Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Metabolic profiling of single hematopoietic stem cells for developing novel ex vivo culture strategies
(USC Thesis Other)
Metabolic profiling of single hematopoietic stem cells for developing novel ex vivo culture strategies
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Metabolic Profiling of Single Hematopoietic Stem Cells for Developing Novel Ex Vivo Culture
Strategies
by
Hao Zhou
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
(BIOMEDICAL ENGINEERING)
December 2022
ii
Acknowledgments
Special thanks go to Dr. Keyue Shen, Irene Li, Dr. Jia Hao, Dr. Yuta Ando, Dr. Hydari Masuma
Begum, Jeong Min Oh, Winfield Zhao, Daniel P. Yen, Xinyuan Zhang, Shin-Jae Lee, Dr. Rong
Lu, Dr. Lisa Nguyen, Charles Bramlett, Bowen Wang, Dr. Ania Nogalska, Mary Vergel, Dr. Scott
Fraser, Dr. Cosimo Arnesano, Dr. Jason Junge, Dr. Peiyu Wang, Dr. David D’Argenio, Dr. Megan
McCain, Dr. Josh Neman-Ebrahim, Dr. Jeffery Boyd, Bernadette Masinsin;
And my family members, Chonglin Zhou, Peiyu Qu, Peiying Qu, Yan Zhou, and Dr. Mengyun
Zhou.
iii
Table of Contents
Acknowledgments........................................................................................................................... ii
List of Tables .................................................................................................................................. v
List of Figures ................................................................................................................................ vi
Abbreviations ................................................................................................................................ vii
Abstract .......................................................................................................................................... xi
Chapter 1: Introduction ................................................................................................................... 1
1.1 Stem cell research for regenerative medicine ....................................................................... 1
1.1.1 Stem cell plays a central role in regenerative medicine ................................................. 1
1.1.2 Application of stem cells in regenerative medicine requires cell fate monitoring ........ 3
1.2 Pluripotent stem cells ............................................................................................................ 3
1.2.1 Establishment of ESC cell lines ..................................................................................... 4
1.2.2 Establishment of iPSC cell lines .................................................................................... 5
1.2.3 Non-genome integrating methods to generate iPSCs .................................................... 7
1.2.4 Optimized culture conditions for PSCs.......................................................................... 8
1.2.5 Directed differentiation of PSCs .................................................................................. 10
1.3 Adult stem cells with a focus on hematopoietic stem cells ................................................ 12
1.3.1 Different adult stem cell types ..................................................................................... 12
1.3.2 From PSCs or mature cells to HSCs ............................................................................ 14
1.3.3 Efforts to expand HSCs in culture ............................................................................... 16
1.3.4 Symmetric and asymmetric division ............................................................................ 18
1.4 Stem cell characterization and quality control .................................................................... 20
1.4.1 Major characterization techniques ............................................................................... 20
1.4.2 Quality control in stem cell products ........................................................................... 27
1.5 Go with the metabolic perspective ...................................................................................... 29
1.5.1 Metabolism regulates cell fate and can be used as a marker ....................................... 29
1.5.2 Current metabolic study methods and their drawbacks ............................................... 30
1.5.3 Fluorescence lifetime imaging ..................................................................................... 33
1.6 Objective and specific aims ................................................................................................ 34
Chapter 2: Non-invasive optical biomarkers distinguish and track the metabolic status of single
hematopoietic stem cells ............................................................................................................... 38
2.1 Introduction ......................................................................................................................... 38
2.2 Materials and methods ........................................................................................................ 40
2.3 Results ................................................................................................................................. 46
2.3.1 HSCs have a distinct profile of metabolic optical biomarkers .................................... 46
2.2.2 Longer NAD(P)H bound correlates with higher intracellular pH and reflects
enhanced lactate dehydrogenase activity in HSCs................................................................ 51
2.2.3 Higher NAD(P)H αbound in HSCs is contributed by enhanced LDH activity............... 54
2.2.4 HSCs have a larger pool of NADH compared to the differentiated cells .................... 55
iv
2.2.5 MOBs distinguish HSCs from multipotent and oligopotent hematopoietic
progenitors ............................................................................................................................ 57
2.2.6 bound tracks changes in glycolysis during in vitro HSC culture .................................. 61
2.4 Discussion ........................................................................................................................... 64
Chapter 3: MOB score: an endogenous optical metric reveals metabolic inheritance and
asymmetry in stem cell division ................................................................................................... 67
3.1 Introduction ......................................................................................................................... 67
3.2 Materials and methods ........................................................................................................ 69
3.3 Results ................................................................................................................................. 76
3.3.1 MOBs track the metabolic dynamics of HSC differentiation ...................................... 76
3.3.2 MOB score derived from 11 representative features recapitulates HSC
differentiation trajectory ....................................................................................................... 78
3.3.3 MOB score identifies the metabolic asymmetry in PDCs ........................................... 81
3.3.4 MOB score reveals distinct patterns of metabolic dynamics in HSC division ............ 82
3.3.5 MOB score predicts culture conditions supporting HSC expansion by division
pattern analysis...................................................................................................................... 86
3.4 Discussion ........................................................................................................................... 88
Chapter 4: Surface oligo tagging for high throughput gene expression profiling of paired
daughter cells ................................................................................................................................ 92
4.1 Introduction ......................................................................................................................... 92
4.2 Materials and Methods ........................................................................................................ 94
4.3 Results ................................................................................................................................. 96
4.3.1 Direct oligo tag conjugation to cell surface proteins ................................................... 96
4.3.2 Liposome conjugation to the cell surface .................................................................... 99
4.3.3 Conjugation of oligo tags to cell-liposome ................................................................ 101
4.4 Discussion ......................................................................................................................... 103
Chapter 5: Summary and future directions ................................................................................. 105
References ................................................................................................................................... 110
v
List of Tables
Table 1-1. Representative protocols for HSC ex vivo culture. ..................................................... 18
Table 1-2. Comparison of technologies applied in HSC metabolism study. ................................ 33
vi
List of Figures
Figure 1-1. Schematic of HSC self-renewal and differentiation................................................... 19
Figure 1-2. Physiological and physical properties of NAD(P)H and FAD. ................................. 34
Figure 2-1. Gating for hematopoietic populations harvested from bone marrow......................... 47
Figure 2-2. HSCs have a distinct profile of metabolic optical biomarkers (MOBs) at the single-
cell and subcellular levels. ............................................................................................................ 50
Figure 2-3. Longer NAD(P)H bound is correlated with higher intracellular pH (pHi) in HSCs
and reflects lactate dehydrogenase (LDH)/glycolytic activity...................................................... 53
Figure 2-4. Higher NAD(P)H αbound is contributed by LDH activity in HSCs. ............................ 55
Figure 2-5. HSCs have a more reduced pool of NADH. .............................................................. 57
Figure 2-6. MOB profiling distinguishes HSCs from hematopoietic progenitor cells
(HPCs)........................................................................................................................................... 60
Figure 2-7. MOB profiling and maintenance of bound in HSCs during in vitro culture. .............. 63
Figure 3-1. Evaluation of MOBs in profiling HSC differentiation. ............................................. 77
Figure 3-2. MOB score derived from representative features and latent variable model. ............ 80
Figure 3-3. MOB score distinguishes different division patterns. ................................................ 82
Figure 3-4. MOB score identifies the metabolic inheritance and asymmetry in paired daughter
cells (PDCs). ................................................................................................................................. 86
Figure 3-5. MOB score identifies conditions that promote HSPC expansion. ............................. 87
Figure 4-1. Schematic of cell surface oligo tagging for tracking paired daughter cells in
droplet-based single cell RNA-seq. .............................................................................................. 94
Figure 4-2. Oligo tagging to cell surface proteins and the influence on cell fate. ........................ 98
Figure 4-3. Liposome conjugation on the cell surface. ............................................................... 100
Figure 4-4. Oligo release from Jurkat cell-liposome under normal cell culture condition. ........ 102
Figure 4-5. Proposed model of oligo-medium components interaction and oligo protection
strategies. .................................................................................................................................... 103
vii
Abbreviations
AD: asymmetric division
ANOVA: analysis of variance
ASC: adult stem cells
CFU: colony forming unit
CLP: common lymphoid progenitor
CMP: common myeloid progenitor
CP: common progenitors
CRISPR: clustered regularly interspaced short palindromic repeat
EB: embryoid body
ECAR: extracellular acidification rate
ESC: embryonic stem cells
FACS: fluorescence-activated cell sorting
FAD: flavin adenine dinucleotide
FAO: fatty acid oxidation
FLIM: fluorescence lifetime microscopy
GlycoPER: glycolytic proton efflux rate
GMP: granulocyte/macrophage progenitor
viii
HPCs: hematopoietic progenitor cells
HSC: hematopoietic stem cell
HSCT: hematopoietic stem cell transplantation
HSPC: hematopoietic stem and progenitor cell
ICM: inner cell mass
KLS: ckit+sca1+lineage-
LDA: linear discriminant analysis
LDH: lactate dehydrogenase
LIF: leukemia inhibitory factor
MOB: metabolic optical biomarker
MPP: multipotent progenitor
MSC: mesenchymal stem cell
NADH: Nicotinamide adenine dinucleotide
NADPH: Nicotinamide adenine dinucleotide phosphate
NGS: next generation sequencing
NSC: neural stem cell
OCR: oxygen consumption rate
OPP: oligopotent progenitor
ix
ORR: optical redox ratio
OXA: oxamate
OXPHOS: oxidative phosphorylation
MEP: megakaryocyte/erythrocyte progenitor
PCA: principal component analysis
PCR: polymerase chain reaction
PDCs: paired daughter cells
PDH: pyruvate dehydrogenase
PDK: pyruvate dehydrogenase kinase
PDMS: polydimethylsiloxane
PSC: pluripotent stem cells
ROS: reactive oxygen species
SSEA: stage-specific embryonic antigen
SC: symmetric commitment
SCF: stem cell factor
SD: symmetric division
TCA: tricarboxylic acid
TPO: thrombopoietin
x
TRA: tumor-related antigen
1-AA: 1-aminoethylphosphinic acid
2-DG: 2-deoxy-D-glucose
6-NBDG: 6-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-6-Deoxyglucose
xi
Abstract
Hematopoietic stem cell (HSC) transplantation is an effective cure for various human diseases.
However, its clinical applications have been hindered by the lack of HSC sources. HSC ex vivo
biomanufacturing has attracted significant research efforts yet has not been successful.
Biomanufacturing HSCs requires comprehensive knowledge of their expansion and rigorous
quality control. While in vivo transplantation assay is the gold standard for evaluating HSC
regenerative potential (stemness), it is time and resource-consuming. A rapid prediction method
will significantly facilitate screening and optimizing novel ex vivo expansion conditions.
Metabolism was recently reported to be a vital cell fate regulator and reflect HSC stemness. In this
dissertation, we proposed to track HSC metabolism as biomarkers to evaluate and optimize the ex
vivo culture conditions. We demonstrated the optical properties of endogenous fluorophores and
metabolic co-enzymes nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin
adenine dinucleotide (FAD), profiled by fluorescence lifetime imaging microscopy (FLIM), can
be used to distinguish HSCs from differentiated hematopoietic cells in a non-invasive and real-
time manner. We elucidated that the unique FLIM profile of HSCs is associated with their
preferred glycolysis and inhibited oxidative phosphorylation (OXPHOS). We further defined a set
of 205 FLIM features that profile single cell metabolism with subcellular spatial resolution. A
bioinformatic workflow was established to select the features that track HSC differentiation, of
which a combinatorial analysis evaluates the stemness of individual cells. With such metric, we
could track the offspring of HSCs under symmetric and asymmetric divisions, and screen for the
cell culture conditions that promote HSC expansion. Moreover, we also explored an oligo-
barcoding and high-throughput RNA sequencing strategy to track paired daughter cells from the
xii
same parent cell, aiming to understand HSC division at the transcriptomic level. We expect this
strategy to provide complementary information to our optical imaging methodology.
1
Chapter 1: Introduction
1.1 Stem cell research for regenerative medicine
The term “regenerative medicine” was first coined by William Haseltine in 1999 to describe the
emerging interdisciplinary research of tissue engineering, stem cell biology, nanoengineering, and
biomedicine.
1
Today, regenerative medicine is defined as a developing branch of medicine that
aims to repair, replace, or regenerate cells, tissues or organs.
2
Such need comes from tissue or
organ function lost due to age, disease, damage, or congenital disabilities, such as heart failure,
bone marrow failure, and neurodegenerative diseases. To pursue the goals of regenerative
medicine, tremendous efforts have been made in cell-based therapy,
3–5
tissue engineering,
6,7
and
the generation of artificial organs.
8,9
Examples include the transplantation of stem cells or mature
functional cells generated by directed differentiation;
10,11
the injection of antibodies or other
biologically active molecules to induce cell or tissue regeneration;
12
transplantation of organs
grown or fabricated in vitro.
13
1.1.1 Stem cell plays a central role in regenerative medicine
Regenerative medicine was inspired by the fact that all organisms have evolved strategies to
continuously replace the short-lived differentiated cells or repair damage throughout their lifetime.
Examples are found in the blood system, gut epithelium, and skin. Such replacement is enabled by
tissue-specific adult stem cells. Indeed, one of the first applications considered as cell-based
2
therapy is bone marrow transplantation in 1968, essentially hematopoietic stem cell transplantation
(HSCT) to repair the blood system.
14
Differentiated cells can also be transplanted to directly
improve tissue functions. For example, CAR T-cell therapy is FDA-approved to treat aggressive
B-cell lymphomas, B-cell leukemia and relapsed or refractory mantle cell lymphoma. However, it
is challenging to get a considerable number of differentiated cells for such therapy, since they have
limited replication potential. To solve this kind of problem, research has been focused on
differentiating mature cells from stem cells, which can proliferate extensively in vitro.
Genetic disorders, caused by abnormalities in the genome, are usually inheritable. Until
2020, more than 7,800 genetic diseases were discovered yet only ~600 are treatable.
15
Recently,
the combination of stem cell engineering and genome editing techniques, such as the clustered
regularly interspaced short palindromic repeat (CRISPR) system
16–23
, has provided a promising
way to correct the underlying genetic problems and induce stem cells from patient cells for future
autologous cell therapy.
24
Organ transplantation from a donor to a recipient has been possible when the body cannot
heal itself.
9
However, there are much more people awaiting life-saving organ transplants than
available donors. Regenerative medicine raises the possibility of growing tissues and organs in the
laboratory to solve the shortage.
25
Moreover, when the cell source of the artificial organ is derived
from the patient’s own cells, the risk of transplant rejection (immune rejection) is eliminated.
Although current technology yet cannot generate functional, transplantable artificial organs,
significant progress has been made in the induction, expansion, and directed differentiation of stem
cells, raising the hope of organizing them into artificial organs.
3
1.1.2 Application of stem cells in regenerative medicine requires cell fate monitoring
To apply stem cell products to regenerative medicine, the pivot is the manipulation of cell fate,
including the induction and expansion of stem cells, the directed differentiation into functional
mature cell types, and more sophisticatedly, orchestrating cell growth with specific temporal and
spatial clues to form tissues or organs. Naturally, most of such processes (except stem cell
induction) are highly regulated in the human body. There are two significant ways to achieve these
aims, mimicking the developmental process in the body,
26–28
or high throughput screening of
factors that can regulate cell fate (e.g., transcription factors, cytokines, small molecules).
29–31
Both
require precise fate monitoring of the de-dedifferentiation or differentiation process. This chapter
will document and discuss the significant progress in stem cell fate regulation and the monitoring
system adopted during the discovery.
1.2 Pluripotent stem cells
Stem cells are classified according to their differentiation potential, including pluripotent stem
cells (PSCs) and adult stem cells (ASCs). PSCs can give rise to all cell types in the body, while
ASCs, also termed somatic stem cells, can generate a limited number of cell types that are tissue
specific. There are two types of PSCs: embryonic stem cells (ESCs) and induced pluripotent stem
cells (iPSCs).
4
1.2.1 Establishment of ESC cell lines
ESCs are derived from the inner cell mass (ICM) during the development of embryos. Given the
ethical concerns about using human embryos for study, most knowledge about ESCs came from
studies in mouse cells in the early years. In 1981, progressively growing cultures of ESCs from
mouse embryos were established by M.J. Evans and M.H. Kaufman, and G. Martin,
32,33
before
when these stem cells could only be obtained from the in vivo formed teratocarcinoma. Evans et
al. successfully isolated stem cells from in vivo maintained embryos and cultured them in a Petri
dish containing mitomycin C-inactivated STO fibroblasts. They found that these cells can actively
proliferate and form colonies that closely resemble embryonal carcinoma cell appearance. After
removing the feeder cell layer, the cultured cells can spread out and differentiate into a complex
of tissues in the Petri dish. The researchers also found that the cultured ESCs carry the cell-surface
antigens Forssman and lacto-N-iso-octaosyl ceramide, similar to that of embryonal carcinoma cells.
To further demonstrate the pluripotency of the cultured cells, they injected the cultured cells into
the flank of syngeneic mice and observed tumor formation, which was revealed to be
teratocarcinoma by histological examination.
In 1998, Thomson et al. derived the first ESC lines from human blastocysts.
34
Following a
similar protocol they previously established for a nonhuman primate (rhesus monkey) ESCs,
35
they
isolated and cultured ICM–derived cells on irradiated mouse embryonic fibroblasts, maintaining
the continuous undifferentiated proliferation for months. The derived cells showed a high nucleus
to cytoplasm ratio and a colony morphology similar to that of rhesus monkey ESCs. They also
5
noted that the level of telomerase activity is high in the derived cell line. The reason for examining
this is that telomerase expression strongly correlates with immortality and replicative lifespan in
human cell lines. Telomerase is highly expressed in germline and embryonic tissues but not in the
somatic cells, resulting in shortened telomeres and limited proliferative lifespan in the latter. Other
markers used to characterize the cell lines include the cell surface markers known to be expressed
on undifferentiated nonhuman primate and human ESCs, which are stage-specific embryonic
antigen (SSEA)-3, SSEA-4, TRA-l-60, TRA-181, and alkaline phosphatase. Eventually, the
authors examined the differentiation potential of the established cell lines by both in vitro feeder
layer-free culture and in vivo models. By injecting them into severe combined immunodeficient-
beige mice, all tested cell lines formed teratomas including derivatives of endoderm, mesoderm,
and ectoderm.
Research utilizing human ESCs for regenerative medicine has met significant technical,
political, and ethical issues.
36–38
Harvesting process may damage the potentially viable embryos.
39
Furthermore, the capacity for unlimited self-renewal of ESCs raises concerns about dysregulated
growth and tumorigenesis. In addition, the in vitro expansion of ESCs can accumulate mutations,
further increasing the risk of tumor formation.
1.2.2 Establishment of iPSC cell lines
Given the issues of applying ESCs to the clinic, in 2006, Yamanaka’s group developed a strategy
for reprogramming mouse embryonic fibroblasts to PSCs, termed induced pluripotent stem cells
(iPSCs).
40
Assuming factors that are critical in maintaining ESCs are also pivotal in the induction
6
of pluripotency from somatic cells, they screened 4 transcription factors out of 24 candidates. This
elegant experiment benefits from a well-designed drug-resistant system for pluripotency screening,
which is cells from Fbx15
bgeo/bgeo
transgenic mouse. Fbx15 is known to be specifically expressed
in mouse ESCs and early embryos, although dispensable for the maintenance of pluripotency. bgeo
cassette is a fusion of the b-galactosidase and neomycin resistance genes. They demonstrated that,
ESCs carrying this knock-in construct were resistant to G418 (analog of neomycin and has a
similar mechanism), while somatic cells with the same construct are sensitive to G418. Using
G418 resistance and colony morphology as criteria, they found a combination of 4 transcription
factors (Oct3/4, Sox2, Klf4, c-Myc, known as Yamanaka factors) can induce PSCs from mouse
fibroblast cells. Through reverse transcription- polymerase chain reaction (RT-PCR), researchers
also examined the expression level of 13 ESC marker genes and found that the derived iPSCs
express most of them. Eventually, in vivo teratoma formation, a standard previously used to test
ESC pluripotency, was used to examine the pluripotency of the induced cells. Like ESCs, iPSCs
can form tumors containing all three germ layers after subcutaneous injection into nude mice.
One year after the discovery of mouse iPSCs, Yamanaka group successfully induced
human PSCs from somatic cells with the same transfection factors.
41
In this study, they used the
morphology of cells and their colony to distinguish the ESC-like cells and those not successfully
induced. These candidate iPSCs also showed human ESC-specific surface antigens,
42
including
SSEA-3, SSEA-4, tumor-related antigen (TRA)-1-60, TRA-1-81 and TRA-2-49/6E (alkaline
phosphatase). The authors also confirmed that iPSCs have similar global gene expression patterns
with ESCs using DNA microarray analysis, and these cells can differentiate into cell types
representative of the three germ layers in vitro and in teratomas.
7
1.2.3 Non-genome integrating methods to generate iPSCs
It is noteworthy that during the discovery of iPSCs, Yamanaka factors were introduced into cells
by retroviral transduction. Given its ability to integrate genes into the host cell genome, this method
sustained the high expression level of Yamanaka factors. While this is critical for the success of
the weeks-long reprogramming process, many of these transcription factors are well-known
oncogenes.
43
Although the transgenes would eventually be silenced by the native genome-defense
pathways, their existence in the iPSC genome is still a risk for potential clinical applications.
44
To
avoid such risk, Yamanaka group used a repeated plasmid (adenoviruses) transfection strategy to
induce mouse embryonic fibroblast converting to iPSCs.
45
To examine the success of transfection,
they used mouse cells that carry a Nanog-GFP reporter, since Nanog is an ESC-specific gene. The
optimized viral construct and protocol produced GFP-positive colonies with indistinguishable
morphology from mouse ESCs. These iPSCs also showed similar expression levels of ESC marker
genes, as indicated by RT-PCR. Pluripotency of these iPSCs was demonstrated by teratoma
formation in nude mice, which contained cell types from all three germ layers. To eliminate the
possibility of genome integration, Zhou et al. developed a recombinant cell-penetrating
reprogramming protein-based method to induce PSCs with proteins encoded by Yamanaka
factors.
46
With OG2/Oct4-GFP reporter system, they screened out the GFP+ colonies. These cells
expressed typical pluripotency markers, including ALP, Oct4, Nanog, Sox2, and SSEA1, as
determined by immunostaining. RT-PCR, bisulfite genomic sequencing and global gene
expression analysis indicated that the induced cells have similar gene expression and
demethylation profiles as in ESCs. Standard in vitro differentiation assay showed the
8
differentiation potential. Notably, the germline competence assay demonstrated that these cells
could efficiently incorporate into the ICM of embryos and generate chimerism with germline
contribution. Warren et al. developed an mRNA transfection method that also avoids genome
integration.
47
They used the colony morphology of induced cells and their prominent nucleoli as
the first sign of successful conversion. Only cells showing morphology similar to ESCs were
mechanically picked and cultured for further analysis. Notably, mRNA gives very high iPSC
generation efficiency and rapid kinetics. Fibroblasts transfected with the RNA cocktail resulted in
an iPSC conversion efficiency of more than 2%, two orders of magnitude higher than the reported
virus-based conversion. Moreover, the ESC-like colonies formed by day 17, in contrast to that
virus-mediated iPSC derivations typically take ~4 weeks. More recently, Hou et. al. induced PSCs
from mouse somatic cells with a combination of 7 small molecule compounds.
48
Small molecules
have various advantages over DNA or mRNA transfection methods. For example, their dosing and
timing can be precisely controlled, and their effects are reversible. They also usually have the
advantages of easy synthesis, low cost, and long shelf-life. Very recently, the same group has
enabled human iPSC conversion with small molecule cocktails.
49
1.2.4 Optimized culture conditions for PSCs
In the early reports establishing ESC cell lines as well as iPSCs, both mouse and human PSCs
were cultured in similar conditions, which is on mouse embryonic fibroblast feeder cells with
serum in the medium. It is believed that the growth factors in this system are critical to maintaining
PSCs in an undifferentiated state. However, such culture conditions are considered xenogeneic
9
and have raised the concern of transmitting infectious diseases from animals.
4
Indeed, Martin et
al. found that human ESCs express an immunogenic nonhuman sialic acid when cultured on the
mouse feeder.
50
Batch-to-batch inconsistencies are another factor to consider. On the other hand,
elucidating factors that maintain ESC culture is also important to understanding the details of
molecular mechanisms regulating ESC fate. Thus researchers have been trying to define PSC
culture with feeder and serum-free conditions. Efforts started with mouse ESCs. In 1988, a few
groups noticed that leukemia inhibitory factor (LIF) can maintain the self-renewal ability of mouse
ESCs, indicated by the morphology and the potential to form chimeric mice.
51,52
After that, it
turned out that LIF can maintain pluripotency by activating the transcription factor STAT3. Mouse
ESCs have then been largely cultured with serum and LIF.
53,54
In 2008, Ying et al. showed that the
role of serum and growth factors in culture can be replaced by two small molecules, PD0325901
and CHIR99021, which inhibit mitogen-activated protein kinase and glycogen synthase kinase-3
respectively.
55
LIF can further improve pluripotency along with the two small molecules, although
not strictly necessary.
However, LIF alone shows little effect on human ESCs. Even with a high concentration of
LIF, human HSCs would quickly differentiate in the absence of mouse feeder cells.
56
Richards et.
al. developed systems culturing human ESCs on human-sourced feeders.
57
Braam et al developed
a protocol based on Matrigel adaptation in mouse embryonic fibroblast conditioned medium
followed by monolayer culture of hESC.
58
Yet the dependence of human ESCs on a feeder system
or conditioned medium creates problems for their large-scale production and potential clinical
applications. Shariki et al. developed a feeder-free human ESC culture protocol with the medium
supplemented with 15% serum replacement, a combination of growth factors including
transforming growth factor β1, LIF, basic fibroblast growth factor, and fibronectin matrix.
59
This
10
protocol fulfilled the requirements for a well-defined culture system, as well as reduced exposure
of human ESCs to non-human pathogens. More recently, Thomson group developed a chemically
defined condition for human PSC derivation and culture.
60
Using cell morphology and NANOG
expression level as the metric, they found that DMEM/F12 medium supplemented with insulin,
selenium, transferrin, Lascorbic acid, FGF2 and TGFβ (or Nodal), and with pH adjusted with
NaHCO3 can support the undifferentiated proliferation of both human ESCs and iPSCs.
1.2.5 Directed differentiation of PSCs
PSCs preserved the potential to differentiate into all cell types and thus favor clinical usage.
However, to fulfill the therapeutic needs, PSCs usually have to be first differentiated into cells that
are functional for specific therapeutic goals. Thus, methods and protocols for directed
differentiation of PSCs are greatly interesting.
The study of directed differentiation started with ESCs early. Thus, immediately after
human iPSC was established in 2007, knowledge from ESC differentiation was applied to iPSC
differentiation and allowed a burst of exciting results. Zhang et al. compared ESC and iPSC
differentiation to cardiomyocytes and found that the time needed for embryoid bodies (EBs)
development was comparable for the iPSC and ESC lines.
61
Moreover, RT-PCR analysis showed
that iPSC- and ESC-derived cardiomyocytes have similar cardiac gene expression patterns.
Electrophysiology studies indicated that iPSCs have a comparable capacity with ESCs for
differentiation into various cardiomyocyte phenotypes based on action potential characteristics.
To explore the potential applications of iPSCs in genetic disorders, Raya et al. genetically
11
corrected the iPSCs generated from Fanconi-anaemia patients.
62
They proved that hematopoietic
progenitors derived from these iPSCs can give rise to myeloid and erythroid lineages that are
disease free, by in vitro colony forming unit (CFU) assay. In 2008, Dimos et al. successfully
induced PSCs from amyotrophic lateral sclerosis patients.
63
They then generated spinal motor
neurons using a directed differentiation protocol developed for mouse and human ES cells using
two small molecules: an agonist of the sonic hedgehog signaling pathway and retinoic acid,
identified by the neuron-like outgrowths and motor neuron-specific marker HB9.
64,65
One concern of iPSC-derived cells is that it is hard to obtain highly homogeneous
populations due to the low conversion efficiency. Many of the differentiation procedures are also
complex and involve multiple steps. The undifferentiated cells will raise the risk of tumor
formation once applied to therapy. Thus, inducing iPSC differentiation with simple protocols has
been of great interest. In 2013, Zhang et al. discovered that forced expression of one transcription
factor, Neurogenin-2, can convert human ESCs and iPSCs into functional neuron cells with nearly
100% yield and purity in less than 2 weeks.
66
This study enables large-scale studies of human
neurons for questions such as analyses of human diseases, examination of human-specific genes,
and drug screening. Recently, Church group comprehensively tested 1,564 transcription factors
and 1,732 transcription factor splice isoforms and found 290 of them can individually induce
human iPSC differentiation. This work explored the programming landscape mediated by
transcription factors and provided a toolbox for the research of directed differentiation of iPSCs
for regenerative medicine.
31
Meanwhile, it is still extremely important to establish fate tracking
and quality control methods and criteria for iPSC products.
12
1.3 Adult stem cells with a focus on hematopoietic stem cells
Another concern of utilizing iPSCs is that the induction and de-differentiation processes are both
time and resource consuming. Besides, different iPSC cell lines have been reported to have distinct
differential potential into different cell types.
67
An alternative source for stem cell therapy is the
in vitro expanded adult stem cells. In this section, we will talk about different adult stem cell types
with a focus on hematopoietic stem cells (HSCs).
1.3.1 Different adult stem cell types
Adult stem cells (ASCs) reside in most human body tissues, and their mission is to replace the
dead cells and repair tissue damage. There are different types of ASCs, including hematopoietic
stem cells (HSCs), mesenchymal stem cells (MSCs),
68
neural stem cells (NSCs),
69
muscle stem
cells, hair follicle stem cells,
70
and intestinal stem cells.
71
Mesenchymal stem cells
Mesenchymal stem cells (MSCs) can self-renew and differentiate into various cell types including
osteoblasts (bone cells), chondrocytes (cartilage cells), myocytes (muscle cells) and adipocytes
(fat cells).
68,72–74
Different from most adult stem cell types that only exist in specific tissues, MSCs
were found in different organs and tissues (brain, spleen, liver, kidney, lung, bone marrow, muscle,
thymus, pancreas).
75,76
Recent studies on various diseases or disease models have shown that adult
MSCs can home to the sites of injury and help restore tissue function. A unique advantage of MSC-
13
based therapy is that these cells do not express major histocompatibility complex class II antigens
and are regarded as nonimmunogenic.
77–79
Therefore, transplantation into an allogeneic host may
not require immunosuppression. There have also been established protocols expanding MSCs in
vitro.
80
Prochymal, an MSC-based treatment for graft-versus-host disease, was the first stem cell
therapy approved by Canada (approved in 2012).
81
Given these advantages, MSCs are receiving
great attention for their potential use in regenerative medicine.
Neural stem cells
Neural stem cells (NSCs) reside in the hippocampus and the lateral ventricles of the adult brain
and can self-renew and generate mature cells including neurons, astrocytes, and oligodendrocytes.
Neurons are functional in neural signal processing and transmission. Astrocytes and
oligodendrocytes are glia cells that support the proper functions of the nervous system. Although
it was once believed that neurons cannot regenerate if damaged (e.g. a spinal cord injury), the
discovery of neural stem cells in the brain raises the hope of therapeutic repair.
82
Possible
therapeutic applications are in vivo drug stimulation to the NSCs to proliferate and differentiate
into functional neural cells and transplant them to repair the damaged nervous system.
83
The
benefits of this therapeutic approach have been examined in Parkinson’s disease, Huntington’s
disease, and multiple sclerosis. NSCs can be cultured and expanded in vitro in neurospheres and
maintain the differentiation potential to neurons, astrocytes and oligodendrocytes.
84
However, this
culture system only results in a very low frequency of NSCs.
85
Thus, it is usually used as a
functional test of NSC stemness or an in vitro disease model, rather than as a protocol to expand
NSCs.
14
Hematopoietic stem cells
HSCs are the blood stem cells resident in the bone marrow and can reconstitute the whole blood
system, including both myeloid and lymphoid lineages (Fig. 1-1A).
86
Hematopoietic stem cells
(HSCs) are the first stem cell type used for transplantation therapy as well as the most well-
characterized adult stem cell. HSC transplantation (HSCT) has been performed since the 1960s for
treating life-threatening diseases, such as bone marrow failure and hematopoietic malignancies, as
well as in genetic editing-based therapy.
87
These applications include autologous and allogeneic
transplantation.
88
In autologous transplantation HSCs are from the patients themselves while
allogeneic transplantation refers to the case that HSCs are collected from matching donors.
According to the center for international blood & marrow transplant research, every year there are
~18, 000 patients in the U.S. diagnosed with diseases that need an allogeneic HSCT. However,
only 8, 326 cases were performed in 2020, although this number has been steadily increasing since
2000. The practice of allogeneic transplantation is limited by the availability of donor HSCs, since
the chance to find a matching donor is extremely low. Thus, there is a desire to develop methods
of HSCs in vitro biomanufacturing. However, HSCs rapidly differentiate to mature cells once
cultured in vitro and there has not been an established method of expanding human HSCs yet.
1.3.2 From PSCs or mature cells to HSCs
Given the significant therapeutic value of HSCs in regenerative medicine, there have been
tremendous efforts to convert HSCs from PSCs.
89–95
Although progress has been made, attempts
to generate engraftable HSC from PSCs have been largely unsuccessful.
96
Early efforts focused on
15
exploring differentiation protocols mimicking the hematopoiesis in embryo development by
culturing PSCs on feeders and regulating their fate by cytokines. These studies have been using
surface markers as readouts and generated phenotypic HSCs that have multipotent differentiation
potential in vitro. However, most such HSCs are not engraftable when tested with in vivo
transplantation assay. In 2001, Thomson group derived hematopoietic precursor cells from human
ESCs by coculturing them with the murine bone marrow cell line or the yolk-sac endothelial cell
line.
97
These hematopoietic precursor cells expressed the cell surface marker CD34 and the
hematopoietic transcription factors TAL-1, LMO-2, and GATA-2. In vitro colony forming unit
(CFU) assay showed that the derived hematopoietic precursor cells can form myeloid, erythroid,
and megakaryocyte colonies. Later, Zambidis et al. differentiated human ESCs to
erythromyelopoiesis by mimicking the embryo development kinetics.
98
Semi-adherent
mesodermal-hematoendothelial colonies were induced from human embryoid bodies (EBs), and
formed endothelium and yolk-sac like structures, which then generated multipotent primitive
hematopoietic stem progenitor cells. Chadwick et al. demonstrated that the treatment of human
ESCs during EB development with a cytokine cocktail and bone morphogenetic protein-4, a
ventral mesoderm inducer, strongly promotes the generation of hematopoietic progenitors of
multiple lineages.
99
More recently, Kennedy et al. used T lymphocyte potential as criteria to track
the hematopoiesis induced by specific morphogens from human ESCs and iPSCs in serum- and
feeder-free cultures.
100
The derived progenitors showed the capacity to generate myeloid, erythroid,
and T cells.
Inspired by the conversion of iPSC from somatic cells, research has also been carried out
on inducing HSCs from mature cells. In 2014, Rossi group showed that transient expression of six
transcription factors (RUNX1T1, HLF, LMO2, PRDM5, PBX1, and ZFP37) could induce HSCs
16
from committed lymphoid and myeloid progenitors, and myeloid effector cells.
101
Single-cell
analysis revealed that the derived HSCs exhibit a gene expression profile similar to endogenous
HSCs. Importantly, the derived HSCs have multi-lineage differentiation potential and are serially
transplantable. However, the efficiency of such conversion is low and needs improving for further
clinical applications.
1.3.3 Efforts to expand HSCs in culture
Although the ex vivo ESC expansion has been accomplished over 40 years ago, most ASCs cannot
be expanded ex vivo until today. Compared to ESCs, adult stem cells usually reside in different
tissues and are surrounded by a more complex physiological microenvironment, referred to as a
“niche”.
86,102–104
The unique microenvironments may suggest that the self-renewal and
differentiation of adult stem cells undergo more complex regulation of signaling than ESCs.
In the past decades, people have tried numerous methods to expand HSCs ex vivo, using
murine HSCs as a model or directly on human HSCs. Currently reported expansion protocols focus
on a few perspectives: stromal cell co-culture or cytokine supplement, overexpression of functional
proteins or transcription factors, high-throughput screening of chemicals, and metabolic
modulation (Table 1-1).
105
Stromal cell co-culture or cytokine supplement is the first generation
of methods explored. The idea is that HSC can be maintained or expanded in the in vivo niche,
where the supporting stromal cells are believed to play the most crucial role by secreting cytokines
to modulate HSC fate. Consequently, various cytokines capable of maintaining or modulating HSC
functions have been discovered. Ema et al. identified stem cell factor (SCF) and thrombopoietin
17
(TPO) as two essential elements for mouse HSC culture.
106
Based on this, Zhang et al. added
angiopoietin and other cytokines into the system and reached 24~30 folds expansion of mouse
HSCs.
107
Most recently, a systematic optimization of SCF/TPO concentration and other medium
and substrate components reached 236~899 folds expansion in murine HSCs.
108
However, the
detailed mechanism of how cytokines regulate stem cell expansion is not clear, which prevents the
rapid translation to human HSC expansion and clinical applications. Indeed, such protocols in
human HSC expansion only get limited success. Bhatia et al. improved the chimera 4-fold after ex
vivo culturing human HSCs for 4 days with SCF, Flt3L, G-CSF, IL-3, and IL-6.
109
Overexpression
of RNA binding protein Dppa5 achieved 6~10-fold expansion of murine HSCs.
110
Similarly,
overexpression of transcription factor human homeobox B4 (HOXB4) produces 41-fold HSCs
over 14 days.
111
High-throughput screening of chemical compounds has recently received
promising progress in expanding HSCs ex vivo. For example, SR-1 and UM171 increased human
HSC chimera in immune-deficient mice 17-fold and 35-fold, respectively.
112,113
However, the
detailed mechanism of how these compounds facilitate HSC self-renewal and expansion remains
largely unknown. On the other hand, metabolic modulation represents another promising HSC ex
vivo expansion method, since metabolism is closely related to stem cell functions.
Category Species Supplement/Method Result Ref
Cytokine
supplement
Mouse SCF, TPO HSCs existed for
3~6 days.
Ema et al.,
2000
Mouse SCF, TPO, IGF-2, FGF-1;
Angptl2
24~30-fold
expansion in 10
days
Zhang et al.,
2006
Mouse SCF, TPO; systemically
optimized other parameters
including cytokine carrier,
substrate coating and medium
change frequency
236~899-fold
expansion
Wilkinson et
al., 2019
18
Human SCF, Flt3L, G-CSF, IL-3, IL-6 15-fold CFU; 4-
fold chimera in 4
days
Bhatia et al.,
1997
Functional protein
overexpression
Mouse SCF, TPO; overexpression of
RNA binding protein Dppa5
6~10-fold
expansion
Miharada et
al., 2014
Human SCF, IL-6, IL-3; overexpression
of transcription factor Hoxb4
41-fold
expansion over
14 days
Antonchuk et
al., 2002
High-throughput
screening of
chemicals
Human SCF, Flt3L, TPO, IL-6; chemical
SR-1
17-fold chimera Wagner et
al., 2016
Human SCF, Flt3L, TPO; chemical
UM171
35-fold chimera Fares et al.,
2014
Metabolic
modulation
Mouse CHIR99021+rapamycin HSCs were
maintained for 7
days.
Huang et al.,
2012
Mouse SCF, TPO; pyruvate
dehydrogenase inhibitor 1-AA
HSCs were
maintained for 4
weeks.
Takubo et al.,
2013
Human SCF, Flt3L, TPO; glycolysis
antagonist GW9662
4-fold expansion Guo et al.,
2018
Table 0-1. Representative protocols for HSC ex vivo culture.
1.3.4 Symmetric and asymmetric division
Hematopoietic stem cells (HSCs) are defined by their ability of self-renewal and multi-lineage
differentiation into mature cells to maintain the hematopoietic system. To balance these goals at
the population level, HSCs rely on the ability of both asymmetric division (AD) and symmetric
division/commitment (SD/SC).
114,115
Asymmetric division generates both a daughter HSC and a
differentiated progeny, while symmetric division or symmetric commitment give rise to two
daughter HSCs or committed cells, respectively. By definition, paired daughter cells generated
from asymmetric division would have different functional identity.
19
Theoretically, to reach the goal of HSC expansion, it is critical to promote the frequency
of self-renewal division, including symmetric cell division that leads to the generation of two
daughter HSCs, and asymmetric cell division that posits a balance between HSC and differentiated
cell, relative to the frequency of symmetric commitment that produces two differentiated cells (Fig.
1-1B). Thus identifying the mechanisms that promote symmetric division and/or asymmetric
division will provide novel strategies for HSC ex vivo expansion and/or maintenance.
Figure 0-1. Schematic of HSC self-renewal and differentiation.
(A) Hierarchy of hematopoiesis. HSCs can give rise to the whole spectrum of mature blood cells. MPP, multipotent
progenitor; CMP, common myeloid progenitor; CLP, common lymphoid progenitor; MEP,
megakaryocyte/erythrocyte progenitor; GMP, granulocyte/macrophage progenitor. (B) HSC division patterns. Self-
renewal division includes asymmetric division and symmetric division, which maintains and increases HSC number
respectively. Symmetric commitment in contrast causes HSC loss. Asymmetric division relies on asymmetric
apportioning of fate determinants while both symmetric division and commitment undergo symmetric apportioning
of those. Diff: differentiated cells.
However, such understanding has been hindered by a lack of ability to dynamically monitor
the identity/status of daughter cells in real-time. Single cell transplantation is the gold standard to
determine the fate and stemness of daughter cells, which is costly, time consuming and technically
20
challenging.
116–118
Conventional live imaging microscopy in bright field has been used to monitor
HSC divisions.
119
Yet, it is difficult to use morphology to inform the fate of daughter cells in real-
time.
120
Post-division cell cycle analysis is often used to compare the proliferative difference
between the paired daughter cells (PDCs), as HSCs are usually more quiescent than the progenitors.
However, such method is post-hoc in nature and the stemness of PDCs remains difficult to
determine in real-time. On the other hand, it has been reported that the uneven apportioning of
cellular components during HSC division, including proteins and cellular organelles, can
differentially regulate the stemness and/or fate of the daughter cells.
118,121
Thus, the inheritance of
these components during mitosis can be used as surrogate markers for identifying HSC division
patterns. However, most of these determinants are located intracellularly, such as signaling
molecules (Cdc42), adapter proteins (Ap2a2), transcription factors (Numb), mitochondria, and
lysosome.
114,120–124
These analyses thus often require cell fixation and immunostaining, and/or
treatment with invasive dyes, which compromise the real-time, uninterrupted tracking of the
division process. Immunostaining of surface markers has also been utilized to analyze HSC
division patterns.
116,121
However, due to the slow turnover of surface proteins, it remains difficult
to monitor the real-time dynamic changes of cell fate with surface markers.
1.4 Stem cell characterization and quality control
1.4.1 Major characterization techniques
The characterization of stem cells serves two primary aims: (1) to examine the identity and potency
of the stem cells and (2) to reveal the molecular mechanisms that regulate stem cell self-renewal
and differentiation.
21
Reporter systems
Reporter genes or reporter systems are the most used strategies in tracking stem cell identity and
differentiation. It indicates whether the gene of interest is expressed in target cells. To establish
such a system, researchers place the reporter gene and the gene of interest into the same construct,
usually a plasmid, and insert it into cells. The commonly used reporter genes express certain
fluorescent or luminescent proteins, generate certain drug resistance, or metabolize certain
substrates. With this strategy, researchers can tell the expression level of the gene of interest, and
use it as a criterion to sort out subpopulations for further study.
Adopting a good reporter system, especially the selection of the gene of interest, is critical
for successful experiments. Examples include the discovery of iPSC by Takahashi and
Yamanaka.
40
They generated an Fbx15
bgeo/bgeo
transgenic mouse model. Fbx15 is known to be
specifically expressed in mouse ESCs and early embryos. bgeo cassette is a fusion of the b-
galactosidase and neomycin resistance genes. Cells from this mouse strain would show resistance
to neomycin, if they express the pluripotency marker Fbx15. With this reporter system, they were
able to perform several rounds of screening in the candidate transcription factor pool, and finalize
the four transcription factors that can convert somatic cells to iPSCs. Notably, although Fbx15 is
specifically expressed in ESCs, it is dispensable for maintaining pluripotency. Thus, this reporter
system does not guarantee reliability in principle. On special occasions, cells that express Fbx15
may have differentiated. Other examples include Lgr5 for intestinal stem cells
71
and Hoxb5 for
HSCs.
125
However, recently, Zhang et al. reported that transducing Hoxb5 into pro-pre-B cells did
not generate HSCs but early T cell lineage progenitors, suggesting that Hoxb5 alone does not
guarantee stemness of the cells. Indeed, researchers develop the reporter systems with the best
22
knowledge they have while carrying out the research. However, these reporters may be shown not
perfect or replaced later.
Morphology
Some stem cells have unique morphology and may be distinguishable from differentiated cells.
For example, ESCs have distinct morphology to differentiated cells; thus, in establishing iPSC
lines, the morphology of cells was used as a standard of successful induction. However,
morphology only contains limited information. A study by Chan et al. showed that morphology
alone could not reflect the pluripotency of iPSCs. They identified distinct colony types
morphologically resembling ESCs yet differ in molecular phenotype and differentiation
potential.
126
With the conventional image analysis method, the morphological difference usually cannot
be identified in the early induction/differentiation stages, even in those where stem cells are quite
different from differentiated cells (e.g., NSCs vs. neurons). With the recent development of deep
learning and artificial intelligence, morphology has become a more informative perspective to
characterize.
127,128
For example, Zhu et al. reported a deep learning model identifying NSC
differentiation as early as 1 day of culture with only bright field images.
127
Buggenthin et al. also
developed a deep neural network that predicts lineage choice of hematopoietic progenitors using
brightfield microscopy and cellular movement. Intriguingly, lineage choice can be predicted up to
three generations before conventional molecular markers are observed.
Surface markers
23
Surface marker immunostaining is one of the most used techniques in identifying cell fate/
different cell types. Unlike the immuno-identification of intracellular antigens, surface marker
immunostaining has advantages from its non-invasive feature. Antibodies targeting these markers
can be engineered to avoid the influences on cell viability or functions, if they ever have any. For
example, the glycolipid antigens SSEA-3 and SSEA-4 and the keratin sulfate antigens TRA-1-60
and TRA-1-81 are the commonly used antigens for identifying ESC populations.
42
However, to
precisely identify/ isolate a specific type of cells, especially ASCs which are usually rare in tissue,
a complex cocktail containing a few or more than 10 markers is necessary. Such a cocktail panel
requires years or decades of work to establish. For this reason, many stem cell types still do not
have a well-established surface marker panel. For example, it took decades to establish the mouse
HSC surface marker cocktail, yet some argued that the current version only yields an HSC purity
of less than 50%.
108,125
Even worse, the current version of the human HSC surface marker cocktail
yields a purity of less than 10%.
129
Notta et al. showed that by combining surface markers with a
functional marker, rhodamine-123, which reflects the efflux ability, the purity of isolated human
HSCs can be improved to 14%~28%.
130
Surface markers can be monitored by either microscopy or flow cytometry. For isolation
purposes, FACS and imaging cytometry may be employed. It is noteworthy that surface markers
may change during in vitro culture.
131
Thus translation of the markers established for in vivo stem
cell biology to ex vivo cultured stem cells needs extra characterization. The problem is that the
markers that exist on the stem cell surface do not necessarily play a functional role in maintaining
stemness.
Gene expression analysis
24
In the early days, gene expression analysis targeting a limited number of signature genes is usually
used to examine cell identity. For example, when Yamanaka group first performed the iPSC assay,
researchers also examined the expression level of a set of 13 ESC marker genes with RT-PCR.
When whole genome level gene expression profiling is desired, the microarray can be used.
132
In
the past decade, with the development of next generation sequencing (NGS), especially the
decrease in cost and the appearance of single cell sequencing techniques,
133–136
whole
transcriptome level gene expression profiling is available to most laboratories. NGS can run in a
high-throughput manner and provide great details.
Conventional NGS usually do not have a spatial resolution of distinct subcellular regions.
As a complementary method, fluorescence in situ hybridization (FISH) can provide the distribution
of gene expression in tissue slices or single cells with spatial resolution. However, regular FISH
can only profile a limited number of genes.
137
A promising way of profiling subcellular gene
expression at the whole transcriptome level is spatially resolved transcriptomics. Originally
proposed for investigating the tissue slices with spatial resolution, this technique is now under
rapid development
138
and was crowned Method of the Year 2020 by Nature Methods.
139
Very
recently, this technique has reached 100 um
2
resolution.
140
Epigenetic profiling
Epigenetic modification refers to DNA methylation and histone-DNA interactions, which can
regulate the accessibility of DNA to the transcriptional machinery and thus regulate gene
expression.
141
DNA methylation is a biological process where methyl groups are modified on
cytosine (more common) or adenine (less common), which subsequentially changes the activity of
25
a gene without changing its sequence. For example, when bases in a gene promoter are methylated,
gene transcription can be repressed. There are many approaches that can detect DNA methylation,
most relying on bisulfite conversion of DNA to detect unmethylated cytosines.
142
After conversion,
genome-level methylation patterns can be analyzed by PCR, microarrays, or DNA
sequencing.
142,143
Histone-DNA interactions can be detected by chromatin immunoprecipitation
sequencing (ChIP-seq). DNA and the interactive histone are cross-linked, segmented, and pulled
down using antibodies against a specific histone protein or modification. The sequence of DNA
can be analyzed by DNA sequencing.
Studies have shown that pluripotency or multipotency of stem cells is regulated by their
unique epigenetic profile.
144
For example, it is generally agreed that the chromatin of ESCs is
highly euchromatic and that the genome is therefore highly accessible for transcription. Such
structure supports the pluripotent nature of ESCs, while the genome structure is more condensed
and heterochromatic in differentiated cells, leading to loss of pluripotency.
145,146
Newly established
stem cell lines are usually examined and characterized for the epigenetic state of the key genes.
Protein profiling and proteomics
Immunocytochemistry is the most widely adopted method to examine specific intracellular protein
levels in stem cells, because of its low technical and platform requirements.
147
Cells (usually fixed)
are incubated with fluorophore-conjugated antibodies targeting specific proteins before analysis.
Protein level can then be analyzed by flow cytometry or fluorescence imaging, similar to surfacer
marker examination. Utilizing imaging platforms, the subcellular localization or co-localization of
26
certain proteins can be examined in detail. Immunohistochemistry can also be used to track stem
cells in tissues, including the engraftment of stem cells after transplantation. However,
immunocytochemistry usually can only target a limited number of proteins. To obtain a more
comprehensive view of stem cells at the protein level, proteomics has been extensively studied in
both PSCs and ASCs.
148–151
Proteomics is usually studied by mass spectrometry to elucidate
protein type and quantity, their subcellular localization, post-translational modifications, and
protein-protein interactions.
152
Proteomics provides detailed information at the translation level,
in complementary to the transcriptome level information in stem cell biology. Recent work in stem
cells revealed that proteomics and transcriptome analyses do not necessarily align
148
, and that the
full differentiation/developmental program can only be discovered systematically with a
comprehensive study.
In vitro differentiation assay
By definition, stem cells can give rise to various types of functional cells. Such potency may be
tested with in vitro models, as an alternative to in vivo assays. For example, the potency of HSCs
can be tested with the colony forming unit (CFU) assay, also referred to as the methylcellulose
assay.
153
The assay proliferates and differentiates hematopoietic stem and progenitor cells into
colonies in a semi-solid media in response to cytokine stimulation. The derived colonies can be
recognized and characterized according to their unique morphology. However, it is noteworthy
that both HSCs and the hematopoietic progenitor cells without long-term engraft ability can form
colonies, complexing the explanation of results.
In vivo transplantation assay
27
In vivo teratoma formation is a standard firstly used to test ESC pluripotency and is also used to
examine the pluripotency of iPSCs. Similar to ESCs, iPSCs can form tumors containing all three
germ layers after subcutaneous injection into nude mice. Teratomas are tumors containing different
cell types from all three germ layers. Generation of teratomas is a method for functional analysis
of pluripotent stem cells in vivo.
Another example is the in vivo functional characterization of HSCs. Competitive
transplantation assay has long been the gold standard. HSCs of interest from one strain (e.g.
C57BL/6-CD45.1 mice) are transplanted alongside whole bone-marrow competitor cells into
C57BL/6-CD45.2 mice, following irradiation at a lethal dose. Donor chimerism is tracked by
collecting peripheral-blood cells, stained with lineage specific markers and analyzed with flow
cytometry. To demonstrate the multipotency of donor HSCs with more confidence, sometimes a
secondary bone-marrow transplantation assay (i.e. serial transplantation assay) can be performed
by transferring bone-marrow cells from the primary recipient mice into lethally irradiated
C57BL/6-CD45.2 mice.
1.4.2 Quality control in stem cell products
For PSC induction and differentiation, there is a need for reliable methods to evaluate the
reprogramming efficiency and to quality control the derived cells.
154,155
To create iPSC products
suitable for regenerative medicine, an examination must be carried out to ensure that all pluripotent
cells have differentiated, and that the cell genome has not mutated or changed during the whole
de-differentiation and differentiation process, to minimize the possibility of tumor formation.
156–
28
158
On the other hand, the “quality” of engrafted stem cells or stem cell-derived mature cells is
important for a better prognosis.
88
Suitable quality control techniques are critical for examining
the induced stem cells, ex vivo expanded stem cells, and cells recovered from cryopreservation.
Karyotype examination
Some types of stem cells, especially ESCs and iPSCs, are more prone to generate and accumulate
genetic mutations and should be examined frequently for chromosomal changes. Karyotyping is
the examination of chromosomal duplications, insertions, deletions, translocations, or centromere
loss in order to detect abnormalities. Common abnormalities include size, the position of
centromeres, and changes in banding patterns.
159,160
It is generally recommended that a stem cell
line be karyotyped every 10-15 passages to ensure that chromosome numbers and structures have
not mutated.
Quality control for cryopreserved cells
Cryopreservation is a process that preserves cells by cooling the samples to very low
temperatures.
161–163
Stem cells may be cryopreserved for future use in research or clinical
application. However, cryopreservation and recovery may be detrimental to cells because of the
ice crystal formation, osmotic shock, toxicity of cryoprotective agent and membrane damage.
161
One example is that HSCs can be cryopreserved. Several studies demonstrated that under
these storage conditions, human CD34+ HSCs remained viable for up to 19 years.
164
Yet the
quality of recovered HSCs needs to be assessed before use, as well as during storage. Enumeration
of nucleated cells and red blood cells and flow cytometry-based CD34+ cell quantification are
29
usually the aspects to be examined. Besides, no microbial growth and minimum cell viability
of >50% after freezing and thawing are mandatory. In vitro differentiation assays can also be
carried out to further evaluate the function of cryopreserved cells. For example, the colony forming
unit (CFU) assay was used to assess the recovery of hematopoietic stem and progenitor cells from
human cord blood cryopreserved for 15 years.
165
1.5 Go with the metabolic perspective
1.5.1 Metabolism regulates cell fate and can be used as a marker
In hematopoietic stem cells (HSCs), mesenchymal stem cells (MSCs), and neural stem cells
(NSCs), glycolysis has been found to support the quiescence and multipotency, while activated
mitochondrial oxidative phosphorylation (OXPHOS) is required for their differentiation. In fact,
the preference for anaerobic glycolysis over OXPHOS has been proposed as the “metabolic
stemness” for HSCs.
166
Lately, cellular metabolism has been increasingly recognized to regulate the unique
functions and the fate decisions of HSCs.
167
Compared to the progeny, HSCs have lower oxygen
consumption and prefer anaerobic glycolysis, which protects them from excessive reactive oxygen
species (ROS)
166,168–170
and aging.
118
In contrast, mitochondrial respiration is required for HSC
differentiation, and their impairment results in anemia and prenatal death.
171
Moreover, increased
fatty acid oxidation promotes HSC self-renewal
116
and symmetric cell division under
hematopoietic stress
117
, while glutaminolysis supports erythroid differentiation for recovery from
anemia.
172
On the other hand, recent studies have shown that individual HSCs have different
30
abilities to self-renew and to form blood,
173
where metabolic heterogeneity is postulated to play a
significant regulatory role through cell cycle status and ROS level.
174
Indeed, HSCs with lower
mitochondrial membrane potential (ΔΨm) were shown to have better long-term reconstitution
capacity than those with higher ΔΨm.
175
Nakamura-Ishizu et al. showed that mitochondria-rich
HSCs exhibit megakaryocyte-lineage differentiation.
176
Therefore, examining HSC metabolism
can provide (1) crucial information about its functional and stemness identity and (2) novel
strategies for its ex vivo expansion.
1.5.2 Current metabolic study methods and their drawbacks
To date, most knowledge about HSC metabolism was obtained through in vitro analysis of freshly
isolated cells.
167
Such analysis usually relies on conventional methods such as Seahorse
assays,
169,177
mass spectrometry
166
and isotopic labeling,
116
and can be confirmed by gene
expression analysis and/or immunostaining of key genes/enzymes
116,166,178
(Table 1-2). Seahorse
assays measure the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR)
of live cells and track their metabolic adaptation to specific pathway inhibitors (i.e. rotenone and
antimycin for mitochondrial respiration and 2-Deoxy-D-glucose for glycolysis), then
mitochondrial respiration and glycolytic capacity can be quantified accordingly.
169,177
Mass
spectrometry can elucidate the identity of molecules by their mass-to-charge ratio and is usually
used to analyze the concentration of metabolites.
166
Mass spectrometry does not characterize
pathway activity since concentrations and fluxes do not necessarily align, and thus must be
supplemented with other methods for reliable conclusions.
179
Isotype labeling (usually used
together with mass spectrometry), in contrast, can be used for quantifying net flux of specific
31
pathways. Cells are first supplied with stable isotope labeled substrates (i.e. “tracer”) and
intermediate or final products of specific pathways are characterized after a period of time. The
contribution of specific substrates to a pathway or pathway activity can then be quantified by
analyzing the abundance of isotopes in these products.
179
Notably, the above techniques usually
require bulk samples as starting materials and is not suitable for rare cell population analysis (such
as HSCs), and they cannot provide spatial information such as which part of the cell that metabolic
process is happening. Gene expression analysis and immunostaining target pivotal genes or
enzymes in metabolic pathways to estimate/predict their activity. These methods can potentially
reach single cell resolution and provide spatial resolution (e.g. fluorescence in situ hybridization,
FISH). Yet they are destructive, thus prohibiting the dynamic tracking of single HSC metabolism
and intact cell retrieval for in vivo functional studies. Most efforts on measuring live single HSC
metabolism have been focused on determining ΔΨm using fluorescent dyes as a surrogate for
mitochondrial respiration
170,175,180,181
or tracking cellular uptake of fluorescent glucose analogs to
estimate glycolysis.
166
However, these dyes provide limited information on specific metabolic
pathways and may interrupt normal glycolysis.
182
In addition, interpretation of the results must be
careful given the fact that HSCs exhibit strong efflux.
183
All these indicators are also not suited for
long-term tracking of metabolic dynamics due to cytotoxicity. There is thus a significant need for
non-invasive, real-time approaches to assess the metabolic status of single HSCs. Addressing this
need will not only enhance our ability to understand HSC heterogeneity and study their response
to extrinsic/intrinsic stimuli,
174
but also to monitor and preserve the quality of HSCs to improve
the success rate of clinical transplantations
88
and to expand HSCs ex vivo to address the clinical
shortages.
184
32
Method Metabolic
processes
Starting
material
s
Invasivene
ss
Spatial
resoluti
on
Rea
l-
tim
e
In
vivo
Ref
Seahorse
assays
Glycolysis;
OXPHOS
>30k
cells
Destructive No No No Yu et al.,
2013;
Qian et
al., 2016
Mass
spectrometry
Glycolysis;
OXPHOS;
potentially
various
pathways
>500k
cells
Destructive Possible No No Takubo et
al., 2013
Isotope
labelling
Various
pathways
>70k
cells
Destructive No No No Ito et al.,
2012
Immunostai
ning
Various
pathways
Single
cells
Destructive Yes No No Takubo et
al., 2010;
Ito et al.,
2016
Gene
expression
analysis
Various
pathways
Single
cells;
bulk
Destructive Possible No No Takubo et
al., 2013;
Wang et
al., 2014
Dye staining Glycolysis;
OXPHOS
Single
cells
Invasive Yes Yes No Simsek et
al., 2010;
Vannini
et al.,
2016
FLIM Glycolysis;
OXPHOS;
potentially
various
pathways
Single
cells
Non-
invasive
Yes Yes Yes -
33
Table 0-2. Comparison of technologies applied in HSC metabolism study.
OXPHOS: oxidative phosphorylation; FAO: fatty acid oxidation. FLIM: fluorescence lifetime imaging microscopy.
1.5.3 Fluorescence lifetime imaging
Fluorescence lifetime imaging microscopy (FLIM) has been used for label-free, non-invasive
observation of cellular metabolism by monitoring nicotinamide adenine dinucleotide (NADH) and
its phosphate ester (NADPH), and flavin adenine dinucleotide (FAD). NAD(P)H and FAD are
naturally occurring autofluorescent metabolic coenzymes, and involved in almost all metabolic
pathways (Fig. 1-2).
185
Importantly, FLIM can capture the fluorescence lifetime (i.e., the rate of
fluorescence decay) of NAD(P)H and FAD, which changes drastically depending on their binding
status with enzymes. Enzyme-bound NAD(P)H shows a longer lifetime than its enzyme-free
counterpart, and the balance between the two states reflects the dominant metabolic process.
186
Besides, the fluorescence lifetime of enzyme-bound FAD depends on the intracellular level of
NAD+.
187
FLIM also allows the recording of fluorescence intensities, which reflect the quantity
and distribution of the coenzymes and the redox state of cells. The intensity ratio of
FAD/(FAD+NAD(P)H), known as the optical redox ratio (ORR), has been associated with
mitochondrial oxidative phosphorylation (OXPHOS)
188
and coenzyme redox states
189
in cells.
Previously, FLIM has been applied to monitor the metabolic changes in live tissues and some
cancer and stem cell types.
190
Notably, FLIM-based parameters have to be interpreted under a
specific context since NAD(P)H participates in various metabolic pathways.
191
Different
intracellular cues, such as the types of enzyme bound to NAD(P)H, intracellular pH and
viscosity
192
in different cellular systems can also influence FLIM readouts. Thus, applying FLIM
34
to a specific cellular system (i.e. hematopoietic cells here) requires specific experimental
validations for the interpretation of the readouts.
Figure 0-2. Physiological and physical properties of NAD(P)H and FAD.
(A) NAD(P)H and FAD participate in various metabolic pathways. Red arrows: NAD(P)H/FADH2 binds to enzymes
and is oxidized to NAD(P)+/FAD; Blue arrows: NAD(P)+/FAD is reduced to NAD(P)H/FADH2. (B) Fluorescent
properties of NAD(P)H and FAD are influenced by enzyme-bound status and environmental cues.
1.6 Objective and specific aims
We developed imaging-based real-time, non-invasive methods for accessing single HSC
metabolism and identity (aim1) as well as division pattern (aim2), and we are developing high
throughput paired-cell sequencing methods for understanding the molecular details of different
division patterns (aim3).
Central hypothesis: Metabolism regulates HSC functions and reflects its identity at the
single cell level.
Aim 1: Establish a set of FLIM-based, non-invasive metabolic optical biomarkers of HSCs at the
single cell level.
35
While metabolism is increasingly known as an important regulator of HSC functions,
progress in such metabolism-function relationships is obstructed by the lack of tools profiling
single HSC metabolism in real-time. NAD(P)H is a key co-enzyme participating in various
metabolic processes, primarily including glycolysis and mitochondrial respiration. Thus we
hypothesized that the metabolic pattern of single HSCs can be profiled by the physical properties,
including intensity, lifetime, and spatial distribution, of NAD(P)H in real-time. To validate our
hypothesis, we FLIM-imaged freshly isolated primary HSCs and their progeny from the mouse
bone marrow, to establish the NAD(P)H profile for HSCs. We further explored how this profile
can be regulated by major metabolic pathways, mainly by perturbing these pathways with specific
inhibitors. Eventually, we explored the applications of such NAD(P)H profile in predicting
stemness and culturing HSCs. We expected that HSCs have distinct NAD(P)H profile compared
to differentiated hematopoietic cells, which can reflect their unique metabolic pattern, especially
in glycolysis and OXPHOS. We also expected this distinct NAD(P)H profile can be used to predict
HSCs from differentiated cells, as well as to track the metabolic adaptation upon stimuli in vitro.
Aim 2: Establish a FLIM-based, optical metric for tracking and quantifying the metabolic
difference in paired daughter cells.
For ex vivo HSC culture with transplantation purposes, the goal is usually to increase HSC
number by promoting symmetric division (SD, expands the HSC pool) or asymmetric division
(AD, generates more progenitor cells while maintaining the HSC pool), rather than symmetric
commitment (SC) that produces only differentiated cells. Thus examining the HSC division pattern
is important and necessary for deciding if a condition is optimized for HSC maintenance or
expansion. Single cell transplantation of daughter cells, as the gold standard of division pattern, is
costly, time consuming and technically challenging. Metabolic difference (i.e. glycolysis vs.
36
OXPHOS) between paired daughter cells has been reported to be an efficient metric for AD.
121,181
However, previous methods for deciding metabolic asymmetry rely on invasive dyes, which
restricts their applications. We hypothesized that the metabolic difference of paired daughter cells
can be resolved by the difference in their NAD(P)H profile. In this study, we first established a set
of FLIM-based features to quantitatively describe the metabolism at the single cell and subcellular
level. We used statistical and machine learning methods to evaluate the importance of these
features and build a scoring metric profiling the metabolic phenotype using primary HSC and
committed cells as a model. We then scored between paired daughter cells and quantify the
asymmetry accordingly. We compared our results with the well-established dye-dependent
methods for metabolic phenotype and stemness identity. We also evaluated our system in real-
time tracking of metabolic differences in paired daughter cells during HSC division. We expected
that the metabolic difference between paired daughter cells can be reliably quantified and is
associated with the stemness identity of each cell. We expected that this system can track the
adaptation of individual daughter cells to extrinsic stimuli in real-time and provide insights into
novel HSC culture strategies.
Aim 3: Develop an oligo barcoding strategy to enable high throughput analysis of the gene
expression difference between paired daughter cells and to understand the molecular mechanisms
of different division patterns.
Asymmetric and symmetric divisions are unique and critical metabolisms that regulate
HSC progeny output in vivo and in vitro. While we expect the imaging-based methods in the first
two aims to give us non-invasive, real-time, and high-throughput readouts, next generation
sequencing (NGS) and the corresponded multi-omics methods can provide detailed information
from another perspective, regularly in the whole transcriptome level, as well as provide
37
corresponding results in between gene expression level and protein translation level (i.e. multi-
omics). However, the study of division patterns has not been able to fully utilize these technical
advantages. To analyze the paired daughter cells (PDCs) from the same parent, researchers must
manually prepare the sample of each single daughter cell and note the sibling information. This
process is labor-intensive and limits the throughput of PDCs studied (usually a few dozen pairs).
118
To address these challenges, we propose a cell surface-liposome-oligo tagging strategy, which
aims to use liposome as the oligo carrier to prolong the existence of oligos on the cell surface.
With this strategy, PDCs from the same parent will carry the same oligo tags, enabling gene
expression profiling and analysis in a high throughput manner.
38
Chapter 2: Non-invasive optical biomarkers distinguish and track the
metabolic status of single hematopoietic stem cells
Aim 1: Establish a set of FLIM-based, non-invasive metabolic optical biomarkers of HSCs at the
single cell level.
2.1 Introduction
Hematopoietic stem cells (HSCs) can reconstitute the entire blood system, and are widely
used in bone marrow transplantation to treat a variety of life-threatening diseases.
87
Lately, cellular
metabolism has been increasingly recognized to regulate the unique functions and the fate
decisions of HSCs.
167
Known knowledge about HSC metabolism can be centered around its
preferred glycolysis and inhibited oxidative phosphorylation (OXPHOS). Low mitochondrial
activity was first identified as a key metabolic feature of HSCs compared to differentiated
hematopoietic cells, and is believed to be an adaptation to their hypoxic niche in the bone
marrow.
170
Lately, Takubo et al. identified that elevated pyruvate dehydrogenase kinase (Pdk)
expression actively suppresses the influx of pyruvate into mitochondria and promotes glycolysis
by thus forced pyruvate-to-lactate conversion.
166
On the other hand, fatty acid oxidation (FAO)
induced mitophagy promotes clearance of mitochondria, which further limits OXPHOS by
removing mitochondrial content and switches the energy metabolism to glycolysis.
116,117
Low ROS
is a common metabolic feature of stem cells.
102,193
In HSCs, it prevents ROS-induced damage,
differentiation and aging.
118,194
Yet it interacts with HIF1- , which regulates mitophagy through
BCL2/adenovirus E1B 19kDa interacting protein 3 (BNIP3)/PINK1 axis.
195
Conclusively,
glycolysis and OXPHOS are downstream of various HSC stemness-promoting mechanisms. Thus
39
examining glycolysis/OXPHOS should provide critical information on both HSC identity and
functions.
Recent studies have shown tremendous functional heterogeneity within the pure HSC
population in self-renewal and linage bias. Metabolism has been reported to play a role in such
heterogeneity. For example, Nakamura-Ishizu et al. showed that mitochondria-rich HSCs exhibit
megakaryocyte-lineage biased differentiation.
176
However, a more comprehensive understanding
of the metabolism-function relationship has been hindered by the technical challenges of observing
metabolism in HSCs at the single-cell level. Conventional bulk and destructive methods such as
Seahorse assays
116
or mass spectrometry
166
prohibit the dynamic tracking of single HSC
metabolism and intact cell retrieval for in vivo functional studies. Most efforts in measuring single
HSC metabolism have been focused on determining ΔΨm using fluorescent dyes as a surrogate for
mitochondrial respiration.
170,175,181
However, ΔΨm provides limited information on cell
metabolism, and it cannot distinguish HSCs from intermediate progenitors which share similar
ΔΨm with HSCs.
170
Options are even more limited for glycolysis, a core metabolic feature and
gatekeeper of HSC functions,
166
which is often measured by the uptake of fluorescent glucose
analogs.
166
These chemicals do not differentiate glucose demands from different downstream
metabolic pathways, compete against glucose, and may interrupt normal glycolysis.
182
All these
indicators are also not suited for long-term tracking of metabolic dynamics due to cytotoxicity.
There is thus a significant need for non-invasive, real-time approaches to assess the metabolic
status of single HSCs. Addressing this need will not only enhance our ability to understand HSC
heterogeneity and study their response to extrinsic/intrinsic stimuli,
174
but also to monitor and
preserve the quality of HSCs to improve the success rate of clinical transplantations,
88
and to
expand HSCs ex vivo to address the clinical shortages.
184
40
NAD(P)H is a key co-enzyme participating in various metabolic processes, primarily
including glycolysis and mitochondrial respiration. Previous studies showed that metabolic
switches can cause changes in the physical properties of NAD(P)H, including intensity, lifetime,
and spatial distribution, which can be characterized by a label-free imaging technique -
fluorescence lifetime imaging microscopy (FLIM).
196
However, to date, there has not been a
thorough study of the NAD(P)H profile in HSCs with FLIM. In this study, we aimed to establish
a set of FLIM-based, non-invasive metabolic optical biomarkers (MOBs) of HSCs at the single
cell level, using primary HSCs and their progeny isolated from the mouse bone marrow as a model.
We achieved this by comparing HSCs against their progeny at various differentiation stages, and
determining the metabolic features underlying the MOBs that are unique to HSCs. We further
explored the utility of these MOBs in identifying primary HSCs from the differentiated cells
(including the closely-related multipotent progenitors, MPPs), tracking their metabolic response
to metabolic substrates and chemical drugs, as well as monitoring the real-time metabolic changes
of HSCs during conditioned maintenance and expansion in vitro. Our study sets a foundation for
identifying the biological/metabolic status of single HSCs and tracking their functions non-
invasively and in real-time.
2.2 Materials and methods
Fluorescence activated cell sorting
Bone marrow cells were extracted from the crushed bones of 4–6 month-old C57BL/6 or Hoxb5–
tri-mCherry mice and then immunostained for CD45+ and Lin-CD45+ cells, or enriched by
41
cKit/IL7R and immunostained for HSCs and HPCs. FACS sorting was carried out on the BD
SORP FACSAria cell sorter at 4°C, as in our previous publication
103
.
Fluorescent lifetime imaging
FACS-sorted cells were washed and resuspended in StemSpan™ SFEM II medium (STEMCELL
Technologies) supplemented with 50 ng/mL SCF and 50 ng/mL TPO (i.e., standard medium) at
~10
6
/mL and seeded in a 1536-well plate (Corning). For freshly isolated cells, the plate was then
incubated at 37
o
C, 5% CO2 for 1 hour to allow the cellular metabolism to reach a steady state
before imaging. Fluorescence lifetime images were acquired with a Zeiss LSM-780 inverted
microscope with a live cell workstation (37
o
C, 5% CO2). Samples were excited at 740 nm in two-
photon mode; the emission wavelength was 460/80 nm for NAD(P)H and 540/50 nm for FAD.
For image acquisition, the following settings were used: image size of 256 × 256 pixels (pixel size:
0.42 µm) and pixel dwelling of 12.41 μsec. For each picture, 20 frames were acquired and averaged.
Each experiment was repeated at least 3 times.
Calculation of αbound, bound, and free
To calculate the free and bound NAD(P)H lifetime in different cell types, we adopted a previously
established 95% confidence ellipse method to determine their metabolic trajectories
63
. Since every
single cell contains only 200–600 pixels and does not allow accurate fitting, pixels were extracted
42
and plotted on the phasor coordinate at the image level. A 95% confidence ellipse was then
generated and its long axis was utilized as the metabolic trajectory. bound and free were calculated
as the intersections of the metabolic trajectory to the universal semicircle that represents the
fluorescence lifetime values of the enzyme-bound and free co-enzymes with single-exponential
decay
44
(Fig. 2-2B). As we have shown that tfree is not significantly different between different cell
types. we fixed the phasor position of free (0.45 ns) for all the samples, and determined the bound
of the individual cells by extending the line from the coordinate of free through the average (g, s)
coordinate of the cell to the universal semicircle
104
. Single-cell αbound was calculated as the ratio of
the distance of the cellular (g, s) coordinate to free over the total length of the metabolic trajectory
between tbound and tfree on the phasor plot (Fig. 2-2B).
Chemiluminescent NAD(P)H assay
NADH/NADPH levels were measured using NAD/NADH-Glo™ Assay (Promega, G9071) and
NADP/NADPH-Glo™ Assay (Promega, G9081) kits, following the manufacturer’s protocols.
Briefly, 10,000–30,000 cells were washed and resuspended in PBS and lysed by adding an
equivalent volume of 0.2 N NaOH with 1% dodecyltrimethylammonium bromide (Sigma). The
lysates were incubated at 60°C for 15 minutes, equilibrated to room temperature, and neutralized
by 0.2 N HCl and 0.25 M Tris base. The samples were then added to a 384-well plate (Corning),
mixed with the detection reagents and incubated at room temperature for 1 hour. The luminescence
was detected by a Varioskan™ LUX multimode microplate reader (ThermoFisher Scientific).
Measured signal was normalized to the cell number in each sample, which was recorded by FACS
43
and verified by hemacytometer. Standard curves were generated using NADH (Sigma, N7410)
and NADPH standards (Sigma, N8035), respectively.
Intracellular pH measurement and calibration
To load the intracellular pH indicator, cells were incubated in standard medium with 10 µM
SNARF-5F-AM (ThermoFisher Scientific) at 37
o
C for 30 minutes. Cells were then washed by
PBS to remove the excessive SNARF-5F-AM and incubated at 37
o
C for an additional 1 hour to
ensure de-esterification. The fluorescence was excited at 488 nm and the dual-peak emission was
detected at 550/80 nm and 640/40 nm respectively
105
. The calibration of intracellular pH was
carried out with the Intracellular pH Calibration Buffer Kit (ThermoFisher Scientific, P35379). In
vitro cultured HSCs were first loaded with SNARF-5F-AM, washed by PBS, and then resuspended
in the pH 6.5 and 7.5 calibration buffers, which were supplemented with valinomycin and nigericin.
Cells were imaged under the same imaging settings. A two-point pH calibration was performed to
generate the standard curve. To study the influence of pH on the fluorescence lifetime of the
enzyme-bound NAD(P)H, cultured HSCs were washed by PBS, resuspended in the calibration
buffers (pH 6.5 and 7.5) and imaged with FLIM.
Inhibition of efflux and mitochondrial complex I
44
Veramapil (Sigma, V4629) was dissolved in MilliQ water to prepare a 5 mM stock. ~10,000
FACS-sorted HSCs from the same mouse were equivalently split into 2 groups and resuspended
in standard medium. The treatment group was supplemented with 50 μM veramapil, and both
groups were then incubated at 37
o
C for 1 hour. FLIM and pH measurements were then carried out
sequentially. For the study of subcellular NAD(P)H distribution, the treatment group was
supplemented with 200 nM rotenone (Sigma, R8875) to inhibit mitochondrial complex I and
imaged with FLIM after 1 hour of incubation.
Live-cell mitochondria staining and imaging
Freshly isolated HSCs (Hoxb5+ KLS) from Hoxb5–tri-mCherry mice were incubated with 50 μM
Veramapil (Sigma, V4629) and 100nM Mitotracker Green FM (ThermoFisher Scientific, M7514)
for 30 min. Cells were then washed with PBS and immobilized with CyGEL Sustain™ (Abcam,
ab109205). Mitochondria and NAD(P)H were subsequently imaged under normal confocal mode
(Ex: 488 nm, Em: 530/40 nm) and FLIM mode respectively.
Timelapse study on in vitro HSC culture and drug treatment
Sorted HSCs were suspended in standard medium supplemented with 1% penicillin and
streptomycin (ThermoFisher Scientific) and seeded in 1536-well plate at ~1000 cells per 10 μL
per well. For the drug-treatment groups, 3 μM CHIR99021 (Stemgent) and 5 nM rapamycin
45
(Calbiochem) were added, as indicated
76
. Cells were transferred to a 96-well plate (Corning) at 36
hours and cultured until the end of the experiment. Half of the medium in each well was changed
at days 3 and 6.
Image analysis
All images were analyzed with a customized Python code. The background was subtracted based
on intensity. For each pixel, phasor coordinates g and s were calculated based on the phase and
modulation recorded
44
. For single-cell analysis, αbound and bound were calculated with the averaged
g and s values. For subcellular analysis, the central region of individual cells was first isolated by
binary erosion (iteration = 4 or 5, decided by cell size). Taking the central region as a continuum
in terms of NAD(P)H distribution, its surrounding area with similar NAD(P)H intensity was then
compensated into the “center”. The rest of the cellular area was considered as the peripheral region
(Fig. 2-2H). Autofluorescent intensity, αbound and bound were calculated, respectively. In Fig. 2-
2K,L and 2-7H, PCA was performed to reduce the dimension of datasets. LDA was then utilized
to determine the gates that separate HSCs from other populations, or drug-treated cells from the
control group in the PCA space
106
.
Data plot and statistical analysis
46
All plots were made in Prism 7 (GraphPad), Python 2.7 (Python Software Foundation), and
SimFCS 2.0 (Laboratory for Fluorescence Dynamics, University of California, Irvine). All data
presented in the format of mean ± error had the error defined as 95% confidence interval (CI).
All statistical analysis for single-cell scatter or box plots were with Mann-Whitney test (2
conditions) or Kruskal-Wallis test (> 2 conditions) due to non-normal data distribution. All bars
in the scatter plot are median and box plots representing the 10
th
–90
th
percentile. All error bars in
the bar graphs were plotted in standard deviation (SD) and compared with Welch’s t-test (2
conditions) or ordinary one-way analysis of variation (ANOVA) (> 2 conditions). Error bars in
the population-level correlation x-y plots were presented in the standard error of the mean
(SEM). The linear regressions with p-values were tested with the zero-slope hypothesis. P-values
are directly labeled in numbers, or as n.s. (not significant, p > 0.05), * (p < 0.05), ** (p < 0.01),
*** (p < 0.001), and **** (p < 0.0001) on the plots.
2.3 Results
2.3.1 HSCs have a distinct profile of metabolic optical biomarkers
Fluorescence-activated cell sorting (FACS) was used to sort HSCs (Lin-cKit+Sca1+Flk2-
CD34-Slamf1+)
86
, lineage-negative CD45-positive cells (Lin-CD45+), and CD45+ leukocytes
from the bone marrow of adult mice (4~6 months old) based on their surface markers (Fig. 2-
1A,B).
47
Figure 0-1. Gating for hematopoietic
populations harvested from bone
marrow.
(A) The relationship of three sorted
populations (HSC, Lin-CD45+, and
CD45+) from the bone marrow; (B)
Gating scheme for CD45+ and Lin-
CD45+ populations; (C) Gating scheme
for HSCs, multipotent progenitors
(MPPs), and oligopotent progenitors
(OPPs). KLS: cKit+Lin-Sca1+; CLP:
common lymphoid progenitor; CMP:
common myeloid progenitor; GMP:
granulocyte/macrophage progenitor;
MEP: megakaryocyte/erythrocyte
progenitor. Sorting starts from DAPI
singlets. Numbers indicate the
percentage of the gated populations to
the parent population.
We used a two-photon FLIM
197
to evaluate the ability of this technique to distinguish
differences in metabolic status between HSCs and the more differentiated hematopoietic cells from
the bone marrow. After 1h incubation in StemSpan 2 medium supplemented with 50 ng/mL stem
cell factor (SCF) and 50 ng/mL thrombopoietin (TPO) at 37
o
C, 5%CO2 to recover the metabolism,
FACS-sorted cells were FLIM-imaged at 740 nm excitation in two-photon mode; the emission
wavelength/optical filter was 460/80 nm for NAD(P)H and 540/50 nm for FAD. We acquired
fluorescence intensity and/or fluorescence lifetime images of FAD and NAD(P)H in the three
populations (Fig. 2-2A). ORR was calculated as an indicator of mitochondrial OXPHOS.
188
A
phasor approach was used to transform the complex multi-exponential lifetime data into 2-
dimensional plots to represent fluorescence decay at each pixel of the FLIM image.
190
By
averaging the clusters of pixels and determining the trajectory of pixel distribution in the phasor
plot, we computed the ratio of enzyme-bound NAD(P)H vs. total NAD(P)H (αbound), and the
48
fluorescence lifetime values (in nanoseconds, ns) of the bound and free NAD(P)H ( bound and free)
at the single-cell or image levels (Fig. 2-2A, right). As initial experiments showed that free is not
significantly different between different cell types (Fig. 2-2B, left), we fixed the phasor position
of free (0.45 ns) for all the samples, and determined the bound of the individual cells by extending
the line from the coordinate of free through the average (g, s) coordinate of the cell to the universal
semicircle (Fig. 2-2B, right).
HSCs had a uniformly low level of ORR compared to the Lin-CD45+ and CD45+
populations (p < 0.0001), whereas Lin-CD45+ and CD45+ cells had similar (p > 0.9999), but
heterogeneous ORR levels (Fig. 2-2C,D). In contrast, HSCs showed significantly higher αbound of
NAD(P)H, while the Lin-CD45+ cells were statistically indistinguishable from the CD45+ cells
(Fig. 2-2C,E). Moreover, HSCs had the highest bound, while Lin-CD45+ cells had the lowest (Fig.
2-2C,F). We also identified at the subcellular level a distinct polar distribution of NAD(P)H at the
edge of HSCs (Fig. 2-2G), which is co-localized with mitochondria (Fig. 2-2H). When segmenting
individual cells into “edge” and “center” areas (Fig. 2-2I), more than 96.0% of the HSCs exhibited
an accumulation of NAD(P)H at the periphery (i.e., above the edge/center intensity ratio = 1;
dotted line, Fig. 2-2I), whereas the differentiated cells had a more even distribution (54.33% of
Lin-CD45+ cells and 49.61% CD45+ cells). Another spatial feature of NAD(P)H in HSCs was the
asymmetric/polar distribution of NAD(P)H. To quantify this, we developed a polarity indicator,
defined as the distance between the center of “mass” (NAD(P)H autofluorescence intensity) and
the geometrical center of a given cell, normalized to its size (Fig. 2-2J). The NAD(P)H
autofluorescence were significantly more polarized in HSCs than in Lin-CD45+ and CD45+
populations (p < 0.0001).
49
Next, we investigated whether these FLIM-based parameters can distinguish the
metabolic/biological status of HSCs from those of the Lin-CD45+ and CD45+ populations. We
combined the five above mentioned parameters (ORR, abound, bound, edge/center ratio, and polarity)
using principal component analysis (PCA). This was followed by linear discriminant analysis
(LDA) to determine the segregation of the three populations. By combining the two subcellular
parameters (Fig. 2-2I,J) with the three single-cell parameters (Fig. 2-2D~F), collectively termed
the metabolic optical biomarkers (MOBs), a planar gate in the 3-D PCA plot can be determined
by LDA to separated ~94% of HSCs while included only 2.4% of Lin-CD45+ and 1.6% of CD45+
cells (Fig. 2-2K). Interestingly, the center of the three hematopoietic populations consistently
appeared in the same regions of the PCA plot (Fig. 2-2L, from five independent experiments),
where HSCs were either sorted by surface markers from the wild-type mice (Fig. 2-2L, empty
symbols), or by a recently reported genotypic marker, Hoxb5, from a mouse model with a Hoxb5-
tri-mCherry reporter
125
(Fig. 2-2L, filled symbols), validating the reproducibility of the results.
50
Figure 0-2. HSCs have a distinct profile of metabolic optical biomarkers (MOBs) at the single-cell and
subcellular levels.
(A) Calculation of ORR (optical redox ratio), α
bound
(ratio of enzyme-bound NAD(P)H vs. total NAD(P)H) and
bound
(fluorescence lifetime of enzyme-bound NAD(P)H) from single cells; (B) Calculation of bound in individual cells. Left:
free in different populations, calculated by 95% confidence ellipse fitting all the pixels from the cells in each image;
n = 4 pictures for each population. Error bars: standard deviation. Error bars: SD. P values: ordinary one-way ANOVA.
51
Right: Schematics of bound calculation in individual cells using the phasor plot. (C) Representative pseudo-color
images of HSCs (Lin-cKit+Sca1+Flk2-CD34-Slamf1+), Lin-CD45+ and CD45+ populations for ORR, α bound and
bound. Scale bar: 100 µm. Single-cell quantification of (D) ORR, (E) α bound and (F) bound in the three populations. Each
dot represents the average ORR, α bound or bound value of an individual cell; (G) Representative images of subcellular
NAD(P)H distribution. Scale bar: 10 µm; (H) Pseudo-color images of NAD(P)H and mitochondria staining. Top:
NAD(P)H autofluorescence signal imaged with FLIM; middle: mitochondrial staining imaged with standard confocal
microscopy; bottom: color merge. Scale bar: 10 µm; (I) Ratio of NAD(P)H fluorescence intensity at the cellular edge
vs. center; (J) Polarity of NAD(P)H fluorescence intensity (M.C.: mass center; G.C.: geometric center); (K)
Segregation of HSCs from the differentiated populations in a 3-D PCA plot utilizing both single-cell (ORR, α bound and
bound) and subcellular MOB parameters (edge/center ratio and polarity of NAD(P)H intensity). n = 127 single cells in
each population; (L) Population level MOB profiles of HSCs sorted from different mice and by different markers vs.
Lin-CD45+ cells and CD45+ cells. Each point represents the average value of the population isolated from an
individual mouse. PC: principal component. P-values: Kruskal-Wallis test.
2.2.2 Longer NAD(P)H bound correlates with higher intracellular pH and reflects enhanced
lactate dehydrogenase activity in HSCs
We next examined the metabolic or cellular functions associated with the unique MOB
profile in HSCs. A distinctive feature of HSCs is their longer NAD(P)H bound. To determine what
influences the bound, we examined the intracellular pH (pHi), a previously reported regulator of
bound
192
, in the three populations using a ratiometric pH indicator, SNARF-5F-AM
198
. We noticed
that pHi values followed a similar pattern as the bound among the three populations, with HSCs
and Lin-CD45+ cells having the highest and lowest pHi, respectively (Fig. 2-3A,B). Importantly,
a close linear relationship existed between pHi and bound at the population level (R
2
= 0.8998, Fig.
2-3C). To further examine a potential causal relationship between pHi and bound, we manipulated
the pHi of HSCs maintained in vitro using a nigericin/K+ method enforcing pHi to be the same as
the extracellular buffer.
192
Changing the pHi in a range between 5.5 and 7.5 did not induce
significant changes in free (Fig. 2-3D, bottom). bound increased in a linear relationship with the
forced pHi change (Fig. 2-3D, top, pH 5.5 ~ 7.5); however, the slope was much lower than that of
52
the linear correlation in the three populations (0.1225 in Fig. 2-3D vs. 0.3536 ns/pH in Fig. 2-3C).
Additionally, there was no statistical difference in bound between pH 6.5~7.5, indicating that pHi
contributes little to the bound differences in the physiological pH range. Previous studies have
suggested a correlation between higher pHi and stem cell functions including the increased
glycolysis
199
, and pHi as a messenger for glycolytic flux
200
(Fig. 2-3E). We found that increasing
glucose concentration in the medium indeed enhanced the pHi (Fig. 2-3F, top), while 2-deoxy-D-
glucose (2-DG), a glycolysis inhibitor, caused a significant drop of pHi in HSCs (Fig. 2-3G, left).
Strikingly, neither the change of glucose concentration nor the addition of 2-DG changed bound
(Fig. 2-3F, bottom; Fig. 2-3G, right), suggesting that bound may be controlled by one or a few
specific enzymes instead of all the enzymes involved in glucose metabolism. Lactate
dehydrogenase (LDH) binds to NADH during pyruvate-to-lactate conversion in anaerobic
glycolysis and contributes to bound (Fig. 2-3E). We found that inhibiting LDH by 10mM sodium
oxamate (OXA), a pyruvate analog, led to significant decreases of both bound (from 3.562±0.032
ns to 3.345±0.018 ns in Fig. 2-3H, right) and pHi (Fig. 2-3H, left) in HSCs. Importantly, by
tracking the bound change at the single cell level, we found that the degree of bound decrease by
LDH inhibition is correlated with the initial bound in HSCs (Fig. 2-3I,J). Overall, these data suggest
that the higher bound in HSCs reflects their higher LDH activity, and that bound can be further used
as a biomarker of LDH activity in individual HSCs.
53
Figure 0-3. Longer NAD(P)H bound is correlated with higher intracellular pH (pHi) in HSCs and reflects
lactate dehydrogenase (LDH)/glycolytic activity.
(A) Representative images of pHi in HSCs and differentiated populations. Scale bar: 5 µm; (B) Scatter plot of pHi;
n= 94 single cells in each population; (C) Correlation between pHi and bound at the population level. Error bars:
standard error of the mean (SEM); (D) Correlation between NAD(P)H bound, free in HSCs and the extracellularly
imposed pH. n = 4 images for each data point. Error bars: standard deviation. P values: ordinary one-way ANOVA.
(E) Schematics of pHi and bound regulation by glycolytic activity; (F) pHi and bound changes in HSCs under different
glucose concentrations in the medium. n = 16~24 cells for each condition. Error bars: standard deviation (SD); (G)
pHi and bound changes in HSCs upon 2-DG treatment. n = 103 single cells for each condition. (H) pHi and bound
changes in HSCs upon OXA treatment; n = 91 single cells for each condition; P-values: Mann-Whitney test (2
conditions) or Kruskal-Wallis test (> 2 conditions); (I) Representative pseudo-color images of bound in individual HSC
54
before (NTX) and after (OXA) LDH inhibition; (J) Correlation between initial bound and Δ bound after LDH inhibition
in individual HSCs. Each dot represents the average bound and Δ bound values of a single HSC.
2.2.3 Higher NAD(P)H αbound in HSCs is contributed by enhanced LDH activity
Higher αbound has previously been interpreted as higher mitochondrial OXPHOS over
glycolysis in FLIM studies.
190
While the HSCs had higher αbound than the more differentiated
populations (Fig. 2-2E), interpreting it as an indicator of higher OXPHOS in HSCs than their
progeny is contradictory not only to the known fact that HSCs predominantly use glycolysis for
energy production
166,170
, but also to our own ORR data (Fig. 2-2D). In the glycolytic pathway,
LDH binds to NADH (which increases abound) when converting pyruvate to lactate, while pyruvate
dehydrogenase (PDH) releases free NADH (which decreases abound) in the first step of pyruvate
oxidation in mitochondria (Fig. 2-4A). We inhibited LDH and PDH with OXA (10mM) and 1-
aminoethylphosphinic acid (1-AA, 0.2%wt) for 1h, respectively.
166
Upon OXA treatment, HSCs
had a significant increase in NAD(P)H fluorescence intensity (i.e., greater accumulation of NADH)
and drop of αbound (i.e., decreased binding of NADH to enzymes) (Fig. 2-4B,E), suggesting a large
contribution of LDH activity to the high abound in HSCs. In stark contrast, OXA treatment caused
little change in either NAD(P)H fluorescence intensity or abound in Lin-CD45+ cells (Fig. 2-4C,E).
CD45+ cells, a more complex mixture of hematopoietic cells, had an intermediate response to the
treatment, between that of HSCs and Lin-CD45+ cells (Fig. 2-4D,E). Interestingly, PDH inhibition
induced little change in either NAD(P)H fluorescence intensity or abound in HSCs, suggesting
minimal PDH activity and pyruvate shuttling into the tricarboxylic acid (TCA) cycle in HSCs (Fig.
2-4B,F). In contrast, CD45+ cells showed the largest drop in NAD(P)H fluorescence intensity and
55
the greatest increase in αbound of the three populations (Fig. 2-4D,F). Therefore, the higher abound
in HSCs is contributed by the distinct glycolytic preference and higher LDH activity in HSCs.
Figure 0-4. Higher NAD(P)H
α bound is contributed by LDH
activity in HSCs.
(A) Schematics of NAD(P)H
generation (cyan arrow: released as
enzyme-free form) and
consumption (red arrow: consumed
through enzyme-binding) in
different metabolic pathways. ETC:
electron transport chain; TCA:
tricarboxylic acid cycle; (B,C,D)
Representative FLIM images of
NAD(P)H intensity and α bound in
different populations upon LDH
and PDH inhibition with oxamate
(OXA) and 1-
aminoethylphosphinic acid (1-AA),
respectively; (E,F) Quantification
of cellular level NAD(P)H intensity
and α bound changes upon LDH and
PDH inhibition by OXA and 1-AA,
respectively. P-values: one-way
ANOVA. n=5 sets of images.
2.2.4 HSCs have a larger pool of NADH compared to the differentiated cells
ORR is an established indicator for the relative rates of mitochondrial OXPHOS over the
glycolysis.
188
The lower ORR in HSCs (Fig. 2-2D) can be contributed by NADH, NADPH or FAD,
the redox states and fluorescence properties of which are intricately related to each other in cells.
185
56
Measuring the autofluorescence intensities of NAD(P)H and FAD showed that the NAD(P)H level
was the highest in HSCs (Fig. 2-5A), while FAD signals were similar among the three populations
(data not shown). As NAD(P)H autofluorescence signal comes from both NADH and NADPH,
we used a chemiluminescent assay to measure the individual NADPH and NADH levels in the
lysates of the three cell types. Interestingly, the NADH level was the highest in the HSCs, while
the NADPH levels were equivalent among the three populations (Fig. 2-5B). Therefore, it was
NADH, not NADPH, that mainly contributed to the lower ORR and larger NAD(P)H pool in HSCs.
It has lately been found that ORR is proportional to the NAD+/NADH ratio, which reflects the
demand for mitochondrial ATP production through NADH oxidation.
189
To validate this in HSCs,
we directly measured the NAD+/NADH ratio in cell lysates using a chemiluminescent NADH
assay. Consistent with the ORR data (Fig. 2-2D), HSCs indeed showed a significantly lower
NAD+/NADH ratio than the Lin-CD45+ and the CD45+ cells (Fig. 2-5C). We further analyzed
the fluorescence lifetime of enzyme-bound FAD in the three cell populations, which is negatively
regulated by the NAD+ concentration through the Stern-Volmer quenching.
187
At the population
level, HSCs had significantly longer FAD bound than Lin-CD45+ and CD45+ cells (Fig. 2-5D),
which agrees with the chemiluminescent measurement of NAD+. Overall, the lower ORR in HSCs
reflects a larger pool of NADH and its less oxidized redox state.
57
Figure 0-5. HSCs have a more reduced pool of NADH.
(A) NAD(P)H fluorescence intensities in HSCs and
differentiated cells; (B) NADPH and NADH contents and
(C) NAD+/NADH ratio measured in cell lysates (n = 3
biological replicates); (D) Fluorescence lifetime of
enzyme-bound FAD in the three populations; Error bars:
SD. P-values: Kruskal-Wallis test for the scatter plot;
ordinary one-way ANOVA for the bar plots.
2.2.5 MOBs distinguish HSCs from multipotent and oligopotent hematopoietic progenitors
Hematopoietic progenitor cells (HPCs), which include the multipotent progenitors (MPPs,
consisting of MPP
Flk2-
and MPP
Flk2+
) and the oligopotent progenitors (OPPs, consisting of common
lymphoid progenitor (CLP), common myeloid progenitor (CMP), megakaryocyte/erythrocyte
progenitor (MEP) and granulocyte/macrophage progenitor (GMP)), are rare progenitor
populations downstream of HSCs in differentiation and share some similar metabolic features with
HSCs via bulk measurement (Fig. 2-6A). We examined whether the above-established MOBs can
also distinguish the difference between HSCs and these early progenitors. We sorted HSCs, MPPs
and OPPs based on their surface markers (Fig. 2-1C), and acquired FLIM images on the freshly
isolated cells (Fig. 2-6B). Combining the single-cell and subcellular MOBs, we generated a 3-D
PCA plot showing the difference in the MOB profile between HSCs and HPCs (MPPs and OPPs)
at the single-cell resolution (Fig. 2-6B). A planar gate could be drawn between HSCs and the HPC
populations by LDA, which included the majority of HSCs (73.2%), and small fractions of MPP
58
(10.4%) and OPP (3.7%) (Fig. 2-6B). Interestingly, the center of the HSPC populations in the PCA
plot could be visually distinguished into two groups, i.e., the early stem and progenitors (HSC and
MPPs) vs. the OPPs (Fig. 2-6C). These results show that MOBs can resolve the differences
between the metabolic status of HSC and HPC populations. Notably, the linear relationship
between bound and pHi in HSCs and progenitors remained at the population level, where HSCs
had the highest bound and pHi of all the hematopoietic stem and progenitor cells (HSPCs) (Fig. 2-
6D), suggesting an increased level of anaerobic glycolysis and higher LDH activity in HSCs
comparing to HPCs.
MPPs are immediately downstream of HSCs in differentiation (Fig. 2-6A). Existing
protocols of HSC purification usually involve a final step of identifying long-term HSCs from
MPPs through surface protein markers or efflux activities.
201
Here we examined, as a proof-of-
concept, the feasibility of identifying HSCs from MPPs through the MOBs, i.e. their metabolic
characteristics which are directly associated with biological functions and conserved in HSCs from
animal models and humans.
170
Our initial evaluation shows that, HSCs are significantly different
from one or both MPPs in most of the MOBs except edge/center NAD(P)H ratio (Fig. 2-6E~I).
Among those, the most distinct MOBs were ORR, bound, and polarity of NAD(P)H (Fig. 2-
6E,G,H), reflecting a less oxidative, and more glycolytic and polarized phenotype of HSCs than
MPPs. We then trained a support vector machine (SVM, a machine learning model)
202
with all the
MOBs from Fig. 2-6E~I, which predicts whether an unknown cell is an HSC based on its MOB
profile (Fig. 2-6J). To assess the predictive capacity of this model, we FACS-sorted Lin-
cKit+Sca+ cells (KLS, a population composed of HSCs and MPPs) from B6 mice carrying Hoxb5-
tri-mCherry reporter, measured/analyzed the five MOBs of single KLS cells, and determined their
identity through the SVM model (Fig. 2-6J, boxed: predicted HSC, pHSC; unboxed: predicted
59
MPPs, pMPPs). Notably, HSCs in this model can also be identified by their high expression of
Hoxb5 (measured as the positivity of mCherry fluorescence, Fig. 2-6J) through regular
fluorescence microscopy.
125
We then compared our prediction with the Hoxb5 expression of each
cell. Importantly, our model yielded a sensitivity of 78.6 ± 7.5% (or the true positive rate, defined
as the percent of Hoxb5+ cells predicted as HSCs in the Hoxb5+ cells), and a specificity of 70.4 ±
3.6% (or the true negative rate, defined as the percent of Hoxb5- cells predicted as MPPs in all the
Hoxb5- cells) from 3 independent experiments (Fig. 2-6K), suggesting that MOBs can be used to
directly identify HSCs in the KLS cells.
60
Figure 0-6. MOB profiling distinguishes HSCs from hematopoietic progenitor cells (HPCs).
61
(A) Representative images of NAD(P)H fluorescence intensity, ORR, α bound and bound in hematopoietic stem and
progenitor cells (HSPCs). MPP: multipotent progenitor; OPP: oligopotent progenitor; CLP: common lymphoid
progenitor; CMP: common myeloid progenitor; MEP: megakaryocyte/erythrocyte progenitor; GMP:
granulocyte/macrophage progenitor. Scale bar: 5 µm; (B) Separation of HSCs from the MPP and OPP populations in
a 3-D PCA plot using the same MOBs as in Fig. 2-2K; (C) Metabolic shift between the center of the HSPC populations.
Each dot represents the average value of a given HSPC population. n = 82 single cells in each population; (D)
Correlation between pHi and bound at the population level ( bound and pHi values are from independent experiments).
n = 41 and 64 single cells for pHi and bound in each population, respectively. Error bars: SEM; (E)-(I) Individual MOB
parameters of HSC and MPP populations. Box plot: 10–90 percentile. P values: Kruskal-Wallis test. (J) Left: FLIM
images of KLS cells from Hoxb5-mCherry mouse to be fed into the pre-trained support vector machine (SVM) model,
for the prediction of HSC and MPPs. Middle (colormap): the probability of a cell being HSC based on the Platt scaling.
Right: comparison of predicted HSC identity against Hoxb5-mCherry reporter expression. Scale bar: 10 µm; (K)
Percentage of SVM-predicted HSCs and MPPs (pHSC and pMPPs) in Hoxb5+ (blue) and Hoxb5- (red) populations.
n=3 biological replicates. Error bars: SD. P values: paired t-test.
2.2.6 bound tracks changes in glycolysis during in vitro HSC culture
In normal in vitro cultures, loss of stemness and rapid differentiation of HSCs are
accompanied by metabolic reprogramming.
203
As another proof-of-concept application, we
examined whether such changes can be tracked non-invasively by MOBs. We compared the
freshly isolated HSCs incubated under a regular cytokine condition (50 ng/mL each of SCF and
TPO) for 1 hour against those cultured in the same medium for 1.5 and 3.5 days (Fig. 2-7A). A
significant increase in cell size was observed on day 1.5 (Fig. 2-7B). ORR slightly increased over
time, indicating gradually activated OXPHOS (Fig. 2-7C). αbound dropped continuously, suggesting
a shift in the balance of metabolic pathways and/or enzyme activities (Fig. 2-7D). Notably, bound
decreased over time, indicating a decrease of anaerobic glycolysis in vitro (Fig. 2-7E). The
mitochondrial content of HSCs has been reported to increase during in vitro culture.
175
Consistently, we observed an increased accumulation of NAD(P)H at the cellular edge, resulting
62
in a significantly higher edge/center ratio of NAD(P)H intensity on day 1.5 and 3.5 compared to
the freshly isolated HSCs (Fig. 2-7A,F). In contrast, the polarity of NAD(P)H autofluorescence
intensity was minimally affected (Fig. 2-7G).
Next, we investigated whether the MOBs can resolve the differences in HSC metabolism
under different in vitro culture conditions. We treated HSCs with the Wnt activator CHIR99021
(3 μM) and the mTOR inhibitor rapamycin (5 nM) (C+R) that were previously reported to promote
maintenance of HSC stemness in vitro when combined.
204
We used cytokines (50 ng/mL each of
SCF and TPO) for both conditions as the purified HSCs did not survive in the cytokine-free
environment as opposed to the KLS cells.
204
PCA analysis revealed that the majority of C+R
treated cells had a distinct MOB profile compared to the control group (Fig. 2-7H: day 3.5: 98.2%
vs. 0%). As bound is positively correlated with pHi and reflects the LDH activity in freshly isolated
HSCs (Fig. 2-3), we further monitored bound under both culture conditions for a week. The bound
of the non-treated group dropped significantly from day 3.5, while it increased, then stabilized
from day 1.5 after C+R treatment (Fig. 2-7I,J). To validate whether the higher bound in the C+R
treated group is due to enhanced glycolysis, we first inhibited LDH activities in the cultured cells
under both conditions (non-treated control vs. C+R) with oxamate at day 6.5. We confirmed that
the bound dropped significantly more in the C+R treated group than in the non-treated control group
(p = 0.003, Fig. 2-7K). Moreover, C+R treated HSCs had a much larger increase in NAD(P)H
fluorescence intensity and a decrease in αbound than the control group (Fig. 2-7L,M), indicating a
release of NADH from LDH binding upon oxamate treatment. We further performed a Seahorse
assay at day 6.5 to examine the glycolytic activities in the control and C+R treated cells. Indeed,
cells treated with C+R showed significantly higher extracellular acidification rate (ECAR) (Fig.
2-7N) and glycolytic proton efflux rate (GlycoPER) (Fig. 2-7O). Together, our results suggest that
63
the longer bound can serve as a biomarker for enhanced glycolysis, an important metabolic feature
for HSC self-renewal in vitro.
205
Figure 0-7. MOB profiling and maintenance of bound in HSCs during in vitro culture.
64
(A) Representative images of FAD and NAD(P)H fluorescence intensity, ORR, α bound and bound of HSCs during in
vitro culture. Scale bar: 10 µm. Quantification of (B) cell size, (C) ORR, (D) NAD(P)H α bound, (E) NAD(P)H bound,
(F) edge/center ratio and (G) polarity of NAD(P)H intensity of cultured HSCs. n = 35 cells per timepoint. (H) 3-D
PCA analysis of cultured HSCs from CHIR99021+Rapamycin (C+R) treatment and control conditions utilizing MOBs
at day 3.5; n = 113 cells per condition; (I) Representative pseudo-color images of NAD(P)H bound in cultured HSCs
over time. Scale bar: 10 µm; (J) Quantification of bound changes over time; n = 35 cells per condition. Quantification
of NAD(P)H (K) bound (L) intensity, and (M) α bound changes of cultured HSC upon LDH inhibition at day 6.5. n = 4
sets of images; (N) ECAR and (O) GlycoPER normalized by cell number; n = 3 biological replicates. Rot: rotenone;
AA: antimycin. Error bars: SD. P-values: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; Kruskal-Wallis test
in (B-G); Mann-Whitney test in (J); t-test in (K-M); paired t-test in (N and O).
2.4 Discussion
In the present study, we used a set of MOBs derived from the fluorescent properties of
NAD(P)H and FAD to distinguish and monitor the metabolic features/status of single HSCs non-
invasively and in real-time. However, given the fact that NAD(P)H and FAD participate in almost
all the metabolic pathways, it remains difficult to precisely deconvolute the contributions of
different pathways and their changes. Another limitation is that NADH and NADPH are spectrally
indistinguishable, it is therefore technically challenging to further differentiate the contribution of
NADH and NADPH in MOBs and cellular metabolism. Future incorporation of more direct,
invasive single-cell measurements (such as mass spectrometry imaging, MSI) will allow for more
definitive validation or elucidation of the contributing elements at the single cell level. In this study,
we used Hoxb5 model for the proof-of-concept validation of the MOBs and machine learning-
based prediction of HSCs. While the model has been characterized for the enrichment of HSCs,
there is still a lack of published data on more detailed characteristics of Hoxb5+ cells in different
hematopoietic subpopulations in bone marrow. In vivo multilineage reconstitution assays will thus
65
be a more definitive route to validate the MOBs- and machine learning-predicted HSCs. On the
other hand, further development of techniques and platforms to isolate cells based on their MOB
profiles will be needed to perform such in vivo functional analysis.
Nevertheless, our study sets a foundation for accessing HSC metabolism at the single cell
level and in real-time. Since this methodology relies on a label-free, non-invasive imaging
paradigm, it should be adaptable for both in vitro and in vivo applications.
1. Metabolic modulation represents a promising way of novel HSC expansion strategies. However,
there is a lack of tools evaluating the metabolic adaptation of HSCs in response to intrinsic and
extrinsic stimuli in vitro.
206
MOBs can be used to evaluate the influence on HSC metabolism,
especially glycolysis and OXPHOS. Therefore, the above methodology can be used for optimizing
suitable conditions for HSC ex vivo maintenance and/or expansion.
2. Metabolic stemness, namely preferred glycolysis and reduced OXPHOS, is the key feature and
functional regulator of both murine and human HSCs in vitro.
166,205
Thus, our findings may also
benefit translational studies where label-free, non-invasive measurement and tracking of HSC
status are needed, such as quality control for the HSC transplantation.
88
3. Given the advantages of two-photon microscopy, such as deep tissue penetration, minimized
light scattering in tissue, low background signal level and low photobleaching, this methodology
can potentially be adapted for in vivo HSC study through intravital imaging. Currently, most
knowledge about HSC metabolism was obtained through ex vivo measurement such as mass
spectrometry or Seahorse assays
166
, while the in vivo evidence supporting these discoveries is
indirect.
207
Combining other in vivo intravital microscopy (e.g. locating HSCs with Hoxb5
66
fluorescent protein), our methodology will allow direct accessing of HSC metabolic features.
Important physiological and pathological processes, including self-renewal, differentiation
173
,
aging, inflammation
208
, and hematological diseases/malignancies can be further studied using
these established MOBs with improved spatial and temporal resolution.
67
Chapter 3: MOB score: an endogenous optical metric reveals metabolic
inheritance and asymmetry in stem cell division
Aim 2: Establish a FLIM-based, optical metric for tracking and quantifying metabolic difference
in paired daughter cells.
3.1 Introduction
Cell division is a fundamental process that regulates the outcome of stem cell cultures in vitro.
115
Asymmetric and symmetric divisions allow for the self-renewal and expansion of stem cells,
respectively, while symmetric commitment division leads to their loss. However, it remains largely
elusive how division patterns are regulated at the individual stem cell level, e.g., the impact of
parent cell states and extrinsic stimuli, and the dynamic fate changes upon division, which are
critical for the optimization and quality control of stem cell cultures, as well as understanding the
cellular mechanisms. To answer such questions, it requires the ability to track and discriminate the
states of individual stem cells and their progeny in real-time at the single-cell level. The existing
methods often involve examining the inheritance of cell fate-related proteins or cellular
organelles.
209
These targets have slow turnover and require invasive or terminal treatments,
making them inadequate for live tracking or real-time assessment. In contrast, metabolism, as a
critical regulator of stem cell functions, shifts rapidly during the cell cycle and fate transitions.
172
It has been shown that glycolysis supports the quiescence and multipotency while activated
mitochondrial oxidative phosphorylation (OXPHOS) is required for the differentiation of
hematopoietic stem cells (HSCs),
166
mesenchymal stem cells (MSCs)
210
, and neural stem cells.
211
In fact, the preference for anaerobic glycolysis over OXPHOS was proposed as the “metabolic
68
stemness” for HSCs.
212
Therefore, analyzing metabolic states offers a unique opportunity of
monitoring the division patterns of stem cells in real-time.
Live metabolic tracking of single cells usually relies on invasive dyes that mimic metabolic
substrates or accumulate in mitochondria. They can interrupt the native metabolic dynamics, or
interfere with each other to prevent simultaneous monitoring of multiple metabolic pathways.
Fluorescence lifetime imaging microscopy (FLIM) offers a non-invasive solution to probe cellular
metabolism through the autofluorescence of metabolic coenzymes such as nicotinamide adenine
dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD).
186,213
Although
FLIM has been used to characterize metabolic changes of cell populations in tissue cultures,
organoids, and live organisms,
214
its applications in tracking single stem cell division have not
been reported. A major issue is the limited number of FLIM parameters with clear metabolic
meaning and/or associated with fate changes in stem cells.
In the presented work, using murine HSCs as a model, we defined a set of FLIM-derived
metabolic optical biomarkers (MOBs).
215
By evaluating the importance of these features with
statistical and machine learning models, we derived an endogenous metabolic metric, namely the
MOB score, to monitor the status and changes of “metabolic stemness” during HSC differentiation.
Our study thus establishes a MOBs-based conceptual framework and technical platform of
metabolic stemness in HSCs, for understanding the fundamental mechanisms of HSC renewal and
improving ex vivo cultures for therapies and transplantation.
69
3.2 Materials and methods
Fluorescence-activated cell sorting
C57BL/6 mice were purchased from Jackson Laboratories and bred at the Research Animal
Facility of the University of Southern California. Animal procedures were approved by the
Institutional Animal Care and Use Committee of the University of Southern California. Bone
marrow cells were extracted from the crushed bones of 2–6 month-old C57BL/6 mice with a mixed
gender, and then enriched using anti-cKit magnetic beads and immunostained for HSCs following
our previously established protocol.
216
FACS sorting was carried out on a BD SORP FACS Aria
cell sorter at 4°C.
Fluorescence lifetime imaging
FACS-sorted murine cells were washed and resuspended in StemSpan™ SFEM medium
(STEMCELL Technologies) supplemented with 50 ng/mL stem cell factor (SCF) and 50 ng/mL
thrombopoietin (TPO) (i.e., standard medium). For freshly isolated HSC-progenitor cell
comparison, cells were resuspended at ~10
6
/mL, seeded in a 1536-well plate (Corning), and
incubated for 1h before imaging. For HSC in vitro culture and tracking, cells were resuspended at
~5×10
4
/mL, and seeded in a 96-well plate (Corning). Medium was changed every 3 days. For
single HSC division pattern study, cells were resuspended at ~10
5
/mL, and seeded in #1 glass-
bottom Petri dishes micropatterned with anti-CD43 (eBioR2/60, eBioscience) (ref). Fluorescence
lifetime images were acquired with a Leica SP8 inverted microscope with a live cell workstation
70
(37
o
C, 5% CO2). Intracellular NAD(P)H and FAD were excited at 740 nm in the two-photon mode;
the emission wavelength/optical filter was 460/80 nm for NAD(P)H and 540/50 nm for FAD.
FLIM images were acquired at the pixel size of 0.18 μm and pixel dwell time of 15.38 μs, with 8×
line repetitions.
Single cell and subcellular region FLIM data extraction
We collected FAD and NAD(P)H intensity, as well as the phase and modulation information. abound
and tbound of NAD(P)H was then derived at the pixel level using the previously established phasor
approach.
196,215
Individual cells were filtered, segmented and masked based on the NADH intensity
by a customized python code (Python 3.9). After background subtraction, FAD intensity of all
pixels in individual single cells was used to distinguish the mitochondria and cytoplasm regions
using Otsu’s method. ORR was calculated as the FAD/NAD(P)H intensity ratio in the
mitochondria region and normalized by subtracting the intensity ratio in the cytoplasm region. In
both regions of a single cell, NAD(P)H intensity, abound, tbound, and ORR (only in the mitochondria
region) were quantified using a collection of defined parameters describing the signal strength
(average, median, 10 percentile, 90 percentile, etc.), statistical distribution (standard deviation,
skewness, kurtosis, etc.) and spatial distribution/ texture (statistics of grey level co-occurrence
matrix, grey level run length matrix, etc.). The difference of individual parameters between the
two regions was also quantified as:
71
A few morphological parameters (size, major axis length, circularity, etc.) was also developed
using the mitochondria, cytoplasm and whole cell masks. Extraction of all above parameter was
processed using a customized Python code with image processing packages, including scipy,
skimage, Pyradiomics, and SimpleITK.
Parameter selection
“Informative” parameters were selected based on the prediction performance (in a defined machine
learning model) and trending continuousness in both in vitro differentiation and HSPC hierarchy
datasets. For each individual parameter, a logistic regression model was trained with 60% of the
sample data and F1 score was employed to evaluate the prediction performance using the rest of
40% of the data. The threshold of F1 score was set to 0.6. Parameters scored below this value were
considered as behaving similarly in HSC and differentiated cells and thus dropped. The
continuousness of individual parameters was quantified by the trending index (TI) during the
differentiation, defined as
where parameter was calculated as the average of the population, and tn refers to the differentiation
stages of HSC in specific datasets. For example, tn = 0 for HSCs, tn = 1 for MPPs, and tn = 2 for
CPs. If the parameter changes in a continuous trend during HSC differentiation, TI equals 1
72
(continuously increasing) or -1 (continuously decreasing). Parameters with TI in between -0.5 and
0.5 were considered as “fluctuated” and thus dropped. Eventually, each parameter was examined
and only those fulfill (1) F1 score > 0.6 in both datasets and (2) TI both below -0.5 or both above
0.5 in the two datasets were highlighted and selected. In this step, 56 MOBs were screened out
under these criteria. A knowledge graph reflecting the nature of different parameters was further
plotted to select the biologically independent parameters based on the enrichment of above 56
parameters in subcategories. Eventually, 11 representative MOBs were selected.
ROC curve and latent variable model
Individual datasets were first normalized with a robust scaler using HSCs as the training set except
otherwise indicated. The robust scaler adjusts the population median to zero and rescales the data
according to the range between 1
st
quartile (25% quartile) and 3
rd
quartile (75% quartile), such that
median is rescaled to 0, 1
st
quartile is rescaled to -1, and 3
rd
quartile is rescaled to 1.
The ROC curve was employed to evaluate the information that can be inferred by the selected
MOBs. In each dataset, a support vector machine model was trained by 70% of the single cells
with all 205 extracted parameters, the 11 selected MOBs, or cell size. True positive rate, false
positive rate, and area under the curve (AUC) were calculated and plotted using the rest of 30%
single cell data.
For latent variable analysis, the 11 selected MOBs were fitted into a regression model with
customized Python code calling the scikit-learn factor analysis algorithm in the scikit-learn
73
package. The dimensionality of latent space was set as 3 and the initial guess of the noise variance
for each parameter in the model was set as 0. Factor loadings of the primary latent variable were
calculated and used as the weight of each MOB to derive the MOB score. To benchmark the MOB
score with pseudotime analysis established for scRNA-seq, the 205 parameters were normalized
and pseudotime analysis was performed using the trajectory inference algorithm from Scanpy
package with a customized python code.
217
Dimension of the datasets was reduced to the first 6
principal components and the Louvain method was employed for community detection to infer the
differentiation path.
Single HSC micropatterning
#1 glass-bottom Petri dishes were cleaned by detergent, rinsed, and dried before use. Stamps were
designed as 30 μm by 30 μm squares with 30 μm edge-to-edge spacing with AutoCAD (Autodesk,
Inc.). Mold was made by SU8 photoresist with 50 μm height on 3-inch silicon wafers. Sylgard 184
polydimethylsiloxane (PDMS) was fully mixed at 10:1 base-to-curing reagent ratio, poured onto
mold and baked at 80
o
C overnight. Stamps were then cut and peeled from the mold, and primed
with anti-CD43 (eBioR2/60, eBioscience). After 1 hour, stamps were rinsed with detergent, PBS,
and water and dried. Micropatterns were then transferred from stamps to the prepared #1 glass-
bottom Petri dishes by microcontact printing. Freshly isolated HSCs were seeded on at a density
of 20 cells/mm
2
, allowing the majority of the micropatterns accommodating less than 1 cell.
74
TMRM and 6NBDG uptake and imaging
Paired cells on micropatterned antibody array were first imaged under FLIM mode using a Leica
SP8 FALCON inverted microscope with each cell’s location recorded. For TMRM staining, cells
were incubated with 20 nM TMRM for 30 min, and imaged in standard z-stack confocal mode
subsequently under 561 nm excitation and 600/40 nm emission. For 6NBDG staining, cells were
starved in glucose-free DMEM medium for 30 min, and then incubated with 6NBDG for 30 min
and imaged in standard z-stack confocal mode subsequently under 488 nm excitation and 515/20
nm emission. During the whole imaging process, cells were incubated at 37
o
C with 5% CO2.
Cellular 6NBDG uptake was calculated as the integrated fluorescent intensity over all z-stack
slices after subtracting the background. TMRM MFI was calculated as the integrated fluorescent
intensity normalized by the TMRM-positive regions after thresholding.
Immunostaining of surface markers
Paired cells on micropatterned antibody array were first imaged under FLIM mode using a Leica
SP8 FALCON inverted microscope with each cell’s location recorded. Cells were then washed
with PBS, and then fixed with 4% PFA at room temperature for 12 minutes. After washing with
PBS, cells were blocked with 4% BSA in PBS at room temperature for 2 hours. Cells were then
stained with Alexa Fluor 647 conjugated Tie2 antibody (Clone 33, BD Bioscience. 1:100 dilution
in 4% BSA) and PE conjugated CD48 antibody (HM48-1, BioLegend. 1:100 dilution in 4% BSA)
at room temperature for 2 hours. Imaging of surface markers was carried out on the same
microscope under normal confocal mode.
75
Colony-forming unit (CFU) assay
Freshly isolated HSCs were seeded into a 96-well plate (Corning) with the initial density of 200
cells per well and cultured with different metabolic drugs as indicated above for 7 days. Cells were
counted with a hemacytometer and then washed with Iscove’s Modified Dulbecco’s Medium
(IMDM) with 2% Fetal Bovine Serum (FBS). 1/100 cells from each well (equivalent to cells
expanded from 2 initial HSCs) were resuspended in 100 uL IMDM + 2% FBS and then mixed
with 1 mL MethoCult™ GF M3434 medium (STEMCELL Technologies), seeded into a 6-well
clear bottom plate (Corning) and incubated at 37
o
C with 5% CO2. After 7 days and 10 days of
culture, colonies from each well were imaged under a Nikon Inverted Microscope Eclipse Ti-E
with a 2X objective.
Data plot and statistical analysis
All plots were made in Prism 7 (GraphPad) and Python 3.9 (Python Software Foundation). All
data presented as the mean ± error had the error defined as 95% confidence interval (CI). For all
paired daughter cell analysis, statistical analysis was performed with paired t-test. Otherwise
statistical analysis for single-cell scatter or box plots were performed with the Mann-Whitney (2
conditions) or Kruskal-Wallis tests (> 2 conditions) due to non-normal data distribution. All bars
in the scatter plot are median, and box plots represent the 10th–90th percentile. All error bars in
the bar graphs were plotted as standard deviation (SD) and compared with the Welch’s t-test (2
76
conditions) or ordinary one-way analysis of variation (ANOVA) (>2 conditions). Correlation
analysis was carried out with Pearson r and p-values were tested with zero-slope hypothesis. P-
values are indicated as numbers or as n.s. (not significant, p > 0.05) on the plots.
3.3 Results
3.3.1 MOBs track the metabolic dynamics of HSC differentiation
HSCs proliferate and differentiate in vivo, or upon cytokine stimuli in vitro (Fig. 3-1A).
We first obtained high-resolution FLIM images of HSCs and hematopoietic progenitors from
freshly isolated bone marrow or from extended in vitro HSC cultures. Four MOB parameters with
established biological/metabolic meanings
215
were extracted and calculated to indicate cellular
states or changes at the subcellular level (Fig. 3-1B). Next, we established a library of 205
cellular/subcellular MOB-based features from each single cells for downstream machine-learning
and statistical workflow-based feature selection (Fig. 3-1C). These include: 1) the morphological
features (size, shape, etc.); 2) the signal strength, statistical distribution, and spatial variance of the
MOBs in cytoplasm and mitochondria, considering NAD(P)H compartmentalization in the two
regions
218
; and 3) the difference of the MOBs between the cytoplasmic and mitochondrial regions.
We hypothesized that the stemness-related MOB features should behave similarly in trend during
both the in vivo and in vitro differentiation processes. Based on this principle, we narrowed the
feature list down to 56 candidates by highlighting those both (1) with consistent up- or downward
trends in the two differentiation processes (Fig. 3-1D-G), and (2) show good prediction
77
performance in distinguishing HSCs from differentiated cells when trained with a logistic
regression model (for details, see Materials and methods).
Figure 0-1. Evaluation of MOBs in profiling HSC differentiation.
78
(A) Phenotypic profiling and quantification of cells expanded in vitro from 200 HSCs in one week. n = 3 biological
replicates. p-values: one-way ANOVA. (B) Schematic of FLIM profiling the metabolic phenotypes of hematopoietic
stem and progenitor cells (HSPCs) (left) and representative FLIM-derived metabolic images of HSC during in vitro
differentiation and freshly isolated HSPCs along differentiation hierarchy (right). Scale bar: 10 μm. (C) Subcellular
region segmentation and parameter definition in individual FLIM-derived channels. (D) Workflow for parameter
selection and MOB scoring using both in vitro differentiation and HSPC hierarchy paradigms. (E) Representative
parameters in continuous process-based selection. n = 36~100 cells per time point for in vitro differentiation model
and 137~150 cells per population for the HSPC hierarchy. (F) and (G) Heatmap of 56 parameters passing machine
learning and continuous process-based selection in both in vitro differentiation and HSPC hierarchy models.
3.3.2 MOB score derived from 11 representative features recapitulates HSC differentiation
trajectory
To avoid the overrepresentation of any specific MOB features in the final metric, we
generated a knowledge graph to illustrate the biological impact and correlation of the 56
parameters (Fig. 3-2A), and finalized 11 highly independent features with clear biological
meanings. We then trained support vector machine (SVM)-based classification models with a
single feature (cell size), the 11 MOBs, or all the 205 features to evaluate its performance in
predicting HSCs from freshly isolated progenitors or cultured HSCs at day 2~7. The receiver
operating characteristic (ROC) curves
219
indicated that the SVM models trained with the 11 MOBs
reached an accuracy similar to that with all the 205 parameters (Fig. 3-2B). Analyzing the 11
individual MOBs revealed a trend of increased cell size, metabolic switch from anaerobic
glycolysis to OXPHOS, as well as decreased NAD(P)H level, spatial variance of mitochondrial
metabolisms, and metabolic distinction between mitochondrial and cytoplasmic regions during
HSC differentiation. The robustness of the 11 MOBs in tracking HSC differentiation was further
illustrated by their consistent states/trends in three independent isolations and cultures,
respectively (Fig. 3-2C,D), .
79
Next, we used a latent variable model to combine the 11 MOBs into a numeric score (i.e.,
the MOB score), to define the “metabolic stemness” of single cells (Fig. 3-2E-I). This model is
linear and can explain explicitly the contribution of individual MOBs to stem cell differentiation
(Fig. 3-2E,H, right). We benchmarked the MOB score with the state-of-art “pseudotime analysis”
algorithm developed originally for scRNA-seq analysis. Using the 205 parameters but following a
different workflow (dimension reduction by PCA and then non-linear trajectory inference), the
resultant “pseudotime” in differentiation matches our result at the single cell level (Fig. 3-2J).
MOB score thus can precisely reflect cell differentiation status with FLIM data. We also expect
this analysis workflow adaptable to many other imaging modalities for single cell tracking.
80
Figure 0-2. MOB score derived from representative features and latent variable model.
(A) Knowledge graph visualizing the selection process of MOBs. Hierarchy reflects the parameter definition and
relationship. Fractions indicate the number of parameters that pass the F1 score and continuous process-based selection
process. Highlighted edges indicate the subcategories that include the eventually selected MOBs. (B) ROC curve
indicating the prediction accuracy by the support vector machine (SVM) model using different parameters in in vitro
differentiation (n = cells) and HSPC hierarchy (n = cells) paradigms. (C) and (D) Hierarchical clustering showed that
MOBs can track HSC differentiation in repeated experiments for in vitro differentiation and HSPC hierarchy. (E)
81
Heatmap of MOBs in HSPC hierarchy (left) and their weight in MOB score (right). (F) Factor analysis of in vitro
HSC culture undergoing differentiation. The MOB score is defined as the primary latent variable (LV1). n = 36~100
cells per population. (G) Correlation of weight of MOBs derived from HSC in vitro differentiation and HSPC
hierarchy. (H) Heatmap of MOBs in in vitro differentiation (left) and their weight in MOB score (right). (I) Factor
analysis plots of in vitro HSC culture undergoing differentiation. The MOB score is defined as the primary latent
variable (LV1). (J) Correlation between MOB score and pseudotime.
3.3.3 MOB score identifies the metabolic asymmetry in PDCs
We then seek to apply the MOB score on tracking single stem cell fate and division
patterns. To do so, we employed a micropatterning platform that constrains the single suspension
HSCs for continuous observation without disrupting their division
119
(Fig. 3-3A,B). We first
evaluated whether the MOB score is sensitive enough to discriminate the metabolic difference
between the paired daughter cells (PDCs). A direct comparison between the difference of the
MOB scores (ΔMOB score) and that of invasive metabolic dyes (Δ6NBDG or ΔTMRM) in the
same PDCs showed a correlation between the two readouts (Fig. 3-3C,D), suggesting MOB
score can discriminate PDCs at different statuses. To quantitatively describe HSC division
patterns, we analyzed the metabolic asymmetry in PDCs using the ΔMOB score. Notably, we
observed two major peaks in the histogram of ΔMOB scores (Fig. 3-3E), indicating the existence
of two distinct division patterns. To access whether the metabolic asymmetry reflects distinct
fates of PDCs, we immunostained the PDCs of HSCs for Tie2 and CD48, a set of binary surface
markers previously established to identify asymmetric divisions of HSCs
15
. Indeed, the more
stem-like daughter cells (Tie2+CD48-) showed significantly higher MOB scores than their
committed siblings (Tie2-CD48+) (Fig. 3-3F). A logistic regression model on the MOB scores
was employed to determine the threshold between the stem and committed cells (Fig. 3-3F).
82
Figure 0-3. MOB score distinguishes different division patterns.
(A) Schematic of trapping single HSCs with micro-contact printing. (B) Illustration and representative images of
single HSCs trapped on the micropattern array and tracked during division. Scale bar: 50 um. (C) Representative
images of differential 6NBDG uptake in PDCs and the correlation of ΔMOB score and Δ6NBDG uptake. Δ6NBDG
was normalized by average signal in PDCs; i.e. Δ6NBDG = (daughter #1 – daughter #2)/(daughter #1+ daughter
#2). Scale bar: 10 μm. n = 29 cell pairs. p-values: paired t-test. (D) Representative images of differential TMRM
uptake in PDCs and the correlation of ΔMOB score and ΔTMRM MFI. ΔTMRM MFI was normalized by the sum
of PDCs. Scale bar: 10 μm. n = 33 cell pairs. p-values: paired t-test. (E) Histogram of ΔMOB scores in PDCs from
the 1
st
division of HSCs in vitro. (F) Representative images of phenotypically identified PDCs from asymmetric
division and the corresponding MOB images. Scale bar: 10 μm. Correlation between Tie2CD48 phenotypes and
MOB scores in PDCs. n = 17 cell pairs. p-values: paired t-test.
3.3.4 MOB score reveals distinct patterns of metabolic dynamics in HSC division
The ability to distinguish the subtle metabolic difference between single cells makes MOBs
appealing metrics for tracking stem cell division. For example, the MOB score can track the
metabolic transition during HSC cell-cycle activation and division (Fig3-4. A,B). The analysis of
83
time-lapse lapse studies enables the real-time tracking of different division patterns. In single
dividing HSCs, we observed that NAD(P)H intensity (along with other MOBs) and the MOB score
became divergent/asymmetric in the M-asymmetric PDCs 60-minute after the division (Fig. 3-4C-
E). Our results suggest that the observed metabolic asymmetry is regulated by some upstream
determinants rather than the differential inheritance of NAD(P)H itself.
The precise tracking of single cell status enables the study of how daughter cells’ fate is
determined. MOB analysis allows fast identification and comparison of the metabolism of PDCs
and their parent HSCs. Interestingly, there is a significant correlation between the MOB scores of
parent HSCs and their offspring (Fig. 3-4F, G), suggesting an inheritance/continuity of the
metabolic phenotype from parent HSCs. This is consistent with a previous report that HSCs display
an epigenetic memory in clonal behaviors including cell division
16
. Our results here indicate that
the metabolic stemness is a relatively stable feature in the cultured HSPCs, and has an “inertia”
from parent to daughter cells during cell division. Importantly, we observed that, while the HSCs
with higher MOB scores can give rise to stem-like daughter cells regardless of the division patterns,
the HSCs with lower MOB scores have a better chance to regain metabolic stemness in their
daughter cells through M-asymmetric division (78.3%) than M-symmetric division (22.8%) (in
daughter cell #1, the one with higher MOB score) (Fig. 3-4H, I). These results suggest a unique
role of asymmetric division in maintaining the HSC population and directing fate decisions.
A key question we hope to answer with MOB metric is, what causes stem cell loss during
division? We tracked the MOB scores of PDCs after division. Upon division, over half of the
daughter cells from M-symmetric division had MOB scores above the threshold line, suggesting
their initial stem-like status; in contrast, daughter cells from M-asymmetric division show
84
divergence of metabolic stemness. However, 8 hours later, MOB scores of PDCs from M-
symmetric division had dropped below the threshold line to become similar to the committed
daughter cells from M-asymmetric division; in contrast, the MOB scores of the more stem-like
daughter cells from M-asymmetric division remained above the threshold (i.e. a more glycolytic
phenotype; Fig. 3-4J,K). Our results strongly suggest that without further intervention, M-
symmetric division fails to expand the HSC population, while M-asymmetric division prevents the
loss/exhaustion of the stem cell pool under the normal cytokine culture condition.
85
86
Figure 0-4. MOB score identifies the metabolic inheritance and asymmetry in paired daughter cells (PDCs).
(A) Illustration of freshly isolated phenotypic HSC (pHSC) turned activated (aHSC) in culture and divides into
PDCs. (B) MOB score reflects metabolic change during HSC division. (C) Time-lapse images of NAD(P)H
intensity in PDCs from metabolically symmetric and asymmetric divisions. Scale bar: 10 μm. (D) Representative
MOB images of PDCs from metabolically symmetric (M-symmetric) and asymmetric (M-asymmetric) divisions.
Arrows indicate the glycolytic daughter cell from an M-asymmetric division. Scale bar: 10 μm. (E) Metabolic
dynamics of individual PDCs from M-symmetric and M-asymmetric division. (F) Representative NAD(P)H, bound
and ORR images of individual HSCs and their daughter cells. Scale bar: 10 μm. (G) Correlation of MOB scores
between parent HSC and PDCs. (H) and (I) Inheritance of metabolic phenotypes under different division patterns.
Color maps: percentage of daughter cells committed to higher and lower MOB scores from aHSCs with lower or
higher MOB scores in the two division patterns. n = 106 cell pairs from 3 independent experiments. (J) and (K)
Maintenance/loss of metabolic stemness of PDCs from different division patterns. n = 54 cell pairs.
3.3.5 MOB score predicts culture conditions supporting HSC expansion by division pattern
analysis
The above analysis suggests that modulating metabolism in parent stem cells may change
the output of daughter cells. We then shortlisted 7 candidate drugs that were previously reported
to target metabolic pathways. A comparison of ΔMOB score to the non-treated group showed that
only rapamycin promotes M-asymmetric division (Fig. 3-5A, B). M-asymmetric division stably
generates two divergent daughter cells, even under the oxidative stress induced by TBHP. Other
than the previously reported drug rapamycin that promotes HSC expansion, strikingly, pan-PI3K
inhibitors LY294002 and copanlisib promoted MOB score in the PDCs from M-symmetric
division (Fig. 3-5C), suggesting they may rescue the loss of stem cells. To validate, we cultured
freshly isolated HSCs for 1 week. Flow cytometry showed that pan-PI3K inhibition dramatically
increased the fraction of Flk2-CD150+ KLS cells (Fig. 3-5D). We then performed a colony
forming unit (CFU) assay, which showed that LY294002 and copanlisib treated cells formed more
colonies as well as the most primitive GEMM colonies (Fig. 3-5E). These results suggest that the
MOB score can predict the output of stem cell expansion by analyzing their division patterns.
87
Figure 0-5. MOB score identifies conditions that promote HSPC expansion.
(A) Experimental design studying the influence of metabolic treatment in parent HSCs to their daughter cells. (B)
Quantification of metabolic asymmetry induced by candidate drug treatment. (C) MOB score of PDCs under
different division patterns after drug treatment to parent HSCs. (D) Flow cytometry analysis of the CD150+Flk2-
fraction of KSL cells after one-week in vitro culture of freshly isolated HSCs. (E) Colony forming ability per 1000
cells from HSCs cultured under pan-PI3K inhibition. n = 3 biological replicates.
88
3.4 Discussion
Studying asymmetric division can help understand the homeostasis of the hematopoietic system,
and provide guidance to the maintenance and expansion of HSPCs in vitro. However, it has been
challenging to track the dynamic process of asymmetric division. Here, we established a non-
invasive, real-time metric (the MOB score) to quantitatively analyze the stemness of single HSCs
and their daughter cells through a metabolic perspective, i.e. the metabolic stemness, defined by
the preference for anaerobic glycolysis over mitochondrial OXPHOS
12
. We derived the metric
through the analysis of the metabolic trajectory of the cultured HSCs undergoing differentiation
and the freshly isolated HSPCs, which resembles the trajectory inference (or pseudotemporal
ordering) originally used in mapping cell differentiation from single-cell RNA-sequencing data
36,37
.
The similarity of the two trajectories (i.e. the eigenvectors of the PC1s) confirms the glycolysis-
to-OXPHOS shift as a key metabolic feature of the HSC differentiation.
Our metric is advantageous over the conventional protein marker or intracellular staining-
based readouts. Metabolism changes faster than protein markers, as shown with the invasive dyes
8
.
However, the commonly used loading protocol of 2NBDG/6NBDG for monitoring glucose uptake
requires starvation, and the glucose analogs also compete against the glucose, thus interrupting the
normal glycolysis. Moreover, 2NBDG/6NBDG uptake reflects the glucose consumption but
cannot distinguish its downstream utilities in different pathways, e.g. the pyruvate-to-lactate
conversion in anaerobic glycolysis or the pyruvate-to-acetyl-CoA conversion in mitochondria.
Moreover, some progenitors have been recently reported to have higher glucose uptake than
HSCs
38
, while the glycolytic phenotype of HSCs is more defined by anaerobic glycolysis (i.e. the
pyruvate-to-lactate conversion)
12,39
. On the other hand, a proper interpretation of TMRM staining
89
for mitochondrial activity depends on sufficient loading time (typically>30 min) for its
accumulation and equilibration in the active mitochondria
40
. The TMRM signal may thus lag
behind the actual metabolic changes in dynamic situations. We confirmed that, at the steady state,
the MOBs show a consistent trend with 6NBDG and TMRM readouts (Fig. 2G-J). Meanwhile,
since NAD(P)H and/or FAD directly participate in glycolysis and OXPHOS, the MOBs can be
used to monitor metabolic dynamics in real-time.
We also adopted a micropatterning approach
41
to facilitate the tracking of single HSCs and
the PDCs. Notably, other groups have used microwells to trap HSC and monitor their divisions
14,42
.
However, microwells are usually cast in polydimethylsiloxane (PDMS), which often absorbs
proteins (e.g. cytokines) non-specifically
43
; the microwell topography also prevents thorough
changes of medium/buffer during immunostaining or dye loading, leading to inaccurate readouts.
On the other hand, uniformly coated anti-CD43 has been utilized to immobilize HSPCs. However,
even under high coating concentrations, the antibody still fails to restrain cell migration while cell
division starts to get influenced, limiting long-term single-cell tracking of HSCs or their daughter
cells
6
. Micropatterning confines HSCs and corresponding daughter cells in a pre-defined array,
which enables cell immobilization, rapid image registration, and long-term monitoring while
allowing for minimal antibody density to reduce the interference to the normal cell division and
ease the downstream staining protocols.
With the MOB score, we were able to measure the metabolic asymmetry and confirm its
association with different fate choices of PDCs, as reported with the invasive approaches
12,13
.
Interestingly, we found that PDCs have similar metabolism immediately after division even in the
M-asymmetric ones, and the asymmetry did not appear until about an hour later under our time
90
interval settings. The metabolic stemness of the daughter cells also largely mirrored their parent
HSCs. Notably, the functional heterogeneity in HSCs has been shown to be clone-specific as a
result of epigenetic constraints
32
. It was also reported that the difference of the levels of fate
determinants in PDCs rarely exceeds two-fold
8
, and that PDCs have similar transcriptomes after
cell division
5
. In T lymphocytes, which can also commit asymmetric cell division, the expression
of a metabolic regulator, c-Myc, becomes divergent in daughter cells only after the cell division
44
.
Our results are consistent with these observations and suggest that cell metabolism may act as an
effector function downstream of some differentially regulated or inherited fate determinants (e.g.
proteins, epigenetic modifications, cellular organelles, or extrinsic cues). Importantly, we
demonstrated an emergence of metabolic asymmetry in the PDCs after division in a much shorter
timescale than what was reported with protein markers
8
. Such rapid response coupled with the
real-time ability of the MOBs will enable the identification of new fate determinants in the
asymmetric division or the study of the functional dynamics of these regulators.
We have further shown a distinct longer-term metabolic dynamic from pHSCs to aHSCs
and to PDCs. While most pHSCs were glycolytic, they got activated during the in vitro culture and
acquired a transient metabolic state with higher OXPHOS (i.e. lower MOB score), which
resembles the differentiated cells. Surprisingly, the PDCs from M-symmetric division temporarily
regained a more glycolytic phenotype but then quickly lost it over time, while the stem-like
daughter cells from M-asymmetric division can maintain their metabolic stemness. Our discovery
thus suggests M-asymmetric division as a mechanism of rescuing the stemness in the aHSCs.
Indeed, it was previously reported that the AKT/mTORC1 pathway, which governs mitochondrial
OXPHOS, is activated in HSCs during the cell cycle; and a failure of countering such activation
results in HSC differentiation and exhaustion
45
. While the previous studies have focused on the
91
molecular mechanisms that prevent the overactivation
45,46
, our result suggests that such a
refraining process can also be achieved through asymmetric division. Interestingly, cancer stem-
like cells have also been reported to maintain their stemness through asymmetric division while
their daughter cells from symmetric division become differentiated
10
. This metabolic dynamic may
thus be a common mechanism in stem cell self-renewal. Such commonality warrants further
investigation and translation of the metabolic stemness in other stem cell systems, which can be
empowered by our MOBs-based approach with unprecedented sensitivity and temporal resolution.
Finding conditions for maintaining and expanding the HSC population ex vivo is a long-
sought goal in the HSC field, which has met with only limited success. The appropriate balance
between different division patterns is critical for maintaining and expanding the HSC population.
As shown with the proof-of-concept study with C+R treatment, the potential successful strategies
would include promoting the M-asymmetric division or the metabolic stemness of the daughter
cells from M-symmetric division in HSC cultures. Our platform provides rapid feedback on the
metabolic states of the daughter cells and the division patterns, while requiring very few cells for
each condition, making it ideal for screening in vitro culture conditions that favor such goals.
Moreover, given the deep tissue penetration of the two-photon FLIM microscopy, this approach
also has the potential to be extended to in vivo applications.
92
Chapter 4: Surface oligo tagging for high throughput gene expression
profiling of paired daughter cells
Aim 3: Develop an oligo barcoding strategy to enable high throughput analysis of the gene
expression difference between paired daughter cells and to understand the molecular
mechanisms of different division patterns.
4.1 Introduction
Adult stem cells (ASCs) are defined by their ability to maintain or repair specific tissues through
self-renewal and differentiation into mature and functional cells. To fulfill these tasks, ASCs rely
on the ability of asymmetric division (AD) and symmetric division/commitment division
(SD/CD).
114,115
Understanding the mechanisms that regulate ASC division patterns will help
understand the homeostasis and regeneration of tissues or organs, as well providing guidance for
the in vitro maintenance and expansion of ASCs. However, the mechanisms that regulate different
division patterns have not been completely elucidated because of lacking tools to decipher this
process with good details. Only a few clues were discovered to regulate cell division
pattern.
118,120,122,123
We have developed a fluorescence lifetime imaging microscopy (FLIM) based
method tracking single cell division in high throughput with good temporal and spatial details, as
described in Chapter 3.
The development of next generation sequencing (NGS) and the corresponding high
throughput platforms have recolonized biological and biomedical studies in the past decade.
133,220
Although they can only obtain information at certain time points (i.e. cells have to be lysed to
prepare the library for NGS), NGS and NGS-based multi-omics analysis can provide great details
93
of the question interested. However, the study of division patterns has not been able to fully utilize
these technical advantages. To analyze the paired daughter cells (PDCs) from the same parent,
researchers must manually prepare the sample of each single daughter cell and note the sibling
information. This process is labor-intensive and limits the throughput of PDCs studied (usually a
few dozen pairs).
118
Various barcoding strategies have been developed to track single cells and the derived colonies in
vivo and in vitro.
221
Most of such strategies rely on certain transfection systems. Transfection
protocol usually takes 24 hours or longer, and introduces stress on the target cells. Given that the
normal cell division cycle is usually 24 ~ 48 hours, the transfection-based barcoding strategies are
not suitable for analyzing primary cells and their immediate PDCs. Moreover, to guarantee the
uniqueness of barcoding, the multiplicity of infection (MOI, the number of viral particles that can
transfect each cell) is usually low (< 0.1), which results in low barcoding efficiency. This is not
optimal for studying adult stem cells since they are rare cell types (e.g. the number of
hematopoietic stem cells is ~10, 000 per mouse). Several strategies for barcoding the cell surface
have been previously reported for the purpose of sample multiplexing or antigen analysis.
222
MULTI-seq inserts lipid-oligo conjugates into the cell membrane. However, the half-life of lipid-
oligo conjugates on the cell surface is less than 60 minutes. Cell surface barcoding with antibody-
oligo conjugates also suffers from a short half-life, likely caused by antibody detachment or
cellular clearance mechanisms.
223
To address these challenges, we propose a cell surface-liposome-oligo tagging strategy,
which aims to use liposome as the oligo carrier to prolong the existence of oligos on the cell surface.
94
With this strategy, PDCs from the same parent will carry the same oligo tags, enabling gene
expression profiling and analysis in a high throughput manner.
Figure 0-1. Schematic of cell surface oligo tagging for tracking paired daughter cells in droplet-based single
cell RNA-seq.
Every single cell is labeled with unique oligo tags, and pooled cultured. After division, the sibling daughter cells will
inherit the same tags. Both surface oligo tags and mRNA will be captured, barcoded and sequenced, allowing the
correspondence of the sibling daughter cells to their gene expression profile. GEM, Gel Bead-In Emulsions; NGS,
next generation sequencing; GEX, gene expression.
4.2 Materials and Methods
Cell isolation and culture
Mouse CD4+ T cells were isolated from the mouse spleen with a negative selection kit (StemCells).
Briefly, the mouse spleen was meshed and filtered through 100 um porous membrane. Red blood
cells (RBCs) were lysed in RBC lysis buffer for 1 min. The remaining cells were washed,
resuspended in EasySep buffer (StemCells), and subsequently incubated with negative selection
95
antibodies and magnetic beads. CD4+ T cells were collected from the supernatant. Upon isolation,
T cells were cultured on CD3 and CD28 coated Petri dish or with Dynabeads mouse T-activator
CD3/CD28 in RPMI 1640 medium containing 10% FBS, 1% penicillin G and streptomycin (P/S)
and 50 uM b-mercaptoethanol.
HSCs were isolated from mouse bone marrow. Cells were first extracted from the crushed bones
of 4–6 month-old C57BL/6 mice and then immunostained for CD45+ and Lin-CD45+ cells, or
enriched by cKit/IL7R and immunostained for HSCs. FACS sorting was carried out on the BD
SORP FACS Aria cell sorter at 4°C, as in our previous publication.
215
Jurkat cells were a generous gift from Prof. Ping Wang’s lab. They were cultured with the complete
medium (1640 medium containing 10% FBS and 1% P/S).
Microscopy and flow cytometry
For tracking T cell activation and examining liposome conjugation on cell surface, cells were
imaged under a Nikon Inverted Microscope Eclipse Ti-E with 20X objective. Flow cytometry
assays were carried out with a MACSQuant Analyzer 10 Flow Cytometer (Miltenyi Biotec).
Liposome fabrication
DOPC, MPB-PE, and DOPG (Avanti Polar Lipids, Inc) were mixed with a molar ratio of
100:100:1 in chloroform. After evaporating chloroform with compressed air, the vacuum was
applied for 1.5 hours to remove the residual chloroform. The lipid mixture was then resuspended
96
in PBS, vortexed, and sonicated. Liposome was then fabricated by extruding the lipid mixture
through a 200 nm polycarbonate membrane for 20 cycles.
Real time qPCR
After cells were conjugated with oligo or liposome-oligo, 1000 cells (after washing and
resuspended) and the equivalent volume of medium supernatant were collected before and after
culture. Samples were then mixed with the PowerUp SYBR Green Master Mix (Thermo Fisher)
and designed primers in a 384-well plate. qPCR was carried out with the BioRad CFX384 Real
Time PCR System.
4.3 Results
4.3.1 Direct oligo tag conjugation to cell surface proteins
While both the lipid-oligo and antibody-oligo conjugates have a short half-life on the cell surface,
we ask if a covalent conjugation of oligo tags to cell surface protein will maintain the oligo tags
for a longer period. It has been reported that cells have a considerable level of thiol groups on their
surface proteins.
224
We thus employed a thiol-reactive maleimide-PEG4-methyltetrazine (Mal-
PEG4-mTZ) molecule to functionalize the cell surface with mTZ groups, so that the trans-
cyclooctene conjugated oligo tags (TCO-PEG4-oligo) can be linked on the cell surface with a fast,
efficient click chemistry reaction between mTZ and TCO (Fig. 4-2A). The oligos were
97
successfully linked on the surface of Jurkat cells, CD4+ T cells and hematopoietic stem cells
(HSCs), indicated by the signal of Cy5 conjugated on oligos (Fig. 4-2B). However, after 36 ~ 48
hours, most oligo tags only exist with a low level on all three cell types (Fig. 4-2B). A timelapse
tracking on Jurkat cells indicated that oligo tags disappeared sharply with a half-life shorter than
3 hours (Fig. 4-2C). Quantitative polymerase chain reaction (qPCR) confirmed that most oligo
tags on Jurkat cells were removed in the first 24 hours (Fig. 4-2D, left). While qPCR showed that
~50% of the oligo tags still exist after 48 hours on CD4+ T cells (Fig. 4-2D, right), we found that
surface oligo tagging inhibits CD4+ T cell activation, growth, and division (Fig. 4-2E). We thus
concluded that direct conjugation to cell surface protein is not a suitable strategy for oligo tagging.
98
Figure 0-2. Oligo tagging to cell surface proteins and the influence on cell fate.
(A) Schematic of the two-step strategy labeling oligo tags on cell surface proteins. (B) Oligo-Cy5 labeling and
retention on Jurkat cell, T cell, and HSC surface in culture. (C) Oligo-Cy5 signal decay over time on Jurkat cell
surface within 48 hours. (D) Left, oligo labeling and retention on Jurkat cell surface within 48 hours. n = 4 technical
replicates; right, oligo labeling and retention on CD4+ T cell surface within 48 hours. n = 4 technical replicates. (E)
Bright-field microscope images of cultured CD4+ T cell with surface oligo tagging. Scale bar: 10 um.
99
4.3.2 Liposome conjugation to the cell surface
It was previously reported that liposomes can be conjugated on the cell surface and resist
endocytosis.
225
We thus tested if liposomes can work as “satellites” on the cell surface allowing
docking of oligo tags (Fig. 4-3A). We first fabricated liposomes with functional maleimide groups
with an adapted nanopore extrusion method. Dynamic light scattering (DLS) showed liposomes
with ~155 nm diameter (Fig. 4-3B) and -37 mV Zeta potential. 0.25% (molar ratio) DiD (a
fluorescent lipophilic tracer)-liposomes can be conjugated on Jurkat cell and T cell surface, as
indicated by flow cytometry and microscopy (Fig. 4-3C,D). We also confirmed that these
liposome surfaces can be further functionalized by thiol-PEG-tetrazine (thiol-PEG-TZ), as these
liposomes can be visualized by TCO-Cy5 (Fig. 4-3E). After 48 hours in culture, tetrazine
functionalized liposomes on the Jurkat cell surface can still be detected by TCO-Cy5 without
signal decay, indicating their stability under normal cell culture conditions (Fig. 4-3B). To exclude
the possibility that the liposomes on individual cells may exchange in pooled culture, we fabricated
0.25% DiO-liposomes and 0.25% DiD-liposomes, separately conjugated them on cells, and mix
the two populations together in culture. We did not observe obvious liposome exchange within 24
hours (Fig. 4-3F). We confirmed that conjugation of liposomes does not influence CD4+ T cell
activation and division (Fig. 4-3G).
100
Figure 0-3. Liposome conjugation on the cell surface.
(A) Schematic of the three-step strategy labeling oligo tags on cell-liposome. (B) Liposome size distribution. (C)
DiD-liposome conjugation on Jurkat cell surface, imaged by microscope. Scale bar: 10 um. (D) DiD-liposome
conjugation on CD4+ T cells, measured by flow cytometry. (E) TZ functionalized liposome conjugation on Jurkat
cell and retention after 48 hours, detected by TCO-Cy5. (F) Exchange of DiO-liposome and DiD-liposome on the
different cell surface in culture. (G) Bright-field microscope images of cultured liposome conjugated CD4+ T cell.
Scale bar: 10 um. (H) Division profile of liposome conjugated CD4+ T cells, tracked by CellTrace Violet.
101
4.3.3 Conjugation of oligo tags to cell-liposome
We successfully conjugated oligo tags on the cell-liposome, as indicated by qPCR results (Fig. 4-
4A). However, after 48 hours under culture, we found the oligo tags detached from cells and were
detected mostly in the supernatant of the medium (Fig. 4-4B). Previous reports indicated that the
negatively charged oligos can interact with serum (10% FBS) in the culture medium and detach
from liposomes. A time-lapse tracking of oligos on the cell surface and in supernatant indicated
that the oligos were sharply released in the first 3 hours (Fig. 4-4C). Moreover, we found that the
total amount of oligos on the cell surface and in the supernatant also decreased over time,
suggesting that oligos may interact with cells and be endocytosed (Fig. 4-4C). To test the influence
of cell culture medium on oligo release, we incubated cell-liposome-oligo tags in different
mediums for 3 hours and examined the supernatant. Indeed, Ca
2+
/Mg
2+
free PBS incubated cell-
liposome released significantly fewer oligos than the culture medium (Fig. 4-4D).
102
Figure 0-4. Oligo release from Jurkat cell-liposome under normal cell culture condition.
(A) Oligo amount per cell after oligo tagging on cell-liposome. Error bars: standard deviation. (B) Oligo amount on
cell-liposome and in supernatant 48 hours in culture after the initial tagging. (C) Timelapse tracking of oligo amount
on cell-liposome and in supernatant within 16 hours in culture. (D) Oligo amount released from cell-liposome to the
supernatant of different mediums after 3 hours incubation.
103
Figure 0-5. Proposed model of oligo-medium components interaction and oligo protection strategies.
(A) Proposed model of oligos on liposome interact with serum in medium and cell surface with cations as “glue”. (B)
By encapsulating oligos in liposome, they may be protected from interaction with serum or cell membrane. (C) By
modifying liposome surface with long PEG chains, conjugated oligos may be protected from interaction with serum
or cell membrane.
4.4 Discussion
In the presented study, we showed that direct conjugation of oligo tags on the cell surface is not a
suitable strategy for tracking cell division in normal conditions. Indeed, it has been reported that
the half-life of most cell surface glycoproteins is less than 24 hours, while only one-fifth of the
glycoproteins have a half-life of longer than 100 hours.
226
Such a fast turnover rate makes it
challenging to keep oligo tagging on the cell surface in a normal cell cycle. Meanwhile, we found
that surface conjugation of oligo tags inhibits T cell activation, growth, and division. While we did
not see such an effect when conjugating Cy5 using the same thiol sites (data not shown), it is likely
that oligos used in this experiment (70~80 nts long) are large molecules and result in hindering
104
effect on functional sites on cell surface proteins or receptors. While it has been reported that the
lipid-oligo conjugate insertion to the cell surface has an even shorter half-life,
222
we decided to
develop other indirect strategies to reach a longer retention time of oligo tags on the cell surface.
The liposome is a type of nanoparticle consisting of a bilayer of lipids. Its structure is similar to
the cell membrane and is thus cell-compatible and biodegradable. Liposome has been widely used
as drug or oligo (both DNA and RNA) carriers,
225
and its conjugation with oligo has been
thoroughly studied.
227
We elucidated that, maleimide functionalized liposomes can be conjugated
on the surface of different cell types, and that they can retain on the cell surface for 48 hours
without significant loss. While oligo tags can be conjugated further on cell-liposome, it appears to
be released under normal culture conditions, in both serum and serum-free medium. While such a
release does not happen in Ca
2+
/Mg
2+
free PBS, we proposed that the docking oligos on liposomes
need to be protected to retain longer for the application of cell division tracking. We propose the
model of oligo loss in during cell culture (Fig. 4-5 A) and accordingly two promising strategies:
(1) Encapsulating oligos in the liposomes before conjugating them onto the cell surface (Fig. 4-5
B) and (2) after oligo conjugation on liposome surface, functionalize the rest of the area with long
chain PEGs to prevent oligo interaction with serum or cell membrane. Further tests need to be
performed to elucidate whether these strategies will improve oligo stability in culture.
105
Chapter 5: Summary and future directions
Regenerative medicine is a fast-developing multi-disciplinary research area that holds the promise
of curing cell, tissue or organ damage in a way that was impossible in traditional medicine.
2
Stem
cells or cells derived from stem cells are the bricks of regenerative medicine. While a lot of
research efforts have been recently focused on induced pluripotent stem cells (iPSCs), adult stem
cells are another unignorable resource that has its advantages. Hematopoietic stem cell (HSC) is
the first adult stem cell type used in clinical transplantation and the most studied one. However,
only less than 50% of the patients who need an allogenic HSC transplantation can find a matching
donor. Ex vivo expansion of HSCs is believed to solve this dilemma. However, until today, there
has not been a protocol that can successfully expand human HSCs.
Unlike embryonic stem cells, of which the ex vivo expansion methods have been
established since over 40 years ago, HSCs reside in a more complex microenvironment (i.e. the
niche) in the bone marrow. It is thus likely that HSCs need more complex and precise fate
regulation. Scientists have spent tremendous efforts looking for the factors (physical environment,
cytokines, small molecules, etc.) that promote HSC expansion in vitro, yet have not been fully
successful. One of the hurdles preventing such success is the lack of methods to rapidly predict
the stemness of cells in culture. While transplantation assay in mouse models has been the gold
standard examining the functions of ex vivo expanded HSCs, this procedure needs months to
complete and is resource consuming, hindering the efficient screening or validation of candidate
factors. In practice, researchers usually use surface markers as a quick prediction in evaluating or
screening conditions that favor HSC expansion. While this method results in limited success, there
have been studies showing that HSC surface markers may change during ex vivo culture.
131
The
surface markers originally established for HSC isolation do not necessarily reflect cell functions
106
critical for stemness. Thus, markers established with functional reflection would inspire HSC
expansion strategies and be a more reliable metric to optimize the ex vivo culture conditions.
Metabolism in recent years has been shown to play critical roles in regulating stem cell fate
and reflecting the identity of HSCs. However, there has not been a well-established method to
track HSC metabolism in real-time with good details. A recently developed non-invasive imaging
methodology, fluorescence lifetime imaging microscopy (FLIM), can track the endogenous
fluorophores and metabolic coenzymes NAD(P)H and FAD to profile cellular metabolism in live
tissues and some cancer and stem cell types.
190,196
However, its application in identifying HSCs
has not been established. On the other hand, FLIM mainly profiles the binding status of NAD(P)H
to enzymes (fraction of bound NAD(P)H, change of fluorescence lifetime after binding, etc.), of
which the link with specific metabolic pathways has not been fully established, especially in stem
cells. Moreover, in most publications using FLIM to study single cell metabolism, FLIM
parameters were calculated as the average of all pixels in the cell, ignoring the subcellular
heterogeneity. In this dissertation, we developed a fluorescence lifetime imaging microscopy
(FLIM)-based method to identify HSCs from differentiated cells and track their metabolism with
temporal and spatial resolution. We addressed the above limitations by:
In aim 1, we derived a set of FLIM-based, non-invasive metabolic optical biomarkers of
HSCs at the single cell level, and elucidated the biological meaning of /metabolic pathways
correlated with each biomarker. Importantly, we found that the lifetime of enzyme-bound
NAD(P)H, tbound, reflects the activity of lactate dehydrogenase (LDH), an enzyme that regulates
glycolysis and is critical for HSC functions.
178
We showed that these biomarkers could monitor
the status of HSCs during ex vivo culture and expansion.
107
In aim 2, we defined 205 FLIM features from the metabolic optical biomarkers to
quantitively profile the single cell metabolism with subcellular resolution, and identified the ones
that track HSC differentiation by developing a bioinformatic workflow. With these features, we
further employed a regression model to score the “metabolic stemness” of individual cells in
culture. Since HSC division is the critical mechanism that regulates their commitment and
proliferation, we used this model to study metabolic switches during HSC symmetric division and
asymmetric division, and screened out two small molecules that favor HSC ex vivo expansion.
While the FLIM-derived metabolic optical biomarkers and features provide temporal and
spatial details of HSC metabolic stemness and division in a non-invasive manner, we hypothesize
that the transcriptome level analysis of paired daughter cells (PDCs) will provide complementary
information for understanding HSC division. In aim 3, we seek to develop an oligo barcoding
strategy to enable high-throughput analysis of the differential gene expression between paired
daughter cells and to understand the molecular mechanisms of symmetric and asymmetric division.
The establishment of such a protocol will favor a comprehensive understanding of HSC
proliferation and differentiation through division.
We expect the research presented in this dissertation to provide a rapid prediction method
of stemness and a new paradigm for analyzing single cell metabolism in the HSC and regenerative
medicine community. Future directions may include:
(1) Utilizing the 205 defined FLIM features and bioinformatic workflow to study other
stem cell types. FLIM has been employed to profile a few other stem cell types. However, most
such studies only extracted the information of different FLIM channels by averaging the
information from individual single cells. We have demonstrated that, however, different
subcellular regions and organelles have different metabolism, likely due to the
108
compartmentalization of NAD(P)H by the membrane system. HSC behaves differently than
committed cells when examined with spatial resolution. We reasoned that unique signatures may
also exist in other stem cell types. Our study provides a bioinformatic workflow that supports the
screening of these signatures from the comprehensively defined 205 features.
While the study presented here used mouse HSCs as a model, it has been reported that
human HSCs share similar metabolic characteristics with mouse HSCs. Cord blood-derived
CD34+ hematopoietic progenitor cells with low mitochondrial mass and membrane potential
exhibits reconstitution potential.
228
Nicotinamide riboside, an NAD+ precursor that promotes
mitochondrial clearance and reduces mitochondrial metabolism, improves the reconstitution
ability of both mouse HSCs and human CD34+ hematopoietic cells.
181
A major challenge of
human HSC study is that there has not been an established marker cocktail that can isolate the
highly purified HSC population. We propose using the 205 FLIM features to profile CD34+ human
hematopoietic cells and attribute them to different subpopulations using unsupervised clustering.
We can then profile the reconstitution ability of each subpopulation by colony forming unit (CFU)
assay and retrospectively examine the FLIM features that are important to identifying human HSCs.
(2) Utilizing the established FLIM-based methodology to high-throughputly screen more
factors that may favor HSC ex vivo expansion. Adding small molecules in culture has been a
popular way to expand HSCs, given its advantages in accurate dosage, timing, and reversible effect.
Previous studies found two small molecules, SR1 and UM171, that can significantly expand HSCs
in vitro.
112,229
The discovery of such molecules usually relies on high-throughput screening of
chemical compound libraries and surface markers are used as criteria to readily predict HSC
proliferation. However, studies showed that surface markers change dramatically during ex vivo
culture and are not reliable as functional profiling.
129,131
Our FLIM-based optical markers have the
109
advantage that they directly profile the metabolic features of HSCs, which intrinsically regulate
HSC fate and stemness. Thus, these markers should be more reliable in predicting HSC
proliferation in high-throughput screening experiments. We also expect to reveal more details of
the metabolic effect of small molecule treatment on HSCs.
(3) We also hold great hope that the cell surface barcoding strategy currently in
development will provide new perspectives on understanding the division of HSCs and other cell
types (e.g. other adult stem cells and T cells). Two possible directions can be explored once the
methodology is established, including:
(3.1) High-throughput screening of factors (e.g. transcription factors) that promote
symmetric expansion or asymmetric division. The mammalian genome contains ~1,500
transcription factors
230
and little is known about whether/ how these factors regulate cell division.
By combining high-content CRISPR screening technique
231
with the cell surface barcoding
strategy, we expect to comprehensively reveal what transcription factors regulate cell division
patterns. We will further validate if a list of these factors can promote HSC or T cell proliferation
in vitro.
(3.2) Multi-omics analysis utilizing the NGS-based platform, such as the correspondence
of gene expression profile with proteomics. CITE-seq provides a way of profiling transcriptome
and proteins of interest simultaneously.
223
Our cell surface barcoding strategy is designed to be
compatible with such multi-omics techniques. We thus expect to understand the mechanisms of
both transcriptome level and protein level regulation on cell division, and correspond the results
together.
110
References
1. Sampogna, G., Guraya, S. Y. & Forgione, A. Regenerative medicine: Historical roots and
potential strategies in modern medicine. J. Microsc. Ultrastruct. 3, 101–107 (2015).
2. Mao, A. S. & Mooney, D. J. Regenerative medicine: Current therapies and future directions.
Proc. Natl. Acad. Sci. 112, 14452–14459 (2015).
3. Daley, G. Q. & Scadden, D. T. Prospects for Stem Cell-Based Therapy. Cell 132, 544–548
(2008).
4. Halme, D. G. & Kessler, D. A. FDA Regulation of Stem-Cell–Based Therapies. N. Engl. J.
Med. 355, 1730–1735 (2006).
5. Mount, N. M., Ward, S. J., Kefalas, P. & Hyllner, J. Cell-based therapy technology
classifications and translational challenges. Philos. Trans. R. Soc. B Biol. Sci. 370, 20150017
(2015).
6. Guimarães, C. F., Gasperini, L., Marques, A. P. & Reis, R. L. The stiffness of living tissues
and its implications for tissue engineering. Nat. Rev. Mater. 5, 351–370 (2020).
7. Khademhosseini, A. & Langer, R. A decade of progress in tissue engineering. Nat. Protoc. 11,
1775–1781 (2016).
8. Altomare, L. et al. Biopolymer-based strategies in the design of smart medical devices and
artificial organs. Int. J. Artif. Organs 41, 337–359 (2018).
9. Wang, X. Bioartificial Organ Manufacturing Technologies. Cell Transplant. 13.
10. Shi, Y., Kirwan, P. & Livesey, F. J. Directed differentiation of human pluripotent stem
cells to cerebral cortex neurons and neural networks. Nat. Protoc. 7, 1836–1846 (2012).
11. Spence, J. R. et al. Directed differentiation of human pluripotent stem cells into intestinal
tissue in vitro. Nature 470, 105–109 (2011).
12. Vargas, M. E., Watanabe, J., Singh, S. J., Robinson, W. H. & Barres, B. A. Endogenous
antibodies promote rapid myelin clearance and effective axon regeneration after nerve injury.
Proc. Natl. Acad. Sci. 107, 11993–11998 (2010).
13. Rajagopal, K. & Hoeper, M. M. State of the Art: Bridging to lung transplantation using
artificial organ support technologies. J. Heart Lung Transplant. 35, 1385–1398 (2016).
14. Copelan, E. A. Hematopoietic Stem-Cell Transplantation. N. Engl. J. Med. 14 (2006).
15. Bick, D. et al. An online compendium of treatable genetic disorders. Am. J. Med. Genet.
C Semin. Med. Genet. 187, 48–54 (2021).
16. Jinek, M. et al. A Programmable Dual-RNA–Guided DNA Endonuclease in Adaptive
Bacterial Immunity. Science 337, 816–821 (2012).
111
17. Cong, L. et al. Multiplex Genome Engineering Using CRISPR/Cas Systems. Science 339,
819–823 (2013).
18. Hsu, P. D., Lander, E. S. & Zhang, F. Development and Applications of CRISPR-Cas9
for Genome Engineering. Cell 157, 1262–1278 (2014).
19. Doudna, J. A. & Charpentier, E. The new frontier of genome engineering with CRISPR-
Cas9. Science 346, 1258096 (2014).
20. Mali, P. et al. RNA-Guided Human Genome Engineering via Cas9. Science 339, 823–
826 (2013).
21. Ran, F. A. et al. Genome engineering using the CRISPR-Cas9 system. Nat. Protoc. 8,
2281–2308 (2013).
22. Qi, L. S. et al. Repurposing CRISPR as an RNA-Guided Platform for Sequence-Specific
Control of Gene Expression. Cell 152, 1173–1183 (2013).
23. Makarova, K. S. et al. An updated evolutionary classification of CRISPR–Cas systems.
Nat. Rev. Microbiol. 13, 722–736 (2015).
24. Halevy, T. & Urbach, A. Comparing ESC and iPSC—Based Models for Human Genetic
Disorders. J. Clin. Med. 3, 1146–1162 (2014).
25. Kubanov, S. I. et al. Development of 3D Bioprinting Technology Using Modified Natural
and Synthetic Hydrogels for Engineering Construction of Organs. Int. J. Pharm. Phytopharm.
Res. 9, 6.
26. Hulme, A. J., Maksour, S., St-Clair Glover, M., Miellet, S. & Dottori, M. Making
neurons, made easy: The use of Neurogenin-2 in neuronal differentiation. Stem Cell Rep. 17,
14–34 (2022).
27. Lancaster, M. A. & Knoblich, J. A. Organogenesis in a dish: Modeling development and
disease using organoid technologies. Science 345, 1247125 (2014).
28. Barak, M. et al. Human iPSC-Derived Neural Models for Studying Alzheimer’s Disease:
from Neural Stem Cells to Cerebral Organoids. Stem Cell Rev. Rep. 18, 792–820 (2022).
29. Kim, Y., Jeong, J. & Choi, D. Small-molecule-mediated reprogramming: a silver lining
for regenerative medicine. Exp. Mol. Med. 52, 213–226 (2020).
30. Lairson, L. L., Lyssiotis, C. A., Zhu, S. & Schultz, P. G. Small Molecule–Based
Approaches to Adult Stem Cell Therapies. Annu. Rev. Pharmacol. Toxicol. 53, 107–125
(2013).
31. Ng, A. H. M. et al. A comprehensive library of human transcription factors for cell fate
engineering. Nat. Biotechnol. (2020) doi:10.1038/s41587-020-0742-6.
112
32. Evans, M. J. & Kaufman, M. H. Establishment in culture of pluripotential cells from
mouse embryos. Nature 292, 154–156 (1981).
33. Martin, G. R. Isolation of a pluripotent cell line from early mouse embryos cultured in
medium conditioned by teratocarcinoma stem cells. Proc. Natl. Acad. Sci. 78, 7634–7638
(1981).
34. Thomson, J. A. et al. Embryonic Stem Cell Lines Derived from Human Blastocysts.
Science 282, 1145–1147 (1998).
35. Thomson, J. A. et al. Isolation of a primate embryonic stem cell line. Proc. Natl. Acad.
Sci. 92, 7844–7848 (1995).
36. Robertson, J. A. Human embryonic stem cell research: ethical and legal issues. Nat. Rev.
Genet. 2, 74–78 (2001).
37. Volarevic, V. et al. Ethical and Safety Issues of Stem Cell-Based Therapy. Int. J. Med.
Sci. 15, 36–45 (2018).
38. Wert, G. d. Human embryonic stem cells: research, ethics and policy. Hum. Reprod. 18,
672–682 (2003).
39. Klimanskaya, I., Chung, Y., Becker, S., Lu, S.-J. & Lanza, R. Derivation of human
embryonic stem cells from single blastomeres. Nat. Protoc. 2, 1963–1972 (2007).
40. Takahashi, K. & Yamanaka, S. Induction of Pluripotent Stem Cells from Mouse
Embryonic and Adult Fibroblast Cultures by Defined Factors. Cell 126, 663–676 (2006).
41. Takahashi, K. et al. Induction of Pluripotent Stem Cells from Adult Human Fibroblasts
by Defined Factors. Cell 131, 861–872 (2007).
42. The International Stem Cell Initiative. Characterization of human embryonic stem cell
lines by the International Stem Cell Initiative. Nat. Biotechnol. 25, 803–816 (2007).
43. Friedmann‐Morvinski, D. & Verma, I. M. Dedifferentiation and reprogramming: origins
of cancer stem cells. EMBO Rep. 15, 244–253 (2014).
44. Warren, L. & Lin, C. mRNA-Based Genetic Reprogramming. Mol. Ther. 27, 729–734
(2019).
45. Okita, K., Nakagawa, M., Hyenjong, H., Ichisaka, T. & Yamanaka, S. Generation of
Mouse Induced Pluripotent Stem Cells Without Viral Vectors. Science 322, 949–953 (2008).
46. Zhou, H. et al. Generation of Induced Pluripotent Stem Cells Using Recombinant
Proteins. Cell Stem Cell 4, 381–384 (2009).
113
47. Warren, L. et al. Highly Efficient Reprogramming to Pluripotency and Directed
Differentiation of Human Cells with Synthetic Modified mRNA. Cell Stem Cell 7, 618–630
(2010).
48. Hou, P. et al. Pluripotent Stem Cells Induced from Mouse Somatic Cells by Small-
Molecule Compounds. Science 341, 651–654 (2013).
49. Guan, J. et al. Chemical reprogramming of human somatic cells to pluripotent stem cells.
Nature 605, 325–331 (2022).
50. Martin, M. J., Muotri, A., Gage, F. & Varki, A. Human embryonic stem cells express an
immunogenic nonhuman sialic acid. Nat. Med. 11, 228–232 (2005).
51. Smith, A. G. et al. Inhibition of pluripotential embryonic stem cell differentiation by
purified polypeptides. Nature 336, 688–690 (1988).
52. Williams, R. L. et al. Myeloid leukaemia inhibitory factor maintains the developmental
potential of embryonic stem cells. Nature 336, 684–687 (1988).
53. Ohtsuka, S., Nakai-Futatsugi, Y. & Niwa, H. LIF signal in mouse embryonic stem cells.
JAK-STAT 4, 1–9 (2015).
54. Wulansari, N., Sulistio, Y. A., Darsono, W. H. W., Kim, C.-H. & Lee, S.-H. LIF
maintains mouse embryonic stem cells pluripotency by modulating TET1 and JMJD2 activity
in a JAK2-dependent manner. Stem Cells 39, 750–760 (2021).
55. Ying, Q.-L. et al. The ground state of embryonic stem cell self-renewal. Nature 453,
519–523 (2008).
56. Holt, M. LIF/STAT3 Signaling Fails to Maintain Self-Renewal of Human Embryonic
Stem Cells. 10.
57. Richards, M., Fong, C.-Y., Chan, W.-K., Wong, P.-C. & Bongso, A. Human feeders
support prolonged undifferentiated growth of human inner cell masses and embryonic stem
cells. Nat. Biotechnol. 20, 933–936 (2002).
58. Braam, S. R. et al. Feeder-free culture of human embryonic stem cells in conditioned
medium for efficient genetic modification. 9.
59. Amit, M., Shariki, C., Margulets, V. & Itskovitz-Eldor, J. Feeder Layer- and Serum-Free
Culture of Human Embryonic Stem Cells1. Biol. Reprod. 70, 837–845 (2004).
60. Chen, G. et al. Chemically defined conditions for human iPSC derivation and culture.
Nat. Methods 8, 424–429 (2011).
61. Zhang, J. et al. Functional Cardiomyocytes Derived From Human Induced Pluripotent
Stem Cells. Circ. Res. 104, (2009).
114
62. Raya, Á. et al. Disease-corrected haematopoietic progenitors from Fanconi anaemia
induced pluripotent stem cells. Nature 460, 53–59 (2009).
63. Dimos, J. T. et al. Induced Pluripotent Stem Cells Generated from Patients with ALS Can
Be Differentiated into Motor Neurons. 321, 5 (2008).
64. Wichterle, H., Lieberam, I., Porter, J. A. & Jessell, T. M. Directed Differentiation of
Embryonic Stem Cells into Motor Neurons. Cell 110, 385–397 (2002).
65. Li, X.-J. et al. Specification of motoneurons from human embryonic stem cells. Nat.
Biotechnol. 23, 215–221 (2005).
66. Zhang, Y. et al. Rapid Single-Step Induction of Functional Neurons from Human
Pluripotent Stem Cells. Neuron 78, 785–798 (2013).
67. Polo, J. M. et al. Cell type of origin influences the molecular and functional properties of
mouse induced pluripotent stem cells. Nat. Biotechnol. 28, 848–855 (2010).
68. Pittenger, M. F. et al. Multilineage Potential of Adult Human Mesenchymal Stem Cells.
Science 284, 143–147 (1999).
69. Zhao, C., Deng, W. & Gage, F. H. Mechanisms and Functional Implications of Adult
Neurogenesis. Cell 132, 645–660 (2008).
70. Nowak, J. A., Polak, L., Pasolli, H. A. & Fuchs, E. Hair Follicle Stem Cells Are
Specified and Function in Early Skin Morphogenesis. Cell Stem Cell 3, 33–43 (2008).
71. Barker, N. et al. Identification of stem cells in small intestine and colon by marker gene
Lgr5. Nature 449, 1003–1007 (2007).
72. De Bari, C., Dell’Accio, F., Tylzanowski, P. & Luyten, F. P. Multipotent mesenchymal
stem cells from adult human synovial membrane. Arthritis Rheum. 44, 1928–1942 (2001).
73. Uccelli, A., Moretta, L. & Pistoia, V. Mesenchymal stem cells in health and disease. Nat.
Rev. Immunol. 8, 726–736 (2008).
74. Campagnoli, C. et al. Identification of mesenchymal stem/progenitor cells in human first-
trimester fetal blood, liver, and bone marrow. Blood 98, 2396–2402 (2001).
75. da Silva Meirelles, L., Caplan, A. I. & Nardi, N. B. In Search of the In Vivo Identity of
Mesenchymal Stem Cells. Stem Cells 26, 2287–2299 (2008).
76. Meirelles, L. da S., Chagastelles, P. C. & Nardi, N. B. Mesenchymal stem cells reside in
virtually all post-natal organs and tissues. J. Cell Sci. 119, 2204–2213 (2006).
77. Schu, S. et al. Immunogenicity of allogeneic mesenchymal stem cells. J. Cell. Mol. Med.
16, 2094–2103 (2012).
115
78. Shi, Y. et al. How mesenchymal stem cells interact with tissue immune responses. Trends
Immunol. 33, 136–143 (2012).
79. Newman, R., Yoo, D., LeRoux, M. & Danilkovitch-Miagkova, A. Treatment of
Inflammatory Diseases with Mesenchymal Stem Cells. Inflamm. Allergy - Drug Targets 8,
110–123 (2009).
80. Lange, C. et al. Accelerated and safe expansion of human mesenchymal stromal cells in
animal serum-free medium for transplantation and regenerative medicine. J. Cell. Physiol.
213, 18–26 (2007).
81. Cyranoski, D. Canada approves stem cell product. Nat. Biotechnol. 30, 571–571 (2012).
82. Gross, C. G. Neurogenesis in the adult brain: death of a dogma. Nat. Rev. Neurosci. 1,
67–73 (2000).
83. Arvidsson, A., Collin, T., Kirik, D., Kokaia, Z. & Lindvall, O. Neuronal replacement
from endogenous precursors in the adult brain after stroke. Nat. Med. 8, 963–970 (2002).
84. Reynolds, B. A. & Weiss, S. Generation of Neurons and Astrocytes from Isolated Cells
of the Adult Mammalian Central Nervous System. Science 255, 1707–1710 (1992).
85. Reynolds, B. A. & Rietze, R. L. Neural stem cells and neurospheres—re-evaluating the
relationship. Nat. Methods 2, 333–336 (2005).
86. Chotinantakul, K. & Leeanansaksiri, W. Hematopoietic Stem Cell Development, Niches,
and Signaling Pathways. Bone Marrow Res. 2012, e270425 (2012).
87. Wagner, J. E. & Gluckman, E. Umbilical Cord Blood Transplantation: The First 20
Years. Semin. Hematol. 47, 3–12 (2010).
88. Watz, E. et al. Quality of the hematopoietic stem cell graft affects the clinical outcome of
allogeneic stem cell transplantation: GRAFT QUALITY INFLUENCES TRM. Transfusion
(Paris) 55, 2339–2350 (2015).
89. Slukvin, I. I. Hematopoietic specification from human pluripotent stem cells: current
advances and challenges toward de novo generation of hematopoietic stem cells. Blood 122,
4035–4046 (2013).
90. Boulais, P. E. & Frenette, P. S. Making sense of hematopoietic stem cell niches. Blood
125, 2621–2629 (2015).
91. Ivanovs, A. et al. Human haematopoietic stem cell development: from the embryo to the
dish. Development 144, 2323–2337 (2017).
92. Daniel, M. G., Lemischka, I. R. & Moore, K. Converting cell fates: generating
hematopoietic stem cells de novo via transcription factor reprogramming: Transcription factor
induction of hematopoiesis. Ann. N. Y. Acad. Sci. 1370, 24–35 (2016).
116
93. Ackermann, M., Liebhaber, S., Klusmann, J. & Lachmann, N. Lost in translation:
pluripotent stem cell‐derived hematopoiesis. EMBO Mol. Med. 7, 1388–1402 (2015).
94. Wahlster, L. & Daley, G. Q. Progress towards generation of human haematopoietic stem
cells. Nat. Cell Biol. 18, 1111–1117 (2016).
95. Sugimura, R. et al. Haematopoietic stem and progenitor cells from human pluripotent
stem cells. Nature 545, 432–438 (2017).
96. Demirci, S., Leonard, A. & Tisdale, J. F. Hematopoietic stem cells from pluripotent stem
cells: Clinical potential, challenges, and future perspectives. Stem Cells Transl. Med. 9, 1549–
1557 (2020).
97. Kaufman, D. S., Hanson, E. T., Lewis, R. L., Auerbach, R. & Thomson, J. A.
Hematopoietic colony-forming cells derived from human embryonic stem cells. Proc. Natl.
Acad. Sci. 98, 10716–10721 (2001).
98. Zambidis, E. T., Peault, B., Park, T. S., Bunz, F. & Civin, C. I. Hematopoietic
differentiation of human embryonic stem cells progresses through sequential
hematoendothelial, primitive, and definitive stages resembling human yolk sac development.
Blood 106, 860–870 (2005).
99. Chadwick, K. et al. Cytokines and BMP-4 promote hematopoietic differentiation of
human embryonic stem cells. Blood 102, 906–915 (2003).
100. Kennedy, sr et al. T Lymphocyte Potential Marks the Emergence of Definitive
Hematopoietic Progenitors in Human Pluripotent Stem Cell Differentiation Cultures. Cell
Rep. 2, 1722–1735 (2012).
101. Riddell, J. et al. Reprogramming Committed Murine Blood Cells to Induced
Hematopoietic Stem Cells with Defined Factors. Cell 157, 549–564 (2014).
102. Jang, Y.-Y. & Sharkis, S. J. A low level of reactive oxygen species selects for primitive
hematopoietic stem cells that may reside in the low-oxygenic niche. Blood 110, 3056–3063
(2007).
103. Blanpain, C., Lowry, W. E., Geoghegan, A., Polak, L. & Fuchs, E. Self-Renewal,
Multipotency, and the Existence of Two Cell Populations within an Epithelial Stem Cell
Niche. Cell 118, 635–648 (2004).
104. Corominas-Murtra, B. et al. Stem cell lineage survival as a noisy competition for niche
access. Proc. Natl. Acad. Sci. 201921205 (2020) doi:10.1073/pnas.1921205117.
105. Desai, J., Steiger, S. & Anders, H.-J. Molecular Pathophysiology of Gout. Trends Mol.
Med. 23, 756–768 (2017).
106. Ema, H., Takano, H., Sudo, K. & Nakauchi, H. In Vitro Self-Renewal Division of
Hematopoietic Stem Cells. J. Exp. Med. 192, 1281–1288 (2000).
117
107. Zhang, C. C. et al. Angiopoietin-like proteins stimulate ex vivo expansion of
hematopoietic stem cells. Nat. Med. 12, 240–245 (2006).
108. Wilkinson, A. C. et al. Long-term ex vivo haematopoietic-stem-cell expansion allows
nonconditioned transplantation. Nature 571, 117–121 (2019).
109. Bhatia, M., Wang, J. C. Y., Kapp, U., Bonnet, D. & Dick, J. E. Purification of primitive
human hematopoietic cells capable of repopulating immune-deficient mice. Proc. Natl. Acad.
Sci. 94, 5320–5325 (1997).
110. Miharada, K., Sigurdsson, V. & Karlsson, S. Dppa5 Improves Hematopoietic Stem Cell
Activity by Reducing Endoplasmic Reticulum Stress. Cell Rep. 7, 1381–1392 (2014).
111. Antonchuk, J., Sauvageau, G. & Humphries, R. K. HOXB4-Induced Expansion of Adult
Hematopoietic Stem Cells Ex Vivo. Cell 109, 39–45 (2002).
112. Fares, I. et al. Pyrimidoindole derivatives are agonists of human hematopoietic stem cell
self-renewal. Science 345, 1509–1512 (2014).
113. Wagner, J. E. et al. Phase I/II Trial of StemRegenin-1 Expanded Umbilical Cord Blood
Hematopoietic Stem Cells Supports Testing as a Stand-Alone Graft. Cell Stem Cell 18, 144–
155 (2016).
114. Florian, M. C. et al. Cdc42 Activity Regulates Hematopoietic Stem Cell Aging and
Rejuvenation. Cell Stem Cell 10, 520–530 (2012).
115. Morrison, S. J. & Kimble, J. Asymmetric and symmetric stem-cell divisions in
development and cancer. Nature 441, 1068–1074 (2006).
116. Ito, K. et al. A PML-PPAR-δ pathway for fatty acid oxidation regulates hematopoietic
stem cell maintenance. Nat. Med. 18, 1350–1358 (2012).
117. Ito, K. et al. Self-renewal of a purified Tie2+ hematopoietic stem cell population relies on
mitochondrial clearance. Science 354, 1156–1160 (2016).
118. Florian, M. C. et al. Aging alters the epigenetic asymmetry of HSC division. PLOS Biol.
16, e2003389 (2018).
119. Loeffler, D. et al. Mouse and human HSPC immobilization in liquid culture by CD43- or
CD44-antibody coating. Blood 131, 1425–1429 (2018).
120. Pham, K., Sacirbegovic, F. & Russell, S. M. Polarized Cells, Polarized Views:
Asymmetric Cell Division in Hematopoietic Cells. Front. Immunol. 5, (2014).
121. Loeffler, D. et al. Asymmetric lysosome inheritance predicts activation of haematopoietic
stem cells. Nature 1–4 (2019) doi:10.1038/s41586-019-1531-6.
118
122. Ting, S. B. et al. Asymmetric segregation and self-renewal of hematopoietic stem and
progenitor cells with endocytic Ap2a2. Blood 119, 2510–2522 (2012).
123. Katajisto, P. et al. Asymmetric apportioning of aged mitochondria between daughter cells
is required for stemness. Science 348, 340–343 (2015).
124. Hinge, A. et al. Asymmetrically Segregated Mitochondria Provide Cellular Memory of
Hematopoietic Stem Cell Replicative History and Drive HSC Attrition. Cell Stem Cell 26,
420-430.e6 (2020).
125. Chen, J. Y. et al. Hoxb5 marks long-term haematopoietic stem cells and reveals a
homogenous perivascular niche. Nature 530, 223–227 (2016).
126. Chan, E. M. et al. Live cell imaging distinguishes bona fide human iPS cells from
partially reprogrammed cells. Nat. Biotechnol. 27, 1033–1037 (2009).
127. Zhu, Y. et al. Deep learning-based predictive identification of neural stem cell
differentiation. Nat. Commun. 12, 2614 (2021).
128. Buggenthin, F. et al. Prospective identification of hematopoietic lineage choice by deep
learning. Nat. Methods 14, 403–406 (2017).
129. Fares, I. et al. EPCR expression marks UM171-expanded CD34+ cord blood stem cells.
Blood 129, 3344–3351 (2017).
130. Notta, F. et al. Isolation of Single Human Hematopoietic Stem Cells Capable of Long-
Term Multilineage Engraftment. Science 333, 218–221 (2011).
131. Zhang, C. C. & Lodish, H. F. Murine hematopoietic stem cells change their surface
phenotype during ex vivo expansion. Blood 105, 4314–4320 (2005).
132. Kaliyappan, K., Palanisamy, M., Govindarajan, R. & Duraiyan, J. Microarray and its
applications. J. Pharm. Bioallied Sci. 4, 310 (2012).
133. Lan, F., Demaree, B., Ahmed, N. & Abate, A. R. Single-cell genome sequencing at ultra-
high-throughput with microfluidic droplet barcoding. Nat. Biotechnol. advance online
publication, (2017).
134. Macosko, E. Z. et al. Highly Parallel Genome-wide Expression Profiling of Individual
Cells Using Nanoliter Droplets. Cell 161, 1202–1214 (2015).
135. Zhang, X. et al. Comparative Analysis of Droplet-Based Ultra-High-Throughput Single-
Cell RNA-Seq Systems. Mol. Cell 73, 130-142.e5 (2019).
136. Zilionis, R. et al. Single-cell barcoding and sequencing using droplet microfluidics. Nat.
Protoc. 12, 44–73 (2017).
119
137. Levsky, J. M. & Singer, R. H. Fluorescence in situ hybridization: past, present and future.
J. Cell Sci. 116, 2833–2838 (2003).
138. Moses, L. & Pachter, L. Museum of spatial transcriptomics. Nat. Methods 19, 534–546
(2022).
139. Marx, V. Method of the Year: spatially resolved transcriptomics. Nat. Methods 18, 9–14
(2021).
140. Chen, A. et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA
nanoball-patterned arrays. Cell 185, 1777-1792.e21 (2022).
141. Jaenisch, R. & Bird, A. Epigenetic regulation of gene expression: how the genome
integrates intrinsic and environmental signals. Nat. Genet. 33, 245–254 (2003).
142. Umer, M. & Herceg, Z. Deciphering the Epigenetic Code: An Overview of DNA
Methylation Analysis Methods. Antioxid. Redox Signal. 18, 1972–1986 (2013).
143. Mehrmohamadi, M., Sepehri, M. H., Nazer, N. & Norouzi, M. R. A Comparative
Overview of Epigenomic Profiling Methods. Front. Cell Dev. Biol. 9, 714687 (2021).
144. Bibikova, M. et al. Human embryonic stem cells have a unique epigenetic signature.
Genome Res. 16, 1075–1083 (2006).
145. Hanna, J. et al. Human embryonic stem cells with biological and epigenetic
characteristics similar to those of mouse ESCs. Proc. Natl. Acad. Sci. 107, 9222–9227 (2010).
146. Atkinson, S. & Armstrong, L. Epigenetics in embryonic stem cells: regulation of
pluripotency and differentiation. Cell Tissue Res. 331, 23–29 (2008).
147. Haasters, F. et al. Morphological and immunocytochemical characteristics indicate the
yield of early progenitors and represent a quality control for human mesenchymal stem cell
culturing. J. Anat. 214, 759–767 (2009).
148. Cabezas-Wallscheid, N. et al. Identification of Regulatory Networks in HSCs and Their
Immediate Progeny via Integrated Proteome, Transcriptome, and DNA Methylome Analysis.
Cell Stem Cell 15, 507–522 (2014).
149. Hidalgo San Jose, L. et al. Modest Declines in Proteome Quality Impair Hematopoietic
Stem Cell Self-Renewal. Cell Rep. 30, 69-80.e6 (2020).
150. García-Prat, L., Sousa-Victor, P. & Muñoz-Cánoves, P. Proteostatic and Metabolic
Control of Stemness. Cell Stem Cell 20, 593–608 (2017).
151. Hansson, J. et al. Highly Coordinated Proteome Dynamics during Reprogramming of
Somatic Cells to Pluripotency. Cell Rep. 2, 1579–1592 (2012).
120
152. Levchenko, A. Proteomics takes stem cell analyses to another level. Nat. Biotechnol. 23,
828–830 (2005).
153. Page, K. M. et al. Total Colony-Forming Units Are a Strong, Independent Predictor of
Neutrophil and Platelet Engraftment after Unrelated Umbilical Cord Blood Transplantation: A
Single-Center Analysis of 435 Cord Blood Transplants. Biol. Blood Marrow Transplant. 17,
1362–1374 (2011).
154. Sullivan, S. et al. Quality control guidelines for clinical-grade human induced pluripotent
stem cell lines. Regen. Med. 13, 859–866 (2018).
155. Steeg, R. et al. EBiSC best practice: How to ensure optimal generation, qualification, and
distribution of iPSC lines. Stem Cell Rep. 16, 1853–1867 (2021).
156. Kawamata, S., Kanemura, H., Sakai, N., Takahashi, M. & Go, M. Design of a
Tumorigenicity Test for Induced Pluripotent Stem Cell (iPSC)-Derived Cell Products. J. Clin.
Med. 4, 159–171 (2015).
157. Wuputra, K. et al. Prevention of tumor risk associated with the reprogramming of human
pluripotent stem cells. J. Exp. Clin. Cancer Res. 39, 100 (2020).
158. Itakura, G. et al. Fail-Safe System against Potential Tumorigenicity after Transplantation
of iPSC Derivatives. Stem Cell Rep. 8, 673–684 (2017).
159. Carrillo-Ávila, J. A., Catalina, P. & Aguilar-Quesada, R. Quality Control of Cell Lines
Using DNA as Target. DNA 2, 44–55 (2022).
160. Dittmar, K. E. J. et al. Quality of Cell Products: Authenticity, Identity, Genomic Stability
and Status of Differentiation. Transfus. Med. Hemotherapy 37, 2–2 (2010).
161. Jang, T. H. et al. Cryopreservation and its clinical applications. Integr. Med. Res. 6, 12–
18 (2017).
162. Berz, D., McCormack, E. M., Winer, E. S., Colvin, G. A. & Quesenberry, P. J.
Cryopreservation of Hematopoietic Stem Cells. Am. J. Hematol. 82, 463–472 (2007).
163. Yong, K. W. et al. Cryopreservation of Human Mesenchymal Stem Cells for Clinical
Applications: Current Methods and Challenges. Biopreservation Biobanking 13, 231–239
(2015).
164. Fernyhough, L. J., Buchan, V. A., McArthur, L. T. & Hock, B. D. Relative recovery of
haematopoietic stem cell products after cryogenic storage of up to 19 years. Bone Marrow
Transplant. 48, 32–35 (2013).
165. Broxmeyer, H. E. et al. High-efficiency recovery of functional hematopoietic progenitor
and stem cells from human cord blood cryopreserved for 15 years. Proc. Natl. Acad. Sci. 100,
645–650 (2003).
121
166. Takubo, K. et al. Regulation of Glycolysis by Pdk Functions as a Metabolic Checkpoint
for Cell Cycle Quiescence in Hematopoietic Stem Cells. Cell Stem Cell 12, 49–61 (2013).
167. Ito, K. & Ito, K. Hematopoietic stem cell fate through metabolic control. Exp. Hematol.
64, 1–11 (2018).
168. Ito, K. et al. Reactive oxygen species act through p38 MAPK to limit the lifespan of
hematopoietic stem cells. Nat. Med. 12, 446–451 (2006).
169. Qian, P. et al. The Dlk1-Gtl2 Locus Preserves LT-HSC Function by Inhibiting the PI3K-
mTOR Pathway to Restrict Mitochondrial Metabolism. Cell Stem Cell 18, 214–228 (2016).
170. Simsek, T. et al. The Distinct Metabolic Profile of Hematopoietic Stem Cells Reflects
Their Location in a Hypoxic Niche. Cell Stem Cell 7, 380–390 (2010).
171. Ansó, E. et al. The mitochondrial respiratory chain is essential for haematopoietic stem
cell function. Nat. Cell Biol. 19, 614–625 (2017).
172. Oburoglu, L. et al. Glucose and Glutamine Metabolism Regulate Human Hematopoietic
Stem Cell Lineage Specification. Cell Stem Cell 15, 169–184 (2014).
173. Yamamoto, R. et al. Clonal Analysis Unveils Self-Renewing Lineage-Restricted
Progenitors Generated Directly from Hematopoietic Stem Cells. Cell 154, 1112–1126 (2013).
174. Haas, S., Trumpp, A. & Milsom, M. D. Causes and Consequences of Hematopoietic
Stem Cell Heterogeneity. Cell Stem Cell 22, 627–638 (2018).
175. Vannini, N. et al. Specification of haematopoietic stem cell fate via modulation of
mitochondrial activity. Nat. Commun. 7, 13125 (2016).
176. Nakamura-Ishizu, A. et al. Thrombopoietin Metabolically Primes Hematopoietic Stem
Cells to Megakaryocyte-Lineage Differentiation. Cell Rep. 25, 1772-1785.e6 (2018).
177. Yu, W.-M. et al. Metabolic Regulation by the Mitochondrial Phosphatase PTPMT1 Is
Required for Hematopoietic Stem Cell Differentiation. Cell Stem Cell 12, 62–74 (2013).
178. Wang, Y.-H. et al. Cell-State-Specific Metabolic Dependency in Hematopoiesis and
Leukemogenesis. Cell 158, 1309–1323 (2014).
179. Jang, C., Chen, L. & Rabinowitz, J. D. Metabolomics and Isotope Tracing. Cell 173,
822–837 (2018).
180. Rimmelé, P. et al. Mitochondrial metabolism in hematopoietic stem cells requires
functional FOXO3. EMBO Rep. 16, 1164–1176 (2015).
181. Vannini, N. et al. The NAD-Booster Nicotinamide Riboside Potently Stimulates
Hematopoiesis through Increased Mitochondrial Clearance. Cell Stem Cell 24, 405-418.e7
(2019).
122
182. Zhu, C. et al. Near-simultaneous intravital microscopy of glucose uptake and
mitochondrial membrane potential, key endpoints that reflect major metabolic axes in cancer.
Sci. Rep. 7, 13772 (2017).
183. Goodell, M. A., Brose, K., Paradis, G., Conner, A. S. & Mulligan, R. C. Isolation and
functional properties of murine hematopoietic stem cells that are replicating in vivo. J. Exp.
Med. 183, 1797–1806 (1996).
184. Park, B., Yoo, K. H. & Kim, C. Hematopoietic stem cell expansion and generation: the
ways to make a breakthrough. Blood Res. 50, 194 (2015).
185. Ying, W. NAD+/NADH and NADP+/NADPH in Cellular Functions and Cell Death:
Regulation and Biological Consequences. Antioxid. Redox Signal. 10, 179–206 (2007).
186. Lakowicz, J. R., Szmacinski, H., Nowaczyk, K. & Johnson, M. L. Fluorescence lifetime
imaging of free and protein-bound NADH. Proc. Natl. Acad. Sci. 89, 1271–1275 (1992).
187. Heart, D. F. P., Maeda-yorita, K. & Aki, K. Effect of Nicotinamide Adenine Dinucleotide
on the Oxidation-Reduction Potentials of Lipoamide. (1984).
188. Hou, J. et al. Correlating two-photon excited fluorescence imaging of breast cancer
cellular redox state with seahorse flux analysis of normalized cellular oxygen consumption. J.
Biomed. Opt. 21, 060503 (2016).
189. Quinn, K. P. et al. Quantitative metabolic imaging using endogenous fluorescence to
detect stem cell differentiation. Sci. Rep. 3, srep03432 (2013).
190. Stringari, C., Sierra, R., Donovan, P. J. & Gratton, E. Label-free separation of human
embryonic stem cells and their differentiating progenies by phasor fluorescence lifetime
microscopy. J. Biomed. Opt. 17, 0460121–04601211 (2012).
191. Yaseen, M. A. et al. Fluorescence lifetime microscopy of NADH distinguishes alterations
in cerebral metabolism in vivo. Biomed. Opt. Express 8, 2368–2385 (2017).
192. Ogikubo, S. et al. Intracellular pH Sensing Using Autofluorescence Lifetime Microscopy.
J. Phys. Chem. B 115, 10385–10390 (2011).
193. García-Prat, L. et al. Autophagy maintains stemness by preventing senescence. Nature
529, 37–42 (2016).
194. Baumgartner, C. et al. An ERK-Dependent Feedback Mechanism Prevents
Hematopoietic Stem Cell Exhaustion. Cell Stem Cell 22, 879-892.e6 (2018).
195. Zhang, T. et al. BNIP3 Protein Suppresses PINK1 Kinase Proteolytic Cleavage to
Promote Mitophagy. J. Biol. Chem. 291, 21616–21629 (2016).
123
196. Stringari, C. et al. Phasor approach to fluorescence lifetime microscopy distinguishes
different metabolic states of germ cells in a live tissue. Proc. Natl. Acad. Sci. 108, 13582–
13587 (2011).
197. Daud, H., Browne, S., Al-Majmaie, R., Murphy, W. & Al-Rubeai, M. Metabolic profiling
of hematopoietic stem and progenitor cells during proliferation and differentiation into red
blood cells. New Biotechnol. 33, 179–186 (2016).
198. Sheldon, C., Cheng, Y. M. & Church, J. Concurrent measurements of the free cytosolic
concentrations of H+ and Na+ ions with fluorescent indicators. Pflugers Arch. 449, 307–318
(2004).
199. Webb, B. A., Chimenti, M., Jacobson, M. P. & Barber, D. L. Dysregulated pH: a perfect
storm for cancer progression. Nat. Rev. Cancer 11, 671–677 (2011).
200. Dechant, R. et al. Cytosolic pH is a second messenger for glucose and regulates the PKA
pathway through V-ATPase. EMBO J. 29, 2515–2526 (2010).
201. Challen, G. A., Boles, N., Lin, K.-Y. K. & Goodell, M. A. Mouse hematopoietic stem
cell identification and analysis. Cytometry A 75A, 14–24 (2009).
202. Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20, 273–297 (1995).
203. Liu, X. et al. Maintenance of mouse hematopoietic stem cells ex vivo by reprogramming
cellular metabolism. Blood 125, 1562–1565 (2015).
204. Huang, J., Nguyen-McCarty, M., Hexner, E. O., Danet-Desnoyers, G. & Klein, P. S.
Maintenance of hematopoietic stem cells through regulation of Wnt and mTOR pathways.
Nat. Med. 18, 1778–1785 (2012).
205. Guo, B., Huang, X., Lee, M. R., Lee, S. A. & Broxmeyer, H. E. Antagonism of PPAR-γ
signaling expands human hematopoietic stem and progenitor cells by enhancing glycolysis.
Nat. Med. (2018) doi:10.1038/nm.4477.
206. Takizawa, H., Regoes, R. R., Boddupalli, C. S., Bonhoeffer, S. & Manz, M. G. Dynamic
variation in cycling of hematopoietic stem cells in steady state and inflammation. J. Exp. Med.
208, 273–284 (2011).
207. Spencer, J. A. et al. Direct measurement of local oxygen concentration in the bone
marrow of live animals. Nature 508, 269–273 (2014).
208. Pietras, E. M. et al. Chronic interleukin-1 exposure drives haematopoietic stem cells
towards precocious myeloid differentiation at the expense of self-renewal. Nat. Cell Biol. 18,
607–618 (2016).
209. Habib, S. J. et al. A Localized Wnt Signal Orients Asymmetric Stem Cell Division in
Vitro. Science 339, 1445–1448 (2013).
124
210. Pattappa, G., Heywood, H. K., de Bruijn, J. D. & Lee, D. A. The metabolism of human
mesenchymal stem cells during proliferation and differentiation. J. Cell. Physiol. 226, 2562–
2570 (2011).
211. Cavallucci, V., Fidaleo, M. & Pani, G. Neural Stem Cells and Nutrients: Poised Between
Quiescence and Exhaustion. Trends Endocrinol. Metab. 27, 756–769 (2016).
212. Suda, T., Takubo, K. & Semenza, G. L. Metabolic Regulation of Hematopoietic Stem
Cells in the Hypoxic Niche. Cell Stem Cell 9, 298–310 (2011).
213. Digman, M. A., Caiolfa, V. R., Zamai, M. & Gratton, E. The Phasor Approach to
Fluorescence Lifetime Imaging Analysis. Biophys. J. 94, L14–L16 (2008).
214. Skala, M. C. et al. In vivo multiphoton microscopy of NADH and FAD redox states,
fluorescence lifetimes, and cellular morphology in precancerous epithelia. Proc. Natl. Acad.
Sci. 104, 19494–19499 (2007).
215. Zhou, H. et al. Non-invasive Optical Biomarkers Distinguish and Track the Metabolic
Status of Single Hematopoietic Stem Cells. iScience 23, 100831 (2020).
216. Nguyen, L. et al. Functional compensation between hematopoietic stem cell clones in
vivo. EMBO Rep. 19, e45702 (2018).
217. Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression
data analysis. Genome Biol. 19, 15 (2018).
218. Lewis, C. A. et al. Tracing Compartmentalized NADPH Metabolism in the Cytosol and
Mitochondria of Mammalian Cells. Mol. Cell 55, 253–263 (2014).
219. Fawcett, T. An introduction to ROC analysis. Pattern Recognit. Lett. 27, 861–874 (2006).
220. Gole, J. et al. Massively parallel polymerase cloning and genome sequencing of single
cells using nanoliter microwells. Nat. Biotechnol. 31, 1126–1132 (2013).
221. Lu, R., Neff, N. F., Quake, S. R. & Weissman, I. L. Tracking single hematopoietic stem
cells in vivo using high-throughput sequencing in conjunction with viral genetic barcoding.
Nat. Biotechnol. 29, 928–933 (2011).
222. McGinnis, C. S. et al. MULTI-seq: sample multiplexing for single-cell RNA sequencing
using lipid-tagged indices. Nat. Methods 16, 619–626 (2019).
223. Stoeckius, M. et al. Simultaneous epitope and transcriptome measurement in single cells.
Nat. Methods 14, 865–868 (2017).
224. Sahaf, B., Heydari, K., Herzenberg, L. A. & Herzenberg, L. A. Lymphocyte surface thiol
levels. Proc. Natl. Acad. Sci. 100, 4001–4005 (2003).
125
225. Stephan, M. T., Moon, J. J., Um, S. H., Bershteyn, A. & Irvine, D. J. Therapeutic cell
engineering with surface-conjugated synthetic nanoparticles. Nat. Med. 16, 1035–1041
(2010).
226. Xiao, H. & Wu, R. Quantitative investigation of human cell surface N-glycoprotein
dynamics. Chem. Sci. 8, 268–277 (2017).
227. Lopez, A. & Liu, J. DNA Oligonucleotide-Functionalized Liposomes: Bioconjugate
Chemistry, Biointerfaces, and Applications. Langmuir 34, 15000–15013 (2018).
228. Romero-Moya, D. et al. Cord blood-derived CD34+ hematopoietic cells with low
mitochondrial mass are enriched in hematopoietic repopulating stem cell function.
Haematologica 98, 1022–1029 (2013).
229. Boitano, A. E. et al. Aryl Hydrocarbon Receptor Antagonists Promote the Expansion of
Human Hematopoietic Stem Cells. Science 329, 1345–1348 (2010).
230. Vaquerizas, J. M., Kummerfeld, S. K., Teichmann, S. A. & Luscombe, N. M. A census of
human transcription factors: function, expression and evolution. Nat. Rev. Genet. 10, 252–263
(2009).
231. Bock, C. et al. High-content CRISPR screening. Nat. Rev. Methods Primer 2, 8 (2022).
Abstract (if available)
Abstract
Hematopoietic stem cell (HSC) transplantation is an effective cure for various human diseases. However, its clinical applications have been hindered by the lack of HSC sources. HSC ex vivo biomanufacturing has attracted significant research efforts yet has not been successful. Biomanufacturing HSCs requires comprehensive knowledge of their expansion and rigorous quality control. While in vivo transplantation assay is the gold standard for evaluating HSC regenerative potential (stemness), it is time and resource-consuming. A rapid prediction method will significantly facilitate screening and optimizing novel ex vivo expansion conditions. Metabolism was recently reported to be a vital cell fate regulator and reflect HSC stemness. In this dissertation, we proposed to track HSC metabolism as biomarkers to evaluate and optimize the ex vivo culture conditions. We demonstrated the optical properties of endogenous fluorophores and metabolic co-enzymes nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD), profiled by fluorescence lifetime imaging microscopy (FLIM), can be used to distinguish HSCs from differentiated hematopoietic cells in a non-invasive and real-time manner. We elucidated that the unique FLIM profile of HSCs is associated with their preferred glycolysis and inhibited oxidative phosphorylation (OXPHOS). We further defined a set of 205 FLIM features that profile single cell metabolism with subcellular spatial resolution. A bioinformatic workflow was established to select the features that track HSC differentiation, of which a combinatorial analysis evaluates the stemness of individual cells. With such metric, we could track the offspring of HSCs under symmetric and asymmetric divisions, and screen for the cell culture conditions that promote HSC expansion. Moreover, we also explored an oligo-barcoding and high-throughput RNA sequencing strategy to track paired daughter cells from the same parent cell, aiming to understand HSC division at the transcriptomic level. We expect this strategy to provide complementary information to our optical imaging methodology.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Membrane-bound regulation of hematopoietic stem cells
PDF
Machine learning and image processing of fluorescence lifetime imaging microscopy enables tracking and analysis of subcellular metabolism
PDF
Dissecting the heterogeneity of mouse hematopoietic stem cells in vivo
PDF
Functional compensation between hematopoietic stem cell clones in vivo
PDF
Improving adeno-associated viral vector for hematopoietic stem cells gene therapy
PDF
Investigating hematopoietic stem cells via multiscale modeling, single-cell genomics, and gene regulatory network inference
PDF
Physiologically-based modeling to predict monoclonal antibody pharmacokinetics in humans from physiochemical properties
PDF
Application of small molecules as self-reprogramming agents for the reconstruction of musculoskeletal tissues within 3-D tissue-specific scaffolds
PDF
Engineering 2D & 3D microphysiological systems for interrogating skeletal muscle tissues
PDF
Exploring the role of vegfr transcripts during vascular development in the zebrafish embryo
PDF
Non-invasive live-cell imaging for monitoring and evaluating pancreatic islet and beta cell metabolism
PDF
Fibroblastic connective tissue cells: the blastema stem cells and source of large-scale chondrogenesis in the regenerating lizard tail
PDF
Longitudinal assessment of neural stem-cell aging
PDF
The calcium-sensing receptor in the specification of normal and malignant hematopoietic cell localization in the bone marrow microenvironment
PDF
Rational selection of CRISPR/Cas9 guide RNAs for homology directed genome editing and its utility in the development of gene therapies
Asset Metadata
Creator
Zhou, Hao
(author)
Core Title
Metabolic profiling of single hematopoietic stem cells for developing novel ex vivo culture strategies
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Biomedical Engineering
Degree Conferral Date
2022-12
Publication Date
12/14/2026
Defense Date
07/26/2022
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
asymmetric division,ex vivo biomanufacturing,fluorescence lifetime imaging microscopy,glycolysis,hematopoietic stem cells,high-throughput RNA sequencing,machine learning,metabolism,NAD(P)H,OAI-PMH Harvest,oligo-barcoding,OXPHOS,symmetric division
Format
theses
(aat)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Shen, Keyue (
committee chair
), D'Argenio, David (
committee member
), Fraser, Scott E. (
committee member
), Lu, Rong (
committee member
), McCain, Megan L. (
committee member
)
Creator Email
zhouhao@usc.edu,zhouhao1228@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC112620922
Unique identifier
UC112620922
Identifier
etd-ZhouHao-11366.pdf (filename)
Legacy Identifier
etd-ZhouHao-11366
Document Type
Dissertation
Format
theses (aat)
Rights
Zhou, Hao
Internet Media Type
application/pdf
Type
texts
Source
20221214-usctheses-batch-997
(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. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
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
Tags
asymmetric division
ex vivo biomanufacturing
fluorescence lifetime imaging microscopy
glycolysis
hematopoietic stem cells
high-throughput RNA sequencing
machine learning
metabolism
NAD(P)H
oligo-barcoding
OXPHOS
symmetric division