Supplemental Table S4.8A. Reactome functional interaction network analysis of the
upregulated proteins in HD72-MSN to define clusters of proteins that are closely connected. . 166
Supplemental Table S4.8B. Reactome functional interaction network analysis of the
downregulated proteins in HD72-MSN to define clusters of proteins that are closely
connected. .............................................................................................................................. 166
Supplemental Table S4.9. Predicted drugs that modify HD ..................................................... 166
Supplemental Table 5.1. Transcriptomics of Bcl11b deficiency with functional analysis .......... 203
Supplemental Table 5.2. HD mouse transcriptomics overlaps with Bcl11b data set. ............... 203
Supplemental Table 6.1. Definition of F3Ts in human C116 and HD NSCs. Sheet 1
shows the complete list of human genes that are differentially expressed upon FOXO3
induction into the nucleus (IN) compared to no FOXO3 induction (F3T-IN). This table is
annotated with information on FOXO3 binding at promoters (-5 kb/+2 kb) and enhancers
(± 20 kb outside the promoter regions), deregulation in HD NSCs, the same as the ones
used herein, as previously reported (Ring et al., 2015), druggability, overlap with FOXO3
xii

targets in other cell types as previously reported (Eijkelenboom, Mokry, Smits,
Nieuwenhuis, & Burgering, 2013; Paik et al., 2009; Renault et al., 2009; Webb,
Kundaje, & Brunet, 2016), and overlap with RNAi screens in a transgenic nematode
(Lejeune et al., 2012) and human cell (Miller et al., 2012) models of HD pathogenesis.
Sheet 2 shows the complete list of human genes that are differentially expressed
upon cell stress (growth factor deprivation) in a FOXO3-knockdown (KD)-dependent
manner. These genes are those for which the log fold change (LFC) of gene expression
levels in stressed cells treated with FOXO3 siRNAs is no longer significant compared
to the significant LFCs in stressed cells treated with non-targeting control (NTC) pool
of RNAs. These genes also comprise those for which there is a significant difference
between log fold change (LFC) of gene expression levels in stressed cells treated with
FOXO3 siRNAs compared to significant LFCs in stressed cells treated with NTC RNAs.
LFCs were considered significant for a q-value < 0.1 (green cells) as determined using
false discovery rate (FDR) analysis. Differences between LFCs were considered
significant for a p-value < 0.05 as determined using the R function pnorm. A significant
difference or a loss of LFC significance upon FOXO3 knockdown define the subgroup
of F3T-IN-KD targets (blue cells). NA, not applicable. ............................................................. 249
Supplemental Table 6.2. Table S1 extracts showing the list of F3T-INs that are gained
in HD NSCs and their behavior upon silencing of Ryk (sheet 1), those that are lost in
HD NSCs and their behavior upon reduction of Ryk expression (sheet 2) and those
that are conserved in HD NSCs and their behavior upon reduction of Ryk expression
(sheet 3). ................................................................................................................................. 249
Supplemental Table 6.3. Comparison of FOXO3 binding sites in human C116 and
mouse NSCs. See also Figure S3. .......................................................................................... 250
xiii

Supplemental Table 6.4. F3T-IN targets reprogrammed (lost or gained) in human HD
NSCs in a Ryk-independent manner (Table S1 extract). The sub-group of F3T-IN-KD
targets is indicated by blue stars in Figure S4B. ...................................................................... 250
Supplemental Table 6.5. FOXO3 targets reprogrammed (lost or gained) in human HD
NSCs in a Ryk-dependent manner (Table S1 extract). The sub-group of F3T-IN-KD
targets is indicated by blue stars in Figure S4C. ...................................................................... 250
xiv

LIST OF FIGURES
Figure 3.1. Differentiation of Developing MSNs. ........................................................................67
Figure 3.2. Transcriptional characterization of HD72 NSCs and developing MSNs. ..................71
Figure 3.3. Single Cell Transcriptional Characterization of HD72 and C116
Developing MSNs. ....................................................................................................................76
Figure 3.4. Comparison of iPSC Derived C116 and HD72 MSNs with Human Fetal LGE. ........78
Figure 3.5. Developmental Dysregulation in Developing HD72 MSNs. ......................................81
Figure 3.6. Small Molecules Predicted to Reverse HD Dysregulation. ......................................85
Figure 3.7. Effects of Cerulenin Treatment on HD72 MSNs. .....................................................88
Supplemental Figure 3.1. Comparison with proteomics and differential splicing analysis. .........94
Supplemental Figure 3.2. scRNAseq of C116 and HD72 NSCs. ...............................................95
Supplemental Figure 3.3. Proportion of Early Progenitors in Developing MSNs. .......................97
Supplemental Figure 3.4. Expression of DLX Genes and Targets in Different Clusters in
C116 and HD72 Developing MSNs. ..........................................................................................98
Figure 4.1. Schematic representation of HD, isogenic HD-MSN and proteomics workflow. ..... 106
Figure 4.2. Generation and characterization of iPSC-derived MSNs. ...................................... 109
Figure 4.3. Deep proteome coverage using FAIMS gas-phase separation with DDA:
Performance of the FAIMS-DDA MS workflow. ....................................................................... 111
Figure 4.4. Differential analysis of the proteome of isogenic C116- and HD72-MSNs by
FAIMS-DDA MS. ..................................................................................................................... 113
Figure 4.5. De novo sub-network construction and clustering using proteins differentially
upregulated when comparing HD72-MSNs to C116-MSNs. .................................................... 120
Figure 4.6. SEPTIN family members are dysregulated in HD-MSNs. ...................................... 124
Figure 4.7. De novo sub-network construction and clustering using proteins differentially
downregulated when comparing HD72-MSNs to C116-MSNs. ................................................ 127
xv

Figure 4.8. HD-MSN lipid metabolism and its modulation by APOE3. ..................................... 132
Figure 4.9. Autophagy in the HD-MSN. ................................................................................... 134
Figure 4.10. Schematic summarizing the altered triglyceride homeostasis, lipophagy
and lipid droplet formation in HD-MSNs. Image was made with BioRender ....................... 143
Supplemental Fig. 4.1. Performances of the TripleTOF 6600 DIA MS workflow. ..................... 145
Supplemental Fig. 4.2. Validation of the protein groups altered in HD72-MSN vs
C116-MSN. ............................................................................................................................. 147
Supplemental Fig. 4.3. Insulin-like growth factor-binding protein 7 levels are increased
in HD72-MSNs. ....................................................................................................................... 149
Supplemental Fig. 4.4. Functional enrichment map of significantly altered proteins
in HD72-MSNs. ....................................................................................................................... 151
Supplemental Fig. 4.5. IPA analysis of the differentially expressed proteins in
HD72-MSNs compared to controls. ......................................................................................... 153
Supplemental Fig. 4.6. HD-MSN lipid metabolism and its modulation by APOE3. ................... 154
Supplemental Fig. 4.7. HD-MSN lipid metabolism and its modulation. .................................... 155
Supplemental Fig. 4.8. HMG-CoA reductase levels are unchanged. ....................................... 156
Supplemental Fig. 4.9. Quantification of p61, LC3 and LAMP1 in HD MSNs with APOE. ........ 157
Supplemental Fig. 4.10 Quantification of p61, LC3 and LAMP1 in HD MSNs with APOE. ....... 160
Supplemental Fig. 4.11. Modulation of MHC and IFN-γ rescues HD cellular phenotypes. ....... 161
Supplemental Fig. 4.12. Comparison of HD proteomics data sets. .......................................... 163
Supplemental Fig. 4.13. HD protein expression overlaps with HTT interacting proteins. ......... 164
Supplemental Fig. 4.14. IPA analysis of the differentially expressed proteins in
HD72-MSNs compared to controls showing the DNA signaling network. ................................ 165
Figure 5.1. Identification of transcriptome portraits of Bcl11b deletion and WT cells
by RNA-seq analysis in Cre+ and Cre− MSNs populations from the striata of D9-Cre mice. ... 174
xvi

Figure 5.2. Transcriptional profile of Bcl11b deletion is highly correlated with HD
mouse models and postmortem human HD tissue. ................................................................. 177
Figure 5.3. Significantly enriched KEGG terms and IPA signaling. .......................................... 180
Figure 5.4. Gene regulatory network analysis reveals critical up-stream TFs
from the gene signatures altered by Bcl11b deletion. .............................................................. 183
Figure 5.5. Bcl11b deficiency leads to a reduced number of MSNs without microgliosis. ........ 186
Figure 5.6. Bcl11b deficiency in adult mice partly recapitulate HD-associated
motor phenotype. .................................................................................................................... 190
Figure 5.7. Human MSNs derived from HD patient iPSCs reveals mislocalization
of BCL11B into nuclear aggregates. ........................................................................................ 193
Supplemental Fig. 5.1. GO enrichment of MSN Bcl11b deficiency. ......................................... 197
Supplemental Fig. 5.2. Calcium signaling pathways enriched in Bcl11b deficiency. ................ 199
Supplemental Fig. 5.3. IPA analysis of MSN mouse Bcl11b deficiency. .................................. 200
Supplemental Fig. 5.4. Bcl11b deficiency does not induce anxiety-like behaviors. .................. 201
Supplemental Fig. 5.5. Cell-type enrichment analysis. ............................................................ 202
Figure 6.1. FOXO3, ß ‐catenin, and Ryk ‐ICD form a protein complex in HEK293T cells. ......... 207
Figure 6.2. FOXO3 binding and gene regulation in human NSCs expressing
normal or mutant HTT with or without Ryk silencing. ............................................................... 210
Figure 6.3. Gene expression analyses in human NSCs. ......................................................... 218
Figure 6.4. Human HD prepatterned NSCs show increase of p16INK4a and of
SA ‐β ‐gal activity. ..................................................................................................................... 220
Figure 6.5. p16INK4a expression is elevated in human HD MSNs. ......................................... 222
Figure 6.6. FOXO3 and p16INK4a oppositely modulate the vulnerability of
human HD NSCs. ................................................................................................................... 226
Supplemental Fig. 6.1. Gene expression analysis, FOXO3 induction and
efficiency of Ryk silencing in HD and C116 NSCs. Related to Figure 2. .................................. 233
xvii

Supplemental Fig. 6.2. Overview of RNA-seq data upon FOXO3 nuclear induction
in HD and C116 NSCs. ........................................................................................................... 235
Supplemental Fig. 6.3. FOXO3 binding sites are enriched for candidate co-regulator
motifs that are shared across HTT genotypes or unmasked in cells expressing
mutant HTT. ............................................................................................................................ 236
Supplemental Fig. 6.4. Overlap between FOXO3 binding sites in human C116
neural stem cells and mouse neural stem cells. ...................................................................... 237
Supplemental Fig. 6.5. Network of FOXO3 direct targets in human HD NSCs
compared to control cells. ....................................................................................................... 239
Supplemental Fig. 6.6. Gene target expression levels upon treatment with siRNAs
and HTT expression levels upon reduction of FOXO3 or reduction of p16
INK4a
. ....................... 240
Supplemental Fig. 6.7. Evaluation of candidate FOXO3 targets (CDKN2AIP,
SERTAD1) and products of the CDKN2A locus (i.e., p14
ARF
). ................................................. 243
Supplemental Fig. 6.8. Increased levels of p16
INK4a
and elevated SA-ß-gal activity
are also characteristic of other non-isogenic HD NSC lines..................................................... 245
Supplemental Fig. 6.9. Relevant markers of senescence evaluated in HD NSCs
and MSNs. .............................................................................................................................. 246
Supplemental Fig. 6.10. Increase of p16
INK4a
mRNA levels in the striatum of HD
model mice. ............................................................................................................................ 248
Figure 7.1. Overview of the single-nuclei RNA sequencing analysis. ...................................... 256
Figure 7.2. Non-parametric comparison of transcriptional signatures caused by
removal of senescent cells by GCV or 25HC. ......................................................................... 260
Figure 7.3. Correlation of transcriptional signatures caused by removal of
senescent cells by GCV or 25HC. ........................................................................................... 262
Figure 7.4. Non-parametric comparison of transcriptional signatures caused by
aging and removal of senescent cells by GCV or 25HC in neurons. ........................................ 267
xviii

Figure 7.5. Removal of senescent cells reverses age-related transcriptional
changes in glutamatergic neurons. .......................................................................................... 269
Figure 7.6. Non-parametric comparison of transcriptional signatures caused
by aging and removal of senescent cells by GCV or 25HC in glial cells. ................................. 272
Figure 7.7. Non-parametric comparison of pseudo-bulk transcriptional signatures
caused by aging, removal of senescent cells via GCV or 25HC against caloric
restriction (CR) in the hippocampus. ....................................................................................... 274
Supplemental Figure 7.1. Quality control metrics. ................................................................... 277
Supplemental Figure 7.2. Batch and sample integration. ........................................................ 278
Supplemental Figure 7.3. Cell type numbers. .......................................................................... 279
Supplemental Figure 7.4. Transcriptional signature of GABAergic neurons after
25HC or GCV treatment. ......................................................................................................... 280
Supplemental Figure 7.5. Correlation of transcriptional signatures caused by
removal of senescent cells via GCV or 25HC, compared to aging signature in neurons. ......... 281
Supplemental Figure 7.6. Correlation of transcriptional signatures caused by removal
of senescent cells via GCV or 25HC compared to aging signature in glial cells. ...................... 282

xix

ABSTRACT
Aging is a complex, multifaceted biological process that entails a systematic decline in
physiological function over time and is intimately linked to chronic diseases. This link is
attributable to the protracted accumulation of cellular and molecular damage as well as other
physiological changes that can trigger an array of ailments including, but not limited to,
cardiovascular disease, diabetes, and cancer. Of particular concern in this gamut of disorders
are neurodegenerative diseases, hallmarked by a gradual, unrelenting impairment of neuronal
structure and function. Aging stands as a predominant risk factor for such diseases, with
Alzheimer's disease, Parkinson's disease, and Huntington's disease as prime exemplars.
In the past few decades, the development of transcriptomic, proteomic and other omics
technologies has allowed us to characterize the molecular components of biological samples in
unprecedented detail. These technologies allow for the unbiased exploration of the effects of
complex biological processes. These data driven approaches allow for the identification of
dysregulated genes, pathways and biological processes at an unprecedented pace. In addition,
the application of these technologies to individual cells has allowed for the study of biological
samples at extraordinary resolution.
Due to their capacity for large-scale, parallel hypothesis testing, omics technologies prove
exceptionally beneficial in detecting dysregulation within intricate biological processes, like
those observed in aging and neurodegenerative diseases. Here, we utilize different types of
omics technologies to elucidate the molecular changes inherent in neurodegenerative diseases
and brain aging.
The first section examines the molecular alterations induced by Huntington's disease (HD), a
neurodegenerative genetic disorder resulting from the expansion of CAG repeats in the
huntingtin gene, ultimately leading to the production of a mutant HTT protein. Although the HTT
xx

protein is expressed in multiple tissues, HD mainly affects medium spiny neurons (MSNs) in the
striatum, leading in their loss and subsequent motor function impairment. Research indicates
that alterations caused by HD during development can trigger pathology later in life and
counteracting these events can delay HD pathology. To determine what are the molecular
changes driven by HD during MSN development, we used HD72 (72/19 CAG repeats) induced
pluripotent stem cells (iPSCs) and isogenic controls (21CAG/19CAG repeats). We differentiated
these iPSC lines into neuronal stem cells (NSCs) and a population of developing MSNs
containing MSNs at different stages of development. By utilizing both bulk and single-cell RNA
sequencing, we reveal transcriptional changes across multiple stages of MSN development,
most notably, dysregulation in the DLX family of transcription factors, which are vital for MSN
development. Through computational methods utilizing transcriptional data, we identify several
potential HD modifiers, including cerulenin. When HD72 MSNs were treated with cerulenin, we
observe a partial reversal of some HD-associated transcriptional changes as well as a partial
restoration of electric activity and increased levels of DARPP-32. This provides proof of concept
for the viability of our approach in identifying potential interventions for HD.
In the second section, we explore the molecular alterations in iPSC derived HD72 MSNs using
quantitative proteomics. We note a significant agreement between changes driven by HD at the
RNA and protein levels. We also identify dysregulation in lipid metabolism and observe lipid
droplet accumulation in HD72 MSNs, hinting at potential lipid turnover deficiencies, possibly via
lipophagy.
The third section focuses on the transcription factor Bcl11b, which is required for MSN
development. We present evidence that loss of Bcl11b in adult MSNs can induce HD-like
molecular and behavioral phenotypes in mice. We also show that BCL11B forms granules in the
nuclei of developing HD72 MSNs, a phenotype not observed in controls, suggesting potential
xxi

alterations in BCL11B function in HD MSNs. This reveals Bcl11b's critical role in mature MSN
function and suggests a link between BCL11B and HD.
In the fourth section, we demonstrate that HD induces senescent-like features in human HD72
NSCs and MSNs. We also show HD to reprogram FOXO3 targets in HD72 NSCs.
In the final section, we characterize the transcriptional changes driven by senescent cells in the
aged mouse hippocampus. To accomplish this, we performed single-nuclei RNA sequencing on
hippocampus collected from aged mice that have undergone clearance of senescent cells.
Subsequently, we contrasted the transcriptional signature stemming from the removal of
senescent cells in aged mice with the signature resulting from aging. This comparison allowed
us to discern the impact of senescent cell clearance on various hippocampal cell types.
Notably, we detected a reversal in the age-related transcriptomic signature of glutamatergic
neurons, characterized by a marked enrichment of genes associated with synapse function.
This observation aligns with prior research that highlighted an age-related decline in
hippocampal synaptic function, which was found to be reversible following the clearance of
senescent cells.
1

CHAPTER 1: Introduction
Aging
Biological aging is a complex and multifaceted process that leads to a progressive decline in
physiological function and increased vulnerability to diseases and death. This phenomenon is
observed across almost all living organisms, from simple unicellular organisms to complex
multicellular ones like humans (Cohen, 2018; Florea, 2017).
The ultimate cause of aging is not known, but multiple processes denominated as “hallmarks” of
aging have been defined in an effort to categorize the multiple biological processes that
manifest with age and modulate age-related phenotypes (López-Otín et al., 2023). These
include the following:
Genomic instability: Genomic integrity is indispensable for the health and survival of cells.
However, as a cell matures, it confronts an array of both internal and external threats, along with
various chemical agents that jeopardize the integrity of the genome (Moskalev et al., 2013). This
leads to genomic instability, and accumulation of genetic damage with age (Hoeijmakers, 2009;
Moskalev et al., 2013).
Genetic damage can occur in numerous forms, such as covalent modifications to DNA base
pairs, chromosomal rearrangements, errors in DNA replication, as well as single or double
strand breaks, to name a few (Hoeijmakers, 2009). This genomic damage can cause point
mutations, deletions, insertions, and expansions in specific genomic regions, potentially leading
to mutations that directly impinge upon the functionality of proteins controlling cellular processes
(Vijg & Dong, 2020).
Genomic instability can catalyze other hallmarks of aging, including cellular senescence, loss of
proteostasis (Huiting & Bergink, 2021; Lee et al., 2021), epigenetic changes (Yang et al., 2023),
and chronic inflammation (Kawanishi et al., 2017), each harboring its own downstream impacts.
2

Although genomic instability appears to be a significant factor influencing the aging process,
and many have proposed it as the primary driver of aging (Moskalev et al., 2013), it is crucial to
recognize the existence of numerous repair mechanisms within the cell that diligently work to
counteract genomic damage.
For instance, during embryogenesis, a single cell undergoes billions of divisions to produce a
new organism. This process demands a high metabolic rate, relying on oxidative metabolism,
which can generate oxidative molecules that could potentially damage DNA (Leese, 2012).
Moreover, the substantial level of DNA replication and chromosome segregation during this
developmental stage would result in elevated genomic aberrations if not for the presence of
robust molecular systems dedicated to preserving genomic integrity (Khokhlova et al., 2020).
This suggests that multicellular organisms inherently possess the necessary systems to
maintain genomic integrity, even under conditions that could increase the likelihood of genomic
damage. However, these systems seem to lose their effectiveness during the aging process.
This observation prompts intriguing questions regarding the biological and possibly evolutionary
reasons for this decline in efficiency.
Telomere attrition: Telomeres are specialized DNA-protein structures that cap the ends of
chromosomes, functioning as protective buffers to prevent the loss of vital genetic information
during cell division. However, as a result of cell division and DNA damage, these telomeres
shorten—a process known as telomere attrition (Olovnikov, 1996). When telomeres reach a
critically short length, the cell enters a state of senescence which effectively halts cell division
(Olovnikov, 1996). Telomere length decreases with age in multiple tissues and this process is
thought to play a crucial role in the gradual age-related decline in tissue regenerative potential
and function associated with increasing age (Blackburn et al., 2006; Demanelis et al., 2020).
Moreover, telomere shortening has been linked to various age-related diseases, indicating its
significant role in human health and lifespan (Gruber et al., 2021). Due to the correlation
3

between chronological age and telomere length, telomere length has been proposed as a
biomarker of biological age (Yadav & Maurya, 2022). However, the relationship between
chronological age and telomere length varies depending on the tissue examined and reports
have indicated a low correlation between telomere length and other age-related phenotypes
(Harris et al., 2012). An interesting observation is that telomeres tend to be longer in men
compared to women despite women having longer lifespan (Demanelis et al., 2020). It is also
notable that cells possess the ability to restore telomere length through the expression of
telomerase (Tomita, 2018), suggesting that their shortening with age is a reversible process that
the cell fails to counteract.
Epigenetic alterations: Epigenetics involves multiple mechanisms that affect gene expression
and do not involve alterations to the underlying DNA sequence. These changes, which include
processes like DNA methylation and histone modification, can significantly affect cellular
function (Allis & Jenuwein, 2016). As it pertains to aging, it's believed that cumulative epigenetic
alterations over time can lead to changes in gene expression patterns, contributing to the
physiological and functional decline observed in aging (Yang et al., 2023). DNA methylation has
gained a lot of attention due to the development of algorithms capable of predicting
chronological age in multiple tissues based on the methylation levels of specific genomic
regions (Field et al., 2018). These algorithms denominated epigenetic clocks are remarkably
interesting due to the accuracy of their predictions which is maintained across tissues and
species (Robeck et al., 2021). In addition, individual with higher epigenetic ages have been
found in multiple studies to be more susceptible to age related diseases (Fransquet et al., 2019;
Hillary et al., 2020; Liu et al., 2023). However, whether the methylation changes that occur in
tissues are causal or not of aging is still intensely debated. A noteworthy advancement in the
field has established a causal link between the methylation status of a specific genomic site —
utilized by multiple epigenetic clocks — and its adverse effects on tissues. This particular
4

genomic region regulates the expression of the ELOVL2 gene, which has been demonstrated to
influence the aging process in the retina and contribute to age-related macular degeneration
(Chao & Skowronska-Krawczyk, 2020).
Loss of proteostasis: Proteostasis, or protein homeostasis, refers to the cell's ability to
maintain a balanced and properly functioning set of proteins (Hartl et al., 2011; Mizushima et al.,
2008). This involves the precise regulation of protein synthesis, folding, trafficking, and
degradation. As organisms age, the efficiency and capacity of these processes can decline,
leading to what's known as a loss of proteostasis (Koga et al., 2011). This can result in the
accumulation of misfolded or damaged proteins, which can form aggregates that are toxic to
cells and can contribute to aging and age-related diseases, such as neurodegenerative
disorders (Koga et al., 2011). Therefore, the maintenance of proteostasis is crucial for cellular
health and longevity, and its disruption is recognized as a key hallmark of aging.
Disabled macroautophagy: Macroautophagy, often referred to as autophagy, is a cellular
recycling process that aids in the removal and degradation of damaged organelles and proteins
within the cell (Parzych & Klionsky, 2014). This process is essential for maintaining cellular
health and homeostasis. During autophagy, cellular components are enclosed within a double-
membrane vesicle, called an autophagosome, which then fuses with a lysosome, a cellular
organelle that contains enzymes for degradation (Parzych & Klionsky, 2014). The contents of
the autophagosome are then broken down and can be reused by the cell. As organisms age,
the efficiency of autophagy can decline, leading to the accumulation of damaged proteins and
organelles, which can in turn contribute to cellular dysfunction, aging, and age-related diseases
(Aman et al., 2021).
Deregulated nutrient-sensing: Deregulated nutrient sensing refers to the disruption of the
normal cellular mechanisms that sense nutrient availability and respond appropriately to ensure
cellular survival and growth. Key nutrient-sensing pathways include the insulin/IGF-1 signaling
5

pathway, mTOR, and AMPK, among others. These pathways play crucial roles in metabolism,
growth, and longevity, and their deregulation can lead to a host of health issues (López-Otín et
al., 2023). As organisms age, these nutrient-sensing pathways often become deregulated,
contributing to various age-associated diseases, such as diabetes, cancer, and
neurodegenerative disorders (Pignatti et al., 2020). Nutrient sensing pathways are amongst the
most well conserved longevity extending pathways and interventions that modulate them such
as caloric restriction or rapamycin have robust lifespan extending properties (Fontana et al.,
2010).
Mitochondrial dysfunction: Mitochondrial dysfunction refers to the impairment of normal
mitochondrial function, which plays a key role in aging (An et al., 2012). Mitochondria are often
referred to as the 'powerhouses of the cell' because they generate most of the cell's supply of
adenosine triphosphate (ATP), used as a source of chemical energy. They also play a role in
other important processes, such as signaling, cellular differentiation, cell death, as well as the
control of the cell cycle and cell growth (Antico Arciuch et al., 2012; Seo et al., 2018). However,
with age, these crucial organelles can become less efficient or even dysfunctional due to
damage from reactive oxygen species (ROS), loss of mitochondrial DNA integrity, or defects in
the proteins needed for proper mitochondrial function (Chistiakov et al., 2014). This dysfunction
can lead to a decrease in cellular energy production and an increase in harmful oxidative stress,
both of which can contribute to the physiological decline associated with aging (López-Otín et
al., 2023).
Cellular senescence: Cellular senescence is a cellular state in which cells lose their ability to
proliferate and secrete a variety of substances like pro-inflammatory cytokines, growth factors,
and proteases — an occurrence known as the Senescence-Associated Secretory Phenotype
(SASP) (Campisi & d'Adda di Fagagna, 2007). This can trigger local and systemic inflammation,
disrupt the microenvironment of surrounding tissues, and foster the development of age-related
6

diseases (Huang et al., 2022). Cellular senescence is often perceived as a double-edged sword.
On one side, it serves as a defense mechanism against cancer by inhibiting the multiplication of
damaged cells. On the flip side, the buildup of senescent cells over time can contribute to tissue
dysfunction and a range of age-related pathologies (Huang et al., 2022). Moreover, senescent
cells and the SASP are also required for various critical physiological processes (Demaria et al.,
2014; Grosse et al., 2020). Senescent cells can emerge in response to multiple types of stress.
While there is a clear phenotypic definition for senescent cells, recent research indicates that
these cells can exhibit a wide array of phenotypes depending on the cell type, the inducer of
senescence, and the environment in which they reside (Faraonio et al., 2002; Wiley et al.,
2016).
Stem cell exhaustion: Stem cell exhaustion refers to the age-related decline in the number and
function of stem cells, the undifferentiated cells in our bodies with the capacity to divide and
differentiate into various cell types (Brunet et al., 2023). Stem cells play a crucial role in
maintaining tissue homeostasis by replenishing lost or damaged cells throughout life. However,
with aging, stem cells face a reduction in their regenerative potential and a diminished capacity
to respond to stress, injury, or disease (Gruber et al., 2006; Molofsky et al., 2006; Shaw et al.,
2010). This is due, in part, to the intrinsic aging of stem cells, but is also influenced by age-
related changes in the stem cell 'niche'—the local tissue microenvironment that provides
support and regulatory signals (Brunet et al., 2023). Stem cell exhaustion contributes to the
decreased ability of tissues to repair and regenerate in older age, leading to tissue degeneration
and functional decline, hallmarks of aging.
Altered intercellular communication: Altered intercellular communication refers to changes in
the way cells communicate with each other as we age. This includes changes in both direct cell-
to-cell interactions and indirect communication through signaling molecules (Fafián-Labora &
O’Loghlen, 2020). Hormonal signaling is an example of this as it can become dysregulated with
7

age, affecting processes such as metabolism and immune response (Hall, 2007). Furthermore,
aging can alter the communication between different types of cells, such as neurons and glial
cells in the brain, or muscle cells and nerve cells in the body (Cleeland et al., 2019). These
changes in intercellular communication can disrupt tissue function and contribute to the
development of age-related diseases.
Chronic inflammation: Chronic inflammation, also referred to as 'inflammaging', is recognized
as a major hallmark of aging. This term describes the persistent, low-grade inflammation that
develops with advanced age (Salminen et al., 2012). Unlike acute inflammation, which is a
healthy and necessary response to injury or infection, chronic inflammation can lead to tissue
damage. It is thought to be driven by various factors such as cellular senescence, mitochondrial
dysfunction, and immune system dysregulation (Ferrucci & Fabbri, 2018). Chronic inflammation
can affect almost all tissues and organs in the body, and it is linked to a number of age-related
diseases, including heart disease, diabetes, cancer, and neurodegenerative diseases (Ferrucci
& Fabbri, 2018).
Dysbiosis: Dysbiosis refers to an imbalance or maladaptation in the microbiota, the trillions of
microorganisms living in and on our bodies, particularly in the gut. These microorganisms play
crucial roles in digestion, immune function, and even the production of certain vitamins. As we
age, changes occur in the composition and function of the microbiota, leading to dysbiosis
(Haran & McCormick, 2021). This altered microbiota composition can influence the host's health
and has been associated with a number of age-related conditions, including inflammatory bowel
disease, obesity, cardiovascular disease, and even neurodegenerative disorders (Alam et al.,
2021; Bosco & Noti, 2021; Haran & McCormick, 2021; Ragonnaud & Biragyn, 2021). Thus,
dysbiosis is being increasingly recognized as a significant player in the aging process and its
related pathologies.
8

The hallmarks of aging are intricately interconnected, creating a complex network of biological
changes that collectively contribute to the aging process (López-Otín et al., 2023). A disruption
in one area can propagate throughout the system, leading to numerous downstream effects. For
example, genomic instability can lead to mutations and errors in DNA, which can cause cells to
become senescent or initiate apoptosis. This, in turn, can lead to a decrease in the overall
number of functional cells in tissues (stem cell exhaustion), reducing the body's ability to repair
damage and maintain homeostasis (Molofsky et al., 2006).
The intricacy and interconnected nature of these hallmarks of aging pose substantial difficulties
in comprehending the precise causal relationships and sequence of events involved in the aging
process. Aging is a complex process likely resulting from the cumulative influence of these
hallmarks, along with potentially undiscovered ones. For instance, recent studies have found
that as cells age, their ability to produce genes with lengthy coding sequences diminishes. This
leads to a bias towards the production of genes with shorter coding sequences, which appears
to result in alterations in the genes responsible for systems maintaining homeostasis (Stoeger et
al., 2022). Therefore, pinpointing the root cause or primary initiator of aging remains an ongoing
and challenging area of research.
Aging across multiple species
Aging manifests differently on different organisms, however, the time dependent decline in
organismal fitness seems to be a widely shared feature. For instance, unicellular organisms
have been thought to evade aging through their symmetrical cell division. This process was
believed to dilute molecular damage across divisions, effectively staving off the accumulation of
molecular damage (Florea, 2017). However, recent evidence challenges this notion. Studies on
prokaryotes that divide symmetrically, such as E. coli, show that the cell inheriting the old pole
after division has a reduced growth rate, lower offspring production, and an increased risk of
death (Stewart et al., 2005). Similar observations have been made in symmetrically dividing
9

yeast (Spivey et al., 2017) and the time dependent decline in cellular fitness is also seen under
lack of cell division (Nyström, 2003). These findings suggest that even in primitive organisms, a
time-dependent decline in cellular fitness exists. It also implies that cellular systems may lack
the ability to indefinitely repair and maintain their molecular components, suggesting that the
synthesis of new molecular components during cell division is essential for maintaining cellular
fitness.
Multicellular organisms present a solution to the issue of cellular aging by virtue of their multi-
cellular structure. Given the multitude of cells in these organisms, one might anticipate that the
decline in cellular fitness, could be mitigated through a continuous replacement of cells
containing newly synthesized molecular components free from damage. In fact, a handful of
multicellular species have accomplished this feat and exhibit no signs of aging. Among these
are various species of planarian, which are considered immortal and exhibit limitless
regenerative capacity, enabling them to constantly renew their cells (Sahu et al., 2017).
Similarly, some species of Hydra possess a similar process in which old cells are constantly
discarded and replaced by new cells generated from a self-renewing population of stem cells
(Muller, 1996). However, complex organisms do not display these capacities.
In vertebrates, despite the existence of very long-lived organisms, there is evidence for a similar
mechanism granting biological immortality. However, interesting patterns and exceptions in
organismal aging can be observed when we look at the varying maximum lifespans among
different species of vertebrates.
For example, there is a well-established correlation between the maximum lifespan of a species
and its adult body mass. Larger animals like elephants and whales tend to have longer
maximum lifespans compared to smaller animals (Speakman, 2005). This relationship,
however, is not without its exceptions or outliers. Some species live significantly longer than
10

would be expected for their body size. Notable examples include multiple species of bats, the
naked mole rat, and humans (Zhao et al., 2021).
Another interesting relationship is the correlation between the rate of accumulation of DNA
mutations and maximum lifespan which has shown that even species with very different body
mass but similar maximum lifespans, such as naked mole rats and giraffes, have similar
mutation rates (Cagan et al., 2022). Similarly, other correlations between the methylation rate
(Crofts et al., 2023) and protein turnover rate (Swovick et al., 2021) of different species have
been linked to maximum lifespan.
Human Aging
Humans are the longest-lived land mammal with a maximum recorded lifespan of 122 years
(Blagosklonny, 2021). Despite their longevity, humans face numerous physiological changes
during aging that impact overall body function. These changes include a decline in respiratory
efficiency through alveoli dilation, airspace enlargement, and reduced gas exchange surface
area, and diminished kidney functions characterized by lower filtration rates and renal blood
flow, increasing the risk for kidney diseases (Boss & Seegmiller, 1981; Denic et al., 2016;
Sharma & Goodwin, 2006). Similarly, the skin's protective capabilities wane due to decreased
cell turnover and a reduced vascular network, while aging also results in diminished arterial
elasticity, increased arterial stiffness, and a lowered heart rate in the cardiovascular system,
leading to heightened disease susceptibility (Boss & Seegmiller, 1981; Cheitlin, 2003).
Furthermore, aging negatively influences immune efficiency, gastrointestinal motility, blood
glucose levels, bone mass, muscle mass, and joint health (Boss & Seegmiller, 1981; Sharma &
Goodwin, 2006).
Multiple of these physiological changes are common amongst mammals and other vertebrates.
However, there are certain features of human aging that are rare in the animal kingdom. For
instance, menopause is a natural biological process that marks the end of a woman's menstrual
11

cycles and reproductive years. This process has been documented in only three species:
humans, short-finned pilot whales and killer whales (Brent et al., 2015). In humans, menopause
usually occurs between the ages of 45 and 55 and is accompanied by various symptoms such
as hot flashes, sleep disturbances, mood changes, and vaginal dryness (Takahashi & Johnson,
2015). It also results in decreased levels of estrogen, a hormone that helps protect against heart
disease and osteoporosis, hence increasing the risk for these conditions (Takahashi & Johnson,
2015).
Another unique aspect of human aging is the increased susceptibility to certain
neurodegenerative diseases, such as Alzheimer's. This progressive disease, the most common
cause of dementia in older adults, is characterized by memory loss, cognitive difficulties, and
behavioral changes (Knopman et al., 2021). Alzheimer's is believed to be caused by the buildup
of plaques and tangles in the brain, leading to brain cell death and brain shrinkage (Knopman et
al., 2021). As the disease progresses, symptoms intensify, disrupting daily activities and
eventually impairing conversational and environmental responsiveness. While the exact cause
of Alzheimer's remains uncertain, the accumulation of amyloid-beta plaques in the brain is a
leading hypothesis (Selkoe & Hardy, 2016). Notably, while other primates also accumulate
these plaques with age, they don't develop Alzheimer's-like dementia (Heuer et al., 2012).
Considering aging is the primary risk factor for Alzheimer's, understanding the unique elements
of human brain aging that contribute to the disease is crucial.
Brain Aging
Aging in the human brain is a complex process that involves both structural and functional
changes. As we age, our brains naturally shrink in volume, a phenomenon known as brain
atrophy. This shrinkage is particularly notable in the prefrontal cortex and hippocampus, areas
involved in higher cognitive functions and memory respectively (Mattson & Arumugam, 2018).
Additionally, the brain's white matter, which is crucial for efficient communication between
12

different brain regions, undergoes age-related changes, often leading to compromised integrity
(Mattson & Arumugam, 2018). The aging brain also experiences alterations in neurotransmitter
systems, which can affect memory and learning (Karst et al., 2020; Rozycka & Liguz-Lecznar,
2017).
The hippocampus is a focus of study in brain aging due to memory loss being one of the main
symptoms of cognitive decline and the important role the hippocampus plays in memory
formation. The hippocampus contains four distinct regions: the dentate gyrus, the hippocampus
proper, the subiculum and the entorhinal cortex (Knierim, 2015; O'Mara et al., 2001). Loss of
neurons in the human hippocampus with age has been detected with unbiased stereological
methods. Identifying a loss of neurons in the subiculum and in the hilus of the dentate gyrus
(West, 1993). Similar phenotypes have also been observed in non-human primates and rodents
(Rizzo et al., 2015).
In addition to neuronal loss, the hippocampus is also affected by age-related alterations in neuro
transmission and calcium homeostasis (Stephens et al., 2011). Long term potentiation is
phenomenon characterized by an increase in synaptic transmission in response to intense
synaptic activity. Aging decreases the capacity for long term potentiation in aged animals and
correlates with lower memory capacity (Bliss et al., 2003). In contrast with long term
potentiation, long term depression is a process that decreases synaptic transmission. Long term
depression was initially discovered while studying brain development and was thought to be
involved in the shaping of neuronal circuits as the brain matures. However, it has now been
stablished that there is an age-related increase in susceptibility towards long term depression
(Norris et al., 1996). The shift from long term potentiation to long term depression with aging is
thought to contribute age related cognitive decline, promoting decreased synaptic strength in
the hippocampus and weakening neuronal circuits. The mechanisms driving these changes at
the synapse level are not well understood but seem to be driven by changes in calcium
13

dysregulation (Pereda et al., 2019). Interestingly, it has been recently shown that whole body
clearance of senescent cells can partially prevent changes in synaptic function in the
hippocampus (Budamagunta et al., 2023).
Senescent cells and hippocampal function
As previously detailed, senescent cells are those which have experienced a form of molecular
damage, leading to a permanent cessation of cell division and the development of a
senescence-associated secretory phenotype (SASP). The emergence of senescent cells can be
attributed to various factors, including extensive DNA damage, oncogenic mutations, and
epigenetic alterations (Campisi & d'Adda di Fagagna, 2007).
Due to the SASP, senescent cells can exert influence not only on nearby cells but also on
distant tissues. This is achieved through the secretion of SASP factors into the bloodstream,
amplifying their reach and impact.
Various markers have been developed to aid in identifying senescent cells. Among the most
widely used is the expression of senescence-associated beta-galactosidase activity at pH 6.0,
which helps in distinguishing senescent cells in culture and in mammalian tissues (Dimri et al.,
1995). Several other markers have been identified, such as the expression of cell cycle arrest
markers (p16 and p21), a decrease in specific proteins of the nuclear lamina (laminb1),
secretion of SASP markers, and intracellular relocalization of HMGB1 (Davalos et al., 2013;
Huang et al., 2022).
While numerous markers of senescent cells have been identified, most of this work has been
conducted on a limited variety of cell types and under specific cell culture conditions. It is
increasingly evident that there is no universal marker of senescence, as senescent cells
demonstrate varied characteristics depending on the tissue context and cell type. Nevertheless,
numerous studies have used multiple senescence markers to successfully identify senescent
cells in tissues.
14

For instance, research conducted on the aging hippocampus purports to have detected
senescent cells, using abnormal nuclear morphology and diminished levels of laminB1 as
indicators (Matias et al., 2022). However, given that the SASP can affect distant tissues via the
bloodstream, the actual presence of senescent cells within a specific tissue is not essential for
triggering dysfunction.
In fact, studies utilizing various senolytics, each with different abilities to traverse the blood-brain
barrier, have demonstrated that clearing senescent cells from peripheral tissues can counteract
age-related hippocampal phenotypes (Budamagunta et al., 2023). These phenotypes include
maintenance of blood-brain barrier integrity, and prevention of age-associated degradation in
synaptic function within the hippocampus.
A similar study discovered a surge in hippocampal neurogenesis following the elimination of
senescent cells (Fatt et al., 2022). Likewise, clearing senescent cells has been observed to
reduce inflammation within the hippocampus and to minimize the activation of microglial cells,
further underscoring the potentially beneficial effects of targeting aging processes (Ogrodnik et
al., 2021). In chapter 7 of this dissertation, I present work characterizing the transcriptomic
signatures of the different cell types of the hippocampus as a result of ageing and how they
change after clearance of senescent cells.
When deliberating about the presence of senescent cells in the brain, it is commonly assumed
that these cells are of glial origin. Nonetheless, there is emerging evidence suggesting that
neurons may also exhibit features reminiscent of senescence (Herdy et al., 2022). To illustrate,
in chapter 6 of this dissertation, I present work showing how a model medium spiny neurons
harboring the mutation responsible for Huntington’s disease present characteristics similar to
those of senescent cells.
15

Huntington’s Disease
Huntington's disease (HD), an inherited, invariably fatal neurodegenerative disorder,
predominantly afflicts neurons in the striatum and cortex (Cudkowicz & Kowall, 1990; Hedreen
et al., 1991). The disease arises from an expansion in the number of CAG repeats in the
huntingtin gene, leading to an expanded polyglutamine segment in the produced HTT protein.
Symptoms typically emerge in individuals with over 38 CAG repeats, severely affecting motor
control, cognitive function, and emotional stability (Group, 1993).
The mutant HTT protein primarily affects medium spiny neurons (MSNs), leading to their
dysregulation and eventual cell death, which ultimately impairs motor control (Barnat et al.,
2020; Braz et al., 2022; Cirnaru et al., 2021; Cirnaru et al., 2019; Kim et al., 2022; Mehler et al.,
2019; Molero et al., 2016; Molero et al., 2009; Ring et al., 2015; Rodríguez-Urgellés et al., 2022;
Zhang et al., 2016). Despite HD being genetic, its symptoms usually don't appear until middle
age, hinting at an age-related factor contributing to disease pathology. Interestingly, multiple
reports suggest that pathology observed later in life may hinge on early life dysregulation
caused by the mutant protein.
Experiments on mice have shown that temporary expression of the mutant HTT protein during
development can cause neurodegeneration and HD-like symptoms in later life (Molero et al.,
2016). Similarly, reducing the non-mutant huntingtin levels during development can cause HD-
like motor abnormalities in adulthood (Arteaga-Bracho et al., 2016). Furthermore, a recent study
has shown that addressing the circuit abnormalities caused by HD during development can
delay the disease onset and ameliorate pathology (Braz et al., 2022). Therefore, the relatively
uncharted territory of early life molecular events in HD holds promise. It offers a potential
pathway to identify molecular events that can modify the disease's neuropathological trajectory
and pave the way for efficacious therapeutic interventions. Thus, I present work in chapters 3, 4
16

and 5 characterizing molecular events during MSN development that caused by the HD
mutation as well as identifying potential corrective interventions.
Omic Data to Study Brain Aging and Neurodegenerative Diseases
The quest for therapies to treat neurodegenerative disorders and mitigate the effects of brain
aging has taken on critical importance, in response to the rising number of elderly individuals in
populations worldwide (Hou et al., 2019). Innovations in bioanalytical technology have equipped
us with the means to scrutinize the molecular constituents of biological systems with
unprecedented precision and depth. Tools such as microarray technology and RNA sequencing
offer the capability to simultaneously analyze the RNA levels of thousands of genes in parallel
(Dai & Shen, 2022). Additionally, these methods have been adapted to assess epigenetic
modifications across the genome (Mehrmohamadi et al., 2021). Concurrently, advances in mass
spectrometry have facilitated the quantitative assessment of a vast array of proteins within a
single biological specimen (Meissner et al., 2022). More recently, these technologies have
made significant strides towards achieving reliable single-cell resolution, enabling us to examine
the molecular makeup of individual cells within various contexts (Vandereyken et al., 2023). This
progression holds great promise for the characterization and understanding of cellular
heterogeneity and function in health and disease.
In this dissertation, I present my work on exploring the molecular changes in the aging
hippocampus driven by senescent cells as well as studying models of Huntington's disease
(HD) using making use of various omics technologies, including bulk and single cell RNAseq,
mass spec based proteomics, and single nuclei RNAseq.
Chapter 3 focuses on the impact of the Huntington's mutation on neuronal development. Here, I
employed single-cell and bulk RNAseq data from an induced pluripotent stem cell (iPSC) model
of HD MSNs. This approach allowed me to identify transcriptional modifiers of the disease and
demonstrate improvements in several MSN health markers.
17

Chapter 4 centers on a quantitative proteomics analysis of the same HD MSN model. My
contribution to this project entailed bioinformatics analysis to compare our model's data with
proteomic data from other HD datasets. I also identified small molecules that could potentially
reverse protein-level changes and conducted preliminary experiments to delineate the
differences in lipid droplets between HD and control MSNs.
In Chapter 5, I partook on a study centered around the transcription factor BCL11B, whose
deletion during adulthood produces phenotypes akin to HD. My role was to characterize
BCL11B in iPSC-derived HD MSNs. We discovered that in HD MSNs, BCL11B forms granules
in the nucleus, potentially impacting its function and suggesting a potential link between this
transcription factor and HD phenotypes.
Chapter 6 outlines a project where we identified senescent-like characteristics in iPSC-derived
HD MSNs. My primary role was to characterize this phenotype by examining various
senescence markers in this model.
Chapter 7, the final segment of my dissertation, presents a comprehensive investigation that I
spearheaded. This research focused on deciphering age-associated transcriptional alterations
within the hippocampus that are driven by senescent cells. I utilized single-nuclei RNA
sequencing on hippocampal samples from young and aged mice, and two cohort of mice
subjected to senescent cell ablation. This resource will help understand age-related
transcriptional changes driven by senescent cells in the hippocampus's cell types and their
biological implications. As the project leader, I was integrally involved in every facet of this
study, from animal work to initial data generation and data analysis.

18

CHAPTER 2: Materials and methods
Experimental procedures for chapter 3
NSC and developing MSN Differentiation: NSCs and MSNs were differentiated as described
in [38]. Briefly, HD72 and C116 iPSCs were turned into cell aggregates and driven towards a
neuroepithelial fate that produced neural rosettes. Neural rosettes were manually picked and
dissociated into single cells to produce NSCs. NSCs were then expanded in Neurobasal
medium (Thermo Fisher Scientific, 21103049) supplemented with B27-supplement 1 X (Thermo
Fisher Scientific, 17504001), GlutaMAX 1 X (Thermo Fisher Scientific, 35050061), 10 ng/mL
leukemia inhibitory factor (PeproTech, 300-05), and 100 U/mL penicillin-streptomycin.
To produce developing MSNs, NSCs were cultured in Synaptojuice A medium supplemented
with 25 ng/mL of Activin A (PeproTech, AF-120-14E) while changing half of the medium every
other day. After 7 days, the cells were switched to Synaptojuice B medium supplemented with
25 ng/mL of Activin A for 14 days. The cells were harvested or used for assays on day 20

of
differentiation.
Cerulenin Treatment: 2.23 mL of DMSO (Sigma, D2650-5X10mL) were added to 1 mg of
Cerulenin (Cayman Technical, No. 10005647) to make a stock solution of 2 mM. Cerulenin was
diluted in Synaptojuice B to the desired concentration and added to the cells starting at day 11
of differentiation and every time the medium was changed until the cells were harvested. The
cells were used for immunocytochemistry or RNAseq on day 20

of differentiation. For use with
MEA plates, the cells were differentiated for 35 days and cerulenin was added every time the
medium were changed after day 11 of differentiation.
Multielectrode Array Measurements: Cytoview MEA plates (Axion Biosystems, M384-tMEA-
24W) were coated with 1% poly(ethyleneimine) solution (Sigma-Aldrich, 03880-500mL)
overnight at 37 °C. The next day, each well was rinsed with cell-culture grade water 3 times and
allowed to dry for 2 hours in a cell-culture hood. Once the plates were dry, they were coated
19

with 0.25mg/mL laminin (Sigma-Aldrich, L2020-1MG) in DPBS for 3 hours at 37 °C. NSCs were
seeded at a density of 49,586 cells/mm
2
, and they were differentiated as described above until
day 20. After day 20, half the medium was changed every 7 days, and readings were taken at
day 35 of differentiation. The cells were placed inside a Maestro Edge (Axion Biosystems)
instrument with temperature set at 37 °C and 5% CO 2. Recordings were taken for 5 minutes,
and the last minute was utilized for analysis. Electrical activity was assessed using a spike
detector with adaptive thresholding set to 4.5 standard deviations. Metrics for electrical activity
were generated using Axis Navigator and Neural Metric Tool software (Axion Biosystems).
Immunofluorescence: Cells were rinsed with PBS and then fixed in 4% paraformaldehyde in
PBS for 12 minutes at room temperature. Fixation was followed by three washes of PBS for 5
minutes each. Fixed cells were blocked using blocking buffer containing 0.1% Triton-x-100
(Thermo Fisher Scientific, 28313), 4% normal donkey serum (Jackson Immuno Research, 017-
000-121) in PBS for 1 hour. Fixed cells were then incubated overnight with primary antibodies
diluted 1:100 in blocking buffer. Cells were washed three times with PBS containing 0.1%
Triton-x-100 and incubated with secondary antibodies diluted 1:500 in blocking buffer for 2
hours. Cells were washed 3 times for 5 minutes with PBS containing 0.1% Triton-x-100. The
cells were finally placed in PBS and imaged utilizing a Cytation 5 (Biotek). The antibodies used
were mouse anti-DARPP-32 antibody (Santa Cruz Biotechnologies, sc-271111) paired with
donkey-anti-mouse Alexa-647 (Thermofisher, A-31571), Nestin (abcam, ab92391) paired with
donkey-anti-rabbit Alexa-594 (Thermofisher, A-21207), Nestin (Santa Cruz Biotechnologies, sc-
33677) paired with goat-anti-mouse Alexa-488 (Thermofisher, A11006), and SOX2 (Cell
Signaling, 14962S) paired with goat-anti-rabbit Alexa 555 (Thermofisher, A21428).
Bulk RNAseq Processing and Analysis: RNA was extracted from MSNs utilizing an RNA
extraction kit (Bioline, BIO-52073). RNA library preparation was prepared at the UC Davis
Genomics Core or Novogene, utilizing a poly-A library prep. Resulting FASTA files were then
aligned utilizing STAR 2.7.10a to the GRCh38 primary assembly genome reference. Features
20

were counted with the summarizeoverlaps function part of the GenomicAlignments R package
[104]. Differential expression was performed with DESeq2 [105], and volcano plots were done
with ggplot2 . DEGs (adjusted p value < 0.01, mean base > 10 and absolute log2 fold-change >
0.1) were used as input for IPA version 76765844 to obtain prediction of canonical pathways.
The p-values for enrichment analysis were corrected using the Benjamini-Hochberg correction.
Enrichr was used to identify enrichment from the bioplanet database [106, 107]. To compare the
DEGs in HD MSNs with those in other HD models, we obtained DEGs from [39-43] and
performed a Fisher’s exact test to compare HD MSN DEGs with each of the HD models and
adjusted p-values using Benjamini-Hochberg correction. Leafcutter 0.2.9 was used to detect
differentially spliced genes in C116 and HD MSNs.
scRNAseq Processing: NSCs and MSNs were cultured in six well plates coated with 50 µg/mL
of Matrigel (Corning) overnight. Cells were treated with Accutase (Thermo Fisher, A1110501)
for 5 minutes, and the cells were centrifuged and resuspended in DPBS. Cells were
encapsulated, barcoded and transformed into libraries utilizing a Chromium Next GEM Single
Cell 3ʹ Kit v3.1. Resulting libraries were sent for sequencing to the UC Davis genomics core and
sequenced in a Novaseq 6000 lane. FASTQ files were aligned to the genome and used to
generate gene counts utilizing cell ranger 6.0 and the GRCh38 reference genome. The Seurat
workflow was used for filtering, normalization, scaling, PCA, UMAP generation, and clustering.
The Seurat objects and C116 and HD datasets were integrated for visualization using Harmony
[108]. To identify differentially expressed genes, we utilized the function FindMarkers and used
MAST as the differential expression test. The results were used as input for IPA, filtering for
genes with an adjusted p-value under 0.05 and an absolute log2-fold change larger than 0.25.
The set of detected genes was used as background for IPA and p-values for enrichment were
adjusted using the Benjamini-Hochberg correction.
Comparison of MSN and LGE scRNAseq data: We obtained the raw data from the LGE
dataset, which was originally published in Bocchi et al [31], by downloading it from the
21

ArrayExpress database (www.ebi.ac.uk/arrayexpress/) under accession number E-MTAB-8894.
FASTQ were processed as described above. The LGE dataset was down sampled using the
subset function from the Seurat package so that it would match the number of cells found in the
MSN datasets. Clusters of cells not positive for MEIS2, indicating a non LGE lineage, were
removed. C116 and HD developing MSNs were integrated to the LGE dataset using Harmony
[108]. Trajectory inference and pseudotime were produced with VIA [60].
Drug Prediction and MOA Inference: DEGs identified in developing HD and C116 MSNs via
Bulk RNAseq were used as input for the L1000CDS2 web application. The log2-fold change of
the genes was used to indicate magnitude. Whether the predicted small molecules had already
been used in HD models and their canonical targets were determined doing a manual literature
search on PubMed. The predicted MOAs for each small molecule were obtained from the
L1000FWD web application. The probability for each MOA in each small molecule signature
was used to generate a tSNE plot using the function Rtsne from the Rtsne package. We
manually colored small molecules in the tSNE that had a top MOA prediction with more than a
40% probability.
Quantitative PCR: RNA extraction was carried out using the RNeasy Plus Mini Kit (QIAGEN,
74034), followed by its conversion into cDNA with the SensiFAST cDNA Synthesis Kit (Bioline,
BIO-65053). qPCR was conducted using the SensiFAST Probe No-ROX Kit (Bioline, BIO-
86005), UPL probes (Roche) and primers listed in Supplemental Table 3.8, on the LightCycler
480 system (Roche). For quantification purposes, the threshold cycle (Cp) for each amplification
was identified using the 2nd derivative analysis provided by the LightCycler 480 software.
Subsequently, the 2-ΔΔCp method was applied to calculate the relative expression levels of
individual genes, which were normalized against the housekeeping gene β-actin (ACTB).
Experimental procedures for chapter 4
Human Induced Pluripotent Stem Cell–Derived NSC Cultures: C116 and HD72 iPSCs were
maintained in mTeSR™1 (STEMCELL Technology, 05850) medium before the differentiation.
22

To induce iPSCs toward a neuroepithelial fate, we used a monolayer differentiation approach
with modifications. Briefly, iPSCs were manually cleaned by removing any colonies with
spontaneous differentiation. To initiate differentiation (Day 0), iPSCs were passaged with 1
mg/mL collagenase [(Type IV, Thermo Fisher Scientific, 17104019) in Gibco KnockOut
DMEM/F-12 medium (Thermo Fisher Scientific, 12660012)] for 35 min at 37°C. The colonies
were gently detached by scraping, and the cell aggregates were triturated by pipetting 2–3 times
with a 2-mL pipette to yield a uniform suspension of aggregates and avoiding creating a single-
cell suspension. The cell aggregates were transferred onto a Matrigel (1 mL, 50 µg, Corning,
CB-40234)-coated plate containing mTeSR™1 and incubated at 37°C. For neural induction
(Day 2), SMAD signaling was inhibited by adding SB431542 (10 µM, Tocris, 1614) and LDN-
193189 (1 µM, Tocris, 6053) in mTeSR™1 during the medium changing, thus promoting
neuroectodermal differentiation and suppressing mesoderm and endoderm fates (40). From day
4, the mTeSR™1 medium was changed every day in presence of SB431542 (10 µM, Tocris,
1614) and LDN-193189 (1 µM, Tocris, 6053). From day 8, the colonies became organized and
increased in size. The center of colonies became dense and compact, and the peripheral
regions presented elongated cells. At day 10, the colonies were first cleaned to remove
peripheral regions, and then the dense center regions were manually picked by scraping after a
collagenase treatment of 25 min at 37°C, to avoid over-collagenase. Cell aggregates from the
center regions were transferred at low density to minimize merging, into a 10-cm Matrigel (1 mL,
50 µg, Corning, CB-40234)-coated plate containing N2B27 medium [(DMEM/F12, Gibco,
Thermo Fisher Scientific, 11320-033) supplemented with 1 X N2 (Thermo Fisher Scientific,
17502001), 1 X B27 (Thermo Fisher Scientific, 17504001), 1 X GlutaMAX (Thermo Fisher
Scientific, 35050061), 1 X Non-Essential Amino Acids (Thermo Fisher Scientific, 11140050), 25
ng/mL β-FGF (Peprotech, 100-18B) and 100 U/mL penicillin-streptomycin (Thermo Fisher
Scientific, 15140122)]. The cells were cultured in presence of 25 ng/mL Activin A (PeproTech,
AF-120-14E) to induce regional patterning toward a lateral ganglionic eminence (LGE) identity.
23

The N2B27 medium was changed every 2 d in presence of Activin A and β-FGF. At day 12,
neuroepithelial differentiation became apparent with the formation of small neural rosettes
showing a columnar shape that further organized and increased in size at day 14 in forming
neural tube-like structures with a central lumen and three-dimensional growth. Thereafter, the
neural rosette structures were mechanically selected by separating the island from the
surrounding cells with a needle to minimize contamination with non-neural cells. The isolated
rosettes were triturated by pipetting 8–10 times with 1,000-µL pipette tip. At least 15–20 neural
rosettes/well were plated in a Matrigel-coated P12 well plate in presence of Neural Proliferation
Medium [Neurobasal medium (Thermo Fisher Scientific, 21103049), B27-supplement 1 X
(Thermo Fisher Scientific, 17504001), GlutaMAX 1 X (Thermo Fisher Scientific, 35050061), and
10 ng/mL leukemia inhibitory factor (PeproTech, 300-05), 100 U/mL penicillin-streptomycin]
supplemented with 25 ng/mL β-FGF and 25 ng/mL Activin A. The resulting NSCs were
passaged when cell cultures became confluent. The passaging cells were moved gradually from
P6, P12 wells to 6-cm plates with the cells plated at a high density. Nestin, SOX1, SOX2, and
PAX6 staining of NSCs validated the cell type.
MSN Differentiation: Activin A (25 ng/mL, PeproTech, AF-120-14E)-generated C116 and
HD72 NSCs were used to prepare MSNs. Nunc six-well plates were treated with poly-D-lysine
hydrobromide (1 mL, 100 µg/mL by Sigma Aldrich, P6407) and incubated (37 C and 5% CO2)
overnight (ON). Corning cell-culture grade water, 25-055-CVC (1 mL), was used to wash plates,
and the plates were dried for 1 h. Next, the plates were treated with Matrigel (1 mL, 50 µg,
Corning, CB-40234) overnight in a 37°C incubator. MSNs were prepared according to Kemp et
al. (30). Synaptojuice A medium (2 mL) was used for seeding NSCs (1x10
6
per well).
Synaptojuice A was prepared with 10X synaptojuice A supplement (5 mL), advanced
DMEM/F12 medium (44.1 mL, Gibco, 12634010), penicillin/streptomycin (P/S) (450 µL,
Invitrogen, 15140122), and 100X Glutamax (450 µL, Invitrogen, 35050079). Synaptojuice A
24

supplement (10X) contains advanced DMEM/F12 medium (38 mL, Thermo Fisher Scientific,
12634010), MACS NeuroBrew-21 with retinoic acid with final concentrations noted (10 mL,
MACS Miltenyi Biotec, 130-093-566), PD0332991 (20 µM, Tocris Bioscience, 4786), DAPT (100
µM, Tocris Bioscience, 2634), human brain-derived neurotrophic factor (BDNF) (100 ng/mL,
MACS Miltenyi Biotec, 130-096-286), LM22A4 (5 µM, Tocris Biotec, 4607), forskolin (100 µM,
Tocris Bioscience, 1099), CHIR 99021 (30 µM, Tocris Bioscience, 1099), GABA (3 mM, Tocris
Bioscience, 0344), CaCl2 (1.8 mM, Tocris Bioscience, 3148), ascorbic acid (2 mM, Tocris
Bioscience, 4055). Medium was passed through a 0.22-µm filter. Cells were treated with
synaptojuice A (2 mL) for 7 days, performing half-medium changes every other day. On day 8,
full medium changes were completed, and then, the cells were treated with synaptojuice B (2
mL) for the next 7 days. Synaptojuice B was prepared with 10X synaptojuice B (5 mL)
supplement and basal medium (45 mL). Basal medium contains advanced DMEM/F12 medium
(22.5 mL, Thermo Fisher Scientific, 12634010), penicillin/streptomycin (P/S) (450 µL, Invitrogen,
15140122), and 100X Glutamax (450 µL, Invitrogen, 35050079) and Neurobasal A Medium
(22.5 mL, Gibco, 10888022), penicillin/streptomycin (P/S) (450 µL, Invitrogen, 15140122), and
100X Glutamax (450 µL, Invitrogen, 35050079). Synaptojuice B supplement contains advanced
DMEM/F12 medium (19.7 mL, Thermo Fisher Scientific, 12634010), Neurobasal A medium
(19.7 mL, Gibco, 10888022), MACS NeuroBrew-21 with retinoic (10 mL, MACS Miltenyi Biotec,
130-093-566), PD0332991 (100 µL, 20 µM, Tocris Bioscience, 4786), human BDNF (50 µL, 100
ng/mL, MACS Miltenyi Biotec, 130-096-286), LM22A4 (25 µL, 5 µM, Tocris Biotec, 4607), CHIR
99021 (250 µL, 30 µM, Tocris Bioscience, 1099), GABA (500 µL, 3 mM, Tocris Bioscience,
0344), CaCl2 (370 µL, 1.8 mM, Tocris Bioscience, 3148) and ascorbic acid (100 µL, 2 mM,
Tocris Bioscience, 4055). Synaptojuice B medium was filtered through a 0.22-µm filter. Cells
were treated with synaptojuice B (2 mL) for 7 days, and half medium changes were performed
until day 14.
25

MSNs Treatments with IFN-γ: For the IFN-γ experiments, prepatterned Activin A–treated
NSCs from C116 and HD72 were plated at 90,000 cells per well in an eight-well chamber slide
for MSN differentiation. After synaptojuice A and B treatment, MSNs were stimulated for 48 h
with IFN-γ (PeproTech, 300-02-100UG) at different concentrations: 10, 50, 100 and 200 ng/mL.
Non-treated MSNs were used as control. For each treatment, a duplicate was performed.
Cell Immunofluorescence of Human MSNs: Cells were fixed using 4% paraformaldehyde
(Sigma, 158127) in 0.1 M PBS, pH 7.4 (Corning, 21-040-CV) for 30 min. After three washes in
cold PBS, cells were permeabilized and blocked for 1 h at room temperature (RT) using 0.1%
Triton X-100 (Thermo Fisher Scientific, 28313) and 4% normal donkey serum (Jackson Immuno
Research, 017-000-121) in PBS. Primary antibodies were added in the presence of blocking
buffer overnight at 4°C. Secondary antibodies (1:500) were added after three PBS washes in
blocking buffer at RT for 1 h. The following primary antibodies were used for the
immunofluorescence studies: rabbit anti-DARPP-32 (Santa Cruz, sc-271111, 1:100), rabbit anti-
MAP2 (Millipore, AB5622, 1:100), rabbit anti-Nestin (Abcam, ab92391, 1:00), mouse anti-MHC-
class-II (Abcam, ab55152, 1:100), rabbit anti-Cleaved Caspase-3 (CellSignal, 9661, 1:100) and
mouse anti-HTT (Millipore, MAB2166, 1:100). The secondary antibodies were donkey anti-
rabbit, anti-mouse IgG conjugated with Alexa-546 (Invitrogen, A10040 and A10036) or Alexa-
647 (Invitrogen, A-31573 and A-31571). Images were acquired using a Biotek Cytation 5
microscope and were prepared using Fiji software (ImageJ, https://fiji.sc).
Protein Extraction for Proteomic Analysis: Triplicate samples of cultured C116-MSNs and
HD72-MSNs were washed three times with cold PBS 1X, pH 7.4 (Corning, 21-040-CV), and
total proteins were isolated using 300 µL of cold mammalian protein extracting reagent (Thermo
Fisher Scientific, 78501) containing protease inhibitor cocktail (cOmplete, Mini Protease Inhibitor
Cocktail, Roche, 11836170001). The cell lysate was harvested by scraping and transferred
directly into a cold 1.5-mL tube and stored at -80°C.
26

Proteomic Sample Preparation:
Chemicals: Acetonitrile (AH015) and water (AH365) were from Burdick & Jackson (Muskegon,
MI). Iodoacetamide (I1149), DT (D9779), formic acid (94318-50ML-F), and triethylammonium
bicarbonate buffer 1.0 M, pH 8.5 (T7408) were from Sigma Aldrich (St. Louis, MO), urea
(29700) was from Thermo Scientific (Waltham, MA), sequencing grade trypsin (V5113) was
from Promega (San Luis Obispo, CA), and HLB Oasis SPE cartridges (186003908) were from
Waters (Milford, MA).
Protein Precipitation, Digestion and Desalting: Protein samples were precipitated with a
ProteoExtract Protein Precipitation Kit (539180) from MilliporeSigma (Burlington, MA) as per the
manufacturer’s protocol. Samples were resuspended in 50 mM triethylammonium bicarbonate.
Total protein concentration was determined with a BCA kit (23227) from Thermo Fisher
(Waltham, MA). Aliquots of each sample containing ~100 μg of protein were brought to equal
volumes with 50 mM triethylammonium bicarbonate buffer at pH 8. The mixtures were reduced
with 20 mM DTT (37 ̊C for 1 h) and then alkylated with 40 mM iodoacetamide (30 min at RT in
the dark). Samples were diluted 10-fold with 50 mM triethylammonium bicarbonate buffer at pH
8 and incubated overnight at 37
o
C with sequencing grade trypsin (Promega, San Luis Obispo,
CA) at a 1:50 enzyme:substrate ratio (wt:wt). Peptide supernatants were collected and desalted
with Oasis HLB 30-mg Sorbent Cartridges (Waters, Milford, MA; 186003908), concentrated, and
resuspended in a solution containing mass spectrometric “Hyper Reaction Monitoring” retention
time peptide standards (HRM, Biognosys, Schlieren, Switzerland; Kit-3003) and 0.2% formic
acid in water.
Mass Spectrometric Analysis:
Orbitrap Lumos FAIMS DDA MS Analysis: Triplicate samples from corrected C116-MSNs
and HD-MSNs were analyzed by reverse-phase HPLC-ESI-MS/MS on the EASY-nLC 1200
27

system and analytical column (Thermo EASYspray 50 cm x 75 µm ID, PepMap C18 2 µm, 100
Å) coupled to the Orbitrap Lumos mass spectrometer (Thermo Fisher Scientific, San Jose, CA)
with an EASY-Spray source. Column temperature was set to 50ºC. Mobile phase A was 0.1%
formic acid in water, and mobile phase B was 0.1% formic acid in 80% acetonitrile and 19.9%
water. Flow rate was set at 300 nL/min, and a two-stage gradient was used for each sample: 1)
7–30% mobile phase B over 125 min, and 2) 30–45% mobile phase B over 40 min. For each
sample, 2 µg of peptides were injected onto the column. All samples were analyzed by DDA.
For DDA analysis, full MS scans were performed over m/z 380–1,580 with the Orbitrap analyzer
operating at 240,000 resolution and AGC = 400,000 ions with FAIMS settings enabled at three
compensation voltages (CVs): -50 V, -65 V, -85 V. Each of the selected CVs was applied to
sequential survey scans and MS/MS cycles (1 s per CV). Survey scans were followed by MS2
scans of the most intense precursor ions for 1 s. MS2 scans were performed by 0.7 m/z
isolation with the quadrupole, normalized higher-energy collisional dissociation (HCD) collision
energy of 35%. Dynamic exclusion was set to 30 s, mass tolerance to 10 ppm, and intensity
threshold to 5,000. Maximum injection time was set to 30 ms, AGC target was set to 10,000
ions, charge states +1 or >+8 were excluded, and the advanced peak determination was
toggled on.
TripleTOF 6600 DIA MS Analysis: Samples were analyzed by reverse-phase HPLC-ESI-
MS/MS using the Eksigent Ultra Plus nano-LC 2D HPLC system (Dublin, CA) combined with a
cHiPLC system directly connected to an orthogonal quadrupole time-of-flight TripleTOF 6600
mass spectrometer (SCIEX, Redwood City, CA). Typically, mass resolution in precursor scans
was approximately 45,000, and fragment ion resolution was approximately 15,000 in “high
sensitivity” product ion scan mode. After injection, peptide mixtures were transferred onto a C18
pre-column chip (200 μm × 6 mm ChromXP C18-CL chip, 3 μm, 300 Å; SCIEX) and washed at
2 μL/min for 10 min with the loading solvent (H 2O/0.1% formic acid) for desalting. Peptides were
28

transferred to the 75 μm × 15 cm ChromXP C18-CL chip, 3 μm, 300 Å (SCIEX) and eluted at
300 nL/min with a 3-h gradient using aqueous and acetonitrile solvent buffers. All samples were
analyzed by DIA, specifically using variable window DIA acquisitions (41). In these DIA
acquisitions, 64 windows of variable width (5–90 m/z) were passed in incremental steps over
the full mass range (m/z 400–1,250) with an overlap of 1 m/z. The cycle time of 3.2 seconds
included a 250-ms precursor ion scan, followed by acquisition of 64 DIA MS/MS segments,
each with a 45-ms accumulation time. The variable windows were determined according to the
complexity of the typical MS1 ion current observed within a certain m/z range using a SCIEX
“variable window calculator” algorithm (more narrow windows were chosen in “busy” m/z
ranges, wide windows in m/z ranges with few eluting precursor ions) (42). DIA tandem mass
spectra produce complex MS/MS spectra, which are a composite of all the analytes within each
selected Q1 m/z window.
Data Processing: For FAIMS DDA experiments, data analysis was performed with Proteome
Discoverer version 2.3.0.523 (Thermo Fisher Scientific). The database search was performed
using SEQUEST HT (Thermo Fisher Scientific), and parameters were as follows: SwissProt
human protein database (20,417 entries, 09 April 2019), trypsin enzyme digestion allowing two
missed cleavages, 10-ppm precursor ion mass tolerance, and 0.6-Da fragment ion mass
tolerance. Dynamic modifications were methionine oxidation (+15.995 Da) and N-terminal
protein acetylation (+42.011 Da), and a static modification was defined as cysteine
carbamidomethylation (+57.021 Da). Identifications were filtered to 1% FDR (PSM, peptide and
protein levels) with Percolator (43). LFQ was performed within Proteome Discoverer using razor
and unique peptides, and chromatographic alignment was enabled (maximum 10-min retention
time shift and 10-ppm mass tolerance). Abundance was normalized to the total peptide amount
and scaled on control average. Modified peptides were excluded from quantification, and
peptide quantities were summed for protein abundances. Statistical analysis was performed
29

using ProStaR software suite (44). Proteins with less than two unique peptides and proteins with
more than three missing values across all conditions were removed. Data were log2-
transformed, and missing values were replaced by the 2.5 percentile value for the partially
observed values and missing on the entire condition. Pairwise protein statistics were performed
using a Limma t-test, and an absolute log2(fold-change) threshold set at 0.58. Slim (sliding
linear model) method (45) was applied to adjust p-values for multiple testing, and significantly
altered proteins were sorted out using a p-value threshold that guarantees a FDR at 1.04%.
For DIA quantification, all collected data were processed in Spectronaut (version
14.2.200619.47784) using Biognosys (BGS) factory settings. Briefly, calibration was set to non-
linear indexed retention time (iRT) calibration with precision iRT selected. DIA data were
matched against a panhuman library that provides quantitative DIA assays for 10,316 human
proteins (46) and supplemented with scrambled decoys (library size fraction of 0.1), using
dynamic mass tolerances and dynamic extraction windows. The DIA/SWATH data were
processed for relative quantification, comparing peptide peak areas from various different time
points during the cell cycle. For the DIA/SWATH MS2 data sets, quantification was based on
XICs of 3-6 MS/MS fragment ions, typically y- and b-ions, matching to specific peptides present
in the spectral library. Interference correction was enabled on MS1 and MS2 levels. Precursor
and protein identifications were filtered to 1% FDR, estimated using the mProphet algorithm
(47). Quantification was normalized to local total ion chromatogram. Statistical comparison of
relative protein changes was performed with paired t-tests, and p-values were corrected for
multiple testing, specifically applying group wise testing corrections using the Storey method
(48). Finally, proteins identified with less than two unique peptides were excluded from the
assay. The quantification significance level was as follows: q-value less than 0.05, and absolute
Log2(fold-change) greater than 0.58 when comparing HD72-MSNs versus C116-MSNs.
30

Data Accession: Raw data and complete MS data sets have been uploaded to the Center for
Computational Mass Spectrometry, to the MassIVE repository at UCSD, and can be
downloaded using the following
link:https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=3a14708986c7468197598b328d1d
b750 (MassIVE ID number: MSV000088650; ProteomeXchange ID: PXD030786).
Western Blot Analysis: C116 and HD72-MSNs were harvested in mammalian protein
extracting reagent (150 µL, Thermo Fisher Scientific, 78501) mixed with a protease inhibitor
cocktail (1 tablet/10 mL, Roche, 11836170001). Cells were then processed further via
sonification using 5 s of pulsing, 5 s of rest for 5 rounds at 40 mA. Samples were then spun
down at 14,000 rpm at 4 C for 20 min and quantified using a bicinchoninic acid assay (Thermo
Fisher Scientific, 23227). Protein lysates of 10–20 g were added in with DTT (1 µL) and LDS
NuPAGE buffer (6 µL). Proteins were boiled at 95 C for 10 min. Running conditions used were
4–12% Bis-Tris gel in 5% MOPS Running Buffer (Invitrogen, NP0001) at 200 V for 1 h. The
transfer conditions used were a 0.45- m polyvinylidene fluoride membrane transferred in 5%
Transfer Buffer (Invitrogen, NP00061) at 20 mA for 840 min. Primary mouse monoclonal
antibody Septin-2 (Proteintech, 60075-1-Ig, 1:20,000), rabbit polyclonal antibody Septin-9
(Proteintech, 10769-1-AP, 1:500), mouse monoclonal IGFBP7 antibody (Santa Cruz
Biotechnology, sc-365293,1:100) and rabbit polyclonal antibody HMGCR (Sigma-Aldrich,
SAB4200529, 1:100) were incubated at 4 C.
MSNs Treatment with APOE3 and Lipid Metabolism Quantification: Prepatterned Activin A–
treated NSCs from C116 and HD72 were plated at 90,000 cells per well in eight-well chamber
slide (Falcon, 354108) for MSNs differentiation using synaptojuice A and B medium. MSNs
cultured in serum withdrawal were treated for 48 h with APOE3 (PeproTech, 350-02-500UG) at
312 ng/ml. Non-treated cells were used as control. For each treatment, a duplicate was
performed. The cells were fixed using 4% paraformaldehyde (Sigma, 158127) in 0.1 M PBS, pH
31

7.4 (Corning, 21-040-CV) for 30 min. To identify the lipid droplets, the fixed MSNs were stained
with Nile Red (Thermo Fisher Scientific, N1142), a lipophilic dye at a dilution of 1/1000 in PBS
for 30 min. To quantify the lipid droplets, confocal microscopy with online spectral fingerprinting
on a Zeiss LSM 980 laser scanning confocal microscope was used. The fluorescence emission
peak of Nile Red shifts yellow to red, based on increasing polarity of the bound lipid (49). Using
automated spectral component extraction, we observed a shorter wavelength punctate (peak at
593 nm) and a longer wavelength (peak at 611 nm) diffuse fluorescence of Nile Red using
excitation at 514 nm. Using the “Count cellular foci with secondary cell segmentation“ pipeline in
Image Analyst MKII (Image Analyst Software Novato, CA), images taken using a Plan-
Apochromat 63×/1.40 Oil lens were segmented using DAPI-stained nuclei as seeds and finding
cell boundaries using the diffuse, longer wavelength fluorescence of Nile Red. Then the
punctate shorter wavelength foci were counted per cellular area. For analysis of intensities,
images recorded with a Plan-Apochromat 20×/0.8 lens and the Basic fluorescence histometry
using nuclear markers (1-3 labels - basic)” pipeline. For analysis of colocalization of neutral lipid
vesicles with p62, LC3 and LAMP1, Nile Red stained cultures were permeabilized and
immunofluorescence staining was performed with mouse p62 (SQSTM1) (Abcam, ab56416,
1:100), rabbit anti-LC3 (Novus, Nb100-2331, 1:100), mouse anti-LAMP1 (Abcam, ab25630,
1:100). The secondary antibodies were donkey anti-rabbit, anti-mouse IgG conjugated with
Alexa-488 (Invitrogen, A21206 and A21202). The fluorescence spectrum of Alexa-488 was
measured in single-probe stained cultures and then was used for recording using online spectral
fingerprinting in the presence of DAPI and Nile Red staining. Image capture was performed in
an unbiased manner by setting up positions for recording in brightfield preview scans. In
permeabilized cells the long-wavelength component of the Nile Red did not outline the cells,
therefore cell area masks were calculated using Labkit trainable segmentation from the
background immunofluorescence. The short-wavelength, neutral lipid component of Nile Red
fluorescence and the immunofluorescence spots were identified by image segmentation in
32

Image Analyst MKII and gated with the cell area masks. To amplify punctate fluorescence local
background was removed using rolling ball background subtraction for immunofluorescence.
Representative images were smoothed using Wiener filtering. The identified puncta (image
segments) were used for measuring size, density and intensities of lipid droplets,
immunofluorescence puncta and counting droplets that colocalized with the immunostaining
using a modification of the “Fluorescence and absorbance histometry using nuclear and
secondary whole-cell segmentation (1–4 labels - advanced background options)” pipeline and
data processing in Microsoft Excel. All images were analyzed using the same pipeline, and
using global intensity scaling for segmentation. Statistical analysis was performed with
Graphpad Prism (Graphpad Software Inc., La Jolla, CA, USA) using an unpaired t-test with
Welch's correction or a two-way ANOVA with Sidak’s multiple comparison as indicated. Bar
graphs represent the mean ± SEM.
Pathway Analysis and Network Visualization: Pathway enrichment analysis was performed
using g:Profiler with parameters set to Homo sapiens, custom background (all proteins identified
in FAIMS DDA acquisitions: supplemental Table S4.6), Benjamini-Hochberg FDR and
threshold at 0.05. The gene-sets included for the pathway enrichment analyses were obtained
from Gene Ontology (GO) database (GOBP_AllPathways), updated February 01, 2020
(http://download.baderlab.org/EM_ Genesets/). Enrichment results are available in
supplemental Tables S4.7 and S4.8. The pathway analysis and network visualization were
carried out by using Cytoscape 3.7.2 and Cytoscape Enrichment Map application [version 3.2.1
of Enrichment Map software (50) with the following parameters: analysis type =
generic/gProfiler, p-value cutoff = 1, FDR Q-value cutoff = 0,05 and similarity between gene-
sets was filtered by Jaccard plus overlap combined (coefficient: 0.375)]. The network was
manually rearranged to improve layout and clusters of nodes were automatically annotated
33

using the AutoAnnotate Cytoscape App to highlight the prevalent biological functions among a
set of related gene-sets. Data were analyzed using QIAGEN Ingenuity Pathway Analysis.
Comparison with Multiple Datasets and Drug Prediction: Enrichment analysis for GO
biological processes with differentially expressed proteins (FDR < 0.05, logFC > 0.58) was done
utilizing the R package clusterProfiler. Drug prediction was done utilizing the LINCS L1000
characteristic direction signatures search engine (https://maayanlab.cloud/L1000CDS2/#/index)
[23] with upregulated and downregulated proteins as input (51).
Experimental Design and Statistical Rationale: In this study, we used human patient-derived
HD-iPSCs (72CAG/19CAG, HD72) and genetically corrected the cells to a normal repeat length
(21CAG/19CAG, C116), thus creating an isogenic control (31), that was then differentiated into
MSNs. Proteomic experiments were conducted with iPSC-derived cultured C116-MSNs (n = 3)
and HD72-MSNs (n = 3). “Hyper Reaction Monitoring” retention time peptide standards (HRM,
Biognosys; Kit-3003) were spiked into the samples before LC-MS/MS analysis in DDA and DIA
modes. First, as a discovery step, DDA acquisitions were performed on an Orbitrap Lumos
mass spectrometer coupled to a FAIMS device. Identification and MS1 XIC-based LFQ were
performed with Proteome Discoverer, as detailed above. To determine significantly altered
protein groups in HD72-MSNs versus C116-MSNs, pairwise comparison was performed with
ProStaR software suite (44) using a Limma t-test, and correcting the p-values for multiple
testing using the slim method (45). Then, as a validation step, the same samples were acquired
in DIA mode on a TripleTOF 6600 mass spectrometer, and one DIA cycle (3.2 s) was composed
of the acquisition of one MS1 scan, followed by the acquisition of 64 variable windows (5–90
m/z) covering the full mass range (m/z 400–1,250) with an overlap of 1 m/z. DIA data were
processed in Spectronaut v14 using a peptide-centric approach and a panhuman library
containing 149,066 precursors and 10,316 human proteins (46) to retrieve MS2 XIC-based
quantification information, as described above, and significantly altered protein groups in HD72-
34

MSNs versus C116-MSNs were obtained using a paired t-test followed by p-value correction for
multiple testing using the Storey method (48).
Experimental procedures for chapter 5
Mice: Experimental procedures were carried out according to the Institutional Animal and Care
and Use Committee at Icahn School of Medicine at Mount Sinai (LA09-00272, 16-0847
PRYR1). We followed the NIH Guidelines for the Care and Use of Experimental Animals.
Bcl11b-floxed mice were obtained from Dr. Mark Leid (Bcl11
btm1.1Leid
/J, #034469, Jackson
Laboratory, West Grove, PA, USA). D9-Cre mice were created in our laboratory (Bogush et al.,
2005). Cre expression was controlled by a regulatory element of the mouse Ppp1r1b gene-
encoding DARPP-32. The breeding of these lines resulted in a Bcl11b-specific deletion in MSNs
at 5–6 weeks of age. Both females and males were used. Food and water was provided ad
libitum and mice were kept in a 12-h light-dark cycle.
Tissue extraction: Pentobarbital (50 mg/kg) was intraperitoneally injected into mice followed by
perfusion with phosphate-buffered saline (PBS, Fisher Bioreagents, Pittsburgh, PA, USA,
BP399-1, 4 ◦C). The hemispheres of the brains were sagittally separated. The left hemisphere
striatum was flash frozen. The frozen striatum was used for RNA extraction. The right
hemisphere was fixed in 4% paraformaldehyde (Electron Microscopy Sciences, 15710).
Immunofluorescence and image acquisition: Brains were sectioned coronally on a vibratome
(Leica Microsystems, Wetzlar, Germany) at 30 µm. Brain sections were washed with 1X Tris
Buffered Saline (TBS, Fisher Bioreagents; BP2471-1) and incubated in 5% goat serum
(ThermoFisher Scientific, Waltham, MA, USA, 31872) with 0.25% Triton X-100 (Sigma Aldrich,
St. Louis, MO, USA, X100-500 mL) in TBS for 1 h at room temperature, and then incubated
overnight at 4 ◦C with primary antibodies, as follows: mouse anti-DARPP-32 (1:1000, Santa
Cruz Biotechnology, Inc., Santa Cruz, CA, USA, sc-271111), rabbit anti-BCL11B (1:1000,
affinity purified, Bethyl Laboratories, Montgomery, TX, USA; A300-385A), rabbit anti-Iba1
35

(1:500; WAKO Chemicals, Richmond, VA, USA, 019-19741), and rabbit anti-NeuN (1:1000;
Millipore; St. Louis, MO, USA, ABN78). Next, sections were incubated with the appropriate
secondary antibody: anti-rabbit Alexa 594 (1:400, ThermoFisher Scientific, A-11012,
ThermoFisher Scientific), or anti-mouse Alexa 488 (1:400, ThermoFisher Scientific, A-11008).
Images were acquired using a Zeiss 700 confocal microscope (Zeiss, Thornwood, NY, USA).
For colocalization experiments, we acquired four image frames of three independent brain slices
per mouse. Images were also obtained using an Olympus BX61 microscope and processed
using Fiji software (ImageJ v1.51).
RNA-seq: Dissected striatum from 4-month-old mice were subjected to RNA extraction. The
TruSeq RNA Sample Prep Kit v2 protocol (Ilumina, San Diego, CA, USA) was used and the
rRNA-depleted libraries were sequenced on the Illumina HiSeq 2500 System with 100
nucleotide paired-end reads. Bases with a quality score lower than 20 and adapted sequences
of the raw reads from the sequencing experiment were removed using Trim Galore! 0.6.4.
(https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/ (January 2022)). The analysis
was performed, as described in (M.-D. Cirnaru et al., 2021). The deposited raw data for the
transcriptomics are GSE185476. Differential expression analysis used an arbitrary cutoff of
adjusted p-value of less than 0.01.
Terminology enrichment analysis and pathway enrichment analysis: These were
conducted using clusterProfiler, a Bioconductor package in our analysis. All terms in biological
process (BP), molecular function, and cellular component categories in GO, as well as pathway
annotations derived from Kyoto Encyclopedia of Genes and Genomes (KEGG), were chosen to
identify statistically over-represented, biologically meaningful annotations that were enriched
and involved in the deletion of Bcl11b. We conducted both analyses on the differentially
expressed gene and TF clusters with the arbitrary cutoff of the adjusted p-value of less than
36

0.01 and the absolute Log2 fold-change of greater than 0. Data were analyzed using QIAGEN
Ingenuity Pathway Analysis.
Enrichr analysis: We analyzed for the enrichment of TFs, co-expressors, neural tissues, and
GEO sets using Enrichr, an online tool for performing enrichment analysis with a range of
biologically meaningful gene-set libraries. In our analysis, the TF perturbation categories in the
GEO, ChEA, and ENCODE databases were chosen to identify the significant upstream TFs of
those genes differentially expressed in Bcl11
btm1.1Leid
mouse MSNs. The ARCHS4 database was
chosen to identify significant co-expressors of those differentially expressed genes and TFs.
The Mouse Gene Atlas database was chosen to conduct the neural tissue enrichment analysis
to determine the similarity between our mouse D9-Cre-Bcl11
btm1.1Leid
MSNs and other typical
neural tissues. Gene sets extracted from perturbations of single genes, drugs, and diseases and
RNA-seq disease gene and drug signature categories in the GEO database were chosen to
identify synergistic genes of Bcl11b and to determine the similarities and differences in our
RNA-seq data and high-throughput sequencing data from other peer-reviewed publications. An
arbitrary cutoff of an adjusted p-value of less than 0.01 was chosen to determine the
differentially expressed genes.
GeneMANIA gene regulatory network analysis: We used GeneMANIA to conduct the
interaction network inference analysis of TFs that were enriched in D9-Cre-Bcl11
btm1.1Leid
MSNs,
with the arbitrary cutoff of an adjusted p-value of less than 0.01.
Statistical analysis of transcriptomics: Fisher’s exact tests were used to determine the
overlap of the gene expression profile of the Q175 knockin mouse model (Peter Langfelder et
al., 2016) and the conditional D9-Cre-Bcl11
btm1.1Leid
. The RNA-seq dataset of 10-month Q175
knockin mouse striatum (GSE 65774) had 2795 differentially expressed genes, and the D9-Cre-
Bcl11
btm1.1Leid
had 2771 differentially expressed genes. We used 0.01 as the arbitrarily adjusted
p-value (FDR) cutoff. The background gene number was 24,106, according to the miRNA and
37

coding gene data from the UCSC mouse mm10 GRCm38 genome assembly. Notably, instead
of comparing the overlap of the gene name, we designed a method to compare the overlap of
the gene signature, using the Fisher’s exact test twice. For the first test, genes were counted as
overlapping genes only if they were differentially expressed in both datasets and shared the
same pattern of expression (either upregulated or downregulated) and were in direct correlation
between datasets. For the second test, only the genes that were differentially expressed in both
datasets and their patterns of expression that were in inverse correlation in both datasets were
counted as overlapping genes. Fisher’s test was not statistically significant. Cell-type
enrichment analysis was calculated using Chi-square test. Cell-type specific genes were
obtained from (Merienne et al., 2019).
Behavioral Testing of Bcl11b-Deletion Mice - Locomotor activity: Spontaneous locomotor
activity was measured using the Digiscan D-Micropro automated activity monitoring system
(Accuscan, Inc., Columbus, OH, USA). This system consists of transparent plastic boxes (45 ×
20 × 20 inch) set inside metal frames that are equipped with 16 infrared light emitters and
detectors with 16 parallel infrared photocell beams. Breaks were recorded using a computer
interface in 5-min bins. Mice were habituated to the testing chamber for 2 days, and on the third
day, their locomotor activity was recorded for 60 min prior to returning them to their home
cages.
Balance beam test: Balance was assessed by measuring the ability of mice to traverse a
narrow beam as described (Creus-Muncunill et al., 2019; Creus-Muncunill et al., 2021), with
brief modifications. The beam consisted of an 85-cm-long wooden prism, divided into 5-cm
frames, with a 1-cm face, placed 40 cm above the bench surface. During the training session,
mice were allowed to walk on the beam for 2 min. After 4 h, mice were returned to the beam,
and their latency to cover 30 frames and total distance traveled were measured.
38

Vertical pole test: Motor coordination and balance were assessed by measuring the ability of
mice to turn and descend from a narrow pole, as described in (Creus-Muncunill et al., 2019).
The pole consisted of a 60-cm wooden cylinder (1-cm diameter) wrapped in tape to facilitate
walking. Mice were trained for 2 consecutive days and tested on the third day. Mice were placed
just below the top of the pole facing upwards. Time to completely orient the body downward
(time to turn) and time to climb down (time to descend) the pole were measured. An average of
three test trials is shown.
Haloperidol-induced catalepsy: Mice were intraperitoneally injected with 1 mg/kg Haloperidol
(Sigma Aldrich, H-030) or saline vehicle (0.9% NaCl, Teknov, Hollister, CA, USA, S5824). After
30 min, mice were gently positioned in catalepsy position, placing their forelimbs on a 0.5 cm
diameter steel rod, covered with non-slippery tape, that was 5 cm above the surface of the
bench. A researcher measured the time to remove both front paws from the bar (catalepsy
time). Catalepsy was measured every 30 min after the first trial.
Elevated plus maze: Anxiety-related behavior was tested by an elevated plus-maze as
described in (Walf & Frye, 2007).
Differentiation of neural stem cells (NSCs): NSCs were generated from induced pluripotent
stem cells (iPSCs) as described in (Zhang et al., 2001). Collagenase detachment media (Type
IV, ThermoFisher Scientific, 17104019, 1 mg/mL) in Gibco KnockOut DMEM/F-12 medium
(ThermoFisher Scientific, 12660012) was used for iPSC colonies. The cells were transferred to
a 0.1% agarose (Sigma-Aldrich, A9414)-coated low-attachment petri dish. The culture dish
contained embryonic stem (ES) culture medium Gibco KnockOut DMEM/F12,20% Gibco
KnockOut Serum Replacement (ThermoFisher Scientific, 10828028), 100 U/mL penicillin-
streptomycin (ThermoFisher Scientific, 15140122), 2.5 mM L-glutamine (ThermoFisher
Scientific, 25030081), 1X Non-Essential Amino Acids (ThermoFisher Scientific, 11140050), 15
mM HEPES (ThermoFisher Scientific, 15630106), and 0.1 mM β-mercaptoethanol
39

(ThermoFisher Scientific, 31350010). Embryoid body (EB) differentiation medium [DMEM
(Corning, 10-013-CV) supplemented with 20% FBS (ThermoFisher Scientific, 16000036), 1X
Non-Essential Amino Acids, 2 mM L-glutamine, and 100 U/mL penicillin-streptomycin] was used
every two days replacing 25% ES medium. At day 10, the EBs were attached to dishes coated
with poly-L-ornithine (1:1000 in PBS; Sigma-Aldrich, P3655) and laminin (1:100 in KnockOut
DMEM/F-12; Sigma-Aldrich, L2020). The EB were cultured in neural induction medium
[DMEM/F12,1X N2 (ThermoFisher Scientific, 17502001), 100 U/mL penicillin-streptomycin, 25
ng/mL βFGF (Peprotech, 100-18B), and 25 ng/mL Activin A (Peprotech, Cranbury, NJ, USA,
120-14P). Every 2 days medium was changed. Rosettes were harvested after 7–10 days with
the addition of 25 ng/mL Activin A, as described in (Naphade et al., 2017).
Differentiation into MSNs: MSNs were prepared as described in (Kemp et al., 2016). Briefly,
96-well plates were coated with a 50 µg/mL solution of Matrigel (Corning, Corning, NY, USA,
CB-40234) for 24 h. Passage 13 NSCs were plated in NPM medium at a concentration of
90,000 cells per well. To start differentiation, NPM medium was replaced with Synaptojuice A.
Half-medium changes were done every other day for 7 days. On day eight of differentiation,
Synaptojuice A was replaced with Synaptojuice B for 10 days with half-medium changes every
other day. Both Synaptojuice A and Synaptojuice B were supplemented with 25 ng/mL of Activin
A (Peprotech, 120-14P).
BCL11B and DARPP-32 immunostaining in MSNs: Cells were washed with PBS and fixed
with 4% paraformaldehyde for 12 min at room temperature. Cells were incubated in block buffer
containing 1% normal donkey serum and 0.1% Triton-X-100 in PBS for 1 h. Primary antibodies
were diluted at 1:100 in blocking buffer, and the cells were incubated with it overnight. The cells
were then washed with buffer containing 0.1% Triton-X-100 in PBS for 5 min, three times.
Secondary antibodies were diluted at 1:500 in blocking buffer with 300 nM DAPI, and the cells
were incubated in it for 2 h. The cells were then washed three times with 0.1% Triton-X-100 in
40

PBS for 5 min and imaged using a Cytation 5 instrument (Biotek). The antibodies used were
rabbit anti-BCL11B (Novus Biological Littleton, CO, USA, NB100-79809) paired with donkey-
anti-rabbit Alexa-488 (ThermoFisher, A-21206), and mouse anti-DARPP-32 antibody (Santa
Cruz Biotechnology, Inc., Santa Cruz, CA, USA, sc-271111) paired with donkey-anti-mouse
Alexa-647 (ThermoFisher, A-31571). iPSC-derived MSN qPCR: MSNs were differentiated in a
six-well plate, as described before. RNA was extracted utilizing an ISOLATE II RNA extraction
kit (BIO-52071, Bioline). Following the manufacturer’s instructions, 850 ng of RNA per sample
were used for cDNA synthesis (Cat No. BIO-65053, Bioline). qPCR Assays for SLIT3 (Cat No.
4453320, ThermoFisher Scientific, Assay ID: Hs00935843_m1), KCNC3 (Cat No. 4448892,
ThermoFisher Scientific, Assay ID: Hs01085817_m1), and WNT10A (Cat No. 4448892,
ThermoFisher Scientific, Assay ID: Hs05042697_s1) were used. A reaction mix used 1.5 µL of
the cDNA template, 5 µL 2x SensiFAST Probe mix (BIO-86005, Bioline), 0.5 µL of ACTB
endogenous control (4325788, ThermoFisher Scientific), and 2 µL of molecular grade water
(AM9937, Invitrogen). Three technical replicates were done for each sample, and the reactions
were run using a Roche LightCycler 480 II.
Confocal microscopy of nuclear aggregates of BCL11B: MSNs grown in plastic bottomed
microplates immunostained for BCL11B were imaged using a Zeiss LSM980/ Airyscan2 laser
scanning confocal microscope, using an LD LCI Plan-Apochromat 40×/1.2 Imm Korr objective
lens with glycerol immersion, and Airyscan super resolution mode (41nm/pixel resolution).
Images were analyzed in Image Analyst MKII (Version 4.1.3, Image Analyst Software, Novato,
CA, USA) using the “Nuclear foci area measurement” standard pipeline, providing counts of
BCL11B foci per nucleus, and the area of each nucleus based on the DAPI staining.

41

Experimental procedures for chapter 6
Cell culture: Human Embryonic Kidney 293 cells (HEK293T) were cultured in DMEM medium
(Gibco), 10% FBS (Gibco) and 100U/ml penicillin and 100 μg/ml streptomycin (Gibco) at 37° C,
5% CO2. Human iPSCs derived from an HD patient (female – 20 years old: 72Q/19Q) and their
CAG-corrected counterpart (21Q/19Q: C116) (An et al., 2012) were used. iPSCs were
differentiated into NSCs (Ring et al., 2015). The differentiation into NSCs was tested by
immunofluorescence using antibodies against the NSC markers Nestin (Sigma-Aldrich, 1:200)
and SOX1 (Sigma-Aldrich, 1:50) and iPSC marker OCT3/4 (Pierce antibodies, 1:500).
Differentiation into NSCs across all experiments was at least 98%. The iPSC lines were verified
for genome integrity prior to performing experiments using multi-color FISH analysis carried out
by Applied Stemcell Inc. (Menlo Park, CA). To generate pre-patterned Activin A NSCs, the
NSCs generated using above protocol were consistently maintained in 25 ng/ml Activin A
(Peprotech) after EB stage starting at day 10.
Non-isogenic HD and control iPSC lines ND41656 (CAG 57), ND42222 (CAG 109), ND42241
were obtained from Coriell Repository, and MN08i-33114.B line from WiCell. NSC lines were
generated using PSC neural induction medium (Life Technologies) as per instructions in the
manual. Briefly, iPSCs cultured in mTeSR were harvested using 1 mg/ml collagenase. The
colonies were transferred to a 60 mm dish coated with Matrigel (1:60 dilution, BD Biosciences)
and cultured in PSC neural induction medium supplemented with 1 µM LDN-193189 and 10 µM
SB431542 for 7 days to induce neuroepithelial fate. These cells were then harvested and
expanded in neural expansion medium (PSC neural induction medium and DMEM/F12 medium
(1:1), 100U/ml penicillin and 100 μg/ml streptomycin, and 2 mM L-Glutamine) supplemented
with 25 ng/ml bFGF.
Differentiation of human NSCs into MSNs: 60 mm dishes or 6-well plates were coated with
100 μg/ml poly-D-lysine (Sigma-Aldrich, P6407) followed by Matrigel (1:60, Corning) coating.
42

NSCs were plated and cultured in NPM. When confluent, NSCs were treated with Synaptojuice
A medium for 1 week followed by Synaptojuice B medium for 10 d at 37° C (Kemp et al., 2016).
25 ng/ml Activin A was added to both Synaptojuice A and Synaptojuice B media. Half media
change was performed every 2 days. The resulting MSNs were characterized by
immunofluorescence using antibodies (1:50-1:100) against the following: ß-III-tubulin (SCBT,
sc-80005), DARPP-32 (SCBT, sc-11365), Calbindin D-28K (Sigma-Aldrich, C9848), GABA
(Sigma-Aldrich, A2052), MAP2 (EMD Millipore, AB5622) and c-myc antibody (SCBT, sc-40).
MSNs labeled positively for these markers and transductions. DARPP-32 expression was also
determined by RT-PCR. Lentivirus transduction was performed with no p16
INK4a
transduction
and myc-p16
INK4a
from Origene (PS100071V and RC220937L1V) using a multiplicity of infection
(MOI) of 1. After four days of Synaptojuice B MSN differentiation, lentivirus was applied into
Synaptojuice B without CHIR 99021 (Tocris 4423). MSNs were transduced with virus was for 4
days.
Statistics: Statistics were performed using Student’s t-tests or two-way ANOVA. All
experiments were repeated at least three times. P < 0.05 was considered significant. Statistics
used for genomic data analysis, overlap analysis and biological content analysis are described
in the Supporting Information.
Protein co-immunoprecipitation assays: HEK 293T cells were transfected with pcDNA3.1-
Myc-Ryk-ICD or empty vector using JetPrime (PolyPlus Transfection, POL-114-07). 48 h after
transfection, cells were treated with 50 µM H 2O2 (AAT Bioquest, 11004) for 1 h 30 min then
washed once with cold Dulbecco's phosphate buffered saline (DPBS, Life Technologies) and
lysed in 25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 10 mM sodium
pyrophosphate, 10 mM b-glycerophosphate, 1 mM sodium orthovanadate, 10% glycerol,
supplemented with protease inhibitors (Halt™ Protease Inhibitor Cocktail, ThermoFisher
43

Scientific, 78430). Lysates were cleared for 10 min at 16,000 g at 4 °C. BCA assay
(ThermoFisher Scientific, 23225) was performed to determine protein concentration.
The cell lysates were precleared with Dynabeads Protein G (Life Technologies, 10003D) for at
least 1 h at 4 °C. Immunoprecipitations were carried out by incubating equal amounts (2 mg) of
total lysate with either 5 μg of ß-catenin antibody (BD 610154) or 8 µg of FOXO3 antibody
(Santa Cruz, H-144 sc-11351X) at 4°C overnight with gentle rotation. As a negative control we
used normal mouse IgG, ThermoFisher Scientific MA110407 and normal rabbit IgG, Cell
Signaling Technology 2729S, respectively). Immune complexes were captured by incubation
with 25 μl of pre-washed Dynabeads Protein G (Life Technologies) for 4 h at 4 °C. Beads were
washed 3 times for 10 min at 4 °C (with end-over-end rotation) in cold lysis buffer plus protease
inhibitors. The complexes were eluted at 100° C for 10 min in 1x LDS buffer (Life Technologies,
NP0007). Samples were resolved on NuPAGE 4–12% gels in 1X NuPAGE MES SDS buffer
(Life Technologies, NP0335BOX and NP0002 respectively) and transferred to a PVDF
membrane (Bio-Rad, 162-0177). Western blot analysis was performed under standard
conditions using anti-FOXO3 (H-144, Santa Cruz, 1:1000), anti-Myc (Cell Signaling Technology,
2278S, 1:1000) and anti-ß-catenin (Cell Signaling Technology, 9582S, 1:1000).
Deletion mapping assays: To determine the binding site of the Ryk-ICD fragment on ß-
catenin, we constructed deletion mutants of ß-catenin for the Armadillo repeat region. The full-
length human ß-catenin with a C-terminal Flag tag in pcDNA3.1 (Plasmid #16828, Addgene)
was used to generate deletion mutants using a PCR fusion-based approach. We generated
constructs encoding ß-catenin mutants deleted for amino acids 277 to 488 (∆277-488) and 489
to 593 (∆489-593). We designed primers immediately upstream and downstream of the
sequence to be deleted to carry short overlapping ends (22-24 bp). All mutants contained a C-
terminal Flag tag. We used primers as follows: ß-catenin ∆277-488, Forward 5’-
GGATCCCCAGCGTGGACAATGGCTACTC-3’ and Reverse 5’-
44

CTTAACCACAACTGGTAGTCCATAAGCTAAACGCACTGCCATTTT-3’ for amplifying the
region encoding amino acids 1-276 and Forward 5’-TATGGACTACCAGTTGTGGTTAAG-3’ and
Reverse 5’-TCTAGATGCATGCTCGAGCGGCCGCTTACTTGTCATCGTCGTCCTTGTA-3’ for
amplifying the region encoding amino acids 489-781; ß- ∆489-593, Forward 5’-
GGATCCCCAGCGTGGACAATGGCTACTC-3’ and Reverse 5’-
CTGCACAAACAATGGAATGGTATTGTGAAGGCGAACTGCATTCTGGGC-3’ for amplifying the
region encoding amino acids 1-488 and Forward 5’- AATACCATTCCATTGTTTGTGCAG-3’ and
Reverse 5’-TCTAGATGCATGCTCGAGCGGCCGCTTACTTGTCATCGTCGTCCTTGTA-3’ for
amplifying the region encoding amino acids 594-781. For each construct, the two PCR products
were fused by nested PCR using primers 5’-GGATCCCCAGCGTGGACAATGGCTACTC-3’ and
5’-TCTAGATGCATGCTCGAGCGGCCGCTTACTTGTCATCGTCGTCCTTGTA-3’. BamHI and
XbaI sites in sense and antisense primers respectively are underlined. The amplified fragments
were cloned into the pGEM-T easy vector (Promega, A1360), sequenced and subcloned into
the BamHI/XbaI sites (Invitrogen Inc.). All constructs were verified for sequence integrity.
HEK293T cells were transfected with Myc-Ryk-ICD and wild-type or mutant ß-catenin-Flag
constructs ∆277-488 and ∆488-593 or insertless vector using JetPrime. 48 h after transfection,
cells were treated with 50 µM H2O2 for 1 h 30 min, washed once with cold DPBS and with 25
mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 10 mM sodium
pyrophosphate, 10 mM b-glycerophosphate, 1 mM sodium orthovanadate, 10% glycerol,
supplemented with protease inhibitors. Lysates were cleared for 10 min at 16,000 g at 4 °C.
Protein quantification was performed by using BCA assays (ThermoFisher Scientific, 23225).
Lysates were precleared with Dynabeads Protein G for at least 1 h at 4 °C.
Immunoprecipitations were carried out by incubating equal amounts (2 mg) of cell lysates with 5
µg anti-FLAG (ThermoFisher Scientific, PA-1-984B) at 4 °C overnight with gentle rotation. As a
negative control, we used an antibody of the same isotype and species (normal rabbit IgG,
45

ThermoFisher Scientific, 10500C). Immune complexes were captured by incubation with 25 μl
pre-washed dynabeads protein G (Life Technologies, 10003D) for 4 h at 4 °C. Beads were then
washed three times for 10 min each at 4 °C (with end-over-end rotation) in cold lysis buffer
containing protease inhibitors. Immunoprecipitated complexes were eluted at 100° C for 10 min
in 1x LDS sample buffer. Samples were resolved on NuPAGE 4–12% gels in 1X NuPAGE MES
SDS buffer and transferred to PVDF membranes. Western blot analysis was performed under
standard conditions using the following antibodies: anti-Myc (ThermoFisher Scientific, MA1-980,
1:1000) and anti-FLAG (ThermoFisher Scientific, MA1-91878, 1:1000).
Immunofluorescence analysis for quantification of FOXO3 levels: NSCs were grown in
Millicell  EZ 8-chamber Slides (EMD Millipore, PEZGS0816) coated with pO/L at 40000
cells/well. After 48 h, complete NPM was switched to medium without bFGF and LIF for 4 h
followed by treatment with 20 µM LY294002 (Selleckchem, S1105) or vehicle for 1 h 30 min.
Cells were fixed with 4% paraformaldehyde (ThermoFisher Scientific) for 20 min at room
temperature (RT), washed twice with DPBS, permeabilized with 0.2% Triton X-100 (Sigma-
Aldrich) in DPBS for 15 min at RT and washed twice with DPBS. Cells were blocked using 1%
bovine serum albumin (BSA), 0.1% Triton 100-X in DPBS for 1 h at RT, then incubated
overnight at 4° C with FOXO3 antibody (H-144, Santa Cruz, 1:200). Cells were washed with
DPBS thrice, and incubated with goat-anti-rabbit Alexa Fluor 555 (Life Technologies, 1:200) in
the dark for 2 h at RT. Following 3 washes with DPBS, coverslips were mounted using ProLong
Gold antifade with DAPI (Life Technologies, P36941) and allowed to cure overnight at RT in the
dark before imaging. Imaging was performed using an SP5 Leica Confocal Microscope (Leica
Microsystems, Wetzlar, Germany), using the same settings for each condition tested in order to
ensure data comparability. Images were analyzed using ImageJ software
(https://imagej.nih.gov/ij/). First, DAPI images were used to create a binary mask to define
46

regions of interest for analysis. Then, this mask was applied to the FOXO3 images to measure
the fluorescence intensity in nuclei.
Messenger RNA isolation and sequencing: The RNA-seq samples corresponding to FOXO3
induction into the nucleus were prepared in duplicate. Cells were amplified in culture for a
limited number of passages (7−12). Briefly, 15 millions HD (72Q/21Q) and Corrected C116 were
transfected with validated siRNAs (Eurofins MWG Operon). Two Ryk siRNAs and scrambled
RNAs were tested by qRT-PCR. The siRNAs were Ryk siRNA-1 5’-
GCAAGUUAGUAGAGGCCAA-3’ and scrambled RNA-1 5’-AUCGAAGCUAGCGAUGAGA-3’
and Ryk siRNA-2 5’-AGUCUGGUUAAUAACGAGU-3’ and scrambled RNA-2 5’-
GAAGGUUCGCGAUUAAUAU-3’. Cells were transfected by electroporation (Neon

Transfection
System 100µl kit, Life Technologies MK10096) according to the manufacturer’s instructions.
Two days after transfection, complete NPM was switched to medium without bFGF and LIF for 4
h and cells were treated with 20 µM LY294002 for 1 h 30 min to activate endogenous FOXO3
as described (Brunet et al., 1999). Total RNA was isolated and purified using the RNeasy Mini
kit (Qiagen, 74104) and DNase treated using the DNA-free
TM
DNA removal

kit (Life
Technologies, AM1906) as per the manufacturer’s instructions. The mRNA libraries were
prepared using standard protocols from Illumina and sequenced with the Illumina HiSeq
Genome Analyzer for generating single-end 125 bp reads.
The RNA-seq samples corresponding to FOXO3 knockdown were prepared in triplicate. FOXO3
siRNAs (ON-TARGET plus SMART pool, L-003007-00-0020) and negative control siRNAs (ON-
TARGET plus Non-targeting Control pool, D-001810-10-20) were obtained from Dharmacon
(GE-Healthcare). Human NSCs were transfected using the Neon System 100 μl kit (Life
Technologies MK10096) according to the manufacturer’s guidelines. Briefly, cells were
harvested with Stempro Accutase (Life Technologies, A1110501), washed with DPBS and
resuspended in Buffer R at 2 x 10
7
cells/ml. 2 x 10
6
cells were mixed with 250 nM siRNA.
47

Conditions used for the electroporation were pulse voltage 1400 V, pulse width 20 ms and 2
pulses. Cells were seeded in 6-well matrigel-coated plates with 2 ml pre-warmed growth
medium without antibiotics, and incubated at 37 °C. 48 h after transfection, complete NPM was
replaced with medium without bFGF and LIF for 6 h before total RNA extraction. Total RNA was
isolated and purified using the NucleonSpin RNA kit (Machery-Nagel 740955.50) and DNase
treated using the DNA-free
TM
DNA removal

kit (Life Technologies, AM1906) as per the
manufacturer’s instructions. The mRNA libraries were prepared using standard protocols from
Illumina and sequenced with the Illumina NextSeq 500 for generating single-end 150 bp reads.
In both case, preparation and sequencing of mRNA libraries was performed by FASTERIS SA
(Switzerland).
Chromatin immunoprecipitation and sequencing: HD and C116 NSCs were amplified in
culture for a limited number of passages (7−12), and ChIP experiments were performed using
the Agilent Mammalian ChIP-on-chip protocol (Agilent Technologies). For performing FOXO3
ChIP-seq, 80-100 million cells were transfected with Ryk siRNA and scrambled RNA (using
electroporation as described above) and used to generate Illumina single-end libraries, the latter
step performed by Fasteris SA (Switzerland). Two days after transfection, we switched complete
NPM to medium without bFGF and LIF for 4 h and treated cells with 20 μM LY294002 for 1 h 30
min to activate endogenous FOXO3 as described (Brunet et al., 1999). Cells were cross-linked
for 10 min with 1% PFA. Cross-linking was stopped by addition of 0.125 M glycine, followed by
incubation for 5 min at room temperature. Cells were washed with cold DPBS, lysed in Lysis
Buffer 1 (LB1) and rocked for 10 min at 4 °C. Cell lysates were cleared at 1,350 g for 5 min at
4°C and pellets were resuspended in Lysis Buffer 2 (LB2) and rocked for 10 min at 4 °C. Pellet
nuclei were resuspended in Lysis Buffer 3 (LB3) and rocked 10 min at 4° C. Chromatin was
sheared by sonication using a Bioruptor Plus device (Diagenode) at 20 times for 30 sec ON–30
sec OFF and centrifuged at 16,000 g for 15 min at 4 °C. For immunoprecipitation, we used 5 μg
48

of a ChIP grade FOXO3 ‘NFL’ antibody (Brunet et al., 1999) or IgG antibdody (Cell Signaling,
2729) coupled to Dynabeads Protein G (Life Technologies). Chromatin was incubated overnight
with antibody-coupled beads. Beads were washed with RIPA buffer 3 times and once with Tris-
EDTA plus 50 nM NaCl. Chromatin was eluted in Elution Buffer at 65° C for 15 min, then
reversed cross-linked by incubating at 65 °C overnight. DNA was extracted using
phenol/chloroform and purified using a PCR purification kit (Qiagen, 28104). ChIP-seq libraries
were prepared for sequencing using standard Illumina protocols as performed by Fasteris SA,
Switzerland. DNA sequencing was carried out at Fasteris using a Illumina HiSeq Genome
Analyzer sequencer.
Next-generation sequencing data analysis: Next generation sequencing data analyses were
performed using the Galaxy framework developed by the ARTbio bioinformatics platform at the
Institute of Biology Paris-Seine (https://mississippi.snv.jussieu.fr/).
RNA-seq data analysis: The raw sequencing reads corresponding to FOXO3 induction into the
nucleus were pre-processed in order to discard both adapter sequences and low quality reads
using the Trimmomatic and FASTQ Groomer tools. The filtered reads were then mapped to the
human reference genome hg38 (UCSC 2013 release) by using Bowtie2 (Langmead & Salzberg,
2012). Duplicate and unmapped reads were removed using SAM Tools (Li et al., 2009). Read
distributions were calculated using the featureCounts tool (Liao, Smyth, & Shi, 2014). To test for
differentially expressed genes, reads were analyzed using the R/Bioconductor package edgeR
(Robinson, McCarthy, & Smyth, 2010). For genes with very low read counts, a prior filtering step
was performed to keep those with at least ten counts in all samples. Differential expression
analysis was assessed using the exact negative binomial test and a false discovery rate (FDR)
correction of 5%. The raw read counts from the RNA-seq data was first filtered to remove genes
with a read count less than 5 in all samples. Principal component analysis (PCA) was performed
upon normalizing raw RNA-seq readcounts using DESeq2 (1.22.2). PCA analysis was then
49

performed using the R function prcomp (3.5.1). A 3D plot was generated using the R package
rgl (0.100.30) to visualize the first 3 principal components. The volcano plots were generated
using the R package EnhancedVolcano (1.0.1). The log fold change values obtained after the
differential expression analysis between FOXO3 nuclear induction vs control were used as the
input data. The data are available at GSE109873, subseries GSE109872.
The raw sequencing reads corresponding to FOXO3 knockdown were pre-processed in order to
perform Quality control of the reads using fastqc and discard adapter sequences using Clip
adapter (Galaxy Version 2.2.0). The reads were then mapped to the human reference genome
hg38 (UCSC 2013 release) by using HISAT2 (Kim, Langmead, & Salzberg, 2015). Read
distributions were calculated using the featureCounts tool [62]. To test for differentially
expressed genes, reads were analyzed using the R/Bioconductor package DESeq2 version
1.18.1 (Love, Huber, & Anders, 2014). For genes with very low read counts, a prior filtering step
was performed to keep those with a sum of at least ten counts in all samples. An adjusted p-
value was calculated and a false discovery rate (FDR) of 1% was applied. Analysis of FOXO3
dependence for gene expression in basal compared to stressed conditions was performed using
the R function pnorm. The data are available at GSE109873, subseries GSE109869.
ChIP-seq analysis/Processing of ChIP-seq reads: Reads of 50 bp lengths were generated
using and Illumina HiSeq Genome analyzer. Raw sequencing reads were mapped to the human
reference genome (hg38 2013 release) using Bowtie2 (Langmead & Salzberg, 2012). Duplicate
and unmapped reads were removed using SAM tools. Peaks were called using MACS2 (Zhang
et al., 2008) set to default parameters except for read lengths set at 100 and cut-offs on P value
set to 0.05. Deduction of input (whole cell extract) and mock antibody signal was carried out
using bdgcmp available from MACS2. Peaks corresponding to induction of FOXO3 in the
nucleus were then called using the tool bdgpeakcall available from MACS2. Generated peaks
were annotated using the R/Bioconductor package ChIPseeker (Yu, Wang, & He, 2015) and a
50

peak calling set to ±5 kb from a transcription start site (TSS). DeepTools (Ramirez et al., 2016)
were used to perform binding enrichment across TSSs with sorting based on mean and k-
means clustering.
Integrative analysis of RNA-seq and ChIP-seq data: Comparison of RNA-seq and ChIP-seq
data was performed using Monte Carlo simulations with 10,000 repetitions for testing all
overlaps between the 4 experimental conditions. In addition, global and pairwise comparisons of
ChIP-seq signals (or peak score distributions) for the 4 conditions were examined using the Chi-
squared test in R.
Motif discovery and enrichment analysis: Analysis of sequence motifs was performed using
tools of the MEME Suite (Stuge & Ellingsen, 1991) for the MACS2 peaks corresponding to the
promoter regions within ±250 bp. Default parameters were used except for MEME Motif Count
(value: 8), DREME Motif E-value (value: 0.5) and CentriMo Match Score (value: 8). The
significance of detected motifs was assessed using an E-value <0.5. Motif annotation was then
performed using the HOmo sapiens COmprehensive MOdel COllection database (HOCOMOCO
v10) (Kulakovskiy et al., 2013). To determine the family-wide significance of motifs (e.g. the
FOXO family), the top p-value was considered for each family.
Orthology analysis: The human-mouse orthologous gene pairs were extracted from Ensembl
database with the DIOPT tool (https://www.flyrnai.org/cgi-bin/DRSC_orthologs.pl) with
parameters set to best orthologs of mouse genes. Fischer’s exact test were used to test for
statistical significance of overlaps.
Biological content analysis: Biological content analysis was using EnrichR
(http://amp.pharm.mssm.edu/Enrichr/). Additionally, networks containing subgroups of FOXO3
direct targets as seed genes and their level-1 neighbors (for a total number of neighbors set at
60-200) were derived from the high coverage and probabilistic functional network STRING
51

v10.0 (Szklarczyk et al., 2015) and displayed using confidence view, selectively performed
using high-confidence (STRING probability score ≥ 0.7) edges based on information from
databases and experiments. Networks were represented using Cytoscape
(http://www.cytoscape.org/). Networks were annotated with signaling pathway and biological
process information as provided in STRING. Seed genes (F3T-INs) in these networks were
annotated further with information on the magnitude of regulation by FOXO3, as inferred from
log2 fold changes in RNA-seq data, information from FOXO3 knockdown data and with
information on gene deregulation in HD compared to C116 NSCs (Ring et al., 2015). To
prioritize F3Ts for validation studies based on the predicted impact of F3T reprogramming on
signaling-pathway and cellular activity, we selected the short paths interconnecting at least
three categories of F3T (lost or gained, positively or negatively regulated by FOXO3). The
magnitude of regulation by FOXO3 and novelty compared to the literature were also used for
prioritizing F3Ts.
Transfection of human NSCs: FOXO3 siRNAs (ON-TARGET plus SMART pool, L-003007-00-
0020), ETS1 siRNAs (ON-TARGET plus SMART pool, L-003887-00-0005), ETS2 siRNAs (ON-
TARGET plus SMART pool, L-003888-00-0005) and negative control siRNAs (ON-TARGET
plus Non-targeting Control pool, D-001810-10-20) were obtained from Dharmacon (GE-
Healthcare). Previously validated siRNAs targeting exon 1 of CDKN2A were p16
INK4A
siRNA-1
(5’-AACGCACCGAATAGTTACGGT-3’) (Bond et al., 2004) and p16
INK4A
siRNA-2 (5'-
CUGCCCAACGCACCGAAUA-3') (Kan et al., 2012) and non-specific control 47% CG siRNA
were obtained from Eurofins Genomics. Human NSCs were transfected using the Neon System
100 μl kit (Life Technologies MK10096) according to the manufacturer’s guidelines. Briefly, cells
were harvested with Stempro Accutase (Life Technologies, A1110501), washed with DPBS and
resuspended in Buffer R at 2 x 10
7
cells/ml. 2 x 10
6
cells were mixed with 250 nM siRNA.
Conditions used for the electroporation were pulse voltage 1400 V, pulse width 20 ms and 2
52

pulses. Cells were seeded in 6- well matrigel-coated plates with 2 ml pre-warmed growth
medium without antibiotics, and incubated at 37 °C. 48 h after transfection, complete NPM was
replaced with medium without bFGF and LIF for 6 h before total RNA extraction.
Human NSCs were transfected with cDNAs using jetOptimus reagent (PolyPlus Transfection)
according to the manufacturer’s guidelines within the recommended reagent/DNA ratio range.
Briefly, cells were plated at density 5x10
6
cells/well

in 12 well-plates. Twenty-four hours after
seeding, complete NPM in all wells was exchanged for 1 ml of fresh medium following
transfection with the 1µg pcDNA3.1-FLAG-FOXO3 WT (Addgene Plasmid #8360), pcDNA3.1-
FLAG-FOXO3-TM (Addgene Plasmid #10709) or pcDNA3.1. 48 hours after transfection,
complete NPM was replaced with medium without bFGF and LIF for 9 h before total RNA
extraction. Cell viability was detected 48h after transfection using the CellTiter-Glo®
Luminescent Cell Viability Assay (Promega, G7571) according to the manufacturer’s protocol.
Briefly, a volume of CellTiter-Glo Reagent equal to the volume of culture medium was added to
each well, and mixed for 2 minutes on orbital shaker. The plate was then incubated at RT for 10
min and luminescence of each sample measured using the plate-reader FLUOstar Optima
(BMG Labtech).
Gene expression analysis: Total RNA was isolated from cells using the RNeasy kit (Qiagen),
and DNase treated using the DNA-free DNA removal kit according to the manufacturer’s
instructions Kit (Ambion). Equal amounts of total RNA (1 μg) was reverse-transcribed using the
RevertAID First Strand cDNA synthesis kit (Thermo Fisher scientific, K1622), according to the
manufacturer’s instructions. The first strand cDNA was diluted and used as template in the real-
time quantitative-PCR analysis. The LightCycler 480 Real-Time PCR System was used to
perform the qRT-PCR using GoTaq qPCR Master Mix (Promega, A6002). qRT PCR
experiments were performed in triplicate using the following primers: FOXO3: Forward: 5’-
AGGGAGTTTGGTCAATCAGAA-3’, Reverse: 5’- TGGAGATGAGGAATCAAAGTT-3’; Ryk:
53

Forward: 5’-CCACTTCTACGCGTGTGTTT-3’, Reverse: 5’- GCCCTTGGGAACTACTGC-3’;
p16
INK4
: Forward: 5’-CCAACGCACCGAATAGTTACG-3’, Reverse: 5’-
GCGCTGCCCATCATCATG-3’; p14
ARF
: Forward: 5‘-CCCTCGTGCTGATGCTACTG-3’,
Reverse: 5’-CATCATGACCTGGTCTTCTAGGAA-3’; CDKN2AIP: Forward: 5’-
GTGTATAGGGTCGGCCATCAA-3’, Reverse: 5’-CCTGCCGTTGTTACCTGAGAG-3’;
SERTAD1: Forward: 5’- CTCAAGCTCCACCACAGCCT-3’, Reverse: 5’-
AGTGTTCACGACCAGCACCA-3’; ETS2: Forward: 5’-CTGGGCATTCCAAAGAACCC-3’,
Reverse: 5’-CCAGACTGAACTCATTGGTGG-3’; ETS1 Forward: 5’-
GGGAGGACCAGTCGTGGTAAA-3’, Reverse: 5’-CACGCTGCAGGCTGTTGAAAG-3’; p21
CIP1
:
Forward: 5’-CACCGAGGCACTCAGAGGAG-3’, Reverse 5’-CCGCCATTAGCGCATCACAG-3’;
p27
KIP1
: Forward: 5’-TAATTGGGGCTCCGGCTAACT-3’, Reverse: 5’-
TGCAGGTCGCTTCCTTATTCC-3’; HRPT: Forward: 5’-ATGCTGAGGATTTGGAAAGG-3’
Reverse: 5’-CTCCCATCTCCTTCATCACA-3’; ACTB: Forward: 5‘-CCAACCGCGAGAAGATGA
-3’, Reverse: 5’-CCAGAGGCGTACAGGGATAG-3’. QRT-PCR was performed at 95 °C for 10
min, followed by 40 cycles at 95 °C for 15 sec, 60 °C for 30 sec, and 72 °C for 30 sec. Data
were analyzed using the LightCycler 480 software (Roche) and advanced relative quantification
method. Gene expression was quantified by the mean cycle threshold (Ct) value for triplicate
measurements. Target gene expression was normalized to two housekeeping genes (HPRT
and ACTB) according to the 2-ΔΔCt formula. Statistical analyses (2-way ANOVA and t-tests)
were performed using GraphPad Prism v6.
For biochemical analysis of Activin A and MSNs total RNA was isolated from NSCs and MSNs
using ISOLATE II RNA Mini Kit (Bioline). cDNA was prepared from 1 μg of RNA in a total
reaction volume of 20 μl using the SensiFAST cDNA synthesis kit (Bioline). RT-PCR reactions
were setup in a 384-well format using 2X SensiFAST Probe No-ROX kit (Bioline) and 1 μl cDNA
per reaction in a total volume of 10 μl. RT-PCR was performed on the Roche LightCycler 480
54

instrument. For quantification, the threshold cycle, Ct, of each amplification was determined by
using the second derivative maximum method. The 2
–ΔΔCt
method was used to determine the
relative expression levels of each gene normalized against the housekeeping gene b-actin. The
primers used were as follows: p16
INK4a
: Forward: 5’-CAGCAGCATGGAGCCTTC-3’, Reverse:
5’-CGTAACTATTCGGTGCGTTG-3’, Probe 67 and Forward: 5’-CTGCCCAACGCACCGAATA-
3’, Reverse: 5’-GCTGCCCATCATCATGACCT-3’, Probe FAM; FOXO3: Forward: 5’-
CTTCAAGGATAAGGGCGACA-3’, Reverse: 5’-CGACTATGCAGTGACAGGTTG-3’, Probe 11;
MMP3: Forward 5’-GCTGATATAATGATCTCTTTTGCAGT-3’, Reverse: 5’-
CATAGGCATGGGCCAAAA-3’, Probe 85.
Additonal primers for senescent markers are shown below along with Probe number are below.

CDKN2AIP-F-41 41 gcgaaccacgtcttcctc
CDKN2AIP-R-41 41 ttggagcatctgtcactttga
ETS1-F-UPL69 69 aagtcctggaagggagatcg
ETS1-R-UPL69 69 gcatacagcttttattccaagtca
SELL-F-UPL72 72 agttgtgggggtggacaat
SELL-R-UPL72 72 cagcagtcggttccatgat
IGFBP7-F-UPL67 67 actggctgggtgctggta
IGFBP7-R-UPL67 67 tggatgcatggcactcata

For analysis of mRNA levels in brain samples from Hdh-Q175 mice, 3 male mice of each
genotype (heterozygous and corresponding wildtype control) were sacrificed at 15 months of
age by cervical elongation. Brains were immediately dissected and the striatum, cortex and
cerebelum of these mice were snap frozen in liquid nitrogen. RNA was extracted using the
55

Nucleospin kit from Macherey-Nagel, following manufacturer’s instructions. RNA samples were
then incubated with DNAse (DNA-free DNA removal kit, ThermoFischer Scientific) to avoid
genomic DNA contamination. Reverse transcription (RevertAid First Strand Synthesis kit,
ThermoFischer Scientific) and qPCR (GoTaq qPCR Master Mix, Promega) was performed using
manufacturer’s instructions on a Roche LC480 instrument. Primer sequences are: p16
INK4a

forward 5’-AGGGCCGTGTGCATGACGTG-3’, reverse 5’-GCACCGGGCGGGAGAAGGTA-3’;
p19
ARF
forward 5’-CATGTTGTTGAGGCTAGAGAGG-3’, reverse 5’-
TCGAATCTGCACCGTAGTTG-3’; HPRT forward 5’-ATTATGCCGAGGATTTGGAA-3’, reverse
5’-CCCATCTCCTTCATGACATCT-3’.
Immunofluorescence analysis and quantification of p16
INK4a
and HMGB1: NSCs plated
(and differentiated into MSNs) in 8-well Nunc Lab-Tek II Chamber Slides (Thermo Fisher
Scientific) were fixed with 4% paraformaldehyde for 15 min at room temperature (RT), and
washed twice with PBS. Cells were permeabilized with 0.25% Triton X-100 (Sigma-Aldrich) in
PBS for 15 min at RT, then washed twice with PBS. Blocking was performed using 5% donkey
serum and 1% BSA in PBS for 30 min at RT. Cells were washed with PBS, and incubated
overnight at 4 °C with primary antibody, washed with PBS three times, and incubated with
fluorescent secondary antibody in the dark for 2 h at RT. Following three washes with PBS,
coverslips were mounted using ProLong Gold antifade with DAPI (Thermo Fisher Scientific).
Slides were cured for 24 h in the dark at RT, and imaging performed on Nikon Eclipse Ti-U
microscope using the Plan Apo  20X/0.75 objective. Primary antibodies – p16
INK4a
(Abcam,
ab108349), HMGB1 (Abcam, ab18256), and Nestin (SCBT, sc-23927) – were used at a dilution
of 1:100. Secondary Alexa Fluor antibodies were purchased from Invitrogen. Image analysis
was performed using the Gen5 software. TIFF images were converted to monochrome images
and single cell analysis was performed using DAPI-stained nuclei to define the region of
interest. The HMGB1 ICC analysis for lentivirus transduced MSNs was carried out using a
56

Biotek Cytation 4, Gen5 Image Software using advanced features in which the GFP positive
cells were masked for the nuclear staining and the cytoplasmic signal was quantified. MSNs
were transduced with lentivirus for 96 hours and then fixed as described above. HMGB1, Rbt,
Abcam AB18256 at 1:100 for ICC as described above.
Senescence-associated ß-galactosidase (SA-ß-gal) staining: NSCs were cultured as
described above with the addition of 25 ng/ml Activin A (Peprotech, AF-120-14E). NSCs were
stained using the Senescence staining kit (#9860, Cell Signaling Technology). Nuclei were
stained with DAPI, and coverslips were mounted as described above. Images were captured
using the Lionheart FX Automated Microscope and a 10X Plan Fluorite WD 10 NA 0.3 objective.
Image analysis was performed using the Gen5 software. TIFF images were converted to
monochrome images and single cell analysis was performed using DAPI-stained nuclei to
define the region of interest and average SA-ß-gal intensity/cell was quantified.
Cell proliferation assays: Human NSCs were seeded on 24-well plates at 0.5-1 x 10
5
cells per
well, 6 wells for per genotype. After 1, 2, 3, 4 and 5 days at 37 °C and 5% CO 2, the medium
was replaced with 500 µl fresh medium containing 10% v/v AlamarBlue® reagent
(ThermoFisher Scientific, DAL1025) according to the manufacturer’s protocol. The plates were
then incubated at 37° C for 3 hours. 100 µl from each well was transferred to a 96-well plate for
reading. Fluorescence (excitation and emission wavelength 550 and 595 nm) was measured
using the Infinite F500 microplate reader (Tecan Genios). The 100% reduced form of
AlamarBlue®, (i.e., medium containing 10% v/v AlamarBlue® autoclaved at 121°C for 15 min)
were used as positive control. Wells without cells with culture medium containing 10% v/v
AlamarBlue were used as negative controls. The relative fluorescence intensity for each
genotype and each day was calculated as the AlamarBlue® fluorescence signal of the sample
at day X minus the signal of the negative control. Statistical analyses (2-way ANOVA) were
performed using Prism v6.
57

Cellular vulnerability assays: Human NSCs were subjected to 24 h of growth factor
deprivation as performed 48 h after cell transfection (by electroporation, as described above).
Cell viability and caspase-3/7 activity were then detected using the ApoLive-Glo Multiplex Assay
(Promega, G6410). Briefly, 10 µl of reagent (GF-AFC substrate) were added to each well, and
gently mixed for 30 seconds. After incubation for 30 min at 37° C, fluorescence was measured
using the plate-reader FLUOstar Optima (Ex at 360 nm, Em at 490 nm, BMG Labtech). Then,
50 μl of Caspase-Glo
®
3/7 reagent was added to each well, and gently mixed for 30 seconds.
These plates were then incubated at RT for 30 min and luminescence of each sample
measured using the plate-reader FLUOstar Optima (BMG Labtech). Caspase-3/7 assays were
performed using 5 replicates/point and data expressed as Caspase-3/7 activity (RLU) divided by
cell viability (RFU). Statistical analyses (paired t test) were performed using GraphPad Prism v6.
Data availability: RNA-seq and ChIP-seq data are available at GSE109873, subseries
GSE109871, GSE109872 and GSE109869.
Experimental procedures for chapter 7
Animals: Male p16-3MR mice young (4-5 months) and old (24-26 months) were maintained
according to National Institutes of Health guidelines for use of live animals. All experimental
procedures were approved by the Institutional Animal Care and Use Committee at the Buck
Institute.
Treatment: Ganciclovir administration was started when the mice were 4 months old and
continued until they reached 22 months. The regimen involved dissolving ganciclovir (Sigma,
G2536) in PBS (Corning, 21-040-CM) and then delivering a 5-day course of treatment every two
weeks at a dosage of 25 mg/kg via intraperitoneal injection. Simultaneously, a control group
was treated with the vehicle solution, PBS, following the same administration pattern as the
ganciclovir group.
58

25HC (Tocris Bioscience, Product 5741) was dissolved in 22.5% 2-hydroxypropyl-beta-
cyclodextrin (HβCD) (Tocris Bioscience, Product 0708). Mice were subjected to a 7-day
treatment protocol involving daily intraperitoneal injections of either 25HC at a dosage of 50
mg/kg or a vehicle control when the mice were 22 months old. This treatment concluded with
the final injection administered 10 weeks prior to dissection.
Dissections: Mice were subjected to a brief isoflurane exposure in an enclosed chamber,
inducing unconsciousness within approximately 60 seconds, followed by cervical dislocation
and decapitation. The hippocampus was swiftly micro dissected and immediately preserved by
freezing on dry ice for subsequent nuclei extraction and fixation.
In the ganciclovir treatment group, tissue samples underwent immediate processing for nuclei
extraction and fixation prior to freezing. Dissection of ganciclovir-treated mice and respective
PBS control mice spanned two days. This approach, which entailed alternating dissections
between a ganciclovir-treated mouse and a PBS control, minimized potential batch effects.
The same protocol was employed for the cohort of 25HC-treated mice and the corresponding
PBS+HβCD control group and the untreated old and young cohort, with dissections carried out
over two days and alternation between treated/old and control/young groups to mitigate batch
effects.
Brain tissue dissociation and nuclei fixation: Nuclei extraction was conducted using a semi-
automated protocol with a Singulator 100 instrument (S2 Genomics). The hippocampi were
placed in a Singulator cartridge (S2 Genomics, 100-059-446) that contained 2 ml of ice-cold
nuclei extraction buffer. This cartridge was then positioned inside the pre-cooled Singulator
device. A preset program was employed to carry out the extraction process, which comprised a
10-minute incubation period, a 5-minute trituration stage, and a 10-minute dissociation interval.
59

Post-extraction, the derived nuclei were filtered and subsequently fixed using a nuclei fixation kit
(Parse Biosciences, SB1003), adhering to the manufacturer's guidelines.
snRNAseq library preparation: The fixed nuclei were used to prepare single nuclei RNA
sequencing libraries. For the ganciclovir and PBS groups, these libraries were assembled using
a WT kit from Parse Biosciences, adhering strictly to the manufacturer's instructions. From this
procedure we obtained 8 sub libraries with each containing a pool of all samples.
An initial sub-library from the WT kit was sequenced on an Illumina HiSeq Lane, to enable
assessment of library quality, estimation of the number of nuclei captured, and the necessary
sequencing depth for the subsequent sub-libraries. Of the remaining seven sub-libraries, two
were excluded due to inadequate nuclei capture, leaving five for analysis. These five were
pooled and sequenced across two Illumina NovaSeq S4 PE150 lanes.
In the case of the 25HC and PBS+HβCD groups as well as young and old untreated mice,
libraries were compiled using a WT Mega kit from Parse Biosciences, again following the
manufacturer's instructions. As with the previous group, a single sub-library from the WT Mega
kit was submitted for sequencing on a single Illumina HiSeq Lane to assess library quality,
gauge nuclei capture, and estimate the requisite sequencing depth for the remaining sub-
libraries. The remaining 15 sub-libraries were pooled together and sequenced across two
Illumina NovaSeq S4 PE150 lanes. One sub library was omitted due to subpar gene capture,
leaving 14 for further analysis.
Raw data processing and quality control: The sequencing reads were demultiplexed and
aligned to the genome, facilitating the generation of gene counts through the Split-seq pipeline.
The product of this process, a gene count matrix and accompanying cell metadata, were used
to construct a Seurat object [35] in R. Cells with less than 2700 unique molecular identifiers
(UMIs), more than 1% mitochondrial RNA and more than 6000 unique features were removed
60

from the samples in the first batch of sequencing (GCV and PBS). Cells with less than 700
unique molecular identifiers (UMIs), more than 1% mitochondrial RNA and more than 4500
unique features were removed from the samples in the second batch of sequencing (25HC,
PBS+HβCD, untreated old and untreated old).
Data normalization was performed using SCTransform [36], while Harmony [37] was employed
to integrate the different batches. Further processing was performed using Seurat [35], including
principal component analysis (PCA), K-nearest neighbors (KNN) graph construction, clustering,
and UMAP (Uniform Manifold Approximation and Projection) calculations.
Determination of cell type identity and differential expression
Cell types were assigned using a two-pronged approach. Initially, we identified clusters
expressing classical markers of major cell types, as described in prior studies [4, 38]. These
markers included Gad1 and Gad2 for GABAergic Neurons, Slc17a7 and Slc17a6 for
Glutamatergic Neurons, C1qa and Cx3cr1 for Microglia, Mal and Apod for Oligodendrocytes,
Pdgfra for OPCs, Pdgfrb for Pericytes, Gfap and Gja1 for Astrocytes, and Pecam1 and Cldn5
for Endothelial cells.
Markers for the clusters that were not identified within the major cell types were identified via
Seurat's "FindAllMarkers" function. The expression of these markers was then mapped to cell
types in the single cell Allen Brain Mouse Atlas of the cortex and hippocampus
(https://portal.brain-map.org/atlases-and-data/rnaseq) as well as described in other studies [4,
39]. This enabled us to label CR Neurons (Reln and Syt1), ABCs (Slc47a1 and Abcg2), CPCs
(Enpp2 and Ttr), and VLMCs (Slc6a13, Tbx18 and Igfbp2).
Differential expression analysis: Differentially expressed genes for the pairwise comparisons
were calculated using Seurat’s “FindMarkers” function utilizing Seurat’s implementation of the
61

package MAST [40]. Genes with an adjusted p value lower than 0.05 and an absolute log2 fold
change above 0.1 were considered significantly differentially expressed.
Comparisons of transcriptional signatures: To investigate the congruence and divergence
among transcriptomic signatures, we leveraged the RRHO algorithm [30]. This algorithm doesn't
mandate a predefined threshold, effectively pinpointing shared and unique gene expression
patterns.
Initially, genes tested in both experiments were sorted by their log2-transformed fold changes,
from most upregulated to most downregulated. Then, a hypergeometric test assessed the
significance of gene overlap, accounting for multiple hypothesis testing with a False Discovery
Rate (FDR) adjustment.
Results were visualized via an RRHO heatmap, marking the most significant overlaps after all
rank combinations were calculated. This heatmap reflects the varying degrees of overlap under
different conditions, providing a holistic view of transcriptomic similarities and differences.
Two-sided Spearman’s rank correlation was calculated with R function “cor.test”.

62

CHAPTER 3: Transcriptomic Characterization Reveals Disrupted Medium Spiny Neuron
Trajectories in Huntington's Disease and Possible Therapeutic Avenues. (C Galicia
Aguirre et al., 2023)
Abstract
Huntington's disease (HD) is a neurodegenerative disorder caused by an expansion of CAG
repeats in exon 1 of the HTT gene, ultimately resulting in the generation of a mutant HTT
(mHTT) protein. Although mHTT is expressed in various tissues, it significantly affects medium
spiny neurons (MSNs) in the striatum, resulting in their loss and the subsequent motor function
impairment in HD. While HD symptoms typically emerge in midlife, disrupted MSN
neurodevelopment has an important role. To explore the effects of mHTT on MSN development,
we differentiated HD induced pluripotent stem cells (iPSC) and isogenic controls into neuronal
stem cells, and then generated a developing MSN population encompassing early, intermediate
progenitors, and mature MSNs. Single-cell RNA sequencing revealed that the developmental
trajectory of MSNs in our model closely emulated the trajectory of fetal striatal neurons.
However, in the HD MSN cultures, the differentiation process downregulated several crucial
genes required for proper MSN maturation, including Achaete-scute homolog 1 and members of
the DLX family of transcription factors. Our analysis also uncovered a progressive dysregulation
of multiple HD-related pathways as the MSNs matured, including the NRF2-mediated oxidative
stress response and mitogen-activated protein kinase signaling. Using the transcriptional profile
of developing HD MSNs, we searched the L1000 dataset for small molecules that induce the
opposite gene expression pattern. Our analysis pinpointed numerous small molecules with
known benefits in HD models, as well as previously untested novel molecules. A top novel
candidate, Cerulenin, partially restored the DARPP-32 levels and electrical activity in HD MSNs,
and also modulated genes involved in multiple HD-related pathways.
63

Background
Huntington’s disease (HD) is a fatal, dominantly inherited neurodegenerative disorder that
primarily affects neurons in the striatum and cortex [1, 2]. HD is caused by a CAG expansion in
the huntingtin gene that leads to a polyglutamine (polyQ) expansion in the encoded protein
(HTT), and patients with a CAG expansion greater than 38 repeats exhibit symptoms [3]. As a
basal ganglia disease, HD impacts motor learning and control (chorea), executive functions, and
emotions. The major cell type and principal output neuron of the caudate-putamen is the
inhibitory g-aminobutyric acid (GABA)-ergic medium spiny neurons (MSNs). Dysfunction and
developmental alterations of MSNs have been implicated in HD, with MSN subtypes being
differently impacted [4-14]. Despite the mutant HTT (mHTT) protein being present throughout
life, HD symptoms only manifest later in life. This raises the question of whether the early
developmental alterations in MSNs set the stage for developing HD symptoms later in life. As
seen in human imaging studies, pre-onset young HD carriers already have abnormal striatal
volumes [15, 16]. These observations have been know for a number of years, [17] and arose
from findings of abnormalities in immature cells carrying the mutation [18, 19]. Humbert and
colleagues [4] showed clear abnormalities in cortical development in human fetuses with mHTT
and in engineered mice, with many deficits in cell cycling, neuronal differentiation, and
endosome dynamics. The KIDS-HD study (reviewed in [20]), including studies that focused on
the developmental trajectory of the striatum, demonstrated hypertrophy in the early years of
childhood and then a steep decline. These results are in line with studies conducted on HD
developmental models, that showed an increase in neuronal proliferation and premature
maturation [21, 22].
We reported isogenic HD neuronal stem cells (NSCs) derived from induced pluripotent stem
cells (iPSCs) display dysregulated signaling pathways [6, 23, 24]. Specifically, we used human
patient–derived HD-iPSCs (72CAG/19CAG, HD72) and genetically corrected the cells to a
normal repeat length (21CAG/19CAG, C116), thus creating an isogenic control [23]. Our
64

transcriptomic analysis of isogenic HD72 NSCs suggested that HD is linked to developmental
impairments that prevent proper generation of MSNs and subsequent loss of MSNs identity [6,
8, 23-25]. Similarly, HD iPSC derived neurons feature dysregulation of multiple pathways
related to development [21, 26, 27]. We also found that deletion of transcription factors required
for striatum development leads to HD-like phenotypes [28]. This highlights the importance of
correct striatum development and suggests that targeting this pathway has important
therapeutic implications for HD.
In studies in mice, transient expression of mHTT during development leads to
neurodegeneration and HD-like symptoms later in life [7]. Furthermore, lower levels of the non-
mutant HTT during development result in HD-like motoric abnormalities in adulthood [29].
Conversely, correcting early neuronal defects caused by mHTT in mice can delay the onset of
HD pathology [12]. These findings highlight the potential for treating developmental deficits in
HD as a therapeutic approach and underscore the importance of understanding the
mechanisms of dysregulation caused by the mutation during early stages of development.
Since the HD gene, HTT, was discovered in 1993 [30], progress has been made in
understanding the cellular pathways disrupted in HD and in identifying potential therapeutics.
Still, the proximal events in HD pathogenesis remain unclear and there are no treatments
modify the disease progression. Thus, a comprehensive characterization of the very early
proximal events in HD MSNs is critical for the design of alternative therapeutic approaches to
delay disease onset and progression.
In this study, we investigated the effects of mHTT on the developmental trajectory of MSNs. We
utilized our isogenic HD72 and C116 iPSC lines [23] to generate developing MSNs at various
stages of maturation, including early and intermediate progenitors and mature MSNs. To
comprehensively analyze the developmental changes, we used bulk and scRNA sequencing.
By integrating our data with scRNAseq data of the developing human striatum [31], we identified
65

the developmental trajectories affected by HD. Our analysis revealed that multiple pathways
were disrupted during HD72 MSN development, along with alterations in transcription factors
responsible for MSNs identity. Notably, Achaete-scute homolog 1 (ASCL1) and the Distal-less
homeobox (DLX) family of transcription factors, essential for MSN development, were
downregulated in developing HD72 MSNs and we also see downregulation in adult HD patients
and post-natal mouse models. With the transcriptomic data, we identified potential therapeutics,
including several small molecules known to improve HD phenotypes, and untested molecules.
We assessed the effects of Cerulenin, a small molecule not previously used in HD, and found a
partial rescue of DARPP-32 levels and, electrical activity and modulation of multiple HD -related
pathways in HD72 MSNs.
Results
Differentiation and Characterization of HD72 NSCs and Developing MSNs.
We differentiated HD72 and C116 iPSCs into NSCs using a monolayer differentiation approach
[32]. (Fig. 3.1A). After confirming the expression of the NSC marker nestin (NES) (Fig. 3.1B)
[33], the cells were differentiated into a mixed population of early, and intermediate progenitors,
and mature MSNs using a chemical protocol that mimics the developmental stages of the lateral
ganglionic eminence (LGE) [34]. After 21 days of differentiation, DARPP-32 positive cells were
visible among the developing MSNs (Fig. 3.1C). Furthermore, developing MSNs expressed
higher levels of dopamine and cAMP-regulated phosphoprotein Mr 32,000 (DARPP-32,
PPP1R1B), calbindin 1 (CALB1), calbindin 2 (CALB2), dopamine receptor D1 (DRD1),
dopamine receptor D2 (DRD2), BCL11 transcription factor B (CTIP2), opioid receptor mu 1
(OPRM1, also known as MOR1-6TM, MOR1-7TM), and nuclear receptor subfamily 4 group A
member 1 (NR4A1) mRNA compared to NSCs (Fig. 3.1D).These results indicate the presence
of mature MSNs in the population.
66

67

Figure 3.1. Differentiation of Developing MSNs. (A) Diagram of differentiation process for
MSNs. (B) C116 and HD72 NSCs immunolabeled with Nestin and SOX2. (C) DARPP-32
staining on developing MSNs. Scale bar: 200 μm. (D) RNA expression of multiple markers of
MSNs in NSCs and developing MSNs determined with qPCR. P values were calculated using t
tests followed by Benjamini-Hochberg correction for multiple tests. # < 0.1, * < 0.05, ** < 0.01.

68

Using bulk RNA sequencing we compared the transcriptional profiles of HD72 developing MSNs
to a published RNAseq dataset of HD72 NSCs [6]. Principal component analysis (PCA) showed
a distinct separation between HD72 and C116 developing MSNs (Fig. 3.2B), whereas HD72
and C116 NSCs clustered more closely (Fig. 3.2A). We depicted the transcriptional changes
induced by mHTT in NSCs and developing MSNs using volcano plots (Fig. 3.2C, D). A total of
4,533 and 4,082 genes were differentially expressed in MSNs and NSCs, respectively (adjusted
p-value < 0.01, absolute log2-fold change > 0.1) (Supplemental Table 3.1 and 3.2). Notably, in
HD72 NSCs, the top downregulated differentially expressed genes (DEGs) included MSH
homeobox 1 (MSX1), a modifier of age of onset in HD [35]. In HD MSNs, we observed a notable
upregulation of insulin-like growth factor binding protein 7 (IGFBP7), a gene implicated in the
initiation of senescence and apoptosis [36]. These findings align with our previous discovery
that HD MSNs display numerous senescence-like characteristics [37]. Though MSNs and NSCs
shared 1,362 DEGs, their correlation was low (Pearson’s r = 0.06, p-value < 2.2e-16)
(Supplemental Fig. 3.1A). As anticipated, since MSNs are the main cell type affected by HD,
DEGs in MSNs demonstrated a larger fold change compared to those in NSCs.
Next, we sought to assess whether RNA alterations were reflected at the protein level in
developing HD72 MSNs. We compared our RNAseq data with our recently published
proteomics dataset from developing HD72 and C116 MSNs [38]. We identified a considerable
correlation between gene fold change values at both the RNA and protein levels (Pearson’s r =
0.75, p-value < 2.2e-16), highlighting the potential impact of these alterations on cellular
function (Supplemental Fig. 3.1B). We also compared the DEGs in developing HD72 MSNs to
those found in other HD models and data from HD patients. Our comparisons encompassed the
striatum of BACHD-ΔN17 at 2, 7, and 11 months of age [39]; the striatum of Q80, Q92, Q111,
Q140, and Q175 at 2, 6, and 10 months [40]; the striatum from 11-week-old R6/2 mice [41];
DEGs from iPSC-derived MSNs [42]; and data from the caudate and putamen of HD patients
69

[43]. We noted a significant overlap in DEGs across most comparisons (Supplemental Table
3.2). As anticipated, the highest similarity was observed with iPSC-derived MSNs (adjusted p-
value 1.6E-35). The mouse model exhibiting the greatest similarity was the 11-week-old R6/2
(adjusted p-value 2.16E-220). In adult HD mice this likely represents the dedifferentiation of
MSNs [44]. Additionally, a significant overlap was detected with data obtained from Grade 2 HD
caudate postmortem tissue (adjusted p-value 1.70E-50).
We employed Ingenuity Pathway Analysis (IPA) to assess alterations in canonical pathways
within HD72 NSCs and developing MSNs (Supplemental Table 3.1 and 3.2). The top IPA
predictions in HD72 NSCs included, dysregulation of axonal guidance and molecular
mechanisms of cancer, along with activation of integrin-linked kinase (ILK) signaling, and ras
homolog (RHO) family GTPase signaling. Additionally, we observed inactivation of semaphorin-
mediated neuronal repulsion, phosphatase and tensin homolog (PTEN) signaling, and rho GDP
dissociation inhibitor alpha (RHOGDI) signaling (Fig. 3.2E). In developing MSNs, top IPA
predictions included dysregulation of axonal guidance and molecular mechanisms of cancer,
activation of ephrin receptor, integrin, and coordinated lysosomal expression and regulation
(CLEAR) signaling pathways, along with inactivation of semaphorin-mediated neuronal
repulsion, RHOGDI signaling, and the sumoylation pathway (Fig. 3.2F). Furthermore,
enrichment analysis using NCATS BioPlanet identified multiple HD-associated pathways
(Supplemental Table 3.1 and 3.2). Among the top terms enriched in HD72 NSCs and
developing HD72 MSNs were axon guidance, transforming growth factor (TGF-b) regulation of
extracellular matrix, developmental biology, brain derived neurotrophic factor (BDNF) signaling
and focal adhesion [6, 45-47] (Fig. 3.2G, H). Distinct enrichments in developing HD72 MSNs,
which were not present in HD72 NSCs, included transmission across chemical synapses,
neuronal systems, and lysosomes, among others (Supplemental Table 3.2). Importantly, the
BDNF pathway has been found to be dysregulated in the striatum of HD patients, various
70

cellular models expressing mHTT, and multiple mouse models including Hdh 109/109 kock-in,
YAC, N-171 82Q, and R6/2 among others [48].
We used Leafcutter [49] to identify 248 genes that had differential RNA-splicing in developing
HD72 MSNs (Supplemental Table 3.3). We compared these genes with those mis-spliced in the
striatum of human HD patients and the R6/1 HD model [50], revealing 133 and 40 shared mis-
spliced genes, respectively (Supplemental Fig. 3.1C). Enrichment analysis of the mis-spliced
genes common to all three datasets demonstrated an overrepresentation of axon guidance and
long-term potentiation pathways. Notably, both pathways are dysregulated in HD [46, 51]
(Supplemental Fig. 3.1D, Supplemental Table 3.3).

71

Figure 3.2. Transcriptional characterization of HD72 NSCs and developing MSNs.
72

(A,B) PCA plot of NSCs (A) and developing MSNs (B). A HD72 NSC outlier has been removed
only for visualization purposes. (C,D) Volcano plots illustrating DEGs when comparing HD72 vs
C116 NSCs (C) and HD72 MSNs vs C116 MSNs (D). (E,F) IPA of HD72 NSCs (E) and HD72
MSNs (F). Gray dots indicate an undetermined direction for the pathway. (G,H) Bioplanet
enrichment analysis of HD72 NSCs (G) and HD72 MSNs (H).
73

scRNAseq Reveals Dysregulation of Pathways Related to HD Pathology as MSN
Maturation Occurs.
We performed scRNAseq on developing HD72 and C116 MSNs to explore transcriptional
dysregulation at single-cell resolution during HD MSN development. Our analysis revealed a
continuous maturation trajectory encompassing i) early progenitors, identified by the expression
of NES, and Vimentin (VIM); ii) intermediate progenitors, characterized by ASCL1, Sp9
transcription factor (SP9), DLX1 and DLX2 expression; and iii) mature MSNs, distinguished by
forkhead box P1 (FOXP1), early B cell factor-1 (EBF1), and microtubule associated protein 2
(MAP2) (Fig. 3.3A, B). All these clusters expressed MEIS2, indicating an LGE-like lineage.
Notably, despite DARPP-32 being detected at the protein level and RNA level via qPCR and
bulk RNAseq, it was not well detected in scRNAseq, presumably due to low abundance leading
to increased dropout. Similar constrains have previously been observed in scRNA studies of the
developing striatum [31].
Next, we conducted differential gene expression analysis between C116 and HD72 cells at each
developmental stage and utilized IPA analysis to predict the activation state of canonical
pathways (Supplemental Table 3.4). Our results revealed that, as MSNs matured, several
pathways associated with HD pathology exhibited increased activation (Fig. 3.3C). Specifically,
α-adrenergic signaling was predicted to increase, which is consistent with prior research linking
α-adrenergic receptors to heightened neurotoxicity in HD [52]. Additionally, the nuclear factor
erythroid 2-related factor 2 (NRF2)-mediated oxidative stress response, which is increased in
HD in response to oxidative stress, displayed increased activation with MSN maturation [53]. In
contrast, we also observed decreased activation in pathways that are protective against HD as
HD72 MSNs matured (Fig. 3.3D). For example, activation of both sphingosine-1-phosphate and
mitogen-activated protein kinase 1 (ERK/MAPK) signaling diminished with increased
maturation, and their activation has been associated with exerting protective effects in HD [54,
74

55]. Additionally, we observed a reduction in Dopamine-DARPP32 feedback within cAMP
signaling, a pathway known to be downregulated in HD [56]. Among the top differentially
expressed genes, we discovered neuromedin B (NMB) and cyclin D2 (CCND2) in early
progenitors (Fig. 3.3E). In intermediate progenitors, the most dysregulated genes included
DLX6 antisense RNA 1 (DLX6-AS1) and homeobox B5 (HOXB5) (Fig. 3.3F). Meanwhile, in
mature MSNs, crystallin beta A2 (CRYBA2) and transcription elongation factor A-like 7
(TCEAL7) emerged as the top differentially expressed genes (Fig. 3.3G). Notably DLX6-AS1
interacts with other DLX transcription factors important during striatum development [57].
We also used scRNAseq in C116 and HD72 NSCs to show heterogeneous expression of
multiple genes. We separated them into four clusters and displayed the top genes representing
their heterogeneity (Supplemental Fig. 3.2A, B). Despite their heterogeneity, they showed
homogeneous expression of NSC marker NES indicating a single cell type (Supplemental Fig.
3.2C). Differential expression in C116 and HD72 NSCs showed among the top downregulated
genes in HD72 NSCs, Ubiquitin B (UBB), Retinol binding protein 1 (RBP1) and Crystallin beta
B1 (CRYBB1). Among the top upregulated genes in HD72 NSCs, we found OCIA domain
containing 2 (OCIAD2), insulin like growth factor binding protein 5 (IGFBP5) and S100 calcium
binding protein B (S100B) (Supplemental Fig. 3.2D, Supplemental Table 3.4). Interestingly,
OCIAD2 is linked to Alzheimer's [58], and IGFBP5 induces cell senescence [59]. Previously,
we've observed senescence-like traits in HD72 MSNs [37].

75

76

Figure 3.3. Single Cell Transcriptional Characterization of HD72 and C116 Developing
MSNs. (A) UMAP plots of C116 and HD72 MSNs. (B) Heatmap showing levels of expression of
various markers of early progenitors, intermediate progenitors and mature MSNs. (C) Heatmap
showing level of activation of pathways that become more activated with MSNs maturation. (D)
Heatmap showing level of activation of pathways that become less activated with MSNs
maturation. (E) Violin plots showing expression of the top upregulated and downregulated
genes in HD72 early progenitors, NMB and CCND2. (F) Violin plots showing expression of the
top upregulated and downregulated genes in HD72 intermediate progenitors, GPC1 and
DHRS3. (G) Violin plots showing expression of the top upregulated and downregulated genes in
HD72 intermediate progenitors, DLX6-AS1 and TCEAL7.
77

HD Alters the Neurodevelopmental Program of MSNs.
To compare the in vitro maturation of MSNs with their native human development, we analyzed
scRNAseq data from our model after integrating it with a human LGE dataset spanning weeks
7, 9, and 11 post-conception (Fig. 3.4A,B) [29]. During fetal development, MSNs begin as apical
progenitors, progress to basal progenitors, and ultimately mature into fully differentiated MSNs.
Our goal was to assess whether in vitro maturation of MSNs follows a similar trajectory by
comparing the expression patterns of various markers throughout these developmental stages
(Fig. 3.4A-C). We observed that NES, an apical progenitor marker, was expressed in both
apical and early progenitors. In contrast, GS homeobox 2 (GSX2) and transcription factor 7 like
1 (TCF7L1), both markers of apical progenitors, were exclusively expressed in apical
progenitors rather than in early progenitors. This finding highlights a distinct identity between
these two cell populations at this particular developmental stage. On the other hand, ASCL1
and DLX1, markers for basal progenitors, were present in both basal and intermediate
progenitors (Fig. 3.4C). Additionally, FOXP1, a marker of mature MSNs, was detected in both
the LGE and in vitro datasets (Fig. 3.4C). We also observed similar patterns of expressions in
developing MSNs and LGE for MEIS2 a marker of the LGE lineage, and doublecortin (DCX) a
marker of neurons.
We analyzed the frequency of apical progenitors, basal progenitors, and mature MSNs in the
LGE dataset, finding similar ratios for these three cell types (Fig. 3.4D). In a comparison of
C116 and developing HD72 MSNs, we observed a higher percentage of mature MSNs and a
lower percentage of early and intermediate progenitors in developing HD72 MSNs (Fig. 3.4E).
This observation was supported by NES immunostaining, which showed reduced NES
expression in developing HD72 MSNs (Supplemental Fig 3A,B).

78

Figure 3.4. Comparison of iPSC Derived C116 and HD72 MSNs with Human Fetal LGE. (A,
B) UMAP plot of fetal LGE (A)C116 and HD72 developing MSNs (B). (C) Expression of NES,
TCF7L1, GSX2, ASCL1, DLX1, DCX, MEIS2 and FOXP1. (D) Proportion of apical progenitors,
basal progenitors and MSNs in LGE. (E) Proportion of early progenitors, basal progenitors and
mature MSNs in C116 and HD72 MSNs.

79

Next, we aimed to characterize gene expression trajectories during MSN maturation in the three
datasets. We employed VIA to model MSN maturation progression [60]. VIA builds a k-nearest
neighbor graph in which nodes represent cell clusters linked according to their gene expression
similarity. The algorithm takes a user-provided root node as the trajectory's starting point and
infers the trajectory using lazy-teleporting random walks, combined with Markov chain Monte
Carlo refinement. To initialize the algorithm, we designated the clusters for apical and early
progenitors with the highest NES expression as root nodes that, represented the starting point
of development. This approach produced a pseudotime value for each cell that reflects their
progression in maturation. Pseudotime correctly increased, as MSNs moved from early and
apical progenitors towards mature MSNs (Fig. 3.5A). We further evaluated the expression of
several neuronal maturation markers in relation to pseudotime to ensure correspondence with
maturation state. As expected, we observed a decline in NES levels as pseudotime advanced
(Fig. 3.5B). ASCL1, an intermediate progenitor marker, accurately demonstrated peak
expression in the middle phases of fetal and C116 MSN development (Fig. 3.5C). However,
developing HD72 MSNs did not exhibit increased ASCL1 expression at any point in pseudotime.
We also examined the expression of DCX, a marker of developing neurons previously identified
to exhibit elevated expression in basal progenitors and newly born MSNs [31]. As expected,
DCX expression increased along with pseudotime progression (Fig. 3.5D).
ASCL1 is a critical transcription factor that promotes neuronal differentiation [61]. Within the
LGE, ASCL1 regulates the expression of DLX1 and DLX2, which are required for proper striatal
development. [62-67]. The downregulation of ASCL1 during HD MSN maturation raises
concerns about the potential impact on its downstream targets. To investigate this, we assessed
the expression of DLX1/2 and other members of the DLX family as a function of pseudotime.
While DLX1/2/5/6/6-AS1 expression increased during the midst of fetal and C116 MSNs
development, this upregulation failed to occur in HD72 MSNs (Fig. 3.5E-I).
80

In addition, we observed a similar pattern of downregulation for glutamate decarboxylase 1
(GAD1), a GABAergic marker that is regulated by the DLX TFs [68, 69] (Fig. 5J). To explore the
broader impact of DLX TF dysregulation in HD72 MSNs, we examined whether other target
genes were affected. Specifically, we compared the differentially expressed genes in developing
HD72 MSNs with those identified as targets for Dlx1, Dlx2, and Dlx5 during murine LGE
development [69]. Our analysis revealed a significant enrichment of Dlx2 targets among the
DEGs in developing HD72 MSNs, with 505 genes shared between the two sets (adjusted p-
value: 0.012) (Fig. 5K). Furthermore, these 505 genes were bound by Dlx2 at their
transcriptional start site in CHIP-seq experiments of developing mouse striatum [69].
To assess whether the dysregulation of DLX genes persisted into adulthood and disease onset,
we analyzed the expression of DLX TFs, GAD1, and GAD2 in a published scRNAseq dataset
from the caudate and putamen of HD patients and striatum from the R6/2 [70]. DLX1 and DLX2
expression was sparse in iSPNs and dSPNs of human and mouse datasets and did not permit a
relevant comparison of HD and control samples. DLX5 and DLX6 were downregulated in iSPNs
and dSPNs of HD patients. GAD1 and GAD2 were also downregulated (Fig. 3.5G). In R6/2
iSPNs and dSPNs, we observed downregulation of Dlx6, Gad1, and Gad2, although Dlx6 was
only downregulated in R6/2 dSPNs and not in iSPNs (Fig. 3.5G).
In addition to pseudotime plots, we also compared the expression of DLX1, DLX2, DLX5, DLX6,
DLX6-AS1, GAD1 and GAD2 in the various clusters corresponding to different stages of
development. We found a lower proportion of cells expressing these genes in HD72 MSNs
(Supplemental Fig. 3.4A-G). We also observed higher expression of cortical markers
neurogenin 2 (NEUROG2), vesicular glutamate transporter (SLC17A6/VGLUT2), EBF
transcription factor 3 (EBF3), and nescient helix-loop-helix 2 (NHLH2) in HD72 MSNs
(Supplemental Fig. 3.4H-J). These markers suggest a partial loss of cell identity, as reported in
other HD models [44].
81

Figure 3.5. Developmental Dysregulation in Developing HD72 MSNs. (A) Pseudotime in
fetal LGE, C116 MSNs and HD72 MSNs. Expression of neuronal maturation markers, NES (B),
82

ASCL1 (C) and DCX (D). Expression of transcription factors DLX1 (E), DLX2 (F), DLX5 (G),
DLX6 (H), DLX6-AS1 (I), GAD1 (J) in human LGE and developing C116 and HD72 MSNs. (K)
Comparison of DLX2 targets at transcriptional start sites with DEGs in HD72 MSNs. P values
were calculated using Fisher’s exact tests followed by Benjamini-Hochberg correction for
multiple tests. (L) Expression of DLX5, DLX6, GAD1 and GAD2 in iSPNs and dSPNs from
caudate and putamen of HD patients and R6/2.
83

Predicted Developmental HD Modifiers Based on Transcriptional Data can Reverse HD
Phenotypes.
A promising strategy for identifying disease modifiers is to search for small molecules that
induce gene expression patterns that are the opposite of those in the disease state. This
approach has been successfully used in Alzheimer’s and Parkinson’s disease models [71, 72].
To demonstrate the feasibility of this approach in HD, we utilized the L1000CDS2 web
application [73]. Leveraging the L1000 dataset, L1000CDS2 identified small molecules that
produce an opposite or similar gene expression pattern to a given input.
We used DEGs from developing HD72 MSNs identified by bulk RNAseq and their log2-fold
change as input to obtain 50 signatures of small molecules that cause changes opposite to
those seen in HD (Supplemental Table 3.6). Several of the resulting small molecules (e.g.,
withaferin-a, celastrol, trichostatin A, vorinostat, and niclosamide), have beneficial effects in
mouse and cellular HD models (Fig. 3.6A) [74-78]. We further categorized the predicted small
molecules based on their canonical targets and identified two major classes that showed
efficacy in HD models: histone deacetylase (HDAC) inhibitors and epidermal growth factor
receptor (EGFR) inhibitors (Fig. 3.6B) (Supplemental Table 3.6) [75, 76, 79].
To explore potential non-canonical mechanisms of action, we utilized the L1000FWD portal, a
web application to predicted mechanisms of action (MOAs) based on the transcriptional profile
produced by the small molecules (Supplemental Table 3.6) [80]. We used the probability scores
that accompany each predicted MOA to generate a t-distributed stochastic neighbor embedding
(tSNE) plot and colored the small molecules by their top predicted MOA (Fig. 3.6C). Small
molecules without MOAs with probability scores above 40% were not labeled.
The top predicted MOA for Afatinib and Pelitinib correctly labeled them as EGFR inhibitors.
Similarly, Vorinostat, Scriptaid, Abexinostat, and Trichostatin-A are known HDAC inhibitors that
84

were correctly classified. While arachidonyl trifluoromethyl ketone is not typically acknowledged
as an HDAC inhibitor, trifluoromethyl ketones indeed act as inhibitors for HDACs [81].
Manumycin A and Cerulenin were labeled as NF-κB inhibitors and both molecules disrupt in NF-
κB signaling [82-85]. JAK3-inhibitor-VI was classified as a poly(ADP-ribose) polymerase 1
(PARP) inhibitor. Canertinib, although canonically recognized as an EGFR inhibitor, received
high scores as an aurora kinase inhibitor, which has been linked to EGFR activity in cancer [86].
Additionally, STOCK1S-53863 and Sepantronium were predicted to be adrenergic receptor
antagonists, and C646 and Mw-A1-12 were classified as calcium channel blockers. Finally, HG-
5-113-01 was predicted to be a rapidly accelerated fibrosarcoma (RAF) inhibitor (Fig. 3.6C).
85

Figure 3.6. Small Molecules Predicted to Reverse HD Dysregulation. (A) List of small
molecules predicted to reverse gene dysregulation in developing HD72 MSNs. Each dot
represents an individual signature obtained from treatment of a cell line with the small molecule.
(B) Canonical targets for predicted small molecules. (C) tSNE plot based on scores of predicted
mechanisms of action for each small molecule. Molecules with more than a 40% confidence are
color coded to its top mechanism of action.
86

The presence of multiple validated small molecules in cell and mouse models of HD confirms
the effectiveness of this approach in finding HD modifiers. To further investigate, the efficacy of
an untested molecule, we chose to test Cerulenin's ability to reverse phenotypes in developing
HD72 MSNs. Among the predicted molecules, Cerulenin's transcriptomic signature had one of
the highest scores. Although it has not been tested in HD models, its mechanisms of action are
similar to Withaferin A’s, which is beneficial in HD models [77]. Cerulenin’s transcriptional
signature showed modulation of multiple members of the BDNF signaling pathway
(Supplemental Table 3.6). BDNF treatment improves HD phenotypes, including a rescue in the
expression of DARPP-32, a phenotype that our model of HD MSNs recapitulates at the protein
and RNA level (Fig. 3.1C, Supplemental Fig. 3C) [47, 56, 87, 88]. Thus, we treated developing
HD72 MSNs with doses of Cerulenin of 31.25–2000 nM starting on day 11 of differentiation.
After 9 days of treatment, DARPP-32 expression was assessed via immunofluorescence.
Cerulenin caused a dose-dependent increase in the expression of DARPP-32, which plateaued
at 250 nM (Fig. 3.7A). Upon treating C116 and HD72 developing MSNs with 250 nM Cerulenin,
we observed an increase in DARPP-32 expression exclusively in HD72 MSNs, but no significant
change was detected in C116 MSNs (Fig. 3.7B, C). We also tested the electrical activity of
C116 and HD72 MSNs on a multi-electrode array (MEA) plate. C116 MSNs displayed a higher
rate of active electrodes per minute (Fig. 3.7D), in line with previous reports showing decreased
electrical activity in iPSC derived neurons [89, 90]. Treatment with 250nM Cerulenin was able to
increase the rate of active electrodes per minute significantly.
RNAseq of HD72 developing MSNs treated with Cerulenin confirmed upregulation of DARPP-32
(PPP1R1B) (Fig. 3.7E). Cerulenin treatment also induced a reversal in expression levels of
multiple genes dysregulated in HD72 developing MSNs (Fig. 3.7F). This yielded a significant
negative correlation between the expression levels of genes impacted by HD and Cerulenin in
HD72 MSNs (Spearman rank correlation, ρ=-0.38, p-value = 0.0035). Enrichment analysis for
87

differentially expressed genes in HD72 developing MSNs treated with Cerulenin showed the
genes being modulated were associated with FSH regulation of apoptosis, BDNF signaling
pathway, lysosome, p53 singnaling pathway, and cell-extracellular matrix interactions among
others (Fig. 3.7G) (Supplemental Table 3.7). In sum, these data suggest that alteration of these
pathways by Cerulenin underlies its effects in HD72 MSNs and, warrants further in-depth
studies. To identify perturbations that caused transcriptional effects similar to those caused by
Cerulenin on HD72 MSNs, we used a SigCom LINCS signature search [91] with the gene
expression signature obtained from HD72 MSNs treated with Cerulenin. The most similar
chemical perturbation was caused by Whitaferin A, confirming their similarity (Supplemental
Table 3.7). Despite Cerulenin's canonical target being fatty acid synthase (FASN), the signature
search did not find a similar signature between cells treated with FASN siRNA or a FASN
CRISPR knockdown system and those treated with Cerulenin. This suggests that Cerulenin
interacts with targets in addition to FASN as part of its mechanism of action in HD72 MSNs.

88

Figure 3.7. Effects of Cerulenin Treatment on HD72 MSNs. (A) Quantification of DARPP-32
levels from immunostaining of Cerulenin treated HD72 developing MSNs. P values were
89

calculated using Dun’s test followed by Benjamini-Hochberg correction for the comparisons to
HD72 DMSO. (B) Immunostaining of HD72 and C116 MSNs treated with 250nM Cerulenin.
Scale bar: 200 μm. (C) Quantification of DARPP-32 in C116 and HD72 MSNs treated with
250nM Cerulenin. P values were calculated using pairwise Wilcoxon’s tests followed by
Benjamini-Hochberg correction. (D) Active electrodes/minute in cultures of C116, and HD72
developing MSNs treated with DMSO or 250nM Cerulenin. P values were calculated using
pairwise Wilcoxon’s tests followed by Benjamini-Hochberg correction. (E) Top genes modulated
in RNAseq of Cerulenin treated HD72 developing MSNs. (F) Log2 fold changes of genes
significantly changed between HD72 and C116 developing MSNs (x axis) and by treatment of
HD72 developing MSNs with Cerulenin (y axis). (G) Bar plot showing pathways altered by
Cerulenin treatment in HD72 MSNs.
90

Discussion
Recent research has shed light on the impact of HD on neuronal development and its potential
pathological consequences [7, 12, 28]. These findings suggest the possibility of new therapeutic
approaches and underscore the need to gain a better understanding of how mHTT alters brain
development and whether those changes sets the stage for the disorders observed later in life.
In this study, we utilized isogenic HD72 and corrected iPSCs-derived MSNs to investigate the
effects of HD on the development of human MSNs as they differentiate from NSCs to mature
MSNs. Consistent with previous studies [26], our results show that iPSCs-derived developing
MSNs recapitulate multiple aspects of fetal striatum development. Specifically, we observed
similar gene-expression patterns and cell-type compositions between the two systems, including
early and intermediate progenitors and mature MSNs. However, we also noted that iPSC-
derived early progenitors lacked expression of TCF7L1 and GSX2, indicating a difference in the
cellular identity at this stage between iPSC-derived early progenitors and the human LGE.
After establishing the degree to which our model faithfully recapitulates MSNs development, we
investigated how the HD mutation affects this process. We observed abnormal neuronal
maturation in HD72 developing MSNs. Specifically, premature maturation was indicated by an
accelerated decline in NES expression and an increase in neuronal maturations markers. This
finding agrees with previous studies that found accelerated maturation of human HD NSCs
implanted in mice [22] and is consistent with the hypertrophy seen in child HD carriers [15].
These developmental dysregulations appear to be associated with the aberrant expression of
factors that regulate neuronal differentiation. For example, ASCL1, a transcription factor that
promotes neuronal differentiation [61], is upregulated in intermediate MSN progenitors, but its
expression is significantly downregulated in HD72 intermediate progenitors. However, even in
the absence of ASCL1, mature HD72 MSNs express neuronal maturation markers, suggesting
that an atypical maturation process is responsible for their development. Notably, NEUROG2,
91

which is also involved in neuronal maturation, is upregulated in HD72 intermediate progenitors,
indicating an abnormal compensatory MSN maturation program due to the lack of ASCL1. This
is notable since cases of ASCL1 and NGN2 with a compensatory relationship in the central
nervous system have been documented [92].
We found dysregulation of ASCL1 correlated with dysregulation of DLX transcription factors,
that are regulatory targets of ASCL1 and necessary for proper striatum development [66, 67]. A
lack of DLX expression during development leads to ectopic expression of cortical markers in
the striatum and to a reduction of genes essential for striatal function and development [67]. In
agreement with these studies, we observed that HD72 mature MSNs have lower levels of key
MSN genes (e.g., GAD1 and GAD2) and ectopic expression of genes normally expressed in the
cortex (e.g., SLC17A6, EBF3 and NHLH2). In addition, we found that multiple other genes
known to be targets of DLX2 during striatum development [69] are dysregulated in developing
HD72 MSNs. Moreover, an examination of published datasets from both Grade 2–4 HD patient
MSNs and the R6/2 mouse model [70] indicated reductions in DLX5, DLX6, GAD1, and GAD2
expression, suggesting that some of the dysregulation during early development could continue
or recur during disease onset.
Our analysis also revealed changes in multiple signaling pathways related to HD pathology.
Bulk RNAseq of NSCs and MSNs showed enrichment for pathways known to play roles in HD,
such as BDNF signaling, TGF-β signaling, axon guidance, synaptogenesis, and NRF2-mediated
oxidative stress response [46, 53, 93-95]. Dysregulation of the BDNF pathway has been linked
to disturbances in neuronal circuits during development [96]. The disruption of BDNF signaling
and the abnormal maturation of HD MSNs could potentially lead to defects in neuronal circuitry
within the striatum, with pathogenic consequences later on. This phenomenon is observed in
the cortex of HD mice, where defects in neuronal circuitry contribute to the disease's
92

development [12]. Early correction of these changes offers protection against the disease's
progression later in life [12].
scRNAseq of developing MSNs allowed us to examine the dysregulation of signaling pathways
as HD72 MSNs matured. For instance, as HD72 MSNs mature, the NRF2-mediated oxidative
stress response is predicted to increase, suggesting that the HD mutation leads to a greater
increase in oxidative stress with advancing MSN maturation. Conversely, the ERK/MAPK
signaling pathway follows an opposing pattern, with high levels of activation in HD72 progenitors
that decline as MSNs mature. ERK activation is altered by HD, and it’s activation is protective in
multiple models [55]. Thus, immature HD progenitors might use it as a protective mechanism
that is lost as MSNs mature. These findings suggest that dysregulation of these pathways in HD
has its origins during MSN development and highlight the potential for early intervention and
influence disease progression.
To identify HD modifiers from transcriptomic data, we searched for small molecules that induce
gene expression changes opposite to those in developing HD72 MSNs. Our predictions
identified several that improve HD pathology. Notably, HDAC inhibitors, Celastrol, and
Withaferin-A were the most robust interventions and, displayed multiple gene expression
signatures predicted to reverse HD dysregulation. These molecules also demonstrated efficacy
in HD models [75-78].
To assess the effects of a predicted small molecule untested in HD, we treated developing
HD72 MSNs with Cerulenin. Cerulenin is a fatty acid synthesis inhibitor, and modulates NF-kB
and eIF2α signaling [83, 97]. Our predictions indicated that it would affect components of the
BDNF signaling pathway which is linked to the downregulation of DARPP-32 expression [98], a
hallmark of HD [56, 99-101]. HD72 MSNs treated through half of the differentiation period with
Cerulenin displayed a dose-dependent increase in DARPP-32 levels that plateaued at 250 nM.
In addition, levels of electrical activity were lower in iPSCs-derived HD72 MSNs. The low level
93

of electrical activity in HD MSNs has previously been reported [89, 90]. We find Cerulenin
rescues this but not to control levels. Cerulenin did not show similar effects on C116 MSNs,
suggesting that the mechanism is specific for HD. RNAseq analysis of Cerulenin-treated HD72
MSNs revealed changes in p53 signaling and other pathways associated with HD pathology
[102]. However, levels of DLX transcription factors or their targets were not improved.
Interestingly, reduction of HDAC4 increases levels of DLX1/2/5/6 in the R6/2 model [103],
suggesting HDAC inhibitors might rescue those deficits.
In summary, we examined the impact of HD on MSNs development using iPSC and isogenic
controls, uncovering dysregulation of key genes for proper maturation. We discovered aberrant
expression of ASCL1 and DLX transcription factors and dysregulation of HD-related pathways,
such as NRF2-mediated oxidative stress response and ERK/MAPK signaling as MSNs mature.
We also provide proof of concept for identification of HD modifiers from transcriptional data by
showing a partial rescue of DARPP-32 levels and electrical activity in HD MSNs after treatment
with Cerulenin as well as predicting multiple small molecules already confirmed to have
beneficial effects in HD models.

94

Supplemental Figures

Supplemental Figure 3.1. Comparison with proteomics and differential splicing analysis.
(A) Plot showing correlation between log2-fold change of genes in HD72 NSCs and HD72
MSNs. (B) Plot showing correlation log2-fold changes in RNA from HD72 MSNs versus
changes in proteins. (C) Venn diagram showing overlap of differentially spliced genes between
HD72 MSNs, human postmortem striatum and R6/1 striatum. (D) Enrichment for genes
differentially spliced in MSNs that are shared with human HD striatum or R6/1 striatum.

95

Supplemental Figure 3.2. scRNAseq of C116 and HD72 NSCs. (A) C116 and HD72 NSCs
UMAP plots divided into four different clusters. (B) Heatmap showing differentially expressed
genes between the identified clusters. (C) Expression of Nestin in C116 and HD72 NSCs. (D)
96

Violin plots showing levels of top downregulated (UBB, RBP1 and CRYBB1) and upregulated
(OCIAD2, IGFBP5 and S100B) genes in HD72 NSCs clusters.

97

Supplemental Figure 3.3. Proportion of Early Progenitors in Developing MSNs. (A) Nestin
labeled C116 and HD72 developing MSNs. Scale bar: 100 μm. (B) Distribution of Nestin-C116
and HD72 in developing MSNs. P value was calculated using the Kolmogorov-Smirnov test. (B)
Levels of DARPP-32 RNA in C116 and HD72 MSNs measured with qPCR. P value was
calculated with a t test.

98

Supplemental Figure 3.4. Expression of DLX Genes and Targets in Different Clusters in
C116 and HD72 Developing MSNs. Violin plots showing expression of DLX1 (A), DLX2 (B),
99

DLX5 (C), DLX6 (D), DLX6-AS1 (E), GAD1 (F), GAD2 (G), NEUROG2 (H), SLC17A6 (I), EBF3
(J) NHLH2 (K) in C116 and HD early progenitors, intermediate progenitors and mature MSNs.

100

Supplemental Tables
Supplemental Table 3.1. DEGs in HD72 and C116 NSCs. IPA on canonical pathways
predicted to be dysregulated in NSCs. Bioplanet enrichment analysis for HD72 NSCs.
Supplemental Table 3.2. DEGs in HD72 and C116 developing MSNs. IPA on canonical
pathways predicted to be dysregulated in MSNs. Bioplanet enrichment analysis for HD72
developing MSNs. Comparison of DEGs found in HD72 developing MSNs and other HD
models.
Supplemental Table 3.3. Differentially spliced genes in NSCs and developing MSNs.
Enrichment analysis of shared differentially spliced genes in HD72 developing MSNs, R6/1
striatum, and HD patient striatum.
Supplemental Table 3.4. Single cell DEGs in HD72 NSCs, early progenitors, intermediate
progenitors and mature MSNs. IPA of DEGs in early progenitors, intermediate progenitors and
mature MSNs. IPA comparison analysis for canonical pathways dysregulated in early
progenitors, intermediate progenitors and mature MSNs.
Supplemental Table 3.5. Comparison of DLX targets during striatum development with DEGs
in HD72 developing MSNs.
Supplemental Table 3.6. Small molecules predicted to reverse transcriptional dysregulation in
HD72-developing MSNs. Additional information on predicted small molecules and predicted
mechanisms of action.
Supplemental Table 3.7. DEGs found after Cerulenin treatment in HD72-developing MSNs.
IPA of DEGs found after Cerulenin treatment. Signatures of other small molecules with
similarities to Cerulenin.
101

Supplemental Table 3.8. qPCR primers and probes used.

102

CHAPTER 4: Proteomic Analysis of Huntington’s Disease Medium Spiny Neurons
Identifies Alterations in Lipid Droplets (Tshilenge et al., 2023)
Abstract
Huntington’s disease (HD) is a neurodegenerative disease caused by a CAG repeat expansion
in the Huntingtin (HTT) gene. The resulting polyglutamine (polyQ) tract alters the function of the
HTT protein. Although HTT is expressed in different tissues, the medium spiny projection
neurons (MSNs) in the striatum are particularly vulnerable in HD. Thus, we sought to define the
proteome of human HD patient–derived MSNs. We differentiated HD72 induced pluripotent
stem cells and isogenic controls into MSNs and carried out quantitative proteomic analysis.
Using data-dependent acquisitions with FAIMS for label-free quantification on the Orbitrap
Lumos mass spectrometer, we identified 6,323 proteins with at least two unique peptides. Of
these, 901 proteins were altered significantly more in the HD72-MSNs than in isogenic controls.
Functional enrichment analysis of upregulated proteins demonstrated extracellular matrix and
DNA signaling (DNA replication pathway, double-strand break repair, G1/S transition) with the
highest significance. Conversely, processes associated with the downregulated proteins
included neurogenesis-axogenesis, the brain-derived neurotrophic factor-signaling pathway,
Ephrin-A: EphA pathway, regulation of synaptic plasticity, triglyceride homeostasis cholesterol,
plasmid lipoprotein particle immune response, interferon-γ signaling, immune system major
histocompatibility complex, lipid metabolism and cellular response to stimulus. Moreover,
proteins involved in the formation and maintenance of axons, dendrites, and synapses (e.g.,
Septin protein members) were dysregulated in HD72-MSNs. Importantly, lipid metabolism
pathways were altered, and using quantitative image analysis, we found that lipid droplets
accumulated in the HD72-MSN, suggesting a deficit in the turnover of lipids possibly through
lipophagy. Our proteomics analysis of HD72-MSNs identified relevant pathways that are altered
in MSNs and confirm current and new therapeutic targets for HD.
103

Background
Huntington’s disease (HD) is a rare progressive monogenic neurological disorder caused by a
trinucleotide repeat expansion in exon-1 of the Huntingtin gene (HTT) [1-3]. The clinical
hallmark of HD is a chorea that co-exists with cognitive decline and emotional disturbances [4].
There is no cure, and no treatment alters the course of this devasting disease. HD phenotypes
are linked to expression of mutant HTT protein (mHTT) that harbors expanded glutamine
stretches (over 38) in the N-terminal region. HD features neuronal degeneration in the brain,
and medium-spiny projections neurons (MSNs) within the striatum are particularly vulnerable
[5].
While substantial progress has been made towards elucidating how the CAG repeats within
mHTT lead to the clinical outcomes in HD, our understanding of the mechanisms underlying the
motor deficits and striatal degeneration is incomplete [6]. Those mechanisms likely occur in
parallel. For example, mHTT has altered localization, conformation, and protein interactions [7-
11]. Proteolysis of mHTT generates N-terminal fragments containing the polyQ expansion is
found in HD human brain and mouse models [9, 12-18]. The cleaved forms of the protein are
found in multiple cellular compartments, including the nucleus, and cause aberrant interactions
of mHTT with key partners, such as transcription factors, autophagy and mitochondrial proteins,
and thus lead to neuronal death in the striatum and cortex [19-21]. Deciphering how the mHTT
alters the proteome is critical to understanding HD molecular mechanisms, and few studies
have focused on defining the proteomes of human HD patient-derived neurons.
Recent progress in mass spectrometry (MS)-based proteomics has allowed significant
improvements in proteome resolution, sensitivity and depth-coverage of a given biological
system [22, 23]. Previous studies used quantitative MS-based proteomics to measure relative
changes in the protein abundances in human postmortem HD frontal cortex and identified
signaling pathways that are dysregulated in HD, including Rho-mediated, actin cytoskeleton and
104

integrin signaling, mitochondrial dysfunction and axonal guidance [24]. Comprehensive
quantitative proteomics applied to investigate spatiotemporal mechanisms of mHTT in R6/2 HD
mice characterized the insoluble proteome during disease progression and highlighted
extensive dysregulation in brain regions vulnerable to HD [25]. A recent review highlights the
proteomics carried out in HD and some of the critical gaps in the field, including the generation
of robust human HD cell-type-specific proteome data sets [26].
Disease modeling in human induced pluripotent stem cells (iPSCs) allows central nervous
system (CNS)-relevant cells to be generated in vitro and molecular defects to be identified that
contribute to polyQ-expansion disorders [27-30]. In our previous work, we used human patient–
derived HD-iPSCs (72CAG/19CAG, HD72) and genetically corrected the cells to a normal
repeat length (21CAG/19CAG, C116), thus creating an isogenic control [31]. Our group showed
that HD phenotypes manifest in differentiated neural stem cells (NSCs), not in iPSCs [32, 33].
Further, our transcriptomic analysis of isogenic HD72-NSCs suggested that HD is linked to
developmental impairments that prevent the proper generation of MSNs and subsequent loss of
MSN identity [27, 31, 32, 34, 35]. So far, quantitative proteomics of iPSCs modeling HD focused
on undifferentiated stem cells or unrestricted neuronal populations [36, 37]. In contrast, the
proteome in directed HD72-MSNs derived from iPSCs has not been explored.
To define the proteomic signature in isogenic HD72-MSNs and to determine how mHTT
leads to neurotoxicity in MSNs, we performed comprehensive quantitative proteomics by liquid
chromatography-tandem mass spectrometry (LC-MS/MS) using complementary approaches.
First, we compared triplicates of HD72-MSNs and isogenic controls in which the CAG expansion
was genetically corrected to a normal repeat length. We used an unbiased discovery workflow
combining a modern high-field asymmetric waveform ion mobility spectrometry (FAIMS) device
[38] with an Orbitrap Lumos mass spectrometer operating in data-dependent acquisition (DDA)
mode for the identification and label-free quantification MS1-based (LFQ) of significantly
105

changing protein candidates (Discovery). Subsequently, we acquired the same samples by
data-independent acquisition (DIA) on a TripleTOF 6600 mass spectrometer for further
quantification and validation (MS2-based Quantification, DIA) [39]. Our proteomic workflow is
summarized in Fig. 4.1. Findings on altered triglyceride homeostasis were followed up using
quantitative confocal microscopy image analysis relevant to autophagy. This pathway was found
in both the FAIMS-DDA MS and DIA workflow.

106

Figure 4.1. Schematic representation of HD, isogenic HD-MSN and proteomics workflow.
(upper left) HD is a monogenic disease caused by a CAG (coding for glutamine) expansion in
the HTT gene. The striatum, depicted in the upper left panel, is heavily affected in HD. The
inhibitory medium spiny neurons are lost during disease progression. (upper right) Total proteins
were isolated from C116-MSN and HD72-MSN cultures (in triplicates). (bottom) Samples were
digested and subjected to a comprehensive quantitative proteomic analysis with deep coverage
with FAIMS ion mobility separation coupled to data-dependent acquisition mode on an Orbitrap
Lumos mass spectrometer for label-free quantification. Subsequently, significantly changed
proteins were validated with an independent quantitative approach collecting data-independent
acquisitions on a TripleTOF 6600 mass spectrometer. After protein identification and
quantification, bioinformatic analyses were used to identify molecular pathways and networks
relevant in HD.
107

Results
Generation and Characterization of MSNs from HD-iPSCs
The striatum is dramatically impacted in HD. MSNs, GABAergic inhibitory neurons, are one of
the main cell types lost from this region and represent 90% of the striatal neuronal population.
To model HD, we used human patient–derived HD-iPSCs (72CAG/19CAG, HD72) that were
genetically corrected to a normal repeat length (21CAG/19CAG, C116), thus creating an
isogenic control [31]. Then, both cell types were differentiated into MSN-like neurons by a
method that mimics the major brain developmental stages for this neuronal type: neural
induction, regional patterning toward a lateral ganglionic eminence (LGE) identity in presence of
Activin A and terminal differentiation [30, 31, 34, 40, 41] (Fig. 4.2A). Using
immunocytochemistry, we found that the cultures were positive for the MSN marker DARPP-32
and neuronal marker MAP2 (Fig. 4.2B). HD72-MSNs showed less DARPP-32 (p ≤ 0.01) and
MAP2 (p ≤ 0.001) than in C116-MSNs (Fig. 4.2C). This result is consistent with the expression
of these markers in postmortem HD striatum and in mouse models of HD [42-44].
108

109

Figure 4.2. Generation and characterization of iPSC-derived MSNs. A, Schematic of steps
illustrating the generation of neural stem cells and differentiated MSNs. The method used to
differentiate iPSCs into MSNs mimics the major brain developmental stages, including neural
induction, regional patterning toward an LGE identity in presence of Activin A, and terminal
differentiation. The images were acquired using the bright field from the Biotek, on 10X and 20X
magnifications with scale bars of 500 and 200 μm, respectively. B, C116-MSN and HD72-MSN
were immunostained after differentiation into MSNs with DARPP-32 (green) and MAP2 (red).
Scale bars: 50 µm.C, Expression levels of DARPP-32 and MAP2 were determined using Biotek
and Image J analysis. Unpaired t-test with Welch's correction **P ≤ 0.01, ***P ≤ 0.001. Four
micrographs were captured for each genotype. For quantification a minimum of N = 300 nucleus
(DAPI positive) were used.

110

Proteomic Analysis of HD-MSNs with FAIMS-DDA MS on a Orbitrap Lumos system
The HD72- and C116-MSNs were grown in parallel with three replicates for each genotype and
subjected to the proteomic workflow in Fig. 4.1. Intracellular proteins were extracted, digested
and subjected to a comprehensive quantitative proteomic analysis by LC-MS/MS using a
combined approach: protein discovery and LFQ using FAIMS-DDA MS on a Orbitrap Lumos
system, followed by protein candidate validation using DIA MS on a TripleTOF 6600 system,
and data and bioinformatic analysis (Fig. 4.1, supplemental Table S4.1).
First, we coupled an additional gas-phase separation using FAIMS ion mobility to DDA
acquisitions, and specifically applied three internal CV steps, -50 V, -65 V and -85 V. The gas-
phase separation protocol reduces the complexity of the ion population entering the mass
spectrometer, which provides deeper MS/MS sampling and proteome coverage [38, 45, 46]
(Fig. 4.3A). This process allowed us to identify 6,323 unique protein groups (≥ 2 unique
peptides, FDR ≤ 0.01, supplemental Table S4.2), among which 6,294 protein groups were
quantifiable by LFQ algorithms in Proteome Discoverer (Table 4.1 and supplemental Table
S4.3), providing a comprehensive and deep proteomic dataset for the HD72-MSNs. Assessing
the LFQ-MS1-based protein quantification reproducibility within each experimental condition,
using three biological replicates of isogenic C116-MSNs and HD72-MSNs, revealed that the
coefficient of variation for peptide peak areas was under 20% for 74% of all peptides of the
C116-MSN group and 86% of all peptides of the HD72-MSN group (Fig. 4.3B). Reproducibility
of protein group identifications is displayed for three biological replicates of HD72-MSNs (Fig.
4.3C). FAIMS-DDA quantification details for all replicates are shown in supplemental Table
S4.3. Of the 6,294 quantifiable protein groups (using FAIMS-DDA), when comparing HD72-
MSNs to C116-MSNs, 901 proteins were significantly changed (supplemental Table S4.4B): 443
proteins were upregulated, 458 were downregulated (FDR set at 1% and absolute Log2(fold-
change) ≥ 0.58), and 5,393 were unchanged (Fig. 4.4A, Table 4.1).
111

Figure 4.3. Deep proteome coverage using FAIMS gas-phase separation with DDA:
Performance of the FAIMS-DDA MS workflow. A, Total ion chromatogram (TIC) and base
peak chromatogram (BPC), followed by BPC with the three differential CVs (-50 V, -65 V, and -
85 V) for a 2 µg C116-corrected MSN sample injection on a FAIMS-Orbitrap Lumos system
112

operating in DDA mode. B, Coefficients of variation (CV) of peptides quantified in three
biological replicates of C116-corrected-MSNs and HD-MSNs. C, Reproducibility of protein
groups identified in three biological replicates of HD-MSNs.

113

Figure 4.4. Differential analysis of the proteome of isogenic C116- and HD72-MSNs by
FAIMS-DDA MS. A, Summary of the proteins quantified and significantly altered using FAIMS
114

gas-phase separation with DDA. B, Heat map illustrating the abundance of the proteins of the
C116- and HD72-MSNs identified by FAIMS DDA MS with at least two peptides (FDR ≤ 0.01).
The heat map represents more precisely the values of MS peak area for n = 3 C116-MSNs and
n = 3 HD72-MSNs. C, Heat map illustrating the top 50 statistically significant altered proteins in
the HD72- versus C116-MSNs using FAIMS-DDA MS. D, Volcano plot illustrating the proteins
differentially expressed when comparing HD72- versus C116-MSNs (significant proteins: FDR
at 1% and log2 fold-change absolute value > 0.58).

115

Proteomic Analysis of HD-MSNs with DIA-MS
To further validate these protein candidates for HD72-MSNs, we used a comprehensive
quantitative methodology, DIA-MS, in which fragment ions (MS2) were quantified with accurate
relative quantification results. This approach for C116 and HD72-MSN provided protein
quantification for 3,106 protein groups with very high reproducibility (supplemental Fig. S4.1)
with at least two peptides identified (supplemental Table S4.5). In fact, 72% and 80% of the
identified precursor ions presented a coefficient of variation below 20% for C116-MSNs and
HD72-MSNs, respectively (supplemental Fig. S4.1). Out of the 3,106 quantifiable protein
groups, 162 protein groups were significantly altered in HD72-MSNs vs C116-MSNs using DIA-
MS (q-value ≤ 0.05 and absolute Log2(ratio) ≥ 0.58) (supplemental Table S4.5B). Among those,
a panel of 129 protein groups identified in the FAIMS DDA discovery study was thus validated
by the highly quantitative DIA-MS strategy (supplemental Fig. S4.2A, supplemental Table
S4.4B,C). More precisely, 54 protein groups were significantly down-regulated, and 75 protein
groups were significantly up-regulated in both FAIMS DDA and DIA-MS datasets, when
comparing HD72-MSNs to C116-MSNs (supplemental Fig. S4.2B,C). A total of 292 proteins
were common between the significant FAIMS DDA proteins and all measured DIA-MS proteins.
266 agree or trend in the same direction as the FAIMS DDA (see correlation plots supplemental
Fig. S4.2D). The correlation plot shows 91% of the proteins trend in the same direction.
Pathways that had proteins in both data sets were further validated as described below (SEPTs,
APOE and minichromosome maintenance (MCM)).
Visualization of the Proteomic Analysis of HD-MSNs with FAIMS-DDA MS
Using hierarchical clustering of protein abundances, we evaluated the variation in C116-MSNs
and HD72-MSNs. Heat map representation of the proteomics showed distinct clustering of the
two sample groups that depended on the polyQ-repeat length, with HD72-MSN samples being
116

clearly delineated from C116-MSNs (Fig. 4.4B, supplemental Table S4.3). This is consistent
with previous studies that found distinct phenotypes for HD and corrected NSCs [32].
To visualize the clustering, quality and significantly altered proteins in HD72-MSNs and C116-
MSNs, a heatmap is shown in Fig. 4.4C for the top 50 proteins with the highest statistical
significance. The biological function, cellular component, molecular function, KEGG, Reactome
and WIKI pathways of each protein are summarized in supplemental Table S4.2. The analysis
of significantly altered proteins is depicted in the volcano plot showing the estimated Log2(fold-
change) versus -Log10(p-value) for each protein, with significantly differentially regulated
proteins having a p-value that guarantees a 1% FDR and an absolute Log2(fold-change) value
above 0.58 (Fig. 4.4D, Table 4.1, supplemental Table S4.4). The FAIMS-DDA MS workflow
applied for the discovery step resulted in 901 significantly changed protein candidates (Table
4.1), and 129 of these proteins were additionally validated by DIA MS with confidence as
significantly changing (Table 4.1).
Newly discovered and previously implicated proteins in HD are shown in the heatmap and
volcano plots (Fig. 4.4C,D). One of the top upregulated proteins was insulin-like growth factor-
binding protein 7 (IGFBP7). IGFBP7 is released by senescent cells, and cellular senescence is
a pathway we previously identified as activated in HD-MSNs with a multitude of relevant
markers, including IGFBP7 mRNA [47]. IGFBPs are biomarkers for multiple diseases, and their
expression causes neurodegeneration [48-54]. Western blot analysis further validated the
increased levels of IGFBP7 in HD72-MSNs (supplemental Fig. S4.3). OCIAD2, another top
upregulated protein, is implicated in Parkinson’s (PD) and Alzheimer’s disease (AD) and
activates STAT3 [55-57]. GPX7 (glutathione peroxidase) is upregulated in HD72-MSNs. GPX
activity is increased in HD patient blood [58], and GPX7 (related family member, GPX6) is
neuroprotective when overexpressed in HD yeast, Drosophila and mouse models [59, 60].
Proteins involved in lipid metabolism, such as apolipoprotein E (APOE), are downregulated and
117

will be discussed further below (Fig. 4.4D). The levels of the HTT were detected in our studies,
but the levels were not changed (supplemental Table S4.3) between the two genotypes.
Functional Enrichment and Protein Network Analysis Reveal Molecular Hallmarks of HD
Functional enrichment studies [61-63] with the significantly altered proteins in HD72-MSNs
revealed molecular dysregulation in several pathways (supplemental Fig. S4.4). We applied the
up- and downregulated protein lists from FAIMS-DDA MS proteomic workflow to g:Profiler with
custom background proteins (supplemental Table S4.6) [64]. The most significant enriched GO
biological process terms upregulated in HD72-MSNs, include pathways related to the
extracellular matrix (ECM) (e.g., Integrin-Laminin signaling, TGF-beta regulation of ECM,
epithelial-mesenchymal transition (EMT) activation, activation of matrix metalloproteinases),
cardiovascular system, angiogenesis, TAp63 pathway, DNA replication, senescence, organism
development, regulation of cell migration and locomotion, aminoglycan glycosaminoglycan
proteoglycan, organism development, regulation of cell migration and locomotion, growth factor
stimulus and fatty acid processes (supplemental Table S4.7). Conversely, processes associated
with the downregulated proteins include neurogenesis-axogenesis and, more specifically, the
BDNF signaling pathway, Ephrin-A:EphA pathway, regulation of synaptic plasticity, triglyceride
homeostasis cholesterol, plasmid lipoprotein particle immune response, INF-γ signaling,
immune system MHC complex, triglyceride homeostasis, lipid metabolism, lymphocyte
proliferation and cellular response to stimulus (supplemental Fig. S4.4). Pathways involved in
organism development and regulation of cell migration and locomotion, and regulation of
biological and homeostatic processes were both up- and downregulated. Our proteomic
analysis demonstrated that pathways related to cardiomyopathy, cardiovascular and
angiogenesis are upregulated in HD72-MSN consistent with peripheral effects of HD
(supplemental Fig. S4.4, supplemental Table S4.7) [65, 66].
118

The canonical pathways from a complementary analysis using Ingenuity Pathway Analysis (IPA)
are shown in supplemental Fig. S4.5. The top pathways include hepatic fibrosis, semaphorin
neuronal repulsive signaling pathway, regulation of cellular mechanics of calpain protease, IL-4
signaling, axonal guidance signaling, caveolar-mediated endocytosis signaling, SNARE
signaling, estrogen receptor, RHO GTPase signaling, CLEAR signaling and many more that
have been implicated in HD.
Reactome Functional Interaction Network for Isogenic HD-MSNs Upregulated Proteins
Next, we used functional interaction analysis to define clusters of proteins that are closely
connected to each other with ReactomeFIViz, a reactome functional interaction network [67].
We identified an interconnected network of the 443 upregulated proteins (Fig. 4.5, supplemental
Table S4.8A). There were 11 clusters for the upregulated proteins with EMC and DNA signaling
(DNA replication pathway, double-strand break repair, G1/S transition) having the highest
significance (Fig. 4.5A,B). We describe each cluster below, along with its correlation with
functional enrichment and the relevant HD pathogenic mechanisms.

119

120

Figure 4.5. De novo sub-network construction and clustering using proteins differentially
upregulated when comparing HD72-MSNs to C116-MSNs. A, Networks of genes were
constructed using 443 upregulated proteins in HD72-MSNs as determined using FAIMS-DDA
MS. The functional network and clustering were performed using the Reactome Functional
Interaction Network (ReactomeFIViz). Nodes in the network correspond to genes, and edges
correspond to interactions. Shaded ovals represent clusters of genes sharing common enriched
biological functions. B, Classification of clusters based on false discovery rates. C, Quantitative
proteomics reveals MCM3, MCM4, MCM6 and MCM6 are expressed more highly in HD72-
MSNs than C116-MSNs. Unpaired t-test with Welch's correction *P ≤ 0.05, **P ≤ 0.01, ***P ≤
0.001.
121

Clusters 0, 1, 4, 6 – Extracellular Matrix Organization, Replicative Senescence
The upregulated interaction network confirmed the dysregulation of ECM-regulated pathways in
HD72-MSNs. The Cluster 0 protein network was associated with ECM organization,
disassembly and tissue development, highlighting the role of matrix metalloproteinases in
HD72-MSNs: CD44, MMP2, MMP14 and TIMP2 (Fig. 4.5A, supplemental Table S4.8A). We
previously described the role of MMP14 and TIMPs in HD [68, 69].Cluster 1 network identified
key proteins involved in ECM-organization, collagen and integrin signaling: ITGA6, ITGB4,
ITGB3, COL1A1, COL4A1, COL4A2, COL5A2, CAV1, FLN and TNC (Fig. 4.5A). Dysregulation
of ECM components impairs the formation and maintenance of neural circuitry and increases
risk for several neurological pathologies, such as HD, AD, and autism spectrum disorder [70].
In clusters 4 and 6, we also identified proteins involved in cellular senescence, including
CDKN1A (p16), SERPINE1 and IGFBP7. Recent studies in mouse models of AD and PD
suggest cellular senescence is important in disease progression and pathogenesis [71-76].
Clusters 2, 3 – Muscle Contraction, Netrin Signaling and Angiogenesis
Interestingly, clusters 2 and 3, Netrin/SEMA1 signaling, Ras protein, Erbb2 signaling, and
angiogenesis were novel dysregulated pathways in HD72-MSN that were not identified in
pathway enrichment analysis (Fig. 4.5A). IPA analysis also identified these pathways, and the
networks are shown in supplemental Fig. 4.4.
Cluster 5 – Septin Signaling Pathways in HD
Notably, three members of the septin protein family, SEPT2, SEPT6 and SEPT9, were present
in cluster 5. Septin family members were found in the FAIMS DDA and DIA-MS datasets, when
comparing HD72-MSNs to C116-MSNs (supplemental Fig. S4.2B,C). Septin proteins participate
in various physiological processes, such as cytoskeleton regulation, cell division, membrane
122

trafficking, neuronal formation, and maintenance. Septin dysregulation is associated with
diverse diseases, including cancer, infection, and neurological disorders [77, 78]. Using
quantitative proteomics, we found that SEPT2, SEPT3, SEPT4, SEPT5 and SEPT9 are
dysregulated in HD MSN (Fig. 4.6A). Further, western blot analysis on HD72- and C116-MSNs
provide validation of the proteomic results. Both SEPT2 and SEPT9 were upregulated in HD72,
compared to C116-MSNs (Fig. 4.6B and C, p ≤ 0.05, p ≤ 0.01). Importantly, SEPT9 and
phosphoinositides regulate lysosome localization, their association with lipid droplets and lipid
droplet growth [79, 80]. This is relevant to Cluster 8 that identifies dysregulation of lipid
metabolism, including triglyceride and cholesterol pathways in the HD72-MSNs.

123

124

Figure 4.6. SEPTIN family members are dysregulated in HD-MSNs. A, Quantitative
proteomics reveals SEPT2, SEPT3, SEPT4, SEPT5 and SEPT9 are dysregulated in HD-MSNs.
B, Western blot analysis shows that SEPT2 and SEPT9 were upregulated in HD72-MSNs,
compared to C116-MSNs. *indicates non-specific band. C, Quantification of SEPT2 and SEPT9
levels, normalized to vinculin. Unpaired t-test with Welch's correction *P ≤ 0.05, **P ≤ 0.01, ***P
≤ 0.001, ****P ≤ 0.0001, ns, not significant.
125

Cluster 7 – Glycosaminoglycan Biosynthetic Process and Wnt Signaling
Cluster 7 identifies glycosaminoglycan biosynthetic process and Wnt signaling. The Wnt
signaling is altered in human and mouse models of HD and may be an early event in the
pathogenesis of HD [81-83]. The ECM, particularly, the sulfated glycosaminoglycan component,
is structurally and functionally altered in AD [84].
Clusters 8,9 – Oxidation Reduction Process and Fructose Metabolic Process
Clusters 8 and 9 highlight that proteins associated with fatty acid oxidation and fructose
metabolic processes were largely upregulated in the HD-MSNs. Fatty acid metabolism is
dysregulated in HD and linked to reduction of active sterol regulatory element responsive
protein 2 (SREBP-2) [85-87]. Fructose metabolism is linked to levels of uric acid, and levels of
uric acid in biofluids are lower in HD patients than controls [88].
Cluster 10 – DNA Signaling Is a Top Enriched Pathway in HD MSNs and Implicates MCM
Proteins
Cluster 10 is a pre-replicative complex assembly involved in nuclear cell-cycle DNA replication.
Genome-wide association studies suggest genes involved in DNA-damage-repair mechanisms
are modifiers of the age of onset of HD [89]. Further, HTT acts as a stress-response protein to
modulate DNA damage. We identified the MCM proteins 2–7 as a top dysregulated pathway.
The MCM complex regulates DNA replication, cell-cycle and DNA damage responses, and so, it
is likely an important signaling pathway for HD [90-92]. Cluster 10 showed MCM3, 4, 5 and 6 as
dysregulated pathways in HD-MSNs (Fig. 4.5A,B). Analysis of individual MS peaks confirmed
that levels of MCM3, 4, 5 and 6 are significantly greater in HD-MSNs than C116-MSNs (Fig.
4.5C). MCM family members were found in the FAIMS DDA and DIA-MS datasets, when
comparing HD72-MSNs to C116-MSNs (supplemental Fig. S4.2B,C).
126

Reactome Functional Interaction Network for Isogenic HD-MSNs Down-Regulated
Proteins
We identified an interconnected network of the 458 down-regulated proteins (FAIMS-DDA MS)
(Fig. 4.7, supplemental Table S4.8B) using ReactomeFIViz [67]. The down-regulated related
clusters were associated with immune system-related pathways, axon guidance signaling,
MAPK cascade, calcium modulation and the Wnt signaling pathway (Fig. 4.7A).
Correspondingly, for downregulated pathways, antigen processing and presentation, interferon-
gamma signaling and ephrin receptor signaling were the most significant (Fig. 4.7B). We
describe each cluster below, including its correlation with functional enrichment and relevance
to HD pathogenic mechanisms.

127

Figure 4.7. De novo sub-network construction and clustering using proteins differentially
downregulated when comparing HD72-MSNs to C116-MSNs. A, Networks of genes were
constructed using 458 downregulated genes in HD72-MSNs as determined using FAIMS-DDA
MS. The functional network and clustering were performed using the Reactome Functional
Interaction Network (ReactomeFIViz). Nodes in the network correspond to genes, and edges
128

correspond to interactions. Shaded ovals represent clusters of genes sharing common enriched
biological functions. B, Classification of clusters based on false discovery rate.

129

Clusters 0 – VEGFR, Integrin and Fc R Signaling Pathways
Alterations in VEGF and VEGFR are common in several triplet repeat diseases, and modulation
of VEGFR is neuroprotective in HD. Depletion of this growth factor may contribute to the HD-
MSN phenotype. The Fc gamma receptors (Fc Rs) are generally thought to be expressed in
immune cells but are also expressed in neurons where they may mediate excitatory pathways,
and therefore are downregulated in HD-MSNs.
Clusters 1, 2 – Axonal Guidance through Ephrin Signaling and the Stathmin Pathway
Previous transcriptomic studies of HD-derived NSCs identified altered pathways related to
neuronal development, axonogenesis and axonal guidance [32, 33]. Our proteomic analysis of
HD-MSNs is consistent with the hypothesis that mHTT prevents the proper development and
maintenance of MSNs by downregulating processes related to CNS development (Fig. 4.7) [93].
Furthermore, the functional interactions analysis identified the downregulation of axonal
guidance pathway by impairing ephrin proteins, including EPHA5, EPHA7, EPHB2 and EPHA3
(Fig. 4.7A and B). We also found that stathmin-1 (STMN1), a protein belonging to the stathmin
family, is downregulated in HD-MSNs (Fig. 4.7A, Cluster 2). Dysregulation of STMN1 occurs in
neurological disorders, including AD [94], amyotrophic lateral sclerosis [95] and spinal muscular
atrophy (SMA) [96].
Clusters 3, 8 – Dysregulation of APOE Signaling and Lipid Metabolism in HD72-MSNs
Pathway enrichment analysis revealed a dysregulation of lipid metabolism, including triglyceride
and cholesterol pathways in the HD72-MSNs, when compared to the corrected C116-MSNs
(Fig. 4.7). This pathway is particularly interesting as APOE and related lipid metabolism
enzymes were found in both the FAIMS DDA and DIA-MS datasets, when comparing HD72-
MSNs to C116-MSNs (supplemental Fig. S4.2B,C). Correspondingly, APOE expression was
130

downregulated in HD-MSNs and likely modulates the lipid metabolism alterations in HD72-
MSNs (Fig. 4.7A). Further, cluster 3 links HD to altered inositol phosphate metabolism and
phosphatidylinositol metabolic processes. Numerous studies link HD to alterations in lipid
metabolism, and IPA analysis suggested APOE is a top upstream regulator [97-99]. In addition,
several of the lipid metabolism proteins significantly altered in HD-MSNs are regulators of lipid
droplet formation [100]. These include monoacylglycerol lipase, lipid droplet–associated
hydrolase, diacylglycerol O-acyltransferase, low density lipoprotein receptor, NPC intracellular
cholesterol transporter 2 and sodium-coupled neutral amino acid transporter.
To quantify for lipid-rich components, including lipid droplets, we used a lipophilic dye Nile Red
and discriminated for neutral lipid and phospholipid (Fig. 4.8A,B). In serum-withdrawal culture,
we found that HD72-MSNs had significantly higher levels of lipid droplets (Fig. 4.8C) and a trend
towards increases in neutral lipids and phospholipids (supplemental Fig. S4.6), compared to
C116-MSNs. Treatment with APOE3 increased the numbers of lipid droplets in HD72 more than
in C116-MSNs (Fig. 4.8B,C). This suggests HD72-MSNs respond and modify lipid metabolism
distinctly from control cells. In addition, we showed that HD72-MSNs have lower levels of APOE
than C116-MSNs (Fig. 4.8D,E). APOE promotes cholesterol phospholipid efflux via ABCA1 and
ATP binding cassette subfamily G member 1 and uptake (via LDLR and LRP1).
Correspondingly, our unbiased proteomics revealed a significant increase in LDLR and lower
levels of ABCA1 (supplemental Fig. S4.7). Interestingly, the relative expression of HMG-CoA
reductase, a key regulator of sterol synthesis, was unchanged as measured by western blot
analysis (supplemental Fig. S4.8). These data indicate that HD72-MSNs have dysregulated
levels of endogenous APOE, in association with the ability to accumulate a significant number of
lipid droplets.

131

132

Figure 4.8. HD-MSN lipid metabolism and its modulation by APOE3. C116-MSNs and
HD72-MSNs cultured in serum withdrawal were treated for 48 h with 312 ng/mL APOE3. Non-
treated (NT) cells were used as controls. The cells were stained with Nile Red, a lipophilic dye.
Micrographs were cropped and saturated for presentation. A,B, Representative confocal images
for lipid droplets. Scale bar: 10 µm. C, Quantification of the density of the lipid droplets in normal
and APOE3-treated conditions (sum of the neutral and phospholipid Nile Red staining). Two-
way ANOVA comparing the overall effect of genotype (***P ≤ 0.001) and using Sidak’s multiple
comparison for the effect of APOE in each genotype (*P ≤ 0.05). D, Western blot analysis of
APOE in C116-MSNs and HD72-MSNs. E, Quantification of APOE levels normalized to vinculin.
Unpaired t-test with Welch's correction, ****P ≤ 0.0001.

133

We previously reported a deficit in autophagic flux in our isogenic HD iPSCs model and,
therefore, evaluated how markers of autophagy localized with the lipid droplets in the presence
of APOE [10]. We quantified the lipid droplets that co-label with autophagy markers (p62, LC3,
LAMP1) using a combination of Nile Red staining and immunofluorescence (Fig. 4.9A,
supplemental Fig. S4.9). As expected, the number of LC3 puncta were increased in HD MSNs,
compared to C116 (Fig. 4.9A). The number of autophagosomes colocalizing with lipid droplets
was higher for p62 and trending higher for LC3 in HD-MSNs (Fig. 4.9B). This indicates that
marking lipid droplets for autophagy by p62 responds to the increased lipid droplet load.
However, autophagy is insufficient to clear lipid droplets perhaps due to increased lipid uptake
due to the higher levels of LDLR.

134

Figure 4.9. Autophagy in the HD-MSN. C116-MSNs and HD72-MSNs cultured in serum
withdrawal were treated for 48 h with 312 ng/mL APOE3. A, cells were stained with Nile Red, a
lipophilic dye, and only the short-wavelength, neutral lipid bound fluorescence is shown in
representative confocal images (green). In addition, cultures were immunostained for LC3 (red)
135

or p62 and LAMP1 (supplemental Fig. 14). Lipid droplets (indicated by green circles in the other
channel) and immunofluorescence puncta (indicated by red circles) were quantified for co-
labeling (arrows) and density. The white outlines show the boundary of the cells. Micrographs
were cropped, smoothed and saturated for presentation. B, Quantification of lipid droplet and
immunofluorescence puncta density for p62, LC3, LAMP1 or lipid droplet. C, Counts of co-
labeled lipid droplets normalized to cell surface area. Two-way ANOVA with Sidak’s multiple
comparison comparing HD72-MSNs with C116-MSNs, *P ≤ 0.05, ***P ≤ 0.001. Image
quantification was done on total 80 unbiased view fields capturing total 458 DAPI-stained nuclei,
a representative of two culture replicates.

136

Interestingly, APOE3-treated, serum-starved HD72-MSNs have a significantly smaller fraction of
the more abundant neutral lipid droplets co-labeled with the autophagy markers p62, LC3 and
LAMP1 than in C116-MSNs in the same conditions (supplemental Fig. S4.10). In other words,
the few lipid droplets found in the C116-MSNs had relatively more colabeling with the
autophagosome markers as a percentage overall. Notably, the average per spot intensities of
p62, LC3 and LAMP1 labeling were not different between HD72-MSNs and C116-MSNs, and
lipid droplets in HD72-MSNs were slightly smaller than in C116-MSNs but the intensity of
individual lipid droplets did not differ (supplemental Fig. S4.9).
Cluster 4 – TGFβ, SMAD and TCF12 Signaling Pathways
Cluster 4 identifies TGFβ2 and SMAD1 proteins as dysregulated in HD72-MSNs (Fig. 4.7A),
consistent with previous studies in HD-NSCs [32, 33].
Clusters 5, 7 – Downregulation of HLA- and CNS-Related Proteins in HD-MSNs
Intriguingly, pathway enrichment and functional interaction analysis illustrated that CNS-related
pathways are downregulated in HD72-MSNs along with the immune system and IFN-γ
responses (Fig. 4.7). Although the CNS was considered to be immunologically inert, the major
histocompatibility complex class I (MHC-I) is expressed in mouse brain and neurons [101-104].
Recently, using a single-cell transcriptomic approach, Darmanis et al. showed that MHC-I genes
are expressed in a subset of neurons in the human adult brain [105]. Therefore, our results
show the expression of MHC components and the IFN-γ response are dysregulated in HD72-
MSNs. The functional interactions analysis of downregulated proteins within cluster 0 identified
key proteins involved in these pathways: HLA-DPA1, HLA-DPB1, HLA-DMB, HLA-DRA, HLA-
DRB1, HLA-DRB4, HLA-DRB3 and HLA-DQB1 (Fig. 4.7, supplemental Table S4.8B). Immune
system and IFN-γ responses were downregulated in HD72-MSNs, and this raises the intriguing
137

possibility that neurons use this innate immunity signaling system to regulate neuronal
differentiation and neuronal subtype selection during development. Interestingly, MHC-II and
DARPP-32 levels were lower in HD72- than C116-MSNs (supplemental Fig. S4.11).
Clusters 6, 9 – Sodium and Potassium Ion Homeostasis and Proteoglycan Pathways
Cluster 6 was enriched with proteins related to sodium and potassium ion transport across the
plasma membrane, including ATP1A3 and ATP1B1. ATP1A3 is highly expressed in the CNS
and critical during brain development by modulating osmotic equilibrium and membrane
potential [106]. Interestingly, in cluster 9, we found that chondroitin sulfate proteins, such as
VCAN and BCAN, are downregulated in HD-MSNs. Those proteins may illustrate the function of
proteoglycan during important processes of neurodevelopment, such as neuronal migration,
differentiation and maturation [107].
Cluster 11 – Transcriptional Initiation, Elongation, and Termination from the RNA
Polymerase 1 Promoter
Altered RNA polymerase I activity is consistent with studies showing ribosomal transcription is
regulated by PGC-1 alpha, and this process is impaired in HD [108]. HTT is also part of a
transcription-coupled DNA complex formed by RNA polymerase II subunit A, basic transcription
factors, PNKP, ATXN3, DNA ligase 3, CREB protein (CBP, histone acetyltransferase), and this
complex identifies lesions in the template DNA strand and mediates their repair during
transcriptional elongation. mHTT likely disrupts RNA polymerase I activity in a similar complex
[109].
138

Comparison of HD-MSNs Proteomic Data Set to Human and Mouse HD Proteome and
Modifier Data Sets
We compared our proteomics data set to published proteomics of postmortem HD cortex and
knock-in HD mouse model Q175 striatum (supplemental Fig. 4.12A) [24, 93]. Enrichment
analysis of downregulated proteins showed dysregulation in all three data sets for regulation of
small GTPase-mediated signal transduction, regulation of neurotransmitter levels, regulation of
neuronal synaptic plasticity, regulation of GTPase activity, regulation of cell morphogenesis
involved in differentiation, regulation of axonogenesis, neurotransmitter transport, neuron
projection guidance, negative regulation of neuron projection development, negative regulation
of cell projection organization, axonogenesis, and axon guidance. Enrichment analysis of
upregulated proteins showed alterations in regulation of GTPase-mediated signal transduction,
regulation of cell morphogenesis and axonogenesis. Comparison of differentially expressed
proteins in the postmortem cortex of HD patients [24] and those in MSN showed an overlap of
50 proteins. In contrast, compared to the Q175 striatal proteome [93], a well-established HD
model, there was an overlap of 19 proteins. Analysis with Enrichr indicates our data set overlaps
with transcriptomics of HD grade 3 caudate nucleus GSE3790 with a p-value of 2.0e-27 with
147 genes in common.
Drug Signature of HD-MSNs
We utilized the L1000CDS2 website to identify possible drugs that are predicted to reverse the
proteomic signature of HD72-MSNs into C116-MSNs (supplemental Fig. 4.12B) [110].
Trichostatin A, Scriptaid and Vorinostat were among the top hits identified. They belong to
classes of histone deacetylase inhibitors that are beneficial in HD models [111, 112]. An
extensive list of drugs identified can be found in supplemental Table S4.9.
139

HTT Protein Interaction Network Overlap
We also compared our HD72-MSN proteome with known HTT protein interactors. This revealed
over 50 known HTT protein interactors that are altered in the HD MSN proteome (supplemental
Fig. 4.13B). Hub proteins that are highly enriched in our proteomic data set includes YWHAB,
GRIN2B, ITGB1, amyloid precursor protein, and DLG4; these are proteins implicated in the
pathogenesis of HD.
Discussion
Using quantitative proteomic analysis by LC-MS/MS with FAIMS for protein discovery (DDA-
MS), we provide a comprehensive coverage of the MSN proteome with 6,294 quantifiable
proteins identified. Of these proteins, we found ~14% of the identified proteins had altered
expression in HD72-MSNs, compared to isogenic control C116-MSNs. HTT has numerous
cellular functions, and the polyQ expansion in the protein would be expected to disrupt multiple
pathways involved in neuronal homeostasis. Because these studies were carried out on human
neurons derived from HD patient iPSCs, the data may prove useful for further understanding the
biology and therapeutic targets in HD MSNs. The comparison of our proteomics data set to
published proteomics of postmortem HD cortex (24) showed a strong correlation for down-
regulated proteins identifying small GTPase-mediated signal transduction, regulation of
neurotransmitter levels, regulation of neuronal synaptic plasticity, regulation of GTPase activity,
regulation of cell morphogenesis involved in differentiation, regulation of axonogenesis,
neurotransmitter transport, neuron projection guidance, negative regulation of neuron projection
development, negative regulation of cell projection organization, axonogenesis, and axon
guidance.
Our data set has a robust enrichment for processes involved in the brain function, including
neurogenesis-axogenesis, the BDNF-signaling pathway, Ephrin-A:EphA pathway, regulation of
synaptic plasticity, axonal guidance signaling, caveolar-mediated endocytosis signaling, SNARE
140

signaling and RHO GTPase signaling. Septin family members were particularly interesting as
the polyQ expansion in HTT dysregulated multiple family members SEPT2, 3, 4, 5 and 9. Septin
family members form highly organized pre- and postsynaptic supramolecular structures and
regulate synaptic transmission. Misregulation of human septins has been linked to AD and PD
[113]. SEPT5 is a substrate for ubiquitin ligase Parkin, and the PD loss of function mutations in
Parkin leads to accumulation of SEPT5 and dopamine-dependent neurodegeneration.
Correspondingly, SEPT4 knockout mice exhibit reduced dopaminergic neurotransmission.
Interestingly, SEPT9 modulates cargo entry into dendrites by regulating the motility of two
distinct kinesin motors [114]. In HD MSNs, SEPT2 and SEPT9 had increased expression,
whereas SEPT3, 4, and 5 were downregulated in HD MSNs.
A highly enriched pathway found in HD-MSNs is the dysregulation of ECM. Normal HTT has a
role in the construction and regulation of the ECM, and its absence results in disruption of ECM
components [115]. Our results support the hypothesis that the abnormal polyQ expansion within
the mHTT affects the ECM components that ensure the integrity of MSNs in terms of neuronal
identity, architecture, and ability to interact with neighboring cells. This hypothesis is consistent
with the detection of EMT pathways in the upregulated proteins (Fig. 4.6, supplemental Table
S4.7). EMT is a critical cellular process in embryonic development that enables epithelial cells to
acquire the properties of mesenchymal cells. ECM proteins are important in maintaining
epithelial integrity, in addition to initiating and regulating the EMT [116]. HD is characterized by
impairment of specification and maturation of MSNs [117]. mHTT may impair age-dependent
maintenance of striatal MSN identity gene expression [93]. Our proteomic analysis showed a
dysregulation of ECM-related pathways with the increased EMT-related proteins, which may
indicate the inability of HD72-MSNs to acquire and maintain a neuronal signature. Notably, we
found that CNS-related pathways were downregulated in HD72-MSNs, including neurogenesis,
141

axonogenesis, axon guidance, regulation of axonal synapse activity, dopamine, and glutamine
metabolic process (Fig. 4.7, supplemental Table S4.7).
One of the top statistically significantly enriched pathways was DNA signaling (DNA replication
pathway, double-strand break repair, G1/S transition). This fits with numerous studies in the
field suggesting an increase in DNA damage in HD, defects in DNA repair mechanisms and a
multitude of genes that enhance or prevent CAG expansion modifying the age of onset of HD
[118, 119]. We identified the MCM proteins 2–7 as a top dysregulated pathway. The MCM
complex regulates DNA replication, cell-cycle and DNA damage responses, and so, it is likely
an important signaling pathway for HD [90-92]. The IPA pathway of DNA replication,
recombination, repair and DNA metabolism for the HD72-MSN proteome is shown in
supplemental Fig. S4.14.
One of the signaling pathways identified in our analysis is cellular senescence. We
previously showed the development of senescence features in human HD NSCs and MSNs
[47]. p16
INK4a
promotes cellular senescence in these human HD cells. FOXO3, a major cell
survival factor that represses cell senescence, opposing p16
INK4a
expression via the FOXO3
repression of the transcriptional modulator ETS2. In our current study, we find the genes
CDKN1A (p21), IGFBP7, HMGA1 and SERPINE1 are part of the cellular senescence pathway
activated in HD72-MSNs. Senescent cells also have a senescence-associated secretory
phenotype (SASP). Interestingly, when we compare our data set to the recently defined “SASP
Atlas”, a data base of the secretomes of senescent cells [120], we found the following SASP
proteins elevated in HD-MSNs: FLNC, AHNAK, CD44, HSPA1B, HSPA1A, TNC, TIMP2, TKT,
EMILIN1, COL6A1, HSPG2, TPM2, TAGLN2, PLEC, MMP2, PKM, HMGA1, ALDOC, CALD1,
PSAP, YWHAE, and MIF.
Lipid droplets accumulate during aging, inflammation, oxidative stress and in
neurological diseases (ALS, AD, PD). For AD, the interaction of glial cell lipid droplets and
142

neurons play a role in neurodegeneration [121-123]. The roles of lipid droplets in neurological
diseases and the brain are not completely understood [124]. Guided by pathway analysis of the
HD-72 MSNs proteomics identifying lipid metabolism and genes involved in lipid droplet
formation, we found an increase in lipid droplets in HD72-MSNs. Our results indicate that HD72-
MSNs have dysregulated levels of endogenous APOE, in association with the ability to
accumulate a significant number of lipid droplets. Our work and that of others suggest that HD
autophagy impairment comes from a deficit in lysosomal processing reflecting changes in p62,
LC3 levels, flux and cargo loading [10, 100, 125]. Interestingly, SEPT9, which is increased in
HD-MSNs, is directedly linked to lipid droplet biology. SEPT9 and phosphoinositides regulate
lysosome localization, their association with lipid droplets and lipid droplet growth [79, 80]. In
Fig. 4.10, we present a model summarizing our findings that includes increased levels of LDLR,
lower levels of ABCA1, increased lipid droplet formation, possible insufficient lipophagy and
dysregulation of SEPT9, a key regulator of lipid droplet biology.

143

Figure 4.10. Schematic summarizing the altered triglyceride homeostasis, lipophagy and
lipid droplet formation in HD-MSNs. Image was made with BioRender.

144

Our quantitative unbiased proteomics analysis of HD-MSNs provides a comprehensive
understanding of the proteins altered in human HD-MSNs derived from patient iPSCs. We
identified signaling pathways not dysregulated in proteomic of human HD neurons, including
MHC class proteins, IFN- , cellular senescence, ApoE signaling/lipid metabolism and regulation
of cellular response to heat in neurons. Since many of the proteins discovered in our study have
been identified in related neurological diseases, our findings will likely accelerate the
identification of new biomarkers for HD.

145

Supplemental Figures

Supplemental Fig. 4.1. Performances of the TripleTOF 6600 DIA MS workflow. DIA data
were processed using the pan-human spectral library [126]. A, Coefficients of variation (CV) of
precursors identified in three biological replicates of C116-corrected MSN and HD-MSN. 72%
and 80% of the identified precursor ions presented a coefficient of variation below 20% in C116-
corrected MSNs and HD-MSNs, respectively. 42% and 54% of the identified precursor ions
presented a coefficient of variation below 10% in C116-corrected MSN and HD MSN,
respectively. B, Number of missing values obtained for each precursor ion in the dataset.
146

15,531 of a total of 31,183 precursor ions were identified and quantified in all six samples of the
dataset (complete profile).

147

Supplemental Fig. 4.2. Validation of the protein groups altered in HD72-MSN vs C116-
MSN. Venn diagrams showing the overlap of A, all significantly altered, B, all down-regulated,
and C, all up-regulated protein groups obtained by DDA-MS as discovery step and by DIA-MS
148

as validation step, when comparing HD72-MSNs vs C116-MSNs. Common protein groups are
listed in Table S4.4C. D, scatter plot of log2 fold change of all shared proteins between DIA and
DDA. E, scatter plot of log2 fold change of all shared significant proteins (FDR < 0.05, absolute
log2 fold change > 0.58) between DIA and DDA. F, scatter plot of log2 fold change of all
significant proteins (FDR < 0.05, absolute log2 fold change > 0.58) in DDA dataset that are also
found in the DIA dataset.

149

Supplemental Fig. 4.3. Insulin-like growth factor-binding protein 7 levels are increased in
HD72-MSNs. A, Western blot analysis of IGFBP7 in C116-MSNs and HD72-MSNs. B,
Quantification of IGFBP7 protein levels normalized to vinculin. Unpaired t-test with Welch's
correction *P ≤ 0.05.

150

151

Supplemental Fig. 4.4. Functional enrichment map of significantly altered proteins in
HD72-MSNs. Enrichment results from g:Profiler were mapped as a network of gene sets
(nodes) related by mutual overlap (edges) in Cytoscape using its EnrichmentMap. Red color
represents gene sets upregulated in HD72-MSNs, whereas blue represents gene sets
downregulated in HD72-MSNs. Node size is proportional to the gene-set size, and edge
thickness represents the number of overlapping genes between sets. Functional enrichment
map was generated with proteins quantified by FAIMS-DDA MS.

152

153

Supplemental Fig. 4.5. IPA analysis of the differentially expressed proteins in HD72-MSNs
compared to controls.

154

Supplemental Fig. 4.6. HD-MSN lipid metabolism and its modulation by APOE3. C116- and
HD72-MSNs cultured in serum withdrawal were treated for 48 h with 312 ng/mL APOE3. Non-
treated cells were used as controls. The cells were stained with Nile Red, a lipophilic dye. A,B,
Quantifications of the phospholipid (red) and neutral lipid (green) intensity in normal and
APOE3-treated conditions, as illustrated from Fig. 4.8. Unpaired t-test with Welch's correction
*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ns, not significant. Image quantification was done on total 13
unbiased, low-resolution view fields capturing total 599 DAPI-stained nuclei.

155

Supplemental Fig. 4.7. HD-MSN lipid metabolism and its modulation. Quantitative
proteomics reveals LDLR, CYP51A1 and DHCR24 are expressed more highly in HD72-MSNs
than C116-MSNs. Quantitative proteomics reveals ABCA1 and HMGCL are expressed less in
HD72-MSNs than C116-MSNs. Unpaired t-test with Welch's correction *P ≤ 0.05, **P ≤ 0.01,
***P ≤ 0.001.

156

Supplemental Fig. 4.8. HMG-CoA reductase levels are unchanged. 3-Hydroxy-3-
methylglutaryl coenzyme A levels are similar in HD72-MSNs. compared to C116-MSNs. A,
Western blot analysis of HMG-CoA reductase in C116-MSNs and HD72-MSNs. B,
Quantification of HMG-CoA protein levels normalized to vinculin. Unpaired t-test with Welch's
correction ns.

157

Supplemental Fig. 4.9. Quantification of p61, LC3 and LAMP1 in HD MSNs with APOE.
C116-MSNs and HD72-MSNs cultured in serum withdrawal were treated for 48 h with 312
ng/mL APOE3. A, Cells were stained with Nile Red, a lipophilic dye, and only the short-
wavelength, neutral lipid bound fluorescence is shown in representative confocal images
(green). In addition, cultures were immunostained for p62 or LAMP1 (red). Lipid droplets
(indicated by green circles in the other channel) and immunofluorescence puncta (indicated by
158

red circles). Select co-labeling is indicated by arrows. The white outlines show the boundary of
the cells. Micrographs were cropped, Wiener smoothed and saturated for presentation.
Matching intensity scaling was used in all Nile Red and in all immunofluorescence channels,
here and in Fig. 4.9.

159

160

Supplemental Fig. 4.10 Quantification of p61, LC3 and LAMP1 in HD MSNs with APOE.
Immunofluorescence puncta were intensified by rolling ball local background subtraction before
smoothing. A,B,C Quantification of lipid droplet and immunofluorescence spot density,
fluorescence intensity (F.U.) and size for p62, LC3 or LAMP1, corresponding to panel A and
Fig. 4.9. B,C,D Two-way ANOVA with Sidak’s multiple comparison comparing HD72-MSNs with
C116-MSNs, *P ≤ 0.05, ***P ≤ 0.001. Including Fig. 9A, images were selected and cropped from
80 unbiased view fields capturing total 458 DAPI-stained nuclei, representing the quantified
data.

161

Supplemental Fig. 4.11. Modulation of MHC and IFN-γ rescues HD cellular phenotypes.
C116-MSNs and HD72-MSNs were treated for 48 h with IFN-γ at different concentrations. Non-
treated cells were used as control. Representative immunofluorescence images showing that
the expressions of DARPP-32 and MHC-II (A-B) were modulated after IFN-γ treatment. C,D,
Quantifications of the expression levels of each marker based on the analysis of pixel intensity
162

were carried using the Biotek and Image J. Scale bar: 200 µm. One-way ANOVA for multiple
comparisons (Tukey’s) ****P<0.0001.

163

Supplemental Fig. 4.12. Comparison of HD proteomics data sets. A, Comparison of
dysregulated GO Biological Process (BP) terms among MSNs from human cortex of HD
patients [24] and the Q175 mouse model of HD [93]. B, The signature of differentially expressed
proteins in MSNs was used as input on the L1000CDS2 [110] website to identify drugs that
would modify the proteomic signature from HD to corrected. The top drugs identified are shown
and are known modifiers of HD. The heatmap on the top shows the proteomic signature of HD
and how it is predicted to be affected by each of the drugs.

164

Supplemental Fig. 4.13. HD protein expression overlaps with HTT interacting proteins.
The overlapping network between significantly altered proteins from FAIMS-DDA MS and the
HTT protein interactome identified with A, yeast two-hybrid method [127] or B, Weighted
Correlation Network Analysis [128] or C, in vitro affinity pull-down assays using mouse protein
[7] or D, in vitro affinity pull-down assays using human protein [7]. E, The overlapping network
between significantly altered proteins as determined by DIA MS and HTT proteomic interactome
identified with Weighted Correlation Network Analysis [128].

165

Supplemental Fig. 4.14. IPA analysis of the differentially expressed proteins in HD72-
MSNs compared to controls showing the DNA signaling network.
166

Supplemental Tables
Supplemental Table S4.1. DIA window isolation scheme of the DIA MS acquisition method.
Supplemental Table S4.2. Protein, peptide and peptide spectrum match identification results
obtained from the FAIMS-DDA MS data set.
Supplemental Table S4.3. Protein quantification and statistical analysis results obtained from
the FAIMS-DDA MS data set.
Supplemental Table S4.4. Comparison and validation of the significantly changing protein
candidates obtained from the FAIMS-DDA MS data set with the significantly changing proteins
obtained from the DIA MS data set.
Supplemental Table S4.5. Protein quantification and statistical analysis results obtained from
the DIA MS data set.
Supplemental Table S4.6. Custom background of proteins.
Supplemental Table S4.7. Pathway enrichment analysis of the upregulated and downregulated
proteins from the FAIMS-DDA MS proteomic analysis.
Supplemental Table S4.8A. Reactome functional interaction network analysis of the upregulated
proteins in HD72-MSN to define clusters of proteins that are closely connected.
Supplemental Table S4.8B. Reactome functional interaction network analysis of the
downregulated proteins in HD72-MSN to define clusters of proteins that are closely connected.
Supplemental Table S4.9. Predicted drugs that modify HD
167

CHAPTER 5: Postnatal Conditional Deletion of Bcl11b in Striatal Projection
Neurons Mimics the Transcriptional Signature of Huntington’s Disease (Song et
al., 2022)
Abstract
The dysregulation of striatal gene expression and function is linked to multiple
diseases, including Huntington’s disease (HD), Parkinson’s disease, X-linked dystonia-
parkinsonism (XDP), addiction, autism, and schizophrenia. Striatal medium spiny
neurons (MSNs) make up 90% of the neurons in the striatum and are critical to motor
control. The transcription factor, Bcl11b (also known as Ctip2), is required for striatal
development, but the function of Bcl11b in adult MSNs in vivo has not been investigated.
We conditionally deleted Bcl11b specifically in postnatal MSNs and performed a
transcriptomic and behavioral analysis on these mice. Multiple enrichment analyses
showed that the D9-Cre-Bcl11b
tm1.1Leid
transcriptional profile was similar to the HD
gene expression in mouse and human data sets. A Gene Ontology enrichment analysis
linked D9-Cre-Bcl11b
tm1.1Leid
to calcium, synapse organization, specifically including the
dopaminergic synapse, protein dephosphorylation, and HDAC-signaling, commonly
dysregulated pathways in HD. D9-Cre-Bcl11b
tm1.1Leid
mice had decreased DARPP-
32/Ppp1r1b in MSNs and behavioral deficits, demonstrating the dysregulation of a
subtype of the dopamine D2 receptor expressing MSNs. Finally, in human HD isogenic
MSNs, the mislocalization of BCL11B into nuclear aggregates points to a mechanism for
BCL11B loss of function in HD. Our results suggest that BCL11B is important for the
function and maintenance of mature MSNs and Bcl11b loss of function drives, in part,
the transcriptomic and functional changes in HD.
168

Background
The basal ganglia comprise interconnected subcortical nuclei that are responsible for motor
learning and control, executive functions, and emotions. The striatum, composed of the caudate
and putamen in humans, is the largest component of the basal ganglia. It receives and
integrates glutamatergic and dopaminergic inputs from several brain regions, including the
cortex, thalamus, hippocampus, and amygdala. These inputs target inhibitory γ-amino butyric
acid (GABA)-ergic medium spiny neurons (MSNs), the principal output neurons of the striatum,
and make up 90–95% of its total neurons. MSNs are morphologically homogeneous; however,
they can be distinguished by their output targets and their specific gene expression. Direct
MSNs project to the global pallidus internal or substantia nigra pars reticula and express the
dopamine D1 receptor (Drd1), substance P, and dynorphin. Indirect MSNs project to the globus
pallidus external and express the dopamine D2 receptor (Drd2), the adenosine 2A receptor, and
enkephalin.
The dysfunction and developmental alterations of MSNs or their subtypes have been implicated
in several neurological and neuropsychiatric disorders, including Huntington’s disease (HD)
(Ring et al., 2015), dystonia (Domingo et al., 2021) (e.g., X-linked dystonia parkinsonism
(Aneichyk et al., 2018; M. D. Cirnaru et al., 2021)), addiction (Egervari et al., 2021; Van Hees et
al., 2021), and schizophrenia (Meyer-Lindenberg, 2010). Some of these diseases manifest in
late adulthood and include prominent gene transcriptional abnormalities in mature MSNs.
Disrupted levels and/or the activity of specific transcription factors (TFs) involved in
development were described as a potential common pathogenic mechanism underlying cell
type-specific vulnerability (Lennon et al., 2017; Volpicelli et al., 2020). TFs play a key role in
controlling the spatiotemporal expression of cell type-specific genes. For this reason, knowledge
of TF activity in mature MSNs is crucial to understanding striatal-related pathologies. One
critical TF involved in the striatal development of MSNs is B-cell leukemia 11b, Bcl11b (also
known as Ctip2).
169

CTIP1 and CTIP2 were first identified in mouse neurons as Krüppel-like C(2)H(2) zinc finger
proteins interacting with all members of the chicken ovalbumin upstream promoter TF family
(COUP-TF) (Avram et al., 2000). CTIP1, later named BCL11A, is highly enriched in COUP-TF-
negative cells of the immune system and it is also expressed in the cerebral cortex (Canovas et
al., 2015; Leid et al., 2004; Wiegreffe et al., 2015). CTIP2, located on chromosome 14, is the
homologue of CTIP1 and was named BCL11B (Satterwhite et al., 2001). Both are sequence-
specific DNA-binding proteins that repress or induce transcription (Avram et al., 2000). Bcl11b is
highly expressed in the embryonic cortex and striatum. Interestingly, unlike most developmental
striatal TFs, Bcl11b levels remain high throughout life (Arlotta et al., 2008; Chandwani et al.,
2013; Desplats et al., 2006; Golonzhka et al., 2009). Therefore, exploration of the role of Bcl11b
in mature MSNs is essential. Importantly, BCL11B activity is dysregulated in HD patients and
models, suggesting that it participates in the maintenance and function of mature MSNs
(Desplats et al., 2006; Desplats et al., 2008; Etxeberria-Rekalde et al., 2020; Fjodorova et al.,
2019; P. Langfelder et al., 2016; Ring et al., 2015).
Bcl11b is required for the development of corticospinal motor neuron projections and the
differentiation of MSNs (Arlotta et al., 2005; Avram et al., 2000; Onorati et al., 2014), and is
frequently used as a pan-MSN marker, along with DARPP-32/Ppp1r1b, which is expressed later
in development relative to Bcl11b (Arber et al., 2015; Chandwani et al., 2013; Fjodorova et al.,
2019; Fullard et al., 2018; Ivkovic & Ehrlich, 1999). Several in vitro and vivo Bcl11b-null models
have been characterized. Constitutive murine Bcl11b deletion results in altered striatal
compartmentalization as determined by the absence of DARPP-32 in the late embryonic
striatum and in the aberrant expression of genes associated with both direct and indirect
pathways (Arlotta et al., 2008). The loss of function and Bcl11b ChIP-sequencing experiments in
an immortalized striatal cell line indicated that Bcl11b regulates the expression of striatal-
enriched genes (Golonzhka et al., 2009; Tang et al., 2011) and the genes involved in the BDNF
170

signaling pathway (Tang et al., 2011). In MSNs derived from human embryonic stem cells,
BCL11B regulates the expression of genes related to Ca2+ signaling and kinase activity,
playing a crucial role in MSN homeostasis (Fjodorova et al., 2019). A consensus motif for
Bcl11b has been proposed (Avram et al., 2000; Tang et al., 2011), but the binding site is
considered “promiscuous”, and Bcl11b is also part of nucleosome remodeling and deacetylation
complexes, NuRD and SWI/SNF, suggesting that it regulates gene expression through
mechanisms that are not restricted to direct DNA binding. Notably, the functions of Bcl11b
specific to adult MSNs have not been elucidated and are the focus of our current study.
Much is known about TFs and their development, but little is known, particularly in neuronal
subtypes, about TFs and neuronal phenotype maintenance, which is an active process (Hobert,
2016). To determine the role of Bcl11b in adult MSNs in vivo, the regulatory components of the
genomic elements of DARPP-32 (i.e., D9) (Bogush et al., 2005) were used to delete the
expression of Bcl11b, selectively, in post-mitotic, post-migrational MSNs (D9-Cre-Bcl11
btm1.1Leid
).
We provide evidence that Bcl11b deletion results in a decrease in the expression of the genes
involved in general neuronal survival and maintenance, and a decrease in the markers that
characterize the unique MSN subtypes. Furthermore, we found that specific dopamine receptor-
mediated behavior is impacted. Compellingly, the transcriptomic profile of D9-Cre-Bcl11
btm1.1Leid

mice significantly overlaps with the gene expression changes in HD human and mouse striatum.
A Gene Ontology (GO) enrichment analysis links Bcl11b deletion to calcium and HDAC
signaling pathways that are commonly dysregulated in HD, and to MSN-specific pathways,
which are also dysregulated in HD. Finally, in human HD-isogenic MSNs, the mislocalization of
BCL11B into nuclear aggregates points to a mechanism for BCL11b loss of function in HD.
171

Results
Transcriptomic analysis of D9-Cre-Bcl11
btm1.1Leid
mice
We deleted Bcl11b selectively in post-mitotic, post-migrational adult MSNs by using a Cre
mouse with the regulatory components of the genomic elements of DARPP-32 (i.e., D9)
(Bogush et al., 2005). D9-Cre mice were crossed with Bcl11
btm1.1Leid
loxP-flanked mutant mice.
Using immunostaining, we confirmed that D9-Cre-Bcl11
btm1.1Leid
mice express almost no Bcl11b
in the MSNs at 5 weeks of age (Figure 5.1A). We performed RNA-sequencing on striatal tissue
from homozygote floxed Cre− and D9-Cre-Bcl11
btm1.1Leid
mice at 4 months of age. We detected
38,386 mRNAs, including hundreds of long-intergenic non-coding RNAs (lincRNA), and 938
miRNAs. A differential expression analysis revealed that the deletion of Bcl11b resulted in 2771
differentially expressed genes (DEGs) with an adjusted p-value of less than 0.01 (Table S1).
Among these DEGs, 1536 were upregulated and 1235 were downregulated (Table S1). The
PCA plot shows the separation in the clustering of the transcriptome of the Cre− and D9-Cre-
Bcl11
btm1.1Leid
mice (Figure 5.1B). As expected, the levels of Bcl11b were substantially reduced
in the D9-Cre-Bcl11
btm1.1Leid
mice (Figure 5.1C). The volcano plot and heatmap highlight the top
enriched genes in the D9-Cre-Bcl11
btm1.1Leid
mice (Figure 5.1D,E). Although the top
downregulated genes were not specifically correlated with MSN function, importantly, Bcl11b
deletion also resulted in the loss of MSN-enriched markers with a reduced expression of:
forkhead box protein P1 (FoxP1), DARPP-32 (also known as Ppp1r1b); Arpp21, proenkephalin-
A (Penk), 5-hydrotryptamine receptor 1B,1D (Htr1b, Htr1D), ryanodine receptor (Ryr1), GABA
receptor subunit delta; alpha-4 (Gabrd, Gabra4), histamine H3 receptor (Hrh3), Drd1, Drd2, and
metabotropic glutamate receptor 1 (Grm1) (Table S1). Notably, Drd2 (log2fold = −0.60) was
reduced more than Drd1 (log2fold = −0.26). Both striosome and matrix genes were altered.
172

Although many markers specifically associated with MSNs were downregulated, there were
some notable exceptions (Figure 5.1D,E). One of the top upregulated genes was latent
transforming factor beta binding protein (Ltbp2). It is critical in the TGFβ signaling and regulation
of this pathway that it has been linked to MSN developmental processes and HD
neuropathogenesis (Ring et al., 2015). Correspondingly, activin A receptor like type 1 (Acvrl1),
is upregulated. During the development of the lateral ganglionic eminence (LGE), the ligand for
Acvr1, Activin A, plays a critical role in the specification of striatal fate (Feijen et al., 1994; Maira
et al., 2010), and both activin receptors and activated activin are expressed in the developing
LGE (Feijen et al., 1994; Maira et al., 2010). Wnt8b is involved in the caudalization of a regional
identity (Fang et al., 2019). Upon the reduction of Bcl11b, ras-specific guanine-nucleotide
releasing factor 2 (RasGRF2) was also reduced. This calcium-regulated exchange factor
(Easton et al., 2014) alters the ERK-dependent cocaine reward in mice (Bernardi et al., 2019).
Some of the top upregulated genes are involved in cell death signaling pathways (e.g., Clec12a,
a uric acid receptor that potentiates type I interferon responses) (Li et al., 2019). Sstr3 is a G-
protein-coupled receptor (GPCR) whose signaling affects neuronal cilia and apoptosis and is
upregulated after heroin exposure (Barbeito & Garcia-Gonzalo, 2021; Yao et al., 2017). In
addition, glutamate metabotropic receptor 2 (Grm2) is increased in the Bcl11b MSN-deletion
mice. A group of top dysregulated genes are not known to be specifically associated with MSNs
and therefore warrant further investigation (Figure 5.1D,E, Table S1).

173

174

Figure 5.1. Identification of transcriptome portraits of Bcl11b deletion and WT cells by
RNA-seq analysis in Cre+ and Cre− MSNs populations from the striata of D9-Cre mice. (A)
D9-Cre- Bcl11
btm1.1Leid
mice specifically knocked down expression Bcl11b in MSNs by using a
D9-Cre under the control of regulatory elements of the mouse Ppp1r1b gene-encoding DARPP-
32. D9-Cre-Bcl11
btm1.1Leid
and Cre-negative mice (4-months-old) were analyzed using
immunohistochemistry (IHC) with a BCL11B antibody. Scale bar is 500 mm. (B) PCA plots using
the rlog-transformed values indicate a significant difference in the transcriptome Bcl11b deleting
MSNs, and controls. (C) Scatter plot shows that Bcl11b gene expression is much less in D9-
Cre-Bcl11
btm1.1Leid
mice than Cre– control mice. (D) Volcano plot shows differences in Cre+ and
Cre− gene expression. Genes with an adjusted p-value below 0.01 with absolute log2 fold ratio
greater than 1 are highlighted. Genes in red are relatively decreased in expression in the Cre−
population (i.e., enriched in the WT population), those in green are relatively increased in
expression in the Bcl11
btm1.1Leid
mice, and those in grey are equally distributed among the two
populations. (E) Heatmap of relative normalized count values across samples. Top 20 up and
downregulated genes that have the highest product of log fold-change and base mean are
reported, respectively. Top downregulated genes include: Bcl11b, free fatty acid receptor
(Ffar3), spermatogenesis and oogenesis-specific basic helix-loop-helix 1 (Sohlh1), beta
tropomyosin (Tmp2), myosin IIIB (Myo3b), wnt family member 8B (Wnt8b), R-spondin-1
(Rspo1), Obscurin (Obscn), glutathione peroxidase 6 (Gpx6), dermokine (Dmkn),
serine/threonine-protein kinase receptor, R3, activin A receptor-like type 1 (Acvrl1), anoctamin-2
(Ano2), kelch-like protein 1 (Klhl1), synaptotagmin-2 (Syt2), 1-phosphatidylinositol 4,5-
bisphosphate phosphodiesterase eta-2 (Plch2), Sec14l3 (uncharacterized protein), and C-type
lectin domain family 12 member A (Clec12a).

175

We conclude that the reduction in Bcl11b affects multiple genes involved in MSN maintenance
and identity and signal transduction.
Bcl11b reduction results in differentially expressed genes that correlate with pathways
dysregulated in HD.
We used Enrichr (Chen et al., 2013) to identify the co-expression networks that most overlap
with the transcriptomics of the Bcl11b reduction in MSNs (Figure 5.2). Strikingly, the expression
changes for the D9-Cre-Bcl11
btm1.1Leid
mice overlapped with expression profiles of HD mouse
models, postmortem HD tissue and/or knockout mice (KO), including the genes Pde10a, Sirt1,
Htra2, Npc1, and Ppargc1a (Figure 5.2A,B). Many of the overlapping genes are markers of
MSNs, and the loss of striatal MSN identity overlaps with the D9-Cre-Bcl11
btm1.1Leid

transcriptomics. The D9-Cre-Bcl11
btm1.1Leid
mice transcriptional profile had drug perturbations
from the GEO database that overlapped with soman, morphine, resveratrol, heroin,
dexamethasone, coenzyme Q, creatine, levetiracetam, methamphetamine, and nicotinamide
riboside (Figure 5.2C). These drug perturbations correlate with striatal function or known drug
targets in HD. We also evaluated the overlap of the gene expression profiles of the 10-month-
old Q175 knockin mouse model (Peter Langfelder et al., 2016) and the conditional D9-Cre-
Bcl11
btm1.1Leid
mice. There were 683 genes shared when comparing the transcriptomic data sets
(zQ175, 2795 genes) with a p-value of 3.75E-76, when using the Fisher exact test. The top
KEGG pathway (2021 human) is the dopaminergic synapse, and the protein-protein interaction
hub protein is GRIN1. The shared genes have ontologies for the regulation of neurotransmitter
receptor activity, calcium signaling, potassium channel regulation, protein/threonine kinase
activity, activin receptor activity, glutamate receptor activity, and postsynaptic density. Kinase
regulation includes CAMK4 and the regulation of a glutamate receptor by CK1 and CDK5. As
expected, our data enriches to constitutively deleted Bcl11b mice. Table S2 summarizes the
176

overlap with the majority of known mouse HD transcriptomics data sets and the overlap with the
D9-Cre-Bcl11
btm1.1Leid
mice.
The top canonical pathways identified by a complementary analysis with an Ingenuity Pathway
Analysis (IPA) were the opioid signaling pathway (p-value 6.22E-14), cAMP-mediated signaling
(p-value 3.60E-08) via which dopamine signals are transduced, the synaptogenesis signaling
pathway (p-value 4.07E-8), protein kinase A signaling (p-value 4.11 × 10−8), and calcium
signaling (p-value 1.10E-07). The top upstream regulators were levodopa, CREB1, amino-5-
phosphonovaleric acid, and huntingtin (HTT).
Next, we carried out a term enrichment analysis for GO or KEGG processes or functions
associated with the DEGs for the conditional D9-Cre-Bcl11
btm1.1Leid
mice, compared to the
controls (Figure 5.3, Table S1, Supplemental Figure S5.1). The KEGG term enrichment analysis
for gene signatures altered by D9-Cre-Bcl11
btm1.1Leid
highlighted the axon guidance,
dopaminergic synapses, adrenergic, estrogen, cAMP, MAPK, insulin, oocytes, and
glutamatergic signaling (Figure 5.3A). An IPA analysis summarizes the critical pathway for
dopamine DARPP-32 feedback cAMP signaling that is enriched in the D9-Cre-Bcl11
btm1.1Leid

mice (Figure 5.3B). This is a key pathway disrupted in HD. Interestingly, many of the genes with
a log fold-change >1.0 correlated with genes involved in calcium homeostasis (Table S1,
Supplemental Figure S5.2).

177

Figure 5.2. Transcriptional profile of Bcl11b deletion is highly correlated with HD mouse
models and postmortem human HD tissue. (A) Gene perturbation enrichment analysis. (B)
Disease perturbation enrichment analysis. (C) Drug perturbation enrichment analysis.

178

Enriched BPs included synapse organization, functions, or molecules related with
transmembrane activities, such as a transmembrane transporter or ion channels (Supplemental
Figure S5.1A). Correspondingly, the GO terms enriched for BPs in the downregulated genes
were protein localization, dephosphorylation, the rhythmic process, and the postsynapse. The
cellular components and molecular functions are also shown in Supplemental Figure S5.1B–G.
Top upstream terms and network from the IPA analysis were HTT, NR4A1, CNTF, epilepsy,
dyskinesia, synaptic depression, the organization of cells, and catalepsy (Supplemental Figure
S5.3). The identification of NR4A1 as a top upstream regulator is interesting. Although Bcl11b
appeared to regulate gene expression in both striosomes and in the matrix in our current study,
Nr4a1 is highly enriched in the striosomes and is required for their development (M.-D. Cirnaru
et al., 2021), and striosomes are altered in HD (Friedman et al., 2020; Hedreen & Folstein,
1995; Morigaki et al., 2020). Thus, a decrease in Bcl11b alters the general neuronal and MSN-
specific processes, including in synapse organization and functions, or in molecules related with
transmembrane activities (e.g., transmembrane transporter and ion channels).

179

180

Figure 5.3. Significantly enriched KEGG terms and IPA signaling. (A) KEGG term
enrichment analysis of gene signatures altered by Bcl11b deletion highlighted the axon
guidance, dopaminergic synapses, adrenergic, estrogen, cAMP, MAPK, insulin, oocytes, and
glutamatergic signaling. (B) IPA signaling pathway highlights for dopamine DARPP-32 feedback
cAMP signaling.

181

Transcriptional network impacted by Bcl11b deletion
We evaluated the TFs and networks impacted by the deletion of Bcl11b. Using the mouse TF
database, a total of 287 differentially expressed TFs were altered (Schmeier et al., 2017).
Among these, 109 were upregulated, and 178 were downregulated (Table S1). These
differentially expressed TFs were used as the input for the gene regulatory network analysis to
determine the key upstream regulators in the D9-Cre-Bcl11
btm1.1Leid
MSNs (Figure 5.4). From the
inferred networks, we identified the hub gene, Egr1 (Figure 5.4A), which plays a key role in the
induction of DARPP-32 expression in MSNs (Keilani et al., 2012). The enrichment of Foxo3,
Foxj2, and Foxj3 showed that the decrease of Bcl11b alters the forkhead pathway (Figure 5.4B),
which is important in adult human neurogenesis and cell-cycle inhibition (Genin et al., 2014;
Schmidt et al., 2002).

182

183

Figure 5.4. Gene regulatory network analysis reveals critical up-stream TFs from the
gene signatures altered by Bcl11b deletion. Gene regulatory network analysis. (A)
Differentially expressed TFs that are upregulated (A) or downregulated (B) in Bcl11b-deletion
cells. (C) All differentially expressed TFs in Bcl11b-deletion cells.

184

Additionally, our TF network identifies Kmt2a, Kdm2a, and Ash1l, which are important histone-
modifying enzymes involved in chromatin remodeling (Figure 5.4B,C). Kmt2a modifies H3K4
(Bochynska et al., 2018), Ash1l modifies H3K36 (Zhu et al., 2016), and Kdm2a modifies H3K4
and H3K36 (Dhar et al., 2014; Gao et al., 2013), and these may indicate that a deficiency in
Bcl11b causes abnormal histone modification. Notably, H3 lysine 4 trimethylation (H3K4me3) at
transcriptionally repressed promoters in the brain is considered an early feature of HD
(Vashishtha et al., 2013).
Other noteworthy TFs that were enriched in the networks were Nfat5 and Nfatc3. They regulate
the calcineurin-mediated signaling pathway, and importantly, calcineurin inactivates DARPP-32
by dephosphorylation (Kunii et al., 2019). Moreover, hyperactivated calcineurin dysregulates
BDNF transport in HD (Mukherjee & Soto, 2011; Zhang et al., 2021). Overall, D9-Cre-
Bcl11
btm1.1Leid
mice strongly mimic aspects of HD transcriptional dysregulation.
The decrease of Bcl11b in differentiated MSNs resulted in a decrease in NeuN+/DARPP-32+
cells, motor deficits, and a decreased response to haloperidol. Selective loss of striatal MSNs is
a major hallmark in HD but is poorly recapitulated in mouse models (Francelle et al., 2014). To
determine if Bcl11b deletion compromises neuronal viability, we counted the striatal neurons,
and specifically MSNs, using a NeuN and DARPP-32 immunofluorescence. Fewer NeuN+ and
DARPP-32+ cells were detected in the striatum of D9-Cre-Bcl11
btm1.1Leid
mice than in the
wildtype (WT) mice (Figure 5.5A,B). The decreased neuronal numbers were not accompanied
by increased numbers of microglia, as revealed by Iba1 immunostaining (Figure 5.5C). The
decrease in NeuN+ and DARPP-32+ cells may suggest a loss of neurons, de-differentiation, or
a lack of differentiation.

185

186

Figure 5.5. Bcl11b deficiency leads to a reduced number of MSNs without microgliosis.
Representative striatal images immunostained with cell type-specific markers, and graphs
detailing corresponding quantification. (A). IHC of Bcl11
btm1.1Leid
and control mice IHC
immunostained with NeuN (neurons) and quantification. (B). IHC of Bcl11
btm1.1Leid
and control
mice immunostained with DARPP-32 (MSNs) and quantification. (C). IHC of Bcl11
btm1.1Leid
and
control mice with Iba1 (microglia). Scale bars, 200 mm. Graphs show the number of NeuN,
187

DARPP-32, and Iba1-positive cells in the striatum. Each point represents an individual mouse.
All data are shown as mean ± SEM (WT n = 5; Bcl11b-KO n = 5.) Two-tailed unpaired t-test, * p
< 0.05.

188

The role of the striatum in movement and the overlap of the gene expression changes with HD
prompted us to evaluate the motor behavior of striatal D9-Cre-Bcl11
btm1.1Leid
mice. We found that
spontaneous locomotor activity was reduced in D9-Cre-Bcl11
btm1.1Leid
mice in the first 5 min
(Figure 5.6A). There was a trend towards subtle balance alterations in the balance beam test, in
that D9-Cre-Bcl11
btm1.1Leid
mice and WT mice crossed a similar number of frames, but the KO
mice appeared to require more time to cross 30 frames than WT mice (p-value = 0.096; t-test)
(Figure 5.6B). Importantly, D9-Cre-Bcl11
btm1.1Leid
mice displayed poor performance in the vertical
pole test, requiring more time to turn and descend than WT mice (unpaired t-test, * p < 0.05; **
p < 0.01) (Figure 5.6C). Bcl11b deletion did not alter anxiety-like behavior in the elevated plus
maze (Supplemental Figure S5.4). These results suggest that motor abnormalities after Bcl11b
deletion in adult MSNs overlap to some extent with HD mouse models.

189

190

Figure 5.6. Bcl11b deficiency in adult mice partly recapitulate HD-associated motor
phenotype. (A). Spontaneous locomotor activity measured in the Bcl11b deficiency mice. (B).
Balance beam: from left to right; numbers of frames crossed in 2 min and times to cross 30
frames. (C). Vertical pole: times to turn (left) and times to descend (right) were recorded after
placing the mice upwards to the pole. Three trials were conducted, and data represent the mean
± SEM (WT n = 11, D9-Cre-Bcl11
btm1.1Leid
mice n = 18). Two-tailed unpaired t-test, * p < 0.05; **
p < 0.01. (D). Schematic diagram of catalepsy position (left). Catalepsy time after Haloperidol
treatment (right). Two-Way ANOVA, with Bonferroni as post-hoc test * p < 0.05.

191

Next, RNA-seq analysis pointed to specific alterations of genes involved in the dopaminergic
synapse pathway (adjusted p-value = 3.27E10−7). A cell-type-specific enrichment analysis
showed that Bcl11b deletion caused a downregulation of genes that were enriched in both D1
and D2-MSN subtypes (Supplemental Figure S5.4). Notably, Drd2 (log2fold = −0.60) was
reduced more than Drd1 (log2fold = −0.26). To functionally explore D2R-mediated behavior, we
performed the haloperidol-induced catalepsy test. Haloperidol treatment in mice produces a
behavioral state in which the mice fail to correct externally imposed postures (i.e., catalepsy).
Integrity of postsynaptic dopamine receptors is required to observe this phenotype (Sanberg,
1980). Haloperidol (1 mg/kg) was injected intraperitoneally into WT and D9-Cre Bcl11b-deletion
mice. Catalepsy was measured 30 min after the injection and every 30 min up to 2 h. Catalepsy
time was lower in Bcl11b-deletion mice than WT littermates (Figure 5.6D, two-way ANOVA, with
Bonferroni post-hoc test, * p < 0.05).
Disruption of BCL11B function in a human HD MSN model.
The strong correlation of the D9-Cre-Bcl11
btm1.1Leid
mice transcriptome with HD models prompted
us to determine how the HTT mutation mimics lower levels of Bcl11b. We differentiated isogenic
human patient HD72 iPSCs (CAG repeat size 72) into MSNs (Figure 5.7A). As expected, the
HD72-MSNs had lower levels of DARPP-32 than isogenic control C116-MSNs. Top genes
dysregulated in D9-Cre-Bcl11
btm1.1Leid
mice follow similar trends in expression as measured by
RT-PCR in human HD72-MSNs (Figure 5.7B). KCNC3 and WNT10A were upregulated in
HD72-MSNs, compared to control C116. Like the D9-Cre-Bcl11
btm1.1Leid
mice transcriptomics,
SLIT3 was downregulated in HD72-MSNs. BCL11B was modestly up-regulated but not
statistically significant (data not shown), and may represent the fact that the iPSC-derived MSNs
are relatively immature, compared to mouse adult MSNs in vivo. As our current studies show a
loss of Bcl11b in MSNs correlates with the HD transcriptome, we investigated the mechanism
for how this might occur in HD. BCL11B expression was characterized by immunofluorescence
192

in HD72-MSNs compared to control. In HD72-MSNs, BCL11B was concentrated in internuclear
aggregates, but showed a diffuse pattern in C116-MSNs (Figure 5.7C). Many more internuclear
aggregates were noted in the HD72-MSNs than in controls (Figure 5.7D). We conclude that the
sequestration of BCL11B into nuclear aggregates may lead to loss of transcriptional activity of
BCL11B in HD even in the presence of normal level of expression.
193

Figure 5.7. Human MSNs derived from HD patient iPSCs reveals mislocalization of
BCL11B into nuclear aggregates. (A). Isogenic HD72 and C116 MSNs differentiated from
iPSCs immunostained with DARPP-32. (B). Top genes dysregulated in D9-Cre-Bcl11
btm1.1Leid

194

mice follow similar trends in expression as measured by RT-PCR. KCNC3 and WNT10A are
upregulated in HD72-MSNs, compared to control C116. Like the D9-Cre-Bcl11
btm1.1Leid
mice
transcriptomics, SLIT3 was downregulated in HD-MSNs. (C). IHC with BCL11B antibody show
more large nuclear aggregates in the HD72-MSNs than in C116-MSNs. (D). Quantification of
the BCL11B foci per nuclear area in isogenic HD72 and C116-MSNs. * p < 0.05; **** p <
0.0001, Mann Whitney test.

195

Discussion
We report that selective Cre-mediated deletion of the transcription factor Blc11b/Ctip2 in
differentiated striatal MSNs leads to a transcriptional signature similar to HD and supports the
notion that Bcl11b has a critical role in maintaining key pathways in the biological function and
identity of MSNs in adult mice. Reduction in Bcl11b results in the lower expression of MSN
differentiation markers, including FoxP1, DARPP-32 (also known as Ppp1r1b), Arpp21, Penk,
Htr1b, Htr1D, Ryr1, Gabrd, Gabra4, Hrh3, Drd1, Drd2 and Grm1. Previous studies have shown
that loss of Bcl11b during development results in deficits in MSN birth, migration and
differentiation (Arlotta et al., 2008). Our results, therefore, are consistent with a continued role of
Bcl11b in MSN differentiation and/or maintenance of identity in adult mice.
A cell-type-specific enrichment analysis showed Bcl11b deletion caused a downregulation of
genes that were enriched in both D1 and D2-MSN subtypes. Notably, Drd2 was reduced more
than Drd1. We also found alterations in MSN gene expression in both the patch/striosome and
matrix compartments of the striatum, without enrichment for either compartment. This includes
the striosome markers Oprm1, Tac1, Spon1, Lydp1, Kremmen1, Tshz1, and Pdyn for patch and
the matrix marker, Epha4. We functionally validated that the gene expression changes were
large enough to compromise dopamine neurotransmission, as evidenced by an abnormal
haloperidol-induced catalepsy test.
The expression changes for the D9-Cre-Bcl11
btm1.1Leid
mice overlapped with the gene expression
profiles of HD mouse models and the postmortem HD tissue. Interestingly, some gene
expression changes that overlapped with D9-Cre-Bcl11
btm1.1Leid
mice are genes involved in HD
pathophysiology including Pde10a, Sirt1, Htra2, Npc1, and Ppargc1a. Transcriptional
dysregulation has long been described as an important pathological change in HD. Many of the
downregulated genes in HD striatum are enriched for genes that define MSN identity and
function (Becanovic et al., 2010; Brochier et al., 2008; Hervas-Corpion et al., 2018; Peter
196

Langfelder et al., 2016; Le Gras et al., 2017; Novati et al., 2018; Vashishtha et al., 2013).
Further, as in HD, genes whose expression are altered after the depletion of Bcl11b in MSNs
were enriched in calcium and HDAC signaling. The mechanism behind mutant HTT-induced
transcriptional effects is unclear. Our studies using human HD-MSNs suggests that the
sequestration of BCL11B into nuclear aggregates may lead to loss of function of BCL11B in HD
and the loss of MSN identity and function. This is consistent with a physical interaction of
BCL11B with mHTT (Desplats et al., 2008) and altered levels in HD mouse models (Etxeberria-
Rekalde et al., 2020).
Our results highlight a cascade of TFs that are impacted when Bcl11b is deleted in striatal
MSNs. As discussed above, we identified that Egr1, required for DARPP-32 expression, is a
hub gene (Keilani et al., 2012). Th enrichment of Foxo3, Foxj2, and Foxj3 showed that the
deficiency of Bcl11b alters the forkhead pathway which is important in adult human
neurogenesis and cell-cycle inhibition (Genin et al., 2014; Schmidt et al., 2002). Stat1/3 are
enriched in the differentially expressed TFs that are upregulated in the case of Bcl11b deletion
MSNs. We recently identified that Stat1/3 is a TF that is required for striosome development
(M.-D. Cirnaru et al., 2021). Our results suggest that this may be an important pathway in adult
MSNs as well for striosome identity maintenance.
Recent studies have used CRISPR/Cas9 to deplete human embryonic stem cells of BCL11B.
The reduction of BCL11B in human MSNs leads to neuronal vulnerability and dysfunction. In the
human model of MSNs where BCL11B is depleted, cAMP-Ca2+ signaling, which integrates the
PKA pathway, was identified as dysregulated (Fjodorova et al., 2019). These same pathways
were identified in the current study. BCL11B knockdown likely leads to common alterations in
signaling in both mice and human models.

197

Supplemental Figures

Supplemental Fig. 5.1. GO enrichment of MSN Bcl11b deficiency. A. GO term enrichment
analysis on the gene signatures altered by Bcl11b deficiency, with the most significantly
enriched GO terms highlighted including positive regulation of cell projection organization,
neuron-to-neuron synapse, actin binding, synapse organization, synaptic membrane. Hub gene
analysis of the protein- protein interaction network identified in the most significantly enriched
GO terms by different categories and expression patterns. B. Positive regulation of cell
198

projection organization C. Neuron- to-neuron synapse. D. Actin binding. E. Synapse
organization. F. Synaptic membrane. G. Channel activity.

199

Supplemental Fig. 5.2. Calcium signaling pathways enriched in Bcl11b deficiency.
200

Supplemental Fig. 5.3. IPA analysis of MSN mouse Bcl11b deficiency. Top upstream terms
and network from IPA analysis are HTT, NR4A1, CNTF, epilepsy, dyskinesia, synaptic
depression, organization of cells and catalepsy. The interconnecting genes are shown in the
network.

201

Supplemental Fig. 5.4. Bcl11b deficiency does not induce anxiety-like behaviors. Elevated
plus maze. Bcl11btm1.1Lead mice did not exhibit any differences in the total distance, number
of entire sin open arms or time spend in open arms. Data represent the mean ± SEM (WT n =
11, Bcl11b KO n = 18).

202

Supplemental Fig. 5.5. Cell-type enrichment analysis. Proportion of up-regulated (red) and
down-regulated (blue) differentially expressed genes, and non-differentially expressed genes
(white) overlapping with cell-type specific genes.

203

Supplemental Tables
Supplemental Table 5.1. Transcriptomics of Bcl11b deficiency with functional analysis
Supplemental Table 5.2. HD mouse transcriptomics overlaps with Bcl11b data set.

204

CHAPTER 6: Human Huntington's disease neural cells and striatal neurons face
senescence with p16INK4a increase while reprogramming FOXO3 targets (Voisin et al.,
2020)
Abstract
Neurodegenerative diseases (ND) have been linked to the critical process in aging - cellular
senescence. However, the temporal dynamics of cellular senescence in ND conditions is
unresolved. Here, we show senescence features develop in human Huntington's disease (HD)
neural stem cells (NSCs) and medium spiny neurons (MSNs), including the increase of p16
INK4a
,
a key inducer of cellular senescence. We found that HD NSCs reprogram the transcriptional
targets of FOXO3, a major cell survival factor able to repress cell senescence, antagonizing
p16
INK4a
expression via the FOXO3 repression of the transcriptional modulator ETS2.
Additionally, p16
INK4a
promotes cellular senescence features in human HD NSCs and MSNs.
These findings suggest that cellular senescence may develop during neuronal differentiation in
HD and that the FOXO3-ETS2-p16
INK4a
axis may be part of molecular responses aimed at
mitigating this phenomenon. Our studies identify neuronal differentiation with accelerated aging
of neural progenitors and neurons as an alteration that could be linked to NDs.
Background
FOXO (Forkhead Box O) transcription factors are key regulators of longevity that engage
several repair mechanisms to promote the survival of cells facing stress (Martins, Lithgow, &
Link, 2016; Salih & Brunet, 2008). In response to overwhelming stress, FOXO factors may
trigger cell death, e.g. target senescent cells to apoptosis via interaction with the p53 protein
(Baar et al., 2017). In neurodegenerative diseases (ND), FOXO factors such as DAF-16 and
FOXO3 may protect against the cytotoxicity of Huntingtin (HTT) (Parker et al., 2012; Tourette et
al., 2014), SOD1 and p150
Glued
(Mojsilovic-Petrovic et al., 2009), -synuclein (Pino et al., 2014)
and Aß (Cohen et al., 2009). Interestingly, NDs have been linked to cellular senescence,
205

particularly that of glial cells (Bussian et al., 2018; Chinta et al., 2018; Musi et al., 2018; Zhang
et al., 2019). However, the temporal dynamics of cellular senescence in NDs and the role of
FOXO gene regulation in this context are unresolved, limiting our capacity to target the
detrimental effects of cellular senescence in NDs.
We hypothesized that FOXO gene regulation might be able to oppose cellular senescence in
ND conditions. We tested this hypothesis in human cell models of Huntington’s disease (HD), a
genetic yet a primarily late-onset ND caused by CAG expansion in HTT. We focused on
FOXO3, a FOXO factor that is neuroprotective in HD (Tourette et al., 2014). Although FOXO3 is
pivotal to neuronal homeostasis in HD, human FOXO3 targets are unknown, including in ND
conditions. Here, we found that human HD induced pluripotent stem cell (iPSC)-derived neural
stem cells (NSC) reprogram FOXO3 targets in the context of cellular senescence features that
are acquired at the time of neuronal differentiation and that are more pronounced in medium
spiny neurons (MSNs). These features include the increase of p16
INK4a
, a key inducer of cellular
senescence (Baker et al., 2016). Remarkably, FOXO3 target reprogramming represses the
transcription modulator and p16
INK4a
activator ETS2 (Irelan et al., 2009), which antagonizes
p16
INK4a
expression and which may represent an adaptive response as p16
INK4a
promotes the
senescence of human HD NSCs and MSNs. Together, these data reveal that cellular
senescence may develop during neuronal differentiation in HD, affecting striatal neurons, and
that FOXO gene regulation may tip the balance away from the detrimental consequences of
cellular senescence via ETS2-p16
INK4a
, providing a rationale and strategy for targeting cellular
senescence during the early phases of NDs, before the onset of overt neuronal injuries and cell
death, a crucial need in HD and other NDs.
206

Results
Ryk-ICD binds to Armadillo repeats 9-10 of ß-catenin.
In HD, FOXO3 neuroprotection is altered by increased mRNA and protein expression of Ryk
(Tourette et al., 2014), a Wnt receptor important for axon guidance and neurogenesis (Andre et
al., 2012). This effect, a consequence of gene deregulation in HD, is mediated by the Ryk
intracellular domain (Ryk-ICD) in the nucleus where Ryk-ICD binds to the FOXO3 partner ß-
catenin (Tourette et al., 2014). To determine how Ryk-ICD alters FOXO3-ß-catenin
homeostasis, we performed co-immunoprecipitation assays in HEK293T cells. We
overexpressed a Myc-tagged Ryk-ICD fragment, as these cells normally produce relatively
small amounts of this gamma-secretase cleavage product (Tourette et al., 2014). The Myc-
tagged Ryk-ICD fragment co-precipitated with ß-catenin when endogenous FOXO3 was
targeted by the immunoprecipitating antibody (Figure 1A) as well as with FOXO3 when
endogenous ß-catenin was targeted by the immunoprecipitating antibody (Figure 1B). Thus,
Ryk-ICD may be an integral part of the ß-catenin/FOXO3 complex.

207

Figure 6.1. FOXO3, ß ‐catenin, and Ryk ‐ICD form a protein complex in HEK293T cells. The
antibodies used for immunoprecipitation (IP) and for immunoblotting (IB) are indicated across
panels. (a) FOXO3, ß ‐catenin, and Ryk ‐ICD co ‐precipitate in pull ‐down experiments. For
negative control, an IgG isotype was used. Representative Western blots for IP of endogenous
FOXO3. (b) FOXO3, ß ‐catenin and Ryk ‐ICD co ‐precipitate in pull ‐down experiments. For
negative control, an IgG isotype was used. Representative Western blots for IP of endogenous
ß ‐catenin. (c) Deletion mapping of the Armadillo repeat region in ß ‐catenin. Ryk ‐ICD binds to
Armadillo repeats 9 ‐10 of ß ‐catenin. HEK293T cells were transfected with Myc ‐Ryk ‐ICD
construct and the indicated deletion mutants (∆277 ‐488, ∆489 ‐593). For negative control, an
IgG isotype was used. (d) Representative Western blots for IP of wild ‐type and mutant
ß ‐catenin ‐FLAG constructs shown in (c)

208

We mapped Ryk-ICD binding site to ß-catenin. The results suggest the Ryk-ICD binding site
encompasses Armadillo repeats 9-10 on ß-catenin (Figure 1C). The FOXO3 binding site was
previously mapped to ß-catenin Armadillo repeats 1-8 (Essers et al., 2005; Hoogeboom et al.,
2008) (Figure 1C-D), suggesting that Ryk-ICD binds ß-catenin adjacent to FOXO3 to form a
tripartite protein complex.
Human HD NSCs reprogram FOXO3 targets.
Having shown that Ryk signaling may modulate FOXO3 gene regulation through
spatial/allosteric modifications of the ß-catenin/FOXO3 complex, we used massively paralleled
RNA sequencing (RNA-seq) and chromatin immunoprecipitation followed by sequencing (ChIP-
seq) to identify FOXO3 direct targets (F3Ts) in human HD cells. A human induced pluripotent
stem cell (iPSC) model of HD, in which isogenic cells express mutant (72Q/19Q) HTT (HD) or
CAG-corrected (21Q/19Q) HTT (C116) was used (An et al., 2012; Ring et al., 2015). HD and
C116 NSCs were treated with Ryk or scrambled sequence siRNAs. As expected, HD NSCs
showed increased (1.2 fold) Ryk mRNA levels (Figure S1A) and a 2 fold in human HD MSNs
(Figure 6B) whereas ß-catenin and FOXO3 mRNA levels are similar in HD and C116 cells
(Figure S1A, middle and right panels). We then induced FOXO3 nuclear translocation (see
Experimental procedures) in Ryk siRNA-silenced NSCs (Figure S1B and S1D) prior to collecting
RNA-seq and FOXO3 ChIP-seq data.
Much of FOXO3 transcriptional activity can be due to binding enhancers (Eijkelenboom, Mokry,
de Wit, et al., 2013), and there is a significant association between gene regulation and FOXO3
binding up to 20 kb from transcriptional start sites in human cells (Eijkelenboom, Mokry, Smits,
Nieuwenhuis, & Burgering, 2013). We thus defined FOXO3 direct targets as genes that (i) show
FOXO3 binding at promoter and enhancer regions (±20 kb) as determined by ChIP-seq data
and (ii) are up- or down-regulated upon FOXO3 induction into the nucleus as determined by
209

RNA-seq data (Table S1/sheet 1). Additionally, we used RNA-seq data upon FOXO3
knockdown (Table S1/sheet 2). However, ß-catenin transcriptional activity may bypass the
absence of TCF/LEF (Doumpas et al., 2019) and, possibly, that of FOXO3 (Essers et al., 2005).
Hence, FOXO3 nuclear induction and knockdown could differently alter gene regulation, which
calls for caution in using FOXO3 knockdown data to prioritize F3Ts. We thus used these data as
a bona fide criterion for defining two classes of F3Ts, i.e. those identified by (i) FOXO3 nuclear
induction (F3T-IN) and (ii) FOXO3 nuclear induction and FOXO3 knockdown (F3T-IN-KD)
(Table S1).
The F3T-IN data indicated that, of the 219 F3Ts in C116 NSCs, 137 were lost in HD NSCs and
that, among 272 F3T-INs in HD cells, 190 were not present in C116 cells (Figure 2A and 2C,
Figure S2). The gain of F3Ts in HD cells was accompanied by an increase in the proportion of
genes with FOXO3 binding (±20 kb) (Figure 2D, left panel) and in FOXO3 binding levels (Figure
2D, right panel), indicating that FOXO3 occupancy is elevated in HD cells. Silencing Ryk greatly
increased the number of F3T-INs in C116 cells and to a lower extent in HD cells, unrelated to
changes of FOXO3 binding (Figure 2C-D). Silencing Ryk also resulted in some loss of F3T-INs
in C116 cells, regardless of the type of FOXO3 regulation (Table S2/sheet-1). Increased FOXO3
occupancy was also true for F3T-INs that are gained (Table S2/Sheet-2) or conserved
(Table S2/Sheet-3) in HD NSCs. Thus, Ryk signaling may function as a co-repressor or co-
activator of FOXO3, with distinct effects on F3Ts between C116 and HD genotypes. In HD cells,
111 F3Ts are dependent on Ryk (Figure 2C, right panel). Together, these results suggest that
F3Ts are reprogrammed in response to HD during neurogenesis and this response cannot be
fully attributed to higher FOXO3 occupancy. Rather, Ryk signaling may act as a significant
modifier of FOXO3 activity.
210

Figure 6.2. FOXO3 binding and gene regulation in human NSCs expressing normal or
mutant HTT with or without Ryk silencing. (a) Enrichment of FOXO3 binding around the
transcriptional start sites (TSSs) (±2 kb) in human NSCs expressing normal (C116: 19Q/21Q) or
211

mutant HTT (HD: 72Q/21Q) and treated with Ryk siRNA ‐1 (siRyk) or scrambled RNA
(scramble). The color scale is chip signal intensity with maximum set as 8.0. (b) FOXO3 binding
at specific loci in human C116 or HD NSCs. The upper panel is a FOXO3 binding site present in
C116 and HD cells at the TMEM132C locus. The lower panel is a FOXO3 binding site gained in
HD cells at the STK4 locus. (c) Venn diagrams depicting F3 gene regulation across the 4
conditions tested. The left panel shows FOXO3 ‐dependent genes (RNA ‐seq data). The middle
panel shows FOXO3 binding (ChIP ‐seq data). The right panel shows the distribution of F3Ts,
highlighting an increase in the number of F3Ts upon Ryk silencing in C116 (p < 2.2e ‐16) and
HD (p < 2.2e ‐16) cells. (d) FOXO3 binding for the FOXO3 ‐dependent genes. The left panel
shows the percentages of genes with binding or no binding. Multiple chi ‐square tests were
performed using the R function pairwise.prop.test. The right panel shows the signal (peak score)
distributions. Chi ‐squared test was performed for global and pairwise comparisons of the
distributions with the R function chisq.test

212

FOXO3 binding sites are enriched for co-regulator motifs.
FOXO gene regulation involves other transcription factors that synergize with or antagonize
FOXO proteins (Webb et al., 2013). The in silico motif analysis of FOXO3 binding sites in HD or
C116 NSCs with or without Ryk silencing detected commonly enriched motifs such as
Forkhead, as expected, Homeobox, Sp/KLF, ETS, E2F, Pou domain, PAS domain, JUN, zinc
finger, TCF, C/EBP and MEF (Figure S3). Regardless of Ryk silencing, Forkhead motifs showed
a stronger enrichment and higher frequency in HD compared to C116 NSCs (Figure S3),
consistent with higher FOXO3 occupancy (Figure 2C, middle panel), which also applied to other
shared motifs (Homeobox, Sp/KLF, E2F, Pou domain). Some motifs (ASCL, bHLH, p53, SRF)
appeared to be specifically enriched in HD NSCs (Figure S3), which could be due to better
sampling of DNA fragments near FOXO3 binding sites owing to chromatin modifications in HD
cells (Achour et al., 2015). Silencing Ryk did not alter co-regulator motif profiles in either HD or
C116 cells, except for Tbrain factor motifs as detected in HD cells with Ryk silencing
(Figure S3), supporting a model in which Ryk signaling modulates FOXO3 gene regulation by
altering the stoichiometry of the FOXO3/ß-catenin/Ryk-ICD complex (Figure 1, Figure 2).
FOXO3 binding sites overlap between human C116 and mouse NSCs.
Some FOXO3 target families, e.g. those responding to stress and proteotoxicity, may be
conserved across species and cell types (Webb, Kundaje, & Brunet, 2016) while others are not
conserved (Webb et al., 2016). We compared FOXO3 binding sites in human C116 NSCs with
those previously reported in mouse NSCs (Webb et al., 2013) as both studies similarly analyzed
F3Ts. To this end, we considered the best human orthologs of mouse genes bearing FOXO3
binding sites. A significant overlap was detected (20.25%, 446 genes, P = 3.15 x 10
-8
),
suggesting that FOXO3 gene regulation and functions has common features in human and
mouse NSCs (Table S3, Figure S4).
213

F3T reprogramming in human HD NSCs implicates regulators of cell senescence.
We prioritized F3Ts based on the strength and convergence of the reprogramming effects
across F3T-INs categories, including targets that are (i) conserved between HD and C116
NSCs, (ii) reprogrammed in a Ryk-independent manner (no change upon Ryk silencing in HD
cells, regardless of status upon Ryk silencing in C116 cells) and (iii) dependent on Ryk in HD
cells (target status corrected to normal upon Ryk silencing, regardless of status upon Ryk
silencing in C116 cells).
First, we performed Enrichr analyses, retaining the top 1-3 annotations for pathways and
ontologies. Conserved F3T-INs (Table S2/Sheet 3: 82 genes) are enriched for TNF signaling
(KEGG pathway: P = 8.94 10
-05
) and Positive regulation of oxidative stress-induced neuron
death (Gene Ontology Biological Process (GOBP): P = 1.84 10
-05
). F3T-INs lost or gained in HD
NSCs with no effect of Ryk silencing (Table S4: 214 genes) are enriched for PI3-AKT signaling
(KEGG Pathway 2016: P = 3.23 10
-04
) and Positive regulation of protein autoubiquitination
(GOBP: P = 1.88 10
-05
). Finally, Ryk-dependent F3T-INs in HD NSCs (Table S5: 111 genes)
showed no enrichment for pathways, but displayed low-significance enrichment for the GOBPs
Positive regulation of JNK cascade (P = 1.88 10
-04
) and Regulation of cell cycle (P = 0.0054).

To enhance the precision of F3T prioritization, we performed network analysis using F3T-INs as
seeds for extracting high-confidence networks from the STRING database (Szklarczyk et al.,
2015). This analysis highlighted interconnected F3T-INs that implicate core FOXO3 functions
such as for example transcription, translation and protein quality control (Webb et al., 2016).
These F3Ts included those in the conserved (Figure S5A) or reprogrammed (Figure S5B-C)
group(s) that belong to Wnt, Hippo/TGF-ß (e.g., LATS2), Toll-like receptor and mTOR signaling.
This analysis also highlighted Ryk-independent and -dependent F3Ts that in HD NSCs are
relevant to neuron differentiation, synaptic function and cell cycle (Figure S5B-C). In the Ryk-
214

dependent group, network analysis (here, F3T-INs showing the strongest regulation by FOXO3
and at least 3 out of 4 classes of F3T-INs connected to the same node) predicted that, in HD
NSCs, FOXO3 (i) no longer activates CDKN2AIP (also known as CARF), a co-activator of
p14
ARF
(ii) activates SERTAD1 (also known as p34(SEL1)), an inducer of neuronal apoptosis
when in excess that renders CDK4 resistant to inhibition by p16
INK4a
(Li et al., 2005) and (iii)
represses ETS2, a transcription factor that positively regulates p16
INK4a
expression (Ohtani et
al., 2001) and a F3T-IN-KD gene. Together, these FT3 changes suggest suppression of the
CDKN2A locus, particularly the p16
INK4a
segment, in HD NSCs.
FOXO3 represses ETS2 expression in human HD NSCs.
We performed validation studies of FOXO3 regulation of SERTAD1, ETS2 and CDKN2AIP in
human NSCs subjected to stress. FOXO3 silencing (Figure S6A) increased F3T-IN-KD ETS2
mRNA levels in stressed HD NSCs, an effect not detected in unstressed HD and C116 NSCs
(Figure 3A). Thus, ETS2 is negatively-regulated by FOXO3 in HD NSCs in response to stress,
which could partially explain the down-regulation of ETS2 in these cells (Figure 3A, middle and
right panels). In addition, FOXO3 silencing decreased F3T-IN CDKN2AIP mRNA levels in
stressed C116 NSCs (Figure S7A). Thus, CDKN2AIP is a positively-regulated F3T which is lost
in HD cells and could decrease p14
ARF
activity in these cells, potentially promoting cellular
vulnerability (Wadhwa, Kalra, & Kaul, 2017). However, CDKN2AIP is upregulated in HD NSCs
(Figure S6A, middle panel), which could increase the activity of p14
ARF
, a gene also upregulated
in HD NSCs (see below), and promotes cellular resistance in a FOXO3-independent manner. In
contrast, the positive regulation of CDKN2AIP by FOXO3 in C116 NSCs could partially explain
the increase of CDKN2AIP expression upon cell stress (Figure S7A, right panel). Finally,
FOXO3 silencing did not change F3T-IN SERTAD1 mRNA levels in HD NSCs (Figure S6B, left
panel). Additionally, SERTAD1 mRNA levels are slightly decreased in HD NSCs under basal
215

conditions (Figure S7B, middle panel), with no change observed in stressed cells (Figure S7B,
right panel). Thus, SERTAD1 does not appear to regulate stress response in HD NSCs.
ETS2 positively regulates p16
INK4a
expression in human HD NSCs.
ETS2 positively regulates p16
INK4a
expression in human fibroblasts (Ohtani et al., 2001). We
asked whether p16
INK4a
is under ETS2 regulation in HD NSCs. Silencing ETS2 (Figure S6B)
decreased p16
INK4a
mRNA levels in human HD NSCs under basal and stressed conditions, an
effect not detected in C116 NSCs (Figure 3B), suggesting that ETS2 positively regulates
p16
INK4a
expression in HD NSCs, regardless of stress exposure. Furthermore, p16
INK4a
mRNA
levels increased in HD compared to C116 NSCs under basal conditions (Figure 3B, middle left
panel), which declined upon cell stress (Figure 3B, middle right panel), possibly due in part to
decreased ETS2 elicited by FOXO3. FOXO3 knockdown tended to increase p16
INK4a
mRNA
levels (Figure 3B, right panel, P = 0.073) but was not statistically significant, due to the transitive
nature of this regulation via ETS2. The CDKN2A locus encodes p16
INK4a
as well as 14
ARF
.
Silencing ETS2 (see Figure S6B) did not alter p14
ARF
mRNA levels in human HD NSCs (Figure
S7C, left panel), suggesting that ETS2 does not regulate p14
ARF
expression. The p14
ARF
mRNA
levels were higher in HD compared to C116 cells under basal (Figure S7C, middle panel) and
stressed (Figure S7C, right panel) conditions, which could promote cell cycle arrest, however in
a FOXO3-independent manner. ETS1 can also regulate p16
INK4a
expression (Ohtani et al.,
2001). Overexpression of FOXO3, but not that of FOXO3-TM (a non-phosphorylatable mutant),
decreased ETS2 and p16
INK4a
mRNA levels in human HD NSCs subjected to growth factor
deprivation (Figure 3C), suggesting that nucleo-cytoplasmic shuttling of FOXO3 is required for
repressing ETS2 and for ensuring homeostasis of interactions with potential co-repressors (van
der Vos & Coffer, 2008), also suggesting that increased FOXO3-binding to the ETS2 promoter
is part of the general increase of FOXO3 binding in HD NSCs. Silencing ETS1 (Figure S6E)
decreased p16
INK4a
mRNA levels in HD and C116 NSCs under basal conditions, an effect that
216

was lost upon growth factor deprivation in C116 NSCs, but not HD NSCs (Figure 3D),
suggesting that ETS1 remains able to promote p16
INK4a
expression in HD NSCs under stressed
conditions. ETS1 mRNA levels were unchanged in HD compared to C116 NSCs (Figure 3D,
lower left panel) and ETS1 mRNA levels slightly declined in HD NSCs subjected to growth
factor deprivation (Figure 3D, lower right panel). Thus, the ability of ETS1 to promote p16
INK4a

expression in HD NSCs may be dependent upon exposure to external stressors. Collectively,
these results (Figure 3E) suggest that both the decreased expression of ETS1 and FOXO3-
repression of ETS2 may antagonize p16
INK4a
increase in human HD NSCs.
217

218

Figure 6.3. Gene expression analyses in human NSCs. The mRNA levels are normalized to
cells treated with nontargeting control (NTC) siRNAs (siRNA tests) or to C116 cells or cells
without growth factor (GF) deprivation (other experiments). ns, not significant. (a) ETS2 mRNA
levels are increased by FOXO3 reduction in HD NSCs subjected to GF deprivation with no
effect detected in basal conditions nor in normal HTT cells (left panel: *p < .05). ETS2 mRNA
levels are decreased in HD NSCs (middle panel: **p < .01). GF deprivation does not change
ETS2 mRNA levels in C116 and decreases ETS2 mRNA levels in HD NSCs (right panel:
*p < .05). (b) p16INK4a mRNA levels are decreased by ETS2 reduction in HD NSCs in basal
conditions and in cells subjected to stress with no effect detected in normal HTT cells (left panel:
*p < .05, **p < .01). p16INK4a mRNA levels are increased in HD NSCs (middle left panel:
***p < .001). GF deprivation does not change p16INK4a mRNA levels in C116 NSCs and
decrease p16INK4a mRNA levels in HD NSCs (middle right panel: *p < .05). p16INK4a mRNA
levels tend to be increased by FOXO3 knockdown in HD NSCs subjected to GF deprivation
(right panel: not significant with p = .0736). (c) ETS2 and p16INK4a mRNA levels are decreased
by overexpression of FOXO3, but not that of FOXO3 ‐TM, in human HD NSCs subjected to GF
deprivation. The mRNA levels are normalized to cells treated with empty vector. *p < .05 and
**p < .01. (d) p16INK4a mRNA levels are decreased by ETS1 reduction in C116 NSCs in basal
conditions and in HD NSCs in both basal and stress conditions (upper panel: *p < .05,
**p < .01). ETS1 mRNA levels are unchanged in HD compared with C116 NSCs (lower left
panel). GF deprivation does not change ETS1 mRNA levels in C116 NSCs and decreases
ETS1 mRNA levels in HD NSCs (lower right panel: *p < .05). (d) Working model for effect of
FOXO3 target reprogramming on the ETS2 ‐p16INK4a pathway

219

Prepatterned HD NSCs show cellular senescence features in striatal neurons.
Given that p16
INK4a
is a key effector of cellular senescence (Baker et al., 2016), we tested
whether F3T reprogramming in HD NSCs might occur in the context of and respond to cellular
senescence acquired in HD during neuronal differentiation. Using Activin A-induced
dorsoventral prepatterning, which efficiently directs striatal projection neuron differentiation of
human iPSCs (Arber et al., 2015), we observed increase of p16
INK4a
mRNA and protein levels in
HD compared to C116 prepatterned NSCs (Figure 4A-C). In addition, senescence-associated ß-
galactosidase (SA-ß-gal) activity was more abundant in HD compared to C116 NSCs (Figure
4D-F). Importantly, in NSCs derived from additional non-isogenic HD (namely, ND41656 and
ND42222) and control (namely, MIN08i-33114.B and ND42241) iPSC lines, we observed robust
increase in p16
INK4a
expression (Figure S8A-B) and elevated SA-ß-galactosidase activity (Figure
S8C), validating our results across multiple HD patients. We also tested for other markers of
cellular senescence, including increased expression of CDKN1A encoding p21
CIP1
, CDKN1B
encoding p27
KIP1
and MMP3 encoding a matrix metalloproteinase. Under basal conditions,
p21
CIP1
mRNA levels were decreased (Figure S9A), p27
KIP1
mRNA levels were unchanged
(Figure S9B) and MMP-3 mRNA levels were increased (Figure S9C) in HD compared to C116
NSCs. Thus, HD NSCs show increased levels of several markers of cellular senescence
(p16
INK4a
, MMP-3, SA-ß-gal).

220

Figure 6.4. Human HD prepatterned NSCs show increase of p16INK4a and of SA ‐β ‐gal
activity. (a) p16INK4a mRNA levels are increased in HD prepatterned NSCs. Data are
mean ± SD (**p < .01), N = 3. (b) Representative images for modest p16INK4a increase in HD
NSCs. Scale bar in all panels: 100 µm. (c) Quantification of nuclear p16INK4a pixel intensity for
532 C116 ‐NSCs and 1000 HD NSCs. Data are mean ± SD (**p < .01). (d) Representative
images for increase of SA ‐ß ‐gal activity in HD NSCs. Scale bar in all panels: 200 µm. (e)
Quantification of SA ‐ß ‐gal activity for 547 C116 ‐NSCs and 645 HD NSCs. Data are mean ± SD
(****p < .0001). (f) Frequency distribution of SA ‐ß ‐gal signals for data shown in panel (e)

221

Moreover, in HD differentiated MSNs derived from HD NSCs (Figure 5A), p16
INK4a
mRNA levels
were strongly elevated, which was also true for Ryk mRNA levels (Figure 5B). This increase
(~5-fold) was greater in magnitude than that of p16
INK4a
mRNA levels in HD NSCs (~1.7-fold;
Figure 4A) and accompanied by increased p16
INK4a
immunostaining (Figure 5C-E). We also
found that other cellular senescence markers including CDKN2AIP, MMP3, SELL and IGFBP7
were all upregulated in HD MSNs when compared to isogenic control C116 MSNs (Figure 5F).
Further, ETS1 and EST2 are upregulated in HD MSNs compared to isogenic control (Figure 5F)
further confirming the known role of ETS1 and EST2 in transcriptionally increasing p16
INK4a
(Ohtani et al., 2001). Human HD MSNs also showed decreased levels of nuclear HMGB1
(Figure S9D), which relocalizes to the extracellular space in senescent cells (Davalos et al.,
2013), an effect not observed in HD NSCs. The size of a senescent cell increases when
compared to non-senescent cells. We found that human HD MSNs have an increase nuclear
area when compared to C116 MSNs (Figure S9E). Together, these data suggest the
differentiation of NSCs into striatal like neurons is accompanied by increasingly pronounced
features of cellular senescence in HD.
222

Figure 6.5. p16INK4a expression is elevated in human HD MSNs. (a) Representative
images of human NSC ‐derived MSNs using defined enhanced media (Synaptojuice medium).
(b) RT ‐PCR analysis of p16INK4a, FOXO3, and Ryk in C116 and HD MSNs reveals modest
increase of FOXO3 mRNA levels and robust increase of p16INK4a and Ryk mRNA levels in HD
223

MSNs. Data are mean ± SD (*p < .05, ***p < .001). N = 3. (c) Immunofluorescence analysis
reveals dramatic increase of p16INK4a in HD MSNs. Scale bar in all panels: 100 µm. (d)
Quantification of nuclear p16INK4a pixel intensity for N = 596 C116 NSCs and N = 609 HD
NSCs. Data are mean ± SD (****p < .0001). (e) Frequency distribution of nuclear p16INK4a
signals for data shown in Panel (d). (e) RT ‐PCR analysis of CDKN2AIP, MMP3, SELL, IGFBP7,
EST1, and EST2 show increased mRNA levels in HD MSNs compared with C116 MSNs. Data
are mean ± SD (*p < .05, **p < .01, ***p < .001). N = 3

224

FOXO3 and p16
INK4a
oppositely modulate the vulnerability of human HD NSCs.
Next, we investigated whether FOXO3 activity in human HD NSCs might oppose the effects of
p16
INK4a
. In cell growth assays, HD NSCs divided more slowly compared to C116 NSCs (Figure
6A). Reducing FOXO3 (Figure S6F) retarded the growth of HD NSCs (Figure 6B, right panel)
with no change detected in HTT expression (Figure S6H, left panel) and a trend (not significant)
toward reduced growth of C116 NSCs (Figure 6B, left panel), suggesting that FOXO3 promotes
the growth of human HD NSCs. Reducing p16
INK4a
(Figure S6G) slightly increased the growth
of both HD (Figure 6C, right panel) and C116 (Figure 6C, left panel) NSCs, without changing
HTT expression (Figure S6H, right panel), suggesting that p16
INK4a
normally restrains the growth
of human NSCs, regardless of the HTT genotype. Together, these results suggest p16
INK4a
does
not significantly impact the dynamics of the NSC pool in HD.
225

226

Figure 6.6. FOXO3 and p16INK4a oppositely modulate the vulnerability of human HD
NSCs. Significance was tested using two ‐way ANOVA (panels a–c), paired t test (panels d) and
Mann ‐Whitney test (panel g). ns: not significant. (A) Human HD NSCs show reduced rates of
cell growth. Data are mean ± SEM. (b) Reducing FOXO3 does not alter the growth of C116
NSCs (left panel) and strongly reduces the growth of HD NSCs (right panel), with no change
detected in HTT mRNA levels (see Figure S6F, left panel). Data are mean ± SEM. (c) Reducing
p16INK4a slightly increases the growth of C116 (left panel) and HD (right panel) NSCs.
Reducing p16INK4a does not alter HTT mRNA levels in HD NSCs (see Figure S5F, right panel).
Data are mean ± SEM. (d) Reducing FOXO3 increases the mortality of HD NSCs with no effect
detected in C116 NSCs (left: *p < .05). Reducing p16INK4a decreases the mortality of HD
NSCs with no effect detected in C116 NSCs (right: *p < .05). (e) Lenti ‐myc ‐p16INK4a
transduction promotes nuclear release of HMGB1 in cytoplasm of HD and corrected (C116)
MSNs. HD and C116 MSNs transduced for 4 days with lenti ‐myc ‐p16INK4a (red) were
immunostained with HMGB1 (green). NT: transduction without myc ‐p16INK4a. HMGB1
co ‐localizes with the nucleus (DAPI), with low level in cytoplasm. The transduction with
lenti ‐myc ‐p16INK4a significantly relocates HMGB1 into cytoplasm of HD and C116 MSNs
(arrowhead). Scale bars: 100 µm. (f) Upper panel: The quantification of cytoplasmic HMGB1
pixel intensity shows a significant increase of nuclear HMGB1 release in HD vs. C116 MSNs
and in HD vs. C116 MSNs following p16INK4a overexpression (Wilcoxon ranked ‐sum test:
C116 ‐ p16INK4a vs. HD ‐ p16INK4a, p = 6.1e ‐22; C116 ‐p16INK4a vs. HD ‐NT, p = 3.4e ‐06;
C116 ‐p16INK4a vs. C116 ‐NT, p = 1.8e ‐23; HD ‐p16INK4a vs. HD ‐NT, p = 2.3e ‐7; HD ‐NT vs.
C116 ‐NT, p = 5.6e ‐37). Lower panel: data normalized against Myc ‐p16INK4a levels using the
ratio (sum of HMGB1 intensity in ‘cells’ MOI 1/number of cells detected in cells MOI 1)/(sum of
myc ‐tag intensity in ‘cells’ MOI 1/number of cells detected in ‘cells’ MOI 1). The ratios show that
227

HMGB1 relocalization is CAG ‐repeat ‐length ‐dependent. C116 ‐p16INK4a: 1604 cells; C116 ‐NT:
1403 cells; HD ‐p16INK4a: 879 cells; HD ‐NT: 1792 cells

228

In cell vulnerability assays, reducing FOXO3 expression (Figure S5F) strongly potentiates the
mortality of HD NSCs with no effect in C116 cells (Figure 6D). In contrast, reducing p16
INK4a

expression (Figure S6G) decreased the mortality of HD NSCs, with no effect detected in C116
cells (Figure 6D), suggesting that increased p16
INK4a
in human HD NSCs may have deleterious
effects. Thus, FOXO3 transcriptional activity may tip the balance away from the detrimental
effets of cell senescence features such as p16
INK4a
increase on the homeostasis of the NSC
pool in HD.
To further understand the role of p16
INK4a
in differentiated HD MSNs and cellular senescent-like
features, we transduced these cells with lentivirus expressing p16
INK4a
. We tested for HMGB1 is
an early responder to cellular senescence. We quantified cytoplasmic HMGB1 levels as a more
sensitive measure of a senescent-like phenotype compared to nuclear HMGB1 (Figure 6E). We
found increased cytoplasmic HMGB1 basal levels in HD compared to C116 MSNs (Figure 6F).
We also found increased cytoplasmic HMGB1 levels upon p16
INK4a
transduction in HD and C116
MSNs (Figure 6F). These data suggest that p16
INK4a
increase in HD MSNs may promote
senescent-like features in these cells.
p16
INK4a
mRNA levels are increased in the striatum of HD knock-in mice
Analysis of published transcriptomic data (Langfelder et al., 2016) of the Cdkn2a locus
(products p16
INK4
and p19
ARF
) in Hdh mice shows a CAG-repeat- and age-dependent increase
of Cdkn2a (Figure S10A). To assess in vivo relevance of senescence with p16
INK4a
increase as
observed in human HD iPSC-derived cells, we used Hhd-Q175 knock-in mice. Given the lack of
antibodies for a reliable study of p16
INK4a
protein expression in mice (see discussion), we tested
for gene expression and we found that p16
INK4a
mRNA levels are strongly increased in the
striatum of Hdh-Q175 mice at 15 months of age, with a lesser increase detected in the cortex
and no change in the cerebellum (Figure S10B). In HD post-mortem caudate, published data
229

show increase of CDKN2A relative to control (Agus, Crespo, Myers, & Labadorf, 2019) (Figure
S10C). Further studies in mice and human tissues will be needed to confirm cellular
senescence.
Discussion
FOXO factors have widespread anti-aging effects via the transcriptional regulation of stress
response in multiple cell contexts (Martins et al., 2016; Salih & Brunet, 2008). Several cell
maintenance mechanisms under FOXO control are affected in several NDs (e.g., mitochondrial
homeostasis, proteostasis, autophagy, immune system, DNA repair). Understanding how FOXO
gene regulation modulates brain cell maintenance in NDs may thus have important therapeutic
implications. Although FOXO gene regulation has been studied in several cellular contexts
(Webb et al., 2016), including in NSCs (Webb et al., 2013) and neurons (McLaughlin & Broihier,
2017), human FOXO targets in ND conditions are unknown as well as the biology of these
targets in patient-derived cells. Our data identify FOXO3 targets in human NSCs, suggesting a
model in which human NSCs reprogram F3Ts in response to HD. Remarkably, this response
takes place in the context of senescence that develops in these cells, involving the repression of
the ETS2-p16
INK4a
axis, a mechanism that is part of the Ryk-dependent element of F3T
reprogramming. Our data suggest that Ryk signaling is a primary factor that modifies the
FOXO3 target space. However, Ryk may signal through multiple mechanisms, including the
canonical Wnt, PCP and Ryk-ICD pathways (Andre et al., 2012; Lyu, Yamamoto, & Lu, 2008;
Tourette et al., 2014), and the effects of silencing Ryk on the F3T repertoire might also result
from changes in pathways other than the Ryk-ICD pathway. Nonetheless, in HD cells, our data
suggest that pathways that signal onto FOXO3 such as Ryk/Ryk-ICD signaling play a primary
role in modifying the F3T repertoire, rendering FOXO3 able to fine tune the expression of key
inducers of cellular senescence such as p16
INK4a
, whereas the increase of FOXO3 occupancy
may primarily reflect the wide-spread effect of HD on chromatin remodeling (Achour et al.,
2015).
230

Stress response involves p16
INK4a
in several stem cell types, during development or aging
(D'Arcangelo, Tinaburri, & Dellambra, 2017; Oh, Lee, & Wagers, 2014). Human HD NSCs show
senescence features, e.g. p16
INK4a
increase, that are increasingly pronounced as they
differentiate into DARPP-32 positive MSNs. Additionally, p16INK4a promotes the relocalization
of HMGB1 to the cytoplasm, a senescence marker increased in human HD MSNs, in a CAG-
repeat-dependent manner. These results suggest the HD brain faces a continuous cellular
senescence process that affects neurogenesis and adult neurons. Transcriptional
reprogramming by FOXO3 and repression of the ETS2-p16
INK4a
axis may be noticeably
important to promote the robustness of the NSC pool as siRNA-mediated reduction of p16
INK4a

expression decreased the mortality of HD NSCs. Our data thus suggest a model in which
FOXO3 signaling can tip the balance away from cellular senescence in HD. Although siRNA-
mediated reduction of p16
INK4a
expression (about 70%) does not accurately recapitulate the
reduction of p16
INK4a
expression (about 20%) that is elicited by the EST2-p16
INK4a
axis, our data
indicate that reinforcing the outcome of FOXO3 activity in response to HD, i.e. by further
inhibiting p16
INK4a
levels, may have therapeutic potential to avoid the harmful effects
(maladaptation) of a chronic cellular senescence response in human HD neurons. Such an
approach might be of interest for promoting adult neurogenesis in HD as adult-born neurons
may be depleted in the striatum of human HD brains (Ernst et al., 2014) and for targeting the
detrimental consequences of neuronal senescence in other ND contexts. The ability of FOXO3
to tip the balance away from cellular senescence in response to CAG expansion in HTT could
persist in adult neurons as the deregulation of senescence markers may be conserved from
developmental to adult stages. Consistent with this, our data show that p16
INK4a
mRNA levels
are strongly increased in the striatum of Hdh mice. Additionally, p16
INK4a
is increased in HD
NSCs differentiated into MSNs.
231

Our data suggest that neural and neuronal senescence could be set early in HD, a ND
associated with chromatin remodeling (Achour et al., 2015), and has potential to be prosecuted
in view of early drug trials (e.g. during prodromal disease). However, additional studies in the
brain of HD mice and in human HD post-mortem brains are needed to test for the relevance of
senescence to HD. We attempted to test for p16
INK4a
levels in the striatum of Hdh mice using
p16
INK4a
antibodies (i.e. MAS-17142). However, we observed that MAS-17142 recognizes
mouse p16INK4a in a non reliable manner (western blot detection of a band presumably
corresponding to p16
INK4a
in the 60-70 kda range, suggesting p16
INK4a
oxidation/aggegration;
detection of a nucleolus signal that looks like an aggregated signal in immunochemistry
experiments), a problem formerly pointed for several antibodies claimed to properly recognize
mouse p16
INK4a
.
Given the tight links between chromatin remodeling, NDs and cellular senescence (Achour et
al., 2015; Criscione, Teo, & Neretti, 2016; Jakovcevski & Akbarian, 2012), our data raise the
possibility that neural/neuronal senescence could be set early in NDs such as Alzheimer's and
Parkinson's. Although senolytics may positively impact on brain activity in mouse models of NDs
via removing senescent glial cells (Bussian et al., 2018; Zhang et al., 2019), they might have
negative effects by removing neurons and neuronal connections that bear senescence features
but retain a proper activity. Based on our findings, we hypothesize cell-type specific strategies
that can oppose specific detrimental effects of cellular senescence while preserving cellular
homeostasis may be safer, particularly in early drug trials.
In conclusion, our data show that cellular senescence features, including increase of p16
INK4a
,
develop during differentiation of human HD iPSC-derived cells to persist in human HD MSNs.
Our data suggest that FOXO3 may antagonize the progression of cellular senescence in ND
conditions, repressing ETS2 in human HD NSCs, which reduces the expression of p16
INK4a
, in
turn fine tuning stress response. These findings provide a rationale and target, early
232

senescence-like responses, to develop pro-resilience approaches that may be useful for early
intervention in HD and other NDs.
Supplemental Figures

233

Supplemental Fig. 6.1. Gene expression analysis, FOXO3 induction and efficiency of Ryk
silencing in HD and C116 NSCs. Related to Figure 2. (A) mRNA levels of Ryk, ß-catenin and
FOXO3 in HD (72Q/19Q) and C116 (21Q/19Q) neural stem cells as measured by using RT-
PCR. Data are mean±SD. **P < 0.01 compared to C116 NSCs. (B) FOXO3 induction into the
nucleus in HD and C116 NSCs. FOXO3 was induced in the nucleus by growth factor deprivation
complemented by 20 μM LY294002 treatment for 90 minutes (see Methods). Representative
confocal images of FOXO3 intracellular distribution upon induction into the nucleus of HD and
C116 NSCs. (C) FOXO3 induction into the nucleus in HD and C116 NSCs. The nuclear
localization of FOXO3 is increased in HD and C116 cells. Data are mean±SEM for the ratio
Intensity/Area of FOXO3 signal in the nucleus (N = 3). ***P < 0.001 compared to no induction.
(D) Ryk mRNA levels are decreased by siRNA treatment in HD and C116 NSCs. Data are
mean±SD (N = 3). ***P < 0.001 and **P < 0.01 compared to scrambled RNA treatment.
234

235

Supplemental Fig. 6.2. Overview of RNA-seq data upon FOXO3 nuclear induction in HD
and C116 NSCs. (A) Principal component analysis of raw RNA-seq data. Each condition
involves two technical replicates (almost indistinguishable) and two biological replicates. (B) The
log fold change values obtained after the differential expression analysis between FOXO3
nuclear induction vs control were used to generate the volcano plots.

236

Supplemental Fig. 6.3. FOXO3 binding sites are enriched for candidate co-regulator
motifs that are shared across HTT genotypes or unmasked in cells expressing mutant
HTT.
237

Supplemental Fig. 6.4. Overlap between FOXO3 binding sites in human C116 neural stem
cells and mouse neural stem cells. Overlap is shown for FOXO3 binding sites in mouse
NSCs compared to a previously reported and comparable study of FOXO3 binding sites in
mouse NSCs (Webb et al., 2013), using best human orthologs of mouse genes.
238

Cellular'component
Cellular'homeostasis
RAS'superfamily
Non'Coding'RNA
Cellular'metabolism
Transcription'factors
Ribosomes
Nuclear'exosomes'
RNA'degradation
TGF> β'signalling
Hippo'signalling
Circadian'rhythm'
related'gene
Induction'of'apoptosis
Hippo'signalling'(MST1)
Neuromodulation
et'al
Apoptosis
Cell'cycle
Endocytosis
Coiled>coiled'protein
Mitochondrial'carriers
Enzymes
Kinases
Ribosome'
biogenesis
*
*
*
*
*
*
*
*
*
*
*
*
et'al
*
*
*
A
* *
et'al
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
* *
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
B
C
239

Supplemental Fig. 6.5. Network of FOXO3 direct targets in human HD NSCs compared to
control cells. To maximize the assessment of biological relevance, FOXO3 direct targets
defined by FOXO3 nuclear induction (F3T-IN, see Table S1/sheet 1) were used as seed genes
to generate networks that include 200 high confidence neighbors as defined by a probability
score ≥ 0.7 and edge information based on databases and experiments as available in the
STRING database v10.0 (http://string-db.org/). This network is annotated with information on
connectivity between F3T-INs with different regulatory profiles (gained or lost, positively or
negatively regulated), strength of FOXO3 regulation upon FOXO3 nuclear induction, bona fide
information from FOXO3 knockdown (F3T-IN-KD; see Table S1/sheet 2) and gene expression
levels in HD NSCs versus corrected C116 NSCs as previously reported (Ring et al., 2015).
Shown are networks of F3T that are conserved between normal and HD NSCs (A),
reprogrammed (lost or gained) in a Ryk-independent manner in HD NSCs, regardless of their
dependence on Ryk in normal HTT NSCs (B), and reprogrammed (lost or gained) in a Ryk-
dependent manner in HD NSCs (C).

240

Supplemental Fig. 6.6. Gene target expression levels upon treatment with siRNAs and
HTT expression levels upon reduction of FOXO3 or reduction of p16
INK4a
. Data are
mean±SD. (A) FOXO3 mRNA levels are decreased by siRNA treatment HD and normal HTT
NSCs. ***P < 0.001 compared to non-targeting control (NTC) siRNAs. Related to Figure S6 and
Figure 3A. (B) ETS2 mRNA levels are decreased by siRNA treatment HD and normal HTT
NSCs. ***P < 0.001 compared to NTC siRNAs. Related to Figure 3B. (C) FOXO3 mRNA levels
upon overexpression of FOXO3-TM or FOXO3 in normal HTT or HD NSCs. ***P < 0.001 and
**P < 0.01 compared to empty vector. (D) Overexpressing FOXO3 or FOXO3-TM does not
reduce the viability of human NSCs. (E) ETS1 mRNA levels are decreased by siRNA treatment
HD and normal HTT NSCs. *P < 0.05 compared to NTC siRNAs. Related to Figure 3C. (F)
FOXO3 mRNA levels are decreased by siRNA treatment HD and normal HTT NSCs. **P < 0.01
and ***P < 0.001 compared to NTC siRNAs. Related to Figure 6. (G) p16
INK4a
mRNA levels are
decreased by siRNA treatment HD and normal HTT NSCs. ***P < 0.001 compared to non-
241

specific control siRNA treatment. Related to Figure 6. (H) HTT mRNA levels are unchanged by
FOXO3 or p16
INK4a
siRNA treatment in HD NSCs. ns, not significant.
242

SERTAD1
(N# =# 8)
CDKN2AIP
(N# =# 9)
B
A
CDNK2A product#p14
ARF
(N# =# 7) (N# =# 8) (N# =# 8)
(N# =# 9)
0.0
0.5
1.0
1.5
2.0
Relative SERTAD1 mRNA
level/ACTB-HPRT
C116
HD
*
C
(N# =# 8)
(N# =# 6) (N# =# 9)
Figure'S7'related'to'Figure'3
0
1
2
3
4
5
Relative p14
ARF
mRNA
levevel/HPRT-ACTB
***
C116
HD
+ - + -
0
1
2
3
4
5
Relative p14
ARF
mRNA
levevel/ACTB-HPRT
C116
HD
ns
ns
0
1
2
3
Relative CDKN2AIP
mRNA level/ACTB-HPRT
C116
HD
**
+ - + -
0
1
2
3
4
Relative CDKN2AIP
mRNA level/ACTB-HPRT
GF
ns
C116
HD
*
+ - + -
0.0
0.5
1.0
1.5
2.0
2.5
Relative SERTAD1 mRNA
level/ACTB-HPRT
GF
C116
HD
ns
ns
s iN T C
s iF O X O 3
s iN T C
s iF O X O 3
s iN T C
s iF O X O 3
s iN T C
s iF O X O 3
0
1
2
3
Relative CDKN2AIP
mRNA level/ACTB-HPRT
+ + - - GF
ns
ns
*
ns
C116
HD
s iN T C
s iF O X O 3
s iN T C
s iF O X O 3
s iN T C
s iF O X O 3
s iN T C
s iF O X O 3
0
1
2
3
Relative SERTAD1
mRNA level/ACTB-HPRT
+ + - - GF
C116
HD
ns ns
ns
ns
s iN T C
s iE T S 2
s iN T C
s iE T S 2
s iN T C
s iE T S 2
s iN T C
s iE T S 2
0.0
0.5
1.0
1.5
2.0
2.5
Relative p14
ARF
mRNA
level/ACTB-HPRT
HD
C116
ns
ns
+ + - - GF
ns
ns
243

Supplemental Fig. 6.7. Evaluation of candidate FOXO3 targets (CDKN2AIP, SERTAD1)
and products of the CDKN2A locus (i.e., p14
ARF
). Related to Figure 3. (A) CDKN2AIP mRNA
levels are decreased by FOXO3 silencing in stressed C116 NSCs, with no change detected in
HD NSCs (left panel). CDKN2AIP mRNA levels are increased in HD NSCs (middle panel: **P <
0.01 compared to C116 cells). Growth factor (GF) deprivation increases CDKN2AIP mRNA
levels in C116 NSCs (*P < 0.05 compared to C116 cells) with no change detected in HD NSCs
(right panel). ns, not significant. (B) SERTAD1 mRNA levels are not altered by FOXO3 silencing
in HD and C116 NSCs (left panel). SERTAD1 mRNA levels are slightly decreased in HD NSCs
(middle panel: *P < 0.05 compared to C116 cells). GF deprivation does not change SERTAD1
mRNA levels in both cell genotypes (right panel). ns, not significant. (C) p14
ARF
mRNA levels
are not altered by ETS2 reduction in HD and C116 NSCs (left panel). p14
ARF
mRNA levels are
increased in HD NSCs (middle panel: ***P < 0.001 compared to C116 cells). GF deprivation
does not change p14
ARF
mRNA levels in both cell genotypes (right panel). ns, not significant.

244

A
ND42222 ND41656 ND42241 MN08i-33114.B
DAPI p16
INK4a
Nestin Merge
B
Figure S8 related to Figure 4
MN08i-33114.B ND42241 ND41656 ND42222
C
245

Supplemental Fig. 6.8. Increased levels of p16
INK4a
and elevated SA-ß-gal activity are also
characteristic of other non-isogenic HD NSC lines. Related to Figure 4. (A) p16
INK4a
mRNA
levels as determined by RT-PCR analysis in control NSCs (blue bars; MN08i-33114.B and
ND42241) and HD NSCs (red bars; ND41656 - CAG 57 and ND42222 - CAG 109). Data are
mean ± SD (N = 3). ***P < 0.001 (one-way ANOVA; Tukey’s multiple comparison test). (B)
Immunofluorescence analysis reveals robust increase of p16
INK4a
in HD NSCs. Scale bar in all
panels: 100 mm. (C) Representative images showing increased expression of SA-b-gal activity
in HD NSCs (ND41656 - CAG 57; ND42222 - CAG 109) compared to control NSCs (MN08i-
33114.B and ND42241). 10X magnification. Scale bar in all panels: 200 mm.

246

Supplemental Fig. 6.9. Relevant markers of senescence evaluated in HD NSCs and MSNs.
Related to Figure 4 and Figure 5. (A) p21
CIP1
mRNA levels are decreased in HD compared to
247

C116 NSCs (**P < 0.01). (B) p27
KIP1
mRNA levels are unchanged in HD compared to C116
NSCs. ns, not significant. (C) MMP-3 mRNA levels are increased in HD compared to C116
NSCs (***P < 0.001). (D) Immunofluorescence analysis reveals depletion of HMGB1 from nuclei
of HD MSNs compared to C116 MSNs. Scale Bar in all panels: 100 µm. (E) Quantification of
nuclear HMGB1 pixel intensity for N = 437 C116 NSCs and N = 491 HD NSCs. Data are
mean±SD (****P < 0.0001). (F) Quantification of nuclear area in HD (N = 1879 cells) and C116
(N = 2607 cells) MSNs. t-test *P < 0.05. Data are mean±SEM.

248

Supplemental Fig. 6.10. Increase of p16
INK4a
mRNA levels in the striatum of HD model
mice. (A) CAG-repeat-length- and age-dependent increase of Cdkn2a mRNA levels. Graph
representation of data by Langfelder and coll. (Langfelder et al., 2016). (B) p16
INK4a
mRNA
levels are increased in the striatum (STR) of heteroztygote HD knock-in mice zQ175DN (het)
compared to wild-type (wt) mice at 15 months of age, with lesser increases detected in the
cortex (CTX) and in the cerebellum (CVT). Similar features were detected for p19
Arf
mRNA
levels. t-test *P < 0.01 and **P < 0.001. Data are expression calculated as relative to the Hprt
housekeeping gene and then ratioed over the lowest wild-type value. (C) Increase of CDKN2A
mRNA levels in post-mortem prodromal HD caudate nuclei. Graph representation of data by
Agus and coll. (Agus et al., 2019). ***P < 0.001.

249

Supplemental Tables
Supplemental Table 6.1. Definition of F3Ts in human C116 and HD NSCs. Sheet 1 shows the
complete list of human genes that are differentially expressed upon FOXO3 induction into the
nucleus (IN) compared to no FOXO3 induction (F3T-IN). This table is annotated with information
on FOXO3 binding at promoters (-5 kb/+2 kb) and enhancers (± 20 kb outside the promoter
regions), deregulation in HD NSCs, the same as the ones used herein, as previously reported
(Ring et al., 2015), druggability, overlap with FOXO3 targets in other cell types as previously
reported (Eijkelenboom, Mokry, Smits, Nieuwenhuis, & Burgering, 2013; Paik et al., 2009;
Renault et al., 2009; Webb, Kundaje, & Brunet, 2016), and overlap with RNAi screens in a
transgenic nematode (Lejeune et al., 2012) and human cell (Miller et al., 2012) models of HD
pathogenesis. Sheet 2 shows the complete list of human genes that are differentially expressed
upon cell stress (growth factor deprivation) in a FOXO3-knockdown (KD)-dependent manner.
These genes are those for which the log fold change (LFC) of gene expression levels in
stressed cells treated with FOXO3 siRNAs is no longer significant compared to the significant
LFCs in stressed cells treated with non-targeting control (NTC) pool of RNAs. These genes also
comprise those for which there is a significant difference between log fold change (LFC) of gene
expression levels in stressed cells treated with FOXO3 siRNAs compared to significant LFCs in
stressed cells treated with NTC RNAs. LFCs were considered significant for a q-value < 0.1
(green cells) as determined using false discovery rate (FDR) analysis. Differences between
LFCs were considered significant for a p-value < 0.05 as determined using the R function
pnorm. A significant difference or a loss of LFC significance upon FOXO3 knockdown define the
subgroup of F3T-IN-KD targets (blue cells). NA, not applicable.
Supplemental Table 6.2. Table S1 extracts showing the list of F3T-INs that are gained in HD
NSCs and their behavior upon silencing of Ryk (sheet 1), those that are lost in HD NSCs and
250

their behavior upon reduction of Ryk expression (sheet 2) and those that are conserved in HD
NSCs and their behavior upon reduction of Ryk expression (sheet 3).
Supplemental Table 6.3. Comparison of FOXO3 binding sites in human C116 and mouse
NSCs. See also Figure S3.
Supplemental Table 6.4. F3T-IN targets reprogrammed (lost or gained) in human HD NSCs in a
Ryk-independent manner (Table S1 extract). The sub-group of F3T-IN-KD targets is indicated
by blue stars in Figure S4B.
Supplemental Table 6.5. FOXO3 targets reprogrammed (lost or gained) in human HD NSCs in a
Ryk-dependent manner (Table S1 extract). The sub-group of F3T-IN-KD targets is indicated by
blue stars in Figure S4C.

251

CHAPTER 7: Exploring Transcriptional Changes in the Aged Hippocampus Driven by
Whole-Body Clearance of Senescent Cells.
Abstract
Aging is a complex process that leads to a decline in cellular and organismal function that is
reflected in harmful alterations in various tissues. A key alteration is the accumulation of
senescent cells, defined by irreversible cell-cycle arrest and the release of a diverse array of
cytokines, growth factors, and other molecules. This unique secretory cocktail, referred to as the
senescence-associated secretory phenotype (SASP), elicits widespread systemic effects. In the
hippocampus, senescent cells contribute to several harmful age-related phenotypes and are
implicated in multiple neurodegenerative diseases, including Alzheimer’s, Parkinson’s, and
Huntington’s diseases. Despite their harmful effects, senescent cells also partake in vital
physiological functions, highlighting the need for understanding the impacts of their removal.
Here, we analyzed the effect of whole-body senescent cell removal on the hippocampi of aged
mice using single nuclei RNA sequencing (snRNAseq). Two methods were used to remove
senescent cells: treatment with the senolytic agent 25-hydroxycholesterol (25HC), and induction
of apoptosis in the p16-3MR transgenic model by expressing the senescence marker p16 upon
ganciclovir treatment. By comparing these data with snRNAseq data from the hippocampi of
both young and old mice, we identified a reversal of age-related transcriptional alterations
across different cell types following senescent cell removal. Notably, glutamatergic neurons
showed a significant reversal of age-related transcriptional patterns after removal of senescent
cells. Pathway analysis of these transcriptomic signatures suggested downregulation of synaptic
plasticity–associated signaling pathways during aging, with a resurgence after senescent cell
clearance. In addition, we compared the transcriptional changes from senescent cell removal to
caloric restriction, another anti-aging intervention. This study offers a detailed examination of
252

age-related transcriptional changes induced by senescent cells, and evaluates the effects of
different senescent cell-removal strategies.
Background
Aging is a complex process involving multiple processes and characterized by a gradual decline
in cellular and organismal function [1]. The decline results from several processes, such as the
accumulation of senescent cells, chronic inflammation, the loss of protein homeostasis,
dysregulated nutrient sensing, and mitochondrial dysfunction [1]. In the brain, harmful
alterations affect various processes, including inflammation, neurogenesis, myelination,
neurotransmission, blood circulation, oxidative stress response, protein production and cellular
respiration [2, 3]. However, multiple interventions counteract age-related processes, including
heterochronic parabiosis [4-8], caloric restriction [9], rapamycin [10], young bone marrow
transplantation [11], and elimination of senescent cells [12, 13].
Clearance of senescent cells has shown promising results by extending health span and
lifespan and improving outcomes in multiple models of age-related diseases [14, 15]. Senescent
cells are characterized by an irreversible cell-cycle arrest and secretion of the senescence-
associated secretory phenotype (SASP) [16]. The SASP is a complex mixture of cytokines,
growth factors and other molecules that affect the surrounding niche and drive systemic effects
[17-20]. Senescent cells have a role in age-related cognitive decline and neurodegenerative
diseases [21-23]. Whole-body elimination of senescent cells appears to mitigate age-related
hippocampal inflammation, and cognitive impairment and to promote neurogenesis [21, 24, 25].
Senolytics, drugs designed to eliminate senescent cells, are being explored for their potential in
a range of clinical scenarios [16]. However, it's worth noting that, despite the harmful effects
senescent cells may have in aging and diseases, they also play crucial roles in beneficial
physiological processes [26-28]. This emphasizes the need for a comprehensive understanding
of the impact of senescent cell removal.
253

In this study, we utilized single nuclei RNA sequencing (snRNAseq) to examine the
hippocampus of aged mice subjected to whole-body clearance of senescent cells. We utilized
two strategies for senescent cell clearance: treatment with the senolytic agent 25-
hydroxycholesterol (25HC) and use of ganciclovir to induce apoptosis in cells expressing p16 in
the p16-3MR transgenic model. Additionally, snRNAseq on the hippocampus of young and aged
mice allowed us to identify age-related transcriptional changes reversed by clearance of
senescent cells. Moreover, we examined the transcriptional shift prompted by senescent cell
clearance in relation to the signatures of other anti-aging interventions. Overall, our work
elucidates the age-related transcriptional changes induced by senescent cells and gauges the
effects of various interventions to identify potential synergistic strategies.
Results
Using single nuclei profiling to study effects of clearance of senescent cells in the
hippocampus.
We employed high-throughput snRNA-seq to study the transcriptional profile of the aged mouse
hippocampus and subsequent changes after removal of senescent cells (Fig. 7.1A-B).
We utilized two models of whole-body senescent cell clearance. The first model consisted of
p16-3MR transgenic mice treated with ganciclovir [28]. In this model, the promoter of the p16
gene, often used as a senescence indicator, drives expression of the thymidine kinase (TK)
gene. The interaction between TK and ganciclovir triggers cell death. The second model
consisted of mice treated with the senolytic agent 25HC [29]. Notably, these two methods of
senescent cell clearance are different. The transgenic model specifically targets cells
expressing p16, and 25HC induces cell death in p21-positive and p16-positive senescent cells,
indicating differences in the types of senescent cell each treatment targets [29].
254

We established corresponding control groups for each treatment, using the respective vehicles
and old and young untreated controls. All groups used the p16-3MR transgenic model to ensure
a consistent genetic background.
We dissociated hippocampal tissues and extracted nuclei by a semi-automated workflow (see
Methods) and obtained 124,589 single nuclei. After rigorous filtering and evaluation of batch
effects (see Methods), we retained 100,740 nuclei for analysis (Supplemental Figs. 7.1 and
7.2). These nuclei originated from the hippocampi of several groups: six aged mice treated with
GCV (and five PBS-treated controls), six aged mice administered with 25HC (and five
PBS+HβCD-treated controls), and eight untreated mice, evenly divided between young and old.
Using Uniform Manifold Approximation and Projections (UMAP), we identified 13 major cell
types (Fig. 7.1C-E). Neurons formed the largest population, further categorized into
glutamatergic neurons, GABAergic neurons, neurons residing in the dentate gyrus (DG
neurons), and Cajal-Retzius cells (CR neurons). We also found significant populations of
oligodendrocytes, astrocytes, microglia, and oligodendrocyte progenitor cells (OPCs), with
smaller numbers of vascular and leptomeningeal cells (VLMCs), pericytes, arachnoid barrier
cells (ABCs), and choroid plexus capillary endothelial cells (CPCs).

255

256

Figure 7.1. Overview of the single-nuclei RNA sequencing analysis. (A) Schematic
representation of groups of animals and respective treatments in this study. (B) Schematic of
experimental workflow to obtain single nuclei data. (C) UMAP projection of single nuclei
obtained from all groups with identified clusters of astrocytes, endothelial cells, microglia,
GABAergic neurons, neurons located in the dentate gyrus (DG neurons), glutamatergic
neurons, Cajal-Retzius cells (CR neurons), oligodendrocytes, oligodendrocyte progenitor cells
(OPCs), pericytes, choroid plexus capillary endothelial cells (CPCs), pericytes, arachnoid barrier
cells (ABCs), and vascular and leptomeningeal cells (VLMCs). (D) UMAP projection of three
major cell types showing expression of representative cell type markers (Gad2 for GABAergic
Neurons; Gja1 for astrocytes; Mal for oligodendrocytes). (E) Heatmap displaying cell type
markers across the different cell types identified.

257

We focused our analysis on glutamatergic neurons, GABAergic neurons, DG neurons,
astrocytes, oligodendrocytes and microglia as they contained the higher numbers of nuclei
collected. Analysis in the smaller populations of cells was not carried out due to the sparsity in
gene coverage caused by the low number of cells (Supplemental Fig. 7.3).
Comparison of clearance of senescent cells via a transgenic model of p16-positive cell
ablation and 25HC.
To assess the similarity of transcriptomic changes induced by the two methods of senescence
cell clearance, we compared the transcriptional changes caused by GCV treatment with those
caused by 25HC treatment across various hippocampal cell types.
We utilized two approaches to compare the signatures. First, we used rank-rank hypergeometric
overlap (RRHO) analysis [30-32]. This tool facilitates a non-parametric comparison without the
limitations of a fixed threshold. To perform RRHO, we ranked genes detected in both
experiments by their log2 fold-change values and identified overlapping genes between the two
signatures. The number of overlapping genes at different positions in the ranked gene list was
then used as a hypergeometric test to determine the significance of the similarity. After adjusting
the resulting p-values, we used them to generate a heatmap that showcases regions of high
similarity in the two gene expression patterns.
Second, we compared the level of correlation of the log2-transformed fold-change of genes
significantly changing (see Methods) in response to 25HC and GCV treatments (absolute
log2FC > 0.1 and p-value < 0.05) by conducting a two-sided Spearman’s rank correlation test
for each cell type.
The RRHO analysis showed a positive correlation between the effects of 25HC and GCV on
glutamatergic neurons, microglia, DG neurons, and oligodendrocytes marked by high levels of
overlap in the lower left and upper right quadrants (Fig. 7.2A-D). The RRHO plots also provided
nuanced insights. For example, the greatest level of correlation in glutamatergic neurons and
258

microglia was seen in genes downregulated by both 25HC and GCV (lower left quadrant). In
contrast, for oligodendrocytes, the highest correlation was in genes upregulated by both 25HC
and GCV (upper left quadrant). We also identified an unexpected negative correlation between
the effects of GCV and 25HC in GABAergic neurons (Fig. 7.2E), and the RRHO plot for
astrocytes showed a sparse signature, indicating a lack of a correlation (Fig. 7.2F).

259

260

Figure 7.2. Non-parametric comparison of transcriptional signatures caused by removal
of senescent cells by GCV or 25HC. RRHO plots showing overlap between GCV and 25HC
transcriptional signatures in (A) glutamatergic neurons (B), microglia, (C) DG neurons, (D)
oligodendrocytes, (E) GABAergic neurons, and (F) astrocytes. Top right quadrant represents
genes upregulated in both signatures, lower left quadrant represents genes downregulated in
both signatures, top left quadrant represents genes upregulated by 25HC expression but
downregulated by GCV expression and bottom right quadrant represents genes upregulated by
GCV but downregulated by 25HC. The color of each pixel represents the statistical significance
of the overlap at that position in the rank-rank matrix. Numbers in the X and Y axis indicate the
number of sectors in the rank-rank matrix.

261

Two-sided Spearman’s rank correlation showed significant positive correlations between the
changes caused by 25HC and GCV treatments in glutamatergic neurons (ρ=0.51, p value < 2e-
16), microglia (ρ=0.54, p value < 2e-16), DG neurons (ρ=0.32, p value < 1.9e-05) and
oligodendrocytes (ρ=0.6, p value < 2e-16) (Fig 7.3A-D). The previously observed negative
correlation in GABAergic neurons was confirmed (ρ=-0.61, p value < 2e-16) (Fig 7.3E) and a
lack of correlation in astrocytes (Fig 7.3F).

262

Figure 7.3. Correlation of transcriptional signatures caused by removal of senescent
cells by GCV or 25HC. Scatter plots showing genes differentially expressed as a result of
treatment with GCV and 25HC in (A) glutamatergic neurons (B), microglia, (C) DG neurons, (D)
oligodendrocytes, (E) GABAergic neurons, and (F) astrocytes. Top right quadrant represents
263

genes upregulated in both signatures, lower left quadrant represents genes downregulated in
both signatures, top left quadrant represents genes upregulated by 25HC expression but
downregulated by GCV expression, and bottom right quadrant represents genes upregulated by
GCV but downregulated by 25HC. Two-sided Spearman’s rank correlation was performed for
each comparison.

264

A closer look into the DEGs in GABAergic neurons by GCV and 25HC treatments showed that
more than 80% of shared DEGs had an opposite direction (Supplemental Fig. 7.4A). Ingenuity
Pathway Analysis (IPA) of the differentially expressed genes by both conditions predicted an
opposite activation status for synaptic long-term depression and calcium signaling, among other
canonical pathways related to synapse function (Supplemental Fig. 7.4B).
Effects of clearance of senescent cells on age-related transcriptional changes in
neurons.
Whole-body clearance of senescent cells has been associated with a reversal of age-related
dysfunction, underscoring the critical role of senescent cells in age-related functional decline.
Such transformative effects were also observed in the hippocampus, where the removal of
senescent cells counteracted several aging-related phenotypes, such as diminished
neurogenesis, reduced synaptic activity, and alterations in gene expression [21, 24].
To uncover the transcriptional changes in the hippocampal neurons attributable to the presence
of senescent cells and potentially reversed upon their removal, we compared the transcriptional
alterations in major cell types occurring naturally with aging in the hippocampus with those after
senescent cell clearance (e.g., via 25HC or ganciclovir treatment).
In glutamatergic neurons, RRHO analysis showed a strong negative correlation between the
transcriptional profile caused by aging and those caused by either 25HC or GCV treatment (Fig.
7.4 A-B). Notably, both 25HC and GCV demonstrated a higher overlap of genes that were
downregulated due to aging but upregulated after the clearance of senescent cells. We detected
a similar but less pronounced reversal of the age-related signature in DG neurons with both
treatments (Fig. 7.4 C,D). In contrast, GABAergic neurons displayed a partial reversal of genes
downregulated with age and upregulated by GCV treatment but showed a positive correlation
between genes upregulated with age and upregulated after 25HC treatment (Fig. 7.4 E,F). This
265

implies that 25HC could intensify certain age-related transcriptional changes in GABAergic
neurons.

266

267

Figure 7.4. Non-parametric comparison of transcriptional signatures caused by aging
and removal of senescent cells by GCV or 25HC in neurons. RRHO plots showing overlap
between the transcriptional signature caused by aging and both models of senescent cell
clearance in different types of hippocampal neurons. (A) GCV vs aging in glutamatergic
neurons. (B) 25HC vs aging in glutamatergic neurons. (C) GCV vs aging in DG neurons. (D)
25HC vs aging in DG neurons. (E) GCV vs aging in GABAergic neurons. (F) 25HC vs aging in
GABAergic neurons.
268

We also performed Spearman’s correlation tests on genes passing differential expression
thresholds in the aging, 25HC and GCV signatures. In glutamatergic neurons, we found a
significant negative correlation between the transcriptional changes caused by aging and those
resulting from senescent cell clearance, regardless of whether the clearance was achieved via
25HC or GCV treatment (Fig. 7.5A-B). Remarkably, over half of the genes that were altered
due to aging reverted after the removal of senescent cells (Fig. 7.5C). Furthermore, IPA of the
transcriptional signatures impacted by aging, 25HC or GCV supported the reversal of age-
related alterations. Aging was predicted to reduce the activity of several pathways involved in
synaptic plasticity, such as the synaptogenesis signaling pathway and CREB signaling, but
these were increased upon senescent cell clearance. Interestingly, this analysis also suggested
an increase on the pathogen-induced cytokine signaling pathway and mitochondrial dysfunction
with age but a decrease after senescent cell clearance (Fig. 7.5D).

269

Figure 7.5. Removal of senescent cells reverses age-related transcriptional changes in
glutamatergic neurons. Two-sided Spearman’s rank correlation of genes differentially
expressed with aging and (A) GCV or (B) 25HC in glutamatergic neurons. (C) Heatmap
showing scaled log2 fold-changes of genes differentially expressed during aging, 25HC and
270

GCV treatment. Red outline delineates genes affected by age that reverse direction after
clearance of senescent cells. (D) Heatmap of IPA canonical pathways predicted to be
modulated with aging, GCV or 25HC treatment.
For DG neurons and GABAergic neurons, we performed Spearman correlation tests between
genes differentially expressed during aging or a modality of senescent cell clearance due to a
low number of common genes passing significance thresholds for all comparisons. This
identified a modest negative correlation between aging and GCV treatment in GABAergic (ρ=-
0.18, p value = 1.2e-7) and DG neurons (ρ = -0.11, p value = 0.037) (Supplemental Fig. 7.5A
and C). Similarly, small negative correlations were also identified between aging and 25HC in
DG neurons (ρ = -0.25, p value = 4.5e-10) (Supplemental Fig. 7.5B). The comparison of aging
and 25HC in GABAergic neurons yielded a small positive correlation (ρ = 0.17, p value = 0.002),
supporting the results from the RRHO plots (Supplemental Fig. 7.5D).
Effects of clearance of senescent cells on age-related transcriptional changes in glia.
In RRHO plots from glial cells, we observed a negative correlation between the aging signature
and that of 25HC and GCV in microglia, oligodendrocytes and astrocytes (Fig. 7.6 A-D). In
microglia and oligodendrocytes, the overlap of genes was higher between genes upregulated
with age and downregulated by clearance of senescent cells. Interestingly, even though we
observed no significant correlation between 25HC and GCV in astrocytes, we saw a negative
correlation in astrocytes between the aging signature of astrocytes and the one caused by
25HC or GCV (Fig 7.6. E-F). The observed discrepancy was due to 25HC showing a greater
correlation between genes downregulated with aging and those upregulated by 25HC.
Conversely, GCV exhibited a stronger correlation between genes upregulated with age and
those downregulated by GCV.

271

272

Figure 7.6. Non-parametric comparison of transcriptional signatures caused by aging
and removal of senescent cells by GCV or 25HC in glial cells. RRHO plots showing overlap
between the transcriptional signature caused by aging and both models of senescent cell
clearance in different types of hippocampal glial cells. (A) GCV vs aging in microglia. (B) 25HC
vs aging in microglia. (C) GCV vs aging in oligodendrocytes. (D) 25HC vs aging in
oligodendrocytes. (E) GCV vs aging in astrocytes. (F) 25HC vs aging in astrocytes.

273

We also performed Spearman’s correlation tests on genes passing differential expression
thresholds in the 25HC, GCV or aging signatures. We detected no statistically significant
correlation in microglia from GCV- or 25HC-treated mice when compared against the aging
signature (Supplemental Fig. 7.6A and B). This was the case also for oligodendrocytes from
GCV-treated mice when compared with the aging signature (Supplemental Fig. 7.6C).
However, we found a significant negative correlation between the 25HC treatment and aging
comparison in oligodendrocytes (ρ=-0.44, p value = 2e-16) (Supplemental Fig. 7.6D). Similarly,
we also detected statistically significant negative correlations in astrocytes, both in the GCV and
25HC signatures when compared to aging (Supplemental Fig. 7.6E-F)
Comparison of transcriptional signatures caused by clearance of senescent cells and
caloric restriction.
We next sought to investigate how the effects caused by clearance of senescent cells compared
to other anti-aging interventions. We utilized a published bulk RNAseq dataset of hippocampus
from aged mice that had undergone caloric restriction (CR) and their ad libitum (AL) controls
[33]. We generated pseudobulk transcriptomic signatures for aging, 25HC and GCV treatments
and compared them to the that of caloric restriction. We saw a negative correlation between the
signature of aging and that of CR (Fig 7.7. A) as reported [33]. When comparing GCV and CR,
we saw a partial positive correlation concentrated in the top right quadrant, indicating similarity
between genes upregulated with caloric restriction and GCV treatment (Fig 7.7. B). As
expected, the signature of 25HC lacked a positive correlation with CR, probably driven by the
reaction created by 25HC in GABAergic neurons that opposed that of GCV and aging (Fig 7.7.
C).

274

Figure 7.7. Non-parametric comparison of pseudo-bulk transcriptional signatures caused
by aging, removal of senescent cells via GCV or 25HC against caloric restriction (CR) in
the hippocampus. RRHO plots showing overlap between the transcriptional signature caused
by (A) aging and CR. (B) GCV and CR. (C) 25HC CR.

275

Discussion
Extensive research has shed light on the dynamic and modifiable nature of the aging process,
with various interventions demonstrating an ability to rejuvenate tissues, extend health span,
and prolong lifespan [34]. The clearance of senescent cells has emerged as a particularly
effective approach [16]. Here, we conducted a comprehensive snRNAseq analysis to explore
the gene expression changes occurring in the aged mouse hippocampus after elimination of
senescent cells. Our goals were to gain insights into the transcriptional alterations associated
with aging and to determine to what extent these changes were driven by senescent cells and
whether these changes could be reversed through the removal of senescent cells from the
whole body.
We first characterized the cellular complexity of the hippocampus and were unable to detect
statistically significant changes in cell composition. However, slight variations during the
microdissection of the hippocampus could have resulted in different quantities of cells across
samples, potentially increasing the variability in cell proportions and reducing the ability to detect
significant changes. Consequently, the possibility of shifts in specific cell populations cannot be
completely ruled out. Future investigations employing spatially resolved transcriptomics
techniques hold promise for providing further insights into this matter.
We then delved into the gene expression changes associated with aging and the clearance of
senescent cells in specific hippocampal cell types. Our goal was to identify the genes and
pathways affected by the accumulation of senescent cells during aging and determine if they
are rescued after their elimination. We identified the primary cell types undergoing these
changes and revealed cell-type-specific transcriptional signatures and programs. Importantly,
we uncovered the dynamic regulation of the transcriptome in glutamatergic neurons, which was
influenced by both aging and the elimination of senescent cells. We observed a significant
proportion of transcriptional changes associated with aging in glutamatergic neurons, which
276

were subsequently reversed after the clearance of senescent cells. Importantly, we found that
the transcriptional program affected by aging and senescent cell clearance was greatly enriched
for genes involved in synaptic function and plasticity. These findings align with recent evidence
highlighting the beneficial effects of eliminating senescent cells in preventing age-related decline
in synaptic activity within the hippocampus [21].
Intriguingly, we found the transcriptomic reprogramming shared similarities with the
reprogramming reported in mice subjected to CR. These findings suggest that both
interventions—senescent cell clearance and CR—may induce similar transcriptional changes in
aged tissue.
Our study shows that removing senescent cells can reverse age-related changes in various
hippocampus cell types. However, the varied responses of GABAergic neurons to different
treatments hint at potential unpredicted effects from senolytic molecules.
In summary, our comprehensive computational analyses and findings provide a valuable
resource for generating hypotheses and driving future research in this field. By expanding upon
previous studies, our work contributes to a deeper understanding of the underlying mechanisms
involved in aging and provides insights beyond mere characterization of cell states. We
enhanced our understanding of the impact of senescent cells on the aging process and their
critical role in brain function. These advancements have the potential to shape future
investigations and interventions aimed at targeting senescent cells or other age-related
processes for therapeutic purposes. Future research should consider the complex relationships
between senescent cells and other aging processes. This approach will help identify
interventions that simultaneously target various aging mechanisms, leading to combined
beneficial effects.

277

Supplementary Figures

Supplemental Figure 7.1. Quality control metrics. Number of features, UMI counts and
percent of mitochondrial RNA for all groups (A) before and (B) after filtering.

278

Supplemental Figure 7.2. Batch and sample integration. UMAP plots showing cells labeled
by batches (A) before and (B) after integration with Harmony. UMAP plots showing cells labeled
by groups (C) before and (D) after integration with Harmony.
279

Supplemental Figure 7.3. Cell type numbers. (A) Bar plot showing the percentage of different
cell types identified in different groups.

280

Supplemental Figure 7.4. Transcriptional signature of GABAergic neurons after 25HC or
GCV treatment. (A) Heatmap showing differentially expressed genes in GABAergic neurons
after treatment with GCV and 25HC. (B) Heatmap of IPA canonical pathways predicted to be
modulated by removal of senescent cells via GCV or 25HC treatment.

281

Supplemental Figure 7.5. Correlation of transcriptional signatures caused by removal of
senescent cells via GCV or 25HC, compared to aging signature in neurons. Scatter plots
showing (A) Genes differentially expressed as a result of treatment with GCV or aging in
GABAergic neurons. (B) Genes differentially expressed as a result of treatment with 25HC or
aging in GABAergic neurons. (C) Genes differentially expressed as a result of treatment with
GCV or aging in DG neurons. (D) Genes differentially expressed as a result of treatment with
25HC or aging in DG neurons. Two-sided Spearman’s rank correlation was performed for each
comparison.
282

Supplemental Figure 7.6. Correlation of transcriptional signatures caused by removal of
senescent cells via GCV or 25HC compared to aging signature in glial cells. Scatter plots
283

showing two-sided Spearman’s rank correlation between the differentially expressed genes
caused by aging or a models of senescent cell clearance in different types of hippocampal glial
cells. (A) GCV and aging in microglia. (B) 25HC and aging in microglia. (C) GCV and aging in
oligodendrocytes. (D) 25HC and aging in oligodendrocytes. (E) GCV and aging in astrocytes.
(F) 25HC and aging in astrocytes.

Abstract (if available)

Abstract Aging is a complex, multifaceted biological process that entails a systematic decline in physiological function over time and is intimately linked to chronic diseases. This link is attributable to the protracted accumulation of cellular and molecular damage as well as other physiological changes that can trigger an array of ailments including, but not limited to, cardiovascular disease, diabetes, and cancer. Of particular concern in this gamut of disorders are neurodegenerative diseases, hallmarked by a gradual, unrelenting impairment of neuronal structure and function. Aging stands as a predominant risk factor for such diseases, with Alzheimer's disease, Parkinson's disease, and Huntington's disease as prime exemplars.

In the past few decades, the development of transcriptomic, proteomic and other omics technologies has allowed us to characterize the molecular components of biological samples in unprecedented detail. These technologies allow for the unbiased exploration of the effects of complex biological processes. These data driven approaches allow for the identification of dysregulated genes, pathways and biological processes at an unprecedented pace. In addition, the application of these technologies to individual cells has allowed for the study of biological samples at extraordinary resolution.

Due to their capacity for large-scale, parallel hypothesis testing, omics technologies prove exceptionally beneficial in detecting dysregulation within intricate biological processes, like those observed in aging and neurodegenerative diseases. Here, we utilize different types of omics technologies to elucidate the molecular changes inherent in neurodegenerative diseases and brain aging.

The first section examines the molecular alterations induced by Huntington's disease (HD), a neurodegenerative genetic disorder resulting from the expansion of CAG repeats in the huntingtin gene, ultimately leading to the production of a mutant HTT protein. Although the HTT protein is expressed in multiple tissues, HD mainly affects medium spiny neurons (MSNs) in the striatum, leading in their loss and subsequent motor function impairment. Research indicates that alterations caused by HD during development can trigger pathology later in life and counteracting these events can delay HD pathology. To determine what are the molecular changes driven by HD during MSN development, we used HD72 (72/19 CAG repeats) induced pluripotent stem cells (iPSCs) and isogenic controls (21CAG/19CAG repeats). We differentiated these iPSC lines into neuronal stem cells (NSCs) and a population of developing MSNs containing MSNs at different stages of development. By utilizing both bulk and single-cell RNA sequencing, we reveal transcriptional changes across multiple stages of MSN development, most notably, dysregulation in the DLX family of transcription factors, which are vital for MSN development. Through computational methods utilizing transcriptional data, we identify several potential HD modifiers, including cerulenin. When HD72 MSNs were treated with cerulenin, we observe a partial reversal of some HD-associated transcriptional changes as well as a partial restoration of electric activity and increased levels of DARPP-32. This provides proof of concept for the viability of our approach in identifying potential interventions for HD.

In the second section, we explore the molecular alterations in iPSC derived HD72 MSNs using quantitative proteomics. We note a significant agreement between changes driven by HD at the RNA and protein levels. We also identify dysregulation in lipid metabolism and observe lipid droplet accumulation in HD72 MSNs, hinting at potential lipid turnover deficiencies, possibly via lipophagy.

The third section focuses on the transcription factor Bcl11b, which is required for MSN development. We present evidence that loss of Bcl11b in adult MSNs can induce HD-like molecular and behavioral phenotypes in mice. We also show that BCL11B forms granules in the nuclei of developing HD72 MSNs, a phenotype not observed in controls, suggesting potential alterations in BCL11B function in HD MSNs. This reveals Bcl11b's critical role in mature MSN function and suggests a link between BCL11B and HD.

In the fourth section, we demonstrate that HD induces senescent-like features in human HD72 NSCs and MSNs. We also show HD to reprogram FOXO3 targets in HD72 NSCs.

In the final section, we characterize the transcriptional changes driven by senescent cells in the aged mouse hippocampus. To accomplish this, we performed single-nuclei RNA sequencing on hippocampus collected from aged mice that have undergone clearance of senescent cells. Subsequently, we contrasted the transcriptional signature stemming from the removal of senescent cells in aged mice with the signature resulting from aging. This comparison allowed us to discern the impact of senescent cell clearance on various hippocampal cell types.

Notably, we detected a reversal in the age-related transcriptomic signature of glutamatergic neurons, characterized by a marked enrichment of genes associated with synapse function. This observation aligns with prior research that highlighted an age-related decline in hippocampal synaptic function, which was found to be reversible following the clearance of senescent cells.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A single cell time course of senescence uncovers discrete cell trajectories and transcriptional heterogeneity

PDF

Phenotypic and multi-omic characterization of novel C. elegans models of Alzheimer's disease

PDF

Modeling neurodegenerative diseases using induced pluripotent stem cells and identifying therapeutic targets

PDF

Natural divergence of traits in species of mice reveals novel molecular mechanisms of cellular senescence

PDF

Longitudinal assessment of neural stem-cell aging

PDF

The overlap between mTOR signaling, rapamycin and cellular senescence

PDF

Synaptic transmission, nutrient sensors, and aging in Drosophila melanogaster

PDF

Swimming exercise reduces native ⍺-synuclein protein species in a transgenic C. elegans model of Parkinson’s disease and confers downstream protective effects independent of food restriction…

PDF

Perimenopausal transition increases blood brain permeability: implications for neurodegenerative diseases

PDF

Computational approaches to identify genetic regulators of aging and late-life mortality

PDF

Insulin sensitivity in cognition, Alzheimer's disease and brain aging

PDF

Air pollution neurotoxicity throughout the lifespan: studies on the mechanism of toxicity and interactions with effects of sex and genetic background

PDF

Alzheimer’s disease: dysregulated genes, ethno-racial disparities, and environmental pollution

PDF

Blood-brain barrier breakdown and vessel integrity in Alzheimer’s disease

PDF

Mitochondrial dynamics regulate Leydig cell health and integrity

PDF

The regulation, roles, and mechanism of action of mitochondrial-derived-peptides (MDPs) in aging

PDF

Experience-dependent neuroplasticity of the dorsal striatum and prefrontal cortex in the MPTP-lesioned mouse model of Parkinson’s disease

PDF

From risk mitochondrial and metabolic phenotype towards a precision medicine approach for Alzheimer's disease

PDF

Understanding DNA methylation and nucleosome organization in cancer cells using single molecule sequencing

PDF

Comparative transcriptomics: connecting the genome to evolution

Asset Metadata

Creator Galicia Aguirre, Carlos Alberto (author)

Core Title Investigating brain aging and neurodegenerative diseases through omics data

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Biology of Aging

Degree Conferral Date 2023-08

Publication Date 07/17/2025

Defense Date 06/20/2023

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag aging,brain aging,Huntington's disease,neurodegenerative diseases,OAI-PMH Harvest,omics,senescence,single cell RNAseq

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ellerby, Lisa M. (committee chair), Campisi, Judith (committee member), Curran, Sean (committee member), Melov, Simon (committee member)

Creator Email carlos.galicia1492@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113281394

Unique identifier UC113281394

Identifier etd-GaliciaAgu-12100.pdf (filename)

Legacy Identifier etd-GaliciaAgu-12100

Document Type Dissertation

Format theses (aat)

Rights Galicia Aguirre, Carlos Alberto

Internet Media Type application/pdf

Type texts

Source 20230719-usctheses-batch-1069 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu