University of Southern California Dissertations and Theses

Deconvolution of circulating tumor cell heterogeneity and implications for aggressive variant prostate cancer

(USC Thesis Other)

Deconvolution of circulating tumor cell heterogeneity and implications for aggressive variant prostate cancer

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DECONVOLUTION OF CIRCULATING TUMOR CELL HETEROGENEITY AND
IMPLICATIONS FOR AGGRESSIVE VARIANT PROSTATE CANCER

BY
SHOUJIE CHAI

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
(MOLECULAR BIOLOGY)

AUGUST 2022

ii

“A scientist can become consumed by devotion to facts without caring about what their
value is to humanity.”

“Treating cancer as one disease is like treating Africa as one country.”

“For the next quantum leap, fundamentally different strategies have to be developed.
The two immediate steps should be a shift from studying animals to studying humans
and a shift from chasing after the last cancer cell to developing the means to detect the
first cancer cell.”

― Azra Raza, The First Cell: And the Human Costs of Pursuing Cancer to the Last

iii

Acknowledgements
Five years ago, I asked myself a question ‘Will I regret it if I quit my physician job to
pursue a PhD degree?’ There were lots of concerns, worries, and uncertainty at that
time, even though my answer was “No”. Five years later, I really appreciate the decision
I made, thinking about all the advisors and mentors I have met, all the colleagues and
collaborators I have worked with, all the friends I have made, all the opportunities I have
encountered, all the projects I have led, and all the experience I have gained.
So, I want to take this moment to thank those who advised me, helped me, supported
me, and accompanied me along my PhD journey.
First, I would like to thank Peter and Jim, for advising me to learn the biological and
clinical science, to explore the early- and late- phase assay development, and to
understand the technology commercialization of liquid biopsy. Their uniqueness in
personality and expertise, but commonplace in their passion and persistence have
inspired me to keep pursuing my goal “Find Disease Early, Make Decision Precise,
Improve Patient Outcome”.
I would also like to thank my committee members Dr. Jorge Nieva, Dr. Lin Chen, and
Dr. Xiaojiang Chen for their guidance and advice on this research. I want to thank my
collaborator, the Genitourinary team at MD Anderson, and especially Dr. Ana Aparicio
for her mentorship and support.
I am extremely grateful to all the members of CSI-Cancer, past and present, for their
generous help and constant motivation. They have decorated my PhD journey with not
only science, but also happiness. I also want to thank my amazing students, for giving
me the opportunity to mentor them and for their dedicated commitment.
iv

I would like to thank my family for their support and understanding. Even though I did
not get chance to visit them in the past 4 years, I could still easily and virtually share my
joy, appreciation, and progress with them and also hear from them, especially my three
cutest nieces, Anning, Jianing, and Yining, who bring me so much fun and relaxation in
the evening chats.
I would like to thank all my friends. I could go on for 50 pages long to talk about each
of them. I love every one of them so much and could not imagine life without them.
Because of them, this place makes me feel like at home. Because of them, I could smile
every day. I am very grateful for all their help, support, and company along the way.
I want to end this section by thanking my partner, Scott. He is such a sweet, caring,
lovely, and supportive person and I had no idea what I was missing in life until I met
him. I learned so much from him, how to balance life and work, how to avoid
overthinking, how to respect different opinions, how to express my feelings and love,
and so much more. I really enjoy the time we spent together and very much look
forward to the new journey with him in the near future.

v

Table of Contents
Epigraph……………………………………………………………………………………....…ii
Acknowledgements………………………………………………………………………...…iii
List of Tables…………………………………………………………………………………vii
List of Figures..….………………………………………………………………………...…viii
Glossary….……………………………………………………………………………….…….ix
Abstract………………………………………………………………………………………….x
Introduction……………………………………………………………………………………..1
Chapter 1: Circulating Rare Cell Detection, Classification, and Characterization through
High Definition Single Cell Assay 3.0
1.1 Introduction……………………………………………………………………………..7
1.2 Landscape Immunostaining Protocol………………………………………………..8
1.3 OCULAR Computational Pipeline…………………………………………………..10
1.4 Rare Cell Classification and Characterization……………………………………..12

Chapter 2: Identification of Epithelial and Mesenchymal CTCs in the Clonal Lineage of
an Aggressive Prostate Cancer Case
2.1 Abstract……………………………………………………………………………….17
2.2 Introduction…………………………………………………………………………..18
2.3 Material and Methods…………….…………………………………………………21
2.4 Results and Discussion……………………………………..………………………26
2.5 Conclusion………………………………………………………...…………………41
vi

Chapter 3: Identifying the Combination of Platelet-coated CTC and Aggressive Variant
Prostate Cancer Molecular Signature as a Predictive Biomarker
3.1 Abstract………………………………………………………………………………42
3.2 Introduction………..…………………………………………………………………43
3.3 Material and Methods……………………………………………….………………45
3.4 Results and Discussion……………….…………………………………….………51
3.5 Conclusion……………….……………………………………………….……….…62

Chapter 4: Early Detection of ‘Aggressive Variant’ Genotype in CTC and ctDNA from
Men with Treatment-naïve De Novo Metastatic Prostate Cancer
4.1 Abstract……………….………………………………………………….………...63
4.2 Introduction……………….……………………………………………….…………64
4.3 Material and Methods……………….…..……………………………….…………66
4.4 Results and Discussion……………….…………………………….………………72
4.5 Conclusion.………………………………………………….……….………………84

Summary.………………………………………………….……….………………………….86
Future Directions……………………………………………………………………………..88
References……………………………………………….……….………………………....91
Appendix1: Supplementary Information………………………………………………..102
Appendix2: Data Availability……………………………………………………..………….118

vii

List of Tables
Table 1. List of masks and channels used for OCULAR……………………………………10
Table 2. Types of parameters extracted from events……………………………………….10
Table 3. Patient demographic and clinical characteristics………………………………….52
Table 4 Demographics of 23 treatment-naïve de novo M1 PCa patients……………….68

viii

List of Figures
Figure 1. Landscape immunostaining and OCULAR pipeline for rare cell detection….…13
Figure 2. Genomic rearrangement/clonality analysis of rare cell groups………………….14
Figure 3. Rare cell classification and biological categorization for cell type and state…...16
Figure 4. Patient demographic, pathology, and HDSCA3.0 workflow………………….…27
Figure 5. CTC subtype and enumeration in the paired PB and BMA………………….…..29
Figure 6. Clonal lineage and its variation between PB and BMA…………………………..32
Figure 7. Association between genotype and phenotype……………………………..……34
Figure 8. Targeted proteomics of CTC subtypes…………………………………….……...36
Figure 9. Identification of CTC subtypes……………………………………………………..54
Figure 10. AVPC-MS detection in CTCs…………………………………………………..…56
Figure 11. Survival analysis using pc.CTC and AVPC-MS status……………………..…..58
Figure 12. Multi-modal liquid biopsy………………………………………………………….71
Figure 13. Concentrations of multi-modal blood analytes and tumor volume……….……73
Figure 14. Molecular signature derived from both CTC and ctDNA profiles………….…..75
Figure 15. Univariate and Multivariate Progression-Free Survival Analysis…………...…76
Figure 16. Genomic profiling of CTC and ctDNA in four AVPC-MS positive patients……79
Figure 17. Association between genomic alterations and phenotypic features……….….81
Figure 18. A future comprehensive liquid biopsy test to delineate ‘AVPC’ …...……….….90

ix

Glossary
AR Androgen Receptor
AR-V7 Androgen Receptor – Variant 7
AVPC Aggressive Variant Prostate Cancer
AVPC-MS Aggressive Variant Prostate Cancer – Molecular Signature
BMA Bone Marrow Aspirate
CK Cytokeratin
CNA Copy Number Alteration
CTC Circulating Tumor Cell
ctDNA Circulating Tumor DNA
DDR DNA Damage Response and Repair
EMT Epithelial Mesenchymal Transition
epi.CTC epithelial-like CTC
HDSCA High Definition Single Cell Assay
IF Immunofluorescence
IMC Image Mass Cytometry
LST Large Scale Transition
mCRPC metastatic Castrate Resistant Prostate Cancer
mes.CTC mesenchymal-like CTC
ML Machine Learning
NBD Normal Blood Donor
OCULAR Outlier Clustering Unsupervised Learning Automated Report
OS Overall Survival
PB Peripheral Blood
pc.epi.CTC platelet-coated epithelial-like CTC
pc.mes.CTC platelet-coated mesenchymal-like CTC
PCA Principal Component Analysis
PCa Prostate Cancer
PFS Progression Free Survival
PSA Prostate Specific Antigen
TFx Tumor Fraction
VIM Vimentin
WBC White Blood Cell
x

Abstract
Tumor heterogeneity is a consequence of the mutation, selection, and adaptation of
cells along the pathway of disease progression and metastasis. It is considered the
major cause of treatment failure and its decoding may provide insights into the
development of new targeted therapies and their predictive biomarkers. Circulating
tumor cells (CTCs), as important analytes for liquid biopsy, have the potential to offer a
minimally invasive and realtime assessment of the heterogeneous and evolving
landscape of cancer. However, current approaches to detect CTCs mostly rely on one
identifier, such as epithelial markers or the cell’s biophysical traits, which can result in
poor sensitivity or misclassification. This warrants the need for method developments
that allow for the identification of various types of rare events, and then deconvolution of
the heterogeneity. Here, we developed a third generation of liquid biopsy technology to
detect and classify a spectrum of rare events followed with molecular characterization
(Chapter 1). This technology was applied to an index patient with prostate cancer in
which numerous CTCs were detected in the liquid biopsy. Results showed the
phenotypic and genotypic heterogeneity, including epithelial/mesenchymal state and
platelet attachment in CTCs with clonal lineage, which laid the foundation for biomarker
discovery (Chapter 2).
Prostate cancer is becoming recognized not as a single disease, but as many,
ranging from slow-growing tumors to highly aggressive and lethal lesions. A goal of this
work was to improve patient outcomes for lethal prostate cancer, especially the
‘Aggressive Variant’ prostate cancer (AVPC) that poorly responds to androgen receptor
inhibitors and has fast progression. This requires approaches to detect the variant early
xi

and inform the treatment decisions in a noninvasive and timely manner. Applying our
technology, we identified the presence of platelet-coated CTCs in combination to the
AVPC molecular signature (AVPC-MS), i.e., at least 2 defects in PTEN, RB1, TP53, as
a predictive biomarker of treatment selection for AVPC (Chapter 3). Furthermore,
AVPC-MS was detected in CTCs and other liquid biopsy analytes at the treatment-naïve
stage, demonstrating the clinical utility of early detection for AVPC (Chapter 4). Overall,
the development and application of our technology deciphered the CTC heterogeneity
for biomarkers identification and validated their clinical implications for AVPC.
1

Introduction
Liquid biopsy, different from tissue biopsy, is a minimally invasive procedure to detect
rare analytes derived from affected tissues of many systemic or circulating diseases.
Peripheral blood is the main source of liquid biopsy in addition to other body fluids, such
as bone marrow, cerebrospinal fluid, pleural effusion, etc. In the past decade, liquid
biopsy has been widely used for patients with cancers to investigate the source of
cancer metastasis between primary and metastatic tumors. Studies have shown that a
range of rare cells and cell-derived components travel in the blood, giving complexity to
the liquid biopsy (Ignatiadis et al., 2021; Kilgour et al., 2020). The cell-derived
components include cell-free DNA, RNA, protein and extracellular vesicles, while rare
cells include circulating tumor cells (CTCs), circulating tumor microenvironment cells,
and rare immune cells. Here, our focus is the cell-based liquid biopsy, especially CTCs.
CTCs as a key analyte present a variety of enumerations, phenotypes, and genotypes,
which are manifestations of heterogeneity. Biologically, heterogeneity is a consequence
of the mutation, selection, and adaptation of tumor cells during the disease progression
and cancer metastasis (Conteduca et al., 2021; Swanton, 2012). Clinically, it is
considered the major cause of therapeutic failures and highly related to different clinical
outcomes (Iacobuzio-Donahue et al., 2020; Russano et al., 2020). Hence, we aimed for
developing a new generation of liquid biopsy technology that not only detects CTCs but
also deconvolutes heterogeneity, allowing for the identification of prognostic and
predictive biomarkers.
2

Technologies of CTC Detection
Thomas Ashworth, an Australian physician scientist first described tumor cells in the
circulation of a patient with advanced cancer in fall of 1869. In describing the finding in
The Medical Journal of Australia (Ashworth, 1869), Ashworth concluded that “The fact
of cells identical with those of the cancer itself being seen in the blood may tend to
throw some light upon the mode of origin of multiple tumours existing in the same
person.” Over a century later, new technologies developed in the early 2000s enabled
the detection of those cells in the blood, recognized as CTCs. Since then, several CTC
detection technologies have emerged and the main strategy of those technologies is
enrichment-based, using either epithelial markers or biophysical traits. For example,
Cell Search, the first FDA-approved CTC test, utilizes antibody-conjugated magnetic
beads to enrich CTCs with epithelial markers (Cristofanilli et al., 2004; Cristofanilli et al.,
2005), while microfluidic chips or filters are used to isolate larger cells as CTCs appear
to have larger sizes compared to the surrounding immune cells (Karabacak et al.,
2014). However, there are several specific issues of those enrichment-based
approaches that affect their technical performance and clinical utility. The first issue is
cell loss, which significantly reduces the sensitivity and accuracy of the results,
especially when a patient has a minimal number of CTCs (Lin et al., 2021). Some of this
cell loss occurs because CTCs are heterogeneous and often undergo epithelial-
mesenchymal transition (EMT) with downregulation of epithelial markers (Satelli et al.,
2015). Additionally, not all CTCs are large and can have similar sizes to immune cells.
These important cells would be missed with filtering resulting in false negative assays.
Conversely, we have seen false positives occur when circulating endothelial cells
3

express epithelial markers, such as cytokeratin. To overcome these issues, there is a
need for technology development that can achieve a comprehensive analysis of all
cells, what we refer to as “No Cell Left behind”. The goal of this approach is to increase
the sensitivity and accuracy of the assay via simultaneous characterization of all cell
types and states in the sample to understand the complexity and heterogeneity of liquid
biopsy.
CTC Heterogeneity
CTCs disseminate from tumors into the bloodstream. We observe heterogeneity due
to biological adaptions within the circulatory microenvironment, somatic alterations
based on clonal lineage, and subpopulation selection from therapeutic selection. EMT is
a fundamental process whereby cells transition from epithelial to mesenchymal cell
state and constitute a key part of phenotypic heterogeneity. This transition is essential
for tumor cell invasion and migration during the cancer metastasis (Genna et al., 2020;
Lin et al., 2021). Thus, identification of epithelial and mesenchymal states in CTCs is
necessary to understand hematogenous dissemination. Meanwhile, others have
observed that some CTCs display the “Never Travel Alone” phenotype (Heeke et al.,
2019). Groupings of multiple CTCs in clusters (also known as microemobli or
aggregates) and their interactions with other components in the circulatory
microenvironment, such as platelets and neutrophils, may provide information about
tumor viability and/or metastatic capability. For example, research has shown that the
CTC-platelet interaction aids the tumor cells to resist the sheer stress in the blood flow
and escape from immune surveillance or attack. This interaction could also induce
mutual activations through key molecules, such as CD97, leading to the EMT on CTCs
4

and the release of chemokines in platelets (Ward et al., 2018). EMT-shifted CTCs may
also acquire the ability to recruit even more platelets or neutrophils binding which
augments their immune-suppressive and invasive functionalities (Genna et al., 2020). In
addition, molecular features associated with stemness, immune suppression, genomic
instability, or DNA Damage Response and Repair (DDR) have been identified in CTCs,
such as stemness surface proteins of CD44, CD24, and CD133, immunosuppressive
marker of PD-L1, instability indicators of PTEN, RB1, and TP53, and DDR genes of
PARP, BRCA1/2 and ATM (Heitmeir et al., 2022; Mohme et al., 2017). However, the full
range of phenotypic and genotypic heterogeneity of CTCs and their associations have
not been fully characterized due to the difficulties in capturing CTCs for single cell
analysis. To achieve this goal, we sought to decode morpholomics, genomics, and
targeted proteomics of CTCs with a single cell multi-omic approach.
Clinical Applications of CTCs
The clinical utility of CTC enumeration was described in 2004 as a prognostic factor
in newly diagnosed metastatic breast cancer patients in a prospective study (Cristofanilli
et al., 2004). Since then, emerging evidence has indicated that the applications of CTCs
can be extended beyond prognostication to predictive biomarkers for targeted therapies
and diagnosis(Lin et al., 2021). Today, however, the measurement most established,
remains enumeration. Results from the Cell Search platform demonstrated that patients
with CTCs above 5 cells per 7.5mL blood, had worse prognosis in various metastatic
cancers (Cristofanilli et al., 2004; Danila et al., 2007). One successful molecular
measure beyond enumeration has been nuclear detection assay of androgen receptor
variant 7 (AR-V7) in CTCs which predicts the subpopulation of advanced prostate
5

cancer who benefit from taxanes over AR signaling inhibitors (Scher et al., 2018; Scher
et al., 2016). Additionally, a new comprehensive liquid biopsy test, i.e., DefineMBC, has
been developed and implemented for diagnostic value of metastatic breast cancer and
treatment decision making when a tissue biopsy is not available. This test includes both
cell-based and cell-free analysis, allowing for the detection of CTCs, quantification of
ER/HER2 expression, and genomic analysis of single cell and circulating tumor DNA
(ctNDA). While there is excitement over these newer uses of CTCs in clinical
applications, the number of clinically useful CTC tests is still limited, necessitating the
need for additional exploration.
Prostate Cancer and Its ‘Aggressive Variant’ Subset
Prostate cancer is the most common cancer in American men with an estimated
268,490 new cases and 34,500 deaths in 2022 (NCI: SEER). It is becoming recognized
not as one single disease, but as a combination of multiple diseases. Some patients
with prostate cancer have very slow-growing tumors with the sole need of active
surveillance and some patients exhibit good responses to standard treatments, such as
androgen deprivation therapy. Still, around 10~20% of cases present as a highly
aggressive and lethal disease with inherited treatment resistance, fast progression, and
short survival. This is in contrast to the majority of cases which has a 5-year survival
rate of 96.8% (NCI: SEER). Distinguishing risk in prostate cancer is difficult. Presently,
there is no universally accepted molecular biomarker that can stratify the disease into
multiple subsets and inform different treatments. Previous efforts in pre-clinical and
clinical studies have identified a molecular signature of three tumor suppressor genes,
PTEN, RB1, and TP53 and these have been proposed to define a subset of metastatic
6

castrate-resistant prostate cancer (mCRPC) as aggressive variant prostate cancer
(AVPC) (Aparicio et al., 2016; Ku et al., 2017; Mu et al., 2017; Zou et al., 2017). Clinical
results showed that this AVPC subset exhibited the dismal outcomes and unfavorable
responses to androgen receptor (AR) signaling inhibitors (Aparicio et al., 2016). Most
importantly, the presence of this AVPC molecular signature (AVPC-MS) may predict the
vulnerability of mCRPC to taxane-platinum combination therapies relative to taxane
alone (Corn et al., 2019). However, the identification and validation of AVPC-MS is
currently based on tissue biopsy with immunohistochemistry analysis of tumor tissues.
What is lacking in the clinical management is the minimally invasive assessment of the
signature in liquid biopsy samples to identify patients who later evolve an AVPC
phenotype. Since AVPC-MS has been mainly investigated at the treatment-resistant
stage, there are fundamental questions arising whether AVPC-MS could be detected
early at the treatment-naïve stage and if so, how it would be associated with clinical
outcome.
In this thesis, we developed a comprehensive liquid biopsy technology for
simultaneous characterization of various cell types and states, applied it to one index
patient and two clinical trials to deconvolute the CTC heterogeneity and delineate the
biological mechanisms of aggressive disease, and further explored the clinical
implications for treatment decision making and early detection of AVPC.
7

Chapter 1
Circulating Rare Cells Detection, Classification, and Characterization through High
Definition Single Cell Assay 3.0 (HDSCA3.0)

The methods development has resulted in intellectual property that has been filed with
the USPTO:

Peter Kuhn, Carmen Ruiz-Velasco, Shoujie Chai, James Hicks, Anand Kolatkar,
Nicholas Matsumoto, Rafael Nevarez, Benjamin Ormseth. System Methods and Assays
for Outlier Clustering Unsupervised Learning Automated Report. U.S. Serial No.
62/914,763. 10/14/2019

1.1 Introduction
In the blood biopsy, we typically find that the majority of cells are immune cells and
additionally find rare cells that actively escape or passively leak into the circulation from
injured or mutated tissues in different disease settings. Initial assumptions were that
these cells may not survive in the circulation but our early work in patients with coronary
artery disease showed that endothelial cells can survive and be detected, which
motivated us to investigate this further and understand that the blood contains a variety
of cells from the microenvironment of the site of injury including a tumor. The types of
circulating rare cells in cancer patients could be 1) tumor cells as evidenced by cancer
genomic profiles and/or protein markers of epithelial cell or tissue of origin, 2) other rare
cells from tumor microenvironment that leak into circulation, including endothelial cells,
stromal cells, etc., 3) unique immune cells responding to the tumor itself or cancer
treatment, contributing to the tumor complexity (Bertolini et al., 2006; Ortiz-Otero et al.,
2020; Xu et al., 2017). The same type of rare cells could further present various
phenotypic states under mutation, adaptation, and treatment selection, such as
8

epithelial, mesenchymal, stem cell like, immune suppressive state, leading to the tumor
heterogeneity (Heitmeir et al., 2022).
While traditionally the field has focused on certain cell types and states through
various enrichment methods that select one specific target cell at a time based on its
one presumably known phenotype, it is clear that the disease is much more complex
and heterogeneous and that these rare cells in the blood do not neatly fall into specific
categories. A next generation approach would identify various cell types and states that
have been implicated simultaneously in one unified experiment. The classification and
biological categorization of rare cells would be separated by protein markers as well as
cellular and nuclear morphology.
Here, HDSCA3.0 leverages both a new assay protocol reducing the 5 markers into 4
fluorescence channels (Landscape immunostaining protocol, Fig. 1A) and an integrative
computational pipeline (Outlier Clustering Unsupervised Learning Automated Report,
OCULAR, Fig. 1B) for classifying the different rare cells groups via analysis of
fluorescence images followed with single cell genomic characterization.

1.2 Landscape Immunostaining Protocol
Our previously non-enrichment sample preparation approach combined a multi-
parametric assay approach to overlay protein fluorescence with cellular morphology to
distinguish 4 relevant cell types based on 3 markers (Cytokeratins, CD45, DAPI) in a
single experiment, i.e., HD-CTC, CTC-Small, CTC-LowCK, CTC-cfDNA producing
(Thiele et al., 2019). Our Landscape immunostaining protocol is an approach to
distinguish a spectrum of rare cell groups using 5 markers, which are fluorescently
9

labeled antibodies to 4 distinct protein markers (Cytokeratins (CK), CD45, Vimentin
(VIM), CD31) and a nuclear marker (DAPI) based on protein expression and
morphological signatures at the appropriate resolution by fluorescence microscopic
imaging (Chai et al., 2021).
The use of these specific markers allocated across 4 fluorescence channels as
described below enables the protocol to differentiate 8 rare cell groups and infer at least
9 biological cell types or states. We allocate the first immunofluorescence channel to the
fluorescent dye, DAPI, for nuclear segmentation and characterization, a second channel
to cytokeratins for epithelial-like phenotype, a third channel to vimentin, a type III
intermediate filament protein for endothelial/mesenchymal-like phenotype, and the
fourth channel to both CD31 for endothelial-like/platelet-relevant phenotype and CD45
for immune cell phenotype. The understanding of differential morphologies, differential
marker expression and cellular distribution and differential marker specificity/sensitivity
requirements then motivated us to use the four channels in the following way:
1. DAPI has its emission maximum wavelength at 461 nm (DAPI channel, blue
color)
2. CK: Alexa555 (TRITC channel, red color) using a pan-cytokeratin antibody
cocktail and a secondary Alexa555 fluorescent antibody.
3. VIM-Alexa488 (FITC channel, white color) using a directly conjugated antibody
4. CD45-Alexa647/CD31-Alexa647 (Cy5 channel, green color) using a directly
conjugated antibodies.

1.3 OCULAR Computational Pipeline
10

Event Segmentation and Feature Extraction
Using the EBImage R library (Pau et al., 2010), the algorithms could be used for cell
segmentation and feature extraction. The initial set of masks are determined by the
signal in the DAPI channel for DAPI positive events. The EBImage library can detect
these events and propagate from the center of those events into the nuclear mask and
the cell mask. After identifying DAPI positive events and their respective nuclear and
cell masks, OCULAR performs feature extraction, collecting 761 parameters across
channels (Table 1 and 2), using ‘compute Feature’ function. Furthermore, we remove all
DAPI positive events from the images and check if there are any events that are
negative in DAPI but positive in any of the other channels. These events are propagated
into a DAPI negative mask. The mask is then used for feature extraction, similar to the
DAPI positive event.
Table 1 and 2. List of masks and channels used for OCULAR and types of parameters
extracted from events.

11

Rare cell candidate detection
Since there are significant correlation and redundancy in extracted features, principal
component analysis (PCA) with its derivation of eigenvectors is used to reduce the data
from initial 761 features to 350 components, which can still describe 99.9% of the
original data space. OCULAR further performs a hierarchical clustering on cells from
each frame (total 1840 frames per slide for analysis) with 350 principal components
generated above, to group similar cells together. OCULAR combines the agglomerative
approach of hierarchical clustering with the Ward’s minimum variance method to find
compact, spherical clusters. Since PCA and hierarchical clustering are static methods
and introduce no stochasticity, all results are reproducible.
To determine the number of clusters, OCULAR aims for average 30 cells per cluster
in one frame, e.g., if a frame contains 1200 cells, it will produce 40 clusters followed
with the distance calculation for each cell from the median cell in the same cluster. The
rare cells are identified by cluster size and by distance from the median cell. Basically,
after ordering the clusters by size and distance, OCULAR obtains a collection of
clusters that sums to 1.5% of the total cells in the frame as rare cell candidates. Original
761 features for each rare cell candidate are stored individually. The remaining clusters
are considered as common cell clusters and their intra-cluster average value of each
feature will be stored as well as the cluster size. These rare cell candidates and
common cell clusters from each frame are saved as an RDS file for further filtration.
Rare cell identification
To further determine whether the rare cell candidates from each frame are also rare
cells within the slide, the strategy is to calculate the distance of each rare cell candidate
12

from the slide-wide average of all common cell clusters and exclude candidates based
on a cut-off. The distance calculation uses PCA-produced components, rather than
original 761 features. OCULAR then determines a 1% quantile length of all distances in
this matrix as the cutoff to define rare cells in the slide. Given that the distance is within
this cutoff to any of the PCA component of slide-wide common cell clusters, that rare
cell candidate will be considered as a common cell and filtered out. The final slide-wide
rare cells will be clustered, sorted by imaging parameters and displayed for review and
manual curation.
OCULAR Report Generation
The final report is a web application with tabs to visualize common cell clusters, rare
cell clusters, DAPI negative event clusters, and final report. The last step is the manual
curation to separate the ‘interesting’, ‘unsure’, and ‘non-interesting’ rare cells in the
slide. The downstream classification and genomic characterization are performed within
‘interesting’ rare cells.

1.4 Rare Cell Classification and Characterization
Fluorescence Intensity Based Rare Cell Classification
Based on the intensity positivity of CK, VIM, and CD45/CD31, detected rare cells can
be classified them into 4 CK+ and 4 CK- cell groups according to the four possible
combinations with VIM and/or CD45/CD31 (Fig. 1C).

13

Figure 1. Landscape immunostaining and OCULAR pipeline for rare cell detection. A.
Immunostaining steps: 1. CD31/Fabs Ab incubation; 2. panCK/VIM/CD45 Ab incubation;
3. IgG1-A555 secondary Ab incubation. B. OCULAR image data analysis workflow. C.
Rare cell groups. Ms: mouse; Rb: rabbit; Gt: goat; A: Alexa Fluor dye. Blue: DAPI; Red:
CK; White: VIM; Green: CD45/CD31 in merged image (far left images in each group).

Genomic Characterization of Rare Cell Groups
Subsequent to classification, single cell sequencing provides a tool to confirm if the
cell is genomically normal or rearranged based on copy number profiles and if at least
two cells share two or more copy number alterations with same breakpoints from the
same patient, they are further recognized as clonal cells, reflecting a proliferating
lineage (Fig. 2 Left). This genomic analysis was then performed in 877 rare cells from
14

metastatic prostate cancer patients and they were covering 8 different rare cell groups,
i.e., CK: 352 cells, CK|VIM: 49 cells, CK|(CD45/CD31): 57 cells, CK|VIM|(CD45/CD31):
243 cells, VIM|(CD45/CD31): 57 cells, VIM: 74 cells, CD45/CD31: 8 cells, DAPI: 38
cells. Result showed cells with clonal alterations were found in all four CK+ groups (Fig.
2 Right). Non-clonal genomic rearrangements were found in a small portion of cells in
all groups, except for the ‘CD45/CD31+ only’ group (Fig. 2 Right). Among the 701
sequenced cells from 4 CK+ groups, cells from the “CK+ only” group had the highest
frequency of clonal cells (304/352, 86.3%). The frequencies of clonal cells in “double
positive CK|VIM”, “double positive CK|(CD45/CD31)”, and “triple positive
CK|VIM|(CD45/CD31)” groups were 11/49 (22.4%), 20/57 (35.1%), and 2/243 (0.8%),
respectively (Fig. 2 Right). These results suggest the presence of intragroup genomic
heterogeneity as well as intergroup differences in frequency of clonal alterations.

Figure 2. Genomic rearrangement/clonality analysis of rare cell groups. Left, three copy
number profiles of rare cells from a metastatic prostate cancer patient; right, percentile
of clonally rearranged, non-clonally rearranged and normal cells in each rare cell group.

Morphology Integrated Biological Type and State Categorization
Within each rare cell group, high morphological heterogeneity was observed and we
further performed an initial knowledge-based biological categorization to identify
different cell types and states based on both fluorescence intensity and morphology.
15

Currently, we have categorized 7 potential cell types and states (Fig. 3):
1. Epithelial-like CTC: CK positive and CD45/CD31 negative identified from the
“CK
+
only” rare cell group, with distinct appearing nucleus by DAPI morphology as
previously described (Thiele et al., 2019)
2. mesenchymal-like CTC: epithelial-like CTC with additional expression of VIM
from the “double positive CK|VIM” rare cell group
3. platelet-coated CTC: epithelial-like and mesenchymal-like CTC presenting a
punctuated pattern in the Alexa Fluor® 647 (CD45/CD31) corresponding to CD31 signal
from platelets from the “double positive CK|(CD45/CD31)” and “triple positive
CK|VIM|(CD45/CD31)” rare cell groups
4. CK- endothelial cell: CD31 and VIM double positive with elongated shape and no
CK expression from the “double positive VIM|(CD45/CD31)” rare cell group
5. CK+ Endothelial cell: CD31 and VIM double positive with elongated shape and
CK expression from the “triple positive CK|VIM|(CD45/CD31)” rare cell group
6. Megakaryocyte: CD31 positive only with a large cell size and a single, large,
multilobulated, polyploidy nucleus from the “CD45/CD31+ only” rare cell group
7. Fibroblast-like cell: VIM positive only with elongated shape from the “VIM+ only”
rare cell group
In the next couple chapters, we will mainly focus on the molecular characterization of
CTC subtypes from 4 CK+ rare cell groups and explore their clinical application for
metastatic prostate cancer.
16

Figure 3. Rare cell classification and biological categorization for cell type and state.

17

Chapter 2
Identification of Epithelial and Mesenchymal CTCs in Clonal Lineage of an Aggressive
Prostate Cancer Case

This chapter is published as:

Shoujie Chai, Carmen Ruiz-Velasco, Amin Naghdloo, Milind Pore, Mohan Singh,
Nicholas Matsumoto, Anand Kolatkar, Liya Xu, Stephanie Shishido, Ana Aparicio,
Amado Zurita, James Hicks, Peter Kuhn. Identification of Epithelial and Mesenchymal
CTCs in Clonal Lineage of an Aggressive Prostate Cancer Case. NPJ Precis Oncol.
2022 Jun 21;6(1):41. doi: 10.1038/s41698-022-00289-1.

2.1 Abstract
Little is known about the complexity and plasticity of circulating tumor cell (CTC) biology
in different compartments of the fluid microenvironment during tumor metastasis. Here
we integrated phenomics, genomics, and targeted proteomics to characterize CTC
phenotypic and genotypic heterogeneity in paired peripheral blood (PB) and bone
marrow aspirate (BMA) from a metastatic prostate cancer patient following the rapid
disease progression, using the High-Definition Single Cell Assay 3.0 (HDSCA3.0).
Uniquely, we identified a subgroup of genetically clonal CTCs that acquired a
mesenchymal-like state and its presence was significantly associated with one subclone
that emerged along the clonal lineage. Higher CTC abundance and phenotypic diversity
were observed in the BMA than PB and differences in genomic alterations were also
identified between the two compartments demonstrating spatial heterogeneity. Single
cell copy number profiling further detected clonal heterogeneity within clusters of CTCs
(also known as microemboli or aggregates) as well as phenotypic variations by targeted
18

proteomics. Overall, these results identify epithelial and mesenchymal CTCs in the
clonal lineage of an aggressive prostate cancer case and also demonstrate a single cell
multi-omic approach to deconvolute the heterogeneity and association of CTC
phenotype and genotype in multi-medium liquid biopsies of metastatic prostate cancer.

2.2 Introduction
The mutation, selection, and adaptation of tumor cells along the pathway of disease
progression and metastasis results in a spectrum of phenotypic and genomic
heterogeneity(Conteduca et al., 2021; Keller & Pantel, 2019; Miyamoto et al., 2016;
Scher et al., 2017). Distinguishing phenotypic states of tumor cells with genetically
clonal identity and along clonal lineage is important in the investigation of epithelial-
mesenchymal transition (EMT) and for understanding the link between genotype and
phenotype(Dago et al., 2014; Malihi et al., 2018). Dissecting the genetic mechanism of
a phenotypic transition is especially critical in circulating tumor cells (CTCs) with
acquired metastatic capability as the key hallmark of cancer(Hanahan & Weinberg,
2011). The underlying concept behind EMT is that a malignant epithelial cell changes its
phenotypic state to become more mesenchymal-like and motile as a means to increase
metastatic potential. This implies that to clearly identify EMT as a state change, the
genomic alterations in the original epithelial cell must be evident in the resultant
mesenchymal cell. Current liquid biopsy methods for CTC detection usually rely on one
particular phenotypic state, traditionally the epithelial state characterized by cytokeratin
and/or EpCAM expression(de Bono et al., 2008). Next generation approaches utilize
enrichment-free methods such as the third generation of the High Definition Single Cell
19

Assay (HDSCA3.0) which has the ability to identify epithelial, mesenchymal, and
endothelial phenotypic states among circulating rare cells, as well as to identify those
cells in the genetically transformed tumor lineage in the liquid biopsy(Chai et al., 2021).
A number of studies have shown that the presence of mesenchymal-like CTCs is
associated with worse prognosis (Batth et al., 2020; Chen et al., 2019; Horimoto et al.,
2018; Li et al., 2015; Lindsay et al., 2016; Mego et al., 2012; Mego et al., 2019;
Papadaki et al., 2019; Satelli et al., 2015; Wu et al., 2017). However, the identification of
mesenchymal CTCs in those studies was limited to immunostaining (Batth et al., 2020;
Horimoto et al., 2018; Lindsay et al., 2016; Papadaki et al., 2019; Satelli et al., 2015;
Wu et al., 2017), RT-PCR(Mego et al., 2012; Mego et al., 2019), or RNA fluorescence in
situ hybridization (RNA FISH) (Chen et al., 2019; Li et al., 2015) of EMT biomarkers,
without further genomic validation of a genetic lineage with cancer cell identity. In our
previous study (Chai et al., 2021), of peripheral blood (PB) and bone marrow aspirates
(BMA) from 65 metastatic castration-resistant prostate cancer patients in the
“cabazitaxel with or without carboplatin” trial (NCT01505868)(Corn et al., 2019), we
identified a large number epithelial-like (CK+) cells that exhibited clonal genomic
alterations characteristic of prostate cancer cells as well as platelet-coated cells that
comprised a biomarker for therapeutic benefit from additional carboplatin. In that study,
we also identified a small number of mesenchymal-like (CK+|VIM+) cells, however with
the exception of two cases, they were genomically normal and thus may represent cells
from the tumor microenvironment rather than tumor cells that had been transformed
through EMT. Thus, dual identification of phenotypic states and genomic alterations on
20

the same single cell is an essential approach to deconvolute EMT heterogeneity of
CTCs with cancer cell identity confirmation and lineage tracing.
Among the potential patients for that trial, there was a single patient that underwent
pre-enrollment evaluation but did not enter the trial due to rapid disease progression.
Blood and bone specimens were taken as part of the pre-enrollment process and were
thus available for this study. Here, we utilized the multi-omic capabilities of the
HDSCA3.0 workflow to characterize the CTC phenotypes in paired PB and BMA
samples from this unusually aggressive case, followed by single cell copy number
profiling based on low pass whole genome sequencing or targeted proteomics based on
imaging mass cytometry. In contrast to the patients in our previous study(Chai et al.,
2021), this index patient had large numbers of genetically clonal epithelial-like (CK+)
CTCs and a nearly equal number of mesenchymal-like (CK+|Vim+) CTCs that
comprised a genetic subclone of the original genomic alterations. Meanwhile,
differences of CTC abundance and phenotypic diversity were observed between PB
and BMA as well as genomic variations. The observed significant change in phenotypic
state while maintaining clear genetic relationship to the transformed epithelial clone
provides the clear evidence of the EMT state change detected in the liquid biopsy and
further suggests that a multi-omic approach can provide useful information about patient
condition in near real-time to influence treatment decisions.

21

2.3 Methods
Patient selection and sample preparation
The index patient in this study was diagnosed with de novo metastatic prostate
cancer (mPC) and following castrate resistance and further disease progression, paired
PB and BMA samples were collected and shipped to University of Southern California
within 24 hours of the collection for analysis with the HDSCA3.0 workflow(Chai et al.,
2021).This patient was registered under the pre-enrollment of “cabazitaxel with or
without carboplatin” trial (NCT01505868). The study was approved by the Institutional
Review Board and adhered to the principles in the Declaration of Helsinki. The patient
provided written informed consent prior to inclusion in the study.
Sample preparation was performed as previously described(Carlsson et al., 2017;
Dago et al., 2014; Malihi et al., 2018; Marrinucci et al., 2012). Briefly, upon arrival in the
laboratory, both PB and BMA liquid biopsy samples were treated with isotonic
ammonium chloride solution for erythrocyte removal and the isolated nucleated cells
from centrifuging were plated on the cell-adhesive slides (Marienfeld) and stored at -
80 °C.
Immunofluorescent staining and scanning
Two slides with approximately 3 million nucleated cells each from PB and BMA,
respectively, were stained according to the HDSCA3.0 protocol as previously
described(Chai et al., 2021). Briefly, slides were thawed for 1 hour, fixed with 2%
formalin for 20 minutes, and incubated with 10% goat serum for another 20 minutes.
The following staining steps were conducted on an IntelliPATH FLX autostainer
(Biocare Medical LLC) with negative and positive control slides included: 1) a mixture
22

containing an anti-human CD31:Alexa Fluor 647 mouse IgG1 monoclonal Antibody
(BioRad; Cat# MCA1738A647; Clone: WM59; Working concentration: 2.5 μg/mL) and
an anti-mouse IgG goat monoclonal Fab fragment (Jackson ImmunoResearch; Cat#
115–007–003; Working concentration: 100 μg/mL) for 4 hours; 2)100% cold methanol
for permeabilization for 5 minutes; 3) a mixture consisting of an anti-human
cytokeratin(CK) 1,4,5,6,8,10,13,18,19 mouse IgG1/IgG2a monoclonal antibody cocktail
(Sigma; Cat# C2562; Clone: C-11, PCK-26, CY-90, KS-1A3, M20, A53-B/A2; Working
concentration: 210 μg/mL), an anti-human CK 19 mouse IgG1 monoclonal antibody
(Dako; Cat# GA61561–2; Clone: RCK108; Working concentration: 0.2 μg/mL), an anti-
human CD45:Alexa Fluor 647 mouse IgG2a monoclonal antibody (AbD Serotec; Cat#
MCA87A647; Clone: F10–89–4; Working concentration: 1.2 μg/mL), and an anti-human
vimentin (VIM): Alexa Fluor 488 rabbit IgG monoclonal antibody (Cell Signaling
Technology; Cat# 9854BC; Clone: D21H; Working concentration: 3.5 μg/mL) for 2
hours; 4) a mixture including an anti-mouse IgG1: Alexa Fluor 555 goat IgG polyclonal
antibody (Invitrogen; Cat# A21127; Working concentration: 2 μg/mL) and 40,6-
diamidino-2-phenylindole (DAPI) for nuclear DNA (Thermo Fisher Scientific; Cat#
D1306; Dilution: 1: 50,000) for 1 hour. Finally, slides were mounted with a glycerol-
based aqueous mounting media followed with adding coverslips. Slides were then
scanned, as previously described(Chai et al., 2021). Briefly, slides were scanned using
an automated fluorescence scanning microscopy at 10x objective magnification and
generating 2304 frame images for each channel (DAPI: DNA; Alexa Fluor 555: CK;
Alexa Fluor 488: VIM; Alexa Fluor 647: CD45/CD31). Exposure time and gain were
automatically set up to yield the same background intensity level across slides before
23

auto-scanning. Some cells on the slides were further manually re-imaged using a
fluorescence microscopy at 40x objective magnification for higher resolution images.
Rare cell detection and CTC subtype enumeration
Image analysis was performed as previously reported(Chai et al., 2021). In brief, the
EBImage package(Pau et al., 2010) was used to segment DAPI+ (cell) events by
generating nuclear and/or cytoplasm masks. For each segmented event, 761
quantitative cellular and nuclear features were extracted by “computeFeatures” in the
EBImage package and top 350 principal components were further identified from
extracted features by principal component analysis (PCA). Hierarchical clustering was
performed among detected cells with those principal components of features to
separate common cells (mainly leukocytes) and rare cells in each frame image.
Manual classification of rare cells into CTC subgroups was also previously
reported(Chai et al., 2021). Briefly, epithelial-like CTCs (epi.CTCs) were classified as
cells are CK positive, VIM negative and CD45/CD31 negative with distinctive nucleus
morphology and mesenchymal-like CTCs (mes.CTCs) gained VIM expression in
addition to CK. Platelet attachment, featured as punctuated CD45/CD31 signals on cell
surface, further categorized platelet-coated (pc) epi.CTCs (pc.epi.CTCs) and mes.CTCs
(pc.mes.CTCs).
Single cell copy number profiling by low pass whole genome sequencing
Single cell copy number profiling was performed as previously reported(Carlsson et
al., 2017; Chai et al., 2021; Dago et al., 2014; Malihi et al., 2018). Briefly, cells of
interest were relocated using XY coordinates generated from the scanning and 40x
high-quality images were captured before the single cell isolation by micromanipulator.
24

Individual single cells were lysed for whole genome amplification (Sigma-Aldrich; Cat#
WGA4) and libraries were constructed using the DNA Ultra Library Prep Kit (New
England Biolabs; Cat# E7370) and sequenced by Illumina NextSeq 500 at USC for
single-end 50bp read sequencing. Following genome mapping and PCR duplicate
removal, unique reads were seated into ~5000 pre-defined bins and the number of
reads per bin was normalized as “ratio to mean” for constructing copy number profile
and heatmap and identifying copy number alteration (CNA). The dendrogram on the
heatmap was grouped by hierarchical clustering using “cluster-agnes” R package
(metric = Manhattan, method = ward). Large scale transition (LST) was a measurement
of the number of large scale (>10Mb) copy number alterations across the whole
genome(Malihi et al., 2020). Minimum evolution method(Rzhetsky & Nei, 1993) was
used for phylogenetic tree construction of individual cells or clade consensus (the
median of cells from the same clade). The contamination of normal cell in CTC could
compress the copy number profiles and was manually identified and labeled as “c”
standing for “compressed”. The evaluation was based on checking if the “ratio to mean”
of CNA was matched with the theoretical ratio between the integer copy number of
gain/loss region and the ploidy number of one single cell, e.g., in a diploid cell, the
theoretical ratio for one copy loss is 0.5 (1/2) and the value for one copy gain is 1.5
(3/2). In terms of the cell selection criteria for single cell genomics, we sequenced cells
across different subtypes including epi.CTC, mes.CTC, pc.epi.CTC, and pc.mes.CTC to
observe genotypic heterogeneity and aimed for at least 5 cells analyzed per subtype per
sample in consideration of the number of cells detected and the number of cells
successfully isolated and sequenced.
25

Targeted proteomics by imaging mass cytometry
Cells of interest were subjected to in-situ targeted proteomic analyses with the use of
the CyTOF Helios imaging mass cytometer (Fluidigm) as previously
described(Gerdtsson et al., 2018; Malihi et al., 2018). Briefly, sample slides were re-
stained with metal-conjugated antibodies (Prostate-specific: AR-N (Cell Signal
Technology; Cat# 5153; Clone: D6F11; Dilution: 1:200), AR-C (LS Bio; Cat# LS-
C210456-500; Clone: SP242; Dilution: 1:200), PSMA (Novus; Cat# MAB4234; Clone:
460420; Dilution: 1:200); EMT: EpCAM (Fluidigm; Cat# 3144026D; Clone: 9C4; Dilution:
1:200), E-cadherin (Fluidigm; Cat# 3158029D; Clone: 24E10; Dilution: 1:300), Vimentin
(Abcam; Cat# ab193555; Clone: EPR3776; Dilution: 1:300), N-cadherin (Abcam; Cat#
ab19348; Clone: 8C11; Dilution: 1:200); Cell-proliferation: PCNA (Abcam; Cat#
ab18197; Clone: Polyclonal; Dilution: 1:400), β-catenin (Fluidigm; Cat# 3147005A;
Clone: D10A8; Dilution: 1:300) and a DNA intercalator. Antibodies that were not
available with Fludigm were sourced from the third-party vendors and were custom
conjugated in the lab. Maxpar antibody labeling kits were used to label the antibodies
with the metals of choice. Metal labelled antibodies cocktail was applied to the
experiment slide during staining process. A region of interest (ROI) of approximately
400 µm × 400 µm centered on each candidate cell was ablated with a 1 µm diameter
pulsed laser, followed by ionization and quantification in the CyTOF Helios instrument.
Ion mass data were collected and used for reconstruction of the 1 µm
2
ROI spatial
resolution, multi-dimensional images of the ROI. Cell segmentation and ion count per
cell were generated by IMC segmentation pipeline created by the Bodenmiller
Lab(Zanotelli & Bodenmiller, 2019), based on the CellProfiler (version 3.15)(McQuin et
26

al., 2018) and Ilastik (version 1.3.3)(Berg et al., 2019) and images with segmented
masks could be further displayed by HistoCAT(Schapiro et al., 2017).
R packages and statistical analysis
For data visualization, we used t-SNE (version 0.15)(Krijthe, 2015) for dimensionality
reduction, ggplot2 (version 2.8.0)(Wickham, 2016) for scatter or bar plots, ape (version
5.5)(Paradis & Schliep, 2019) and ggtree (version3.0.4)(Yu, 2020) for phylogenetic tree,
and Complex Heatmap (version 3.3.5)(Gu et al., 2016) for heatmaps. Chi-square was
used for categorical data association analysis and Mann-Whitney U test was used for
non-parametric data (i.e., ion count).

2.4 Results and Discussion
Patient demographics and HDSCA3.0 workflow
The index patient was diagnosed with de novo mPC at age 69 with high-volume
prostatic adenocarcinoma, Gleason Score 9 (4+5), PSA 66.4 ng/mL, and bone
metastasis. He had acquired castrate resistance after only 12 months of androgen
deprivation therapy (ADT) and was subsequently treated with sequential treatments of
docetaxel (8 cycles), abiraterone (3 months), and cabazitaxel (8 cycles). Following PSA
progression (from 0.7 to 61.1 ng/mL within two months) from cabazitaxel therapy, paired
PB and BMA were collected for HDSCA3.0 analysis before the 3
rd
line chemotherapy
paclitaxel and carboplatin which maintained progression-free state for 2.8 months.
Unfortunately, the patient passed away in 4.7 months after this progression. (Fig. 4A).
Two slides each from PB and BMA were processed with four-channel staining assay for
CTC detection and phenotypic characterization. Furthermore, single cell copy number
27

profiling by whole genome sequencing or targeted proteomics by image mass cytometry
were utilized on selected CTCs for genotype and tissue of origin analysis. (Fig. 4B)

Figure 4. Patient demographic, pathology, and HDSCA3.0 workflow. A. Patient’s
diagnosis, pathology, disease progression, treatment history, and liquid biopsy
timepoint. B. HDSCA3.0 workflow including immunofluorescence staining and imaging,
rare cell detection, CTC classification, molecular profiling, and downstream analyses.

CTC enumeration and phenotypical characterization
CTC subtypes were classified based on mesenchymal features and platelet
attachment status, including epithelial-like CTCs (epi.CTCs), mesenchymal-like CTCs
(mes.CTCs), platelet-coated epithelial-like CTCs (pc.epi.CTCs), and platelet-coated
mesenchymal-like CTCs (pc.mes.CTCs). Consistent with our previous
publication(Carlsson et al., 2017; Chai et al., 2021; Malihi et al., 2018), the
concentration of total CTCs in BMA was ~200-fold higher than in PB (13.82x10
3
vs 66.8
28

cells per mL). Meanwhile a significant diversity of CTC phenotype was observed,
particularly in BMA, with high incidence of CTC clusters and platelet attachment to
CTCs compared to the PB (Fig. 5A-B & Supplementary Fig. 1). Within the BMA, 32.4%
(4.48x10
3
cells per mL) of CTC expressed VIM and 12.9% (1.78x10
3
cells per mL) of
them were coated with platelets. Interestingly, the fraction of VIM positivity was higher in
the platelet-coated group, compared to non-coated cells (64.0% vs 27.8%) (Fig. 5C).
29

Figure 5. CTC subtype and enumeration in the paired PB and BMA.
A.Immunofluorescence images of CTCs from PB sample. Scale bar: 10μm. B.
Immunofluorescence images of CTCs from BMA sample. Color coding: DAPI (blue); CK
(red); VIM (white); CD45/CD31 (green). Scale bar: 10μm. C. Enumeration of CTC
subtypes in each test of PB and BMA.
30

Based on EBImage-generated features from the four different fluorescence channels,
we further evaluated intensity and morphological variations across four CTC subtypes.
Following the feature selection pipeline (Supplementary Fig. 2A), we firstly identified 10
unique image feature groups and further selected one feature per group as
representatives, including 4 intensity features of DAPI, CK, VIM and CD45/CD31 and 6
morphological features, i.e., cell and nucleus sizes, cell and nucleus eccentricities,
cell/nucleus size ratio, and nucleus location in cell (Supplementary Fig. 2B). As shown
in Supplementary Fig. 3, there were minimal differences observed in morphological
features, DAPI, or CK intensities among CTC subtypes except for the VIM and
CD45/CD31 intensities.
Clonal lineage and its variation between PB and BMA
Within PB and BMA samples, 93 cells were sequenced for single cell copy number
profiling, including 85 CTCs (27 epi.CTCs, 33 mes.CTCs, 5 pc.epi.CTCs, 20
pc.mes.CTCs) and 8 “VIM+ only” rare cells. Of all cells, 88 (94.6%) cells presented
clonal alterations and further hierarchical clustering identified one main clone with
multiple subclones (clade2-5, 74 cells) and one minor clone (clade1, 14 cells) (Fig. 6A).
Unlike common copy number profiling of prostate cancer cells(Taylor et al., 2010), the
minor clone (clade1) had only chromosome 13 and 22 losses and its breakpoints of
chromosome 13 loss were different from those in the main clone (Supplementary Fig.
4C). Clonal lineage tracing indicated that the parental cell of the main clone divided into
three different subclones (clade2, clade3, and clade4/5) with the acquisition of different
CNAs and clade5 was generated from clade4 with the additional chromosome 2p gain
(Fig. 6B left) which was further confirmed by phylogenetic tree analysis of individual
31

cells and subclones (Supplementary Fig. 5A-B). Intrasample comparison showed that
the subclone (clade4/5) was uniquely detected in the BMA while the emerging minor
clone (clade1) was only seen in the PB (Fig. 6B mid & right). As for the fraction of each
subclone, 45 of 74 (60.8%) cells were from the clade4-5 (clade4: 30 cells, clade 5:15
cells), 22 cells were from clade2, and 7 cells were from clade3. Due to the small size
and high heterogeneity of clade3, we mainly compare the differences between clade2
and clade4-5 and identified 17 distinct CNAs with numerous oncogenes and tumor
suppressor genes involved (Supplementary Fig. 4A), e.g., 3q gain (PIK3CA, SOX2), 4q
gain (PDGFR), 6q gain (ROS1), and 9q gain (NOTCH1) in the clade4-5 and 2q loss
(ERCC3, ZEB2), 9p loss (JAK2, CDKN2A), and 14q gain (HSP90) in the clade2. There
was no difference of LSTs observed between clade2 and clade4/5 (Supplementary Fig.
4B).
32

Figure 6. Clonal lineage and its variation between PB and BMA. A. Complex heatmap of
single cell copy number profiles from paired PB and BMA grouped by hierarchical
clustering. The “c” in clade IDs means compressed profile due to normal cell
contamination. Red, copy number gain; blue, copy number loss; white, copy number
33

neutral. VIM+ only refers to “VIM+ only” rare cells. B. Clonal lineage in PB and BMA
combination and separate. The number in parenthesis is the number of cells analyzed.

Association between genomic subclone and mesenchymal phenotype
To characterize the relationship between the phenotype and the genotype of the
CTCs from the index patient, we firstly compared fluorescence intensity between
subclones and observed that the clades4/5 showed higher expression of VIM than
clade2 (Supplementary Fig. 6A). To further test if phenotype could be more directly
related to the genotype, we initially compared the fractions of clades 2/4/5 between
VIM+ and VIM- groups in all, non-coated and platelet-coated population. Results
showed higher fraction of clade4/5 (39/54, 72.2%) observed in VIM+ group while higher
fraction of clade2 (19/32, 59.4%) in the VIM- group independent of platelet status,
despite the small sample size in the platelet-coated population (Fig. 7A). Meanwhile, we
examined the clade fraction difference between non-coated and platelet-coated groups
as well and there was no significant difference in the VIM+ population while the
difference in the CTC population was present due to the sample size imbalance in VIM-
population (Fig. 7B). The associations of VIM positivity, platelet attachment, CTC
subtypes, and cluster status with genomic clonality were further visualized by t-SNE
plots with the input of copy number data (Supplementary Fig. 6B).
Meanwhile, we investigated the association between subclones and cluster
phenotype to see if genotypes of cells from the same cluster are heterogeneous or
homogeneous. Among 43 single cells picked from 14 CTC clusters sequenced from the
BMA sample, 10 of them (71.4%) were homogeneous within the same clusters (3
clusters from clade2, 5 clusters from clade4, and 2 clusters from clade5) and the other 4
34

(28.6%) clusters were heterogeneous, including various combinations, i.e., clade2 and
4, clade2 and 3, and clade3 and 5 (Fig. 7B & Supplementary Fig. 7A-B).

Figure 7. Association between genotype and phenotype. A. Association of genomic
subclones with mesenchymal phenotype or platelet attachment in all CTCs or
35

subpopulations. B. Immunofluorescence images and single cell copy number profiles of
representative heterogeneous CTC clusters in the BMA sample. Images order:
composite, DAPI, CK, VIM, and CD45/CD31. The number is the sequencing cell ID.

Proteomic characterization of CTC subtypes
The results of targeted proteomic analysis by imaging mass cytometry are presented
as an expression heatmap (Fig. 8A) and CTC images from scanned regions of interest
(ROI) (Fig. 8B). The combined results show that EpCAM and prostate-specific
biomarkers including AR and PSMA were abundantly expressed in all CTC subtypes
while negative in white blood cells (WBCs) thus confirming their tissue of origin. VIM
expression was significantly higher in mes.CTCs than epi.CTCs, echoing the results of
the immunofluorescence assay(Fig. 8C). PSMA expression was significantly
downregulated while EpCAM, E-Cadherin, and N-Cadherin were significantly
upregulated in mes.CTCs compared to epi.CTCs, and similar differences were
observed between pc.mes.CTCs and pc.epi.CTCs despite not rising to statistical
significance (Fig.8C). In addition, we observed a trend of increasing AR and PCNA in
pc.CTCs compared to CTCs (Fig.8C). Overall, a spectrum of proteomic heterogeneity
was observed across and within CTC subtypes (Fig. 8A-B). Besides, all 31 CTC
clusters showed various intra-cluster proteomic heterogeneity. Examples include cluster
5871, where AR, PSMA and PCNA were uniquely expressed in top right two cells of
cluster and in cluster 5854, the pc.epi.CTC had higher expression of PSMA, EpCAM,
PCNA and beta-catenin compared to the adjacent epi.CTC.
36

Figure 8. Targeted proteomics of CTC subtypes. A. Heatmap of proteomic expression in
CTC subtypes, including AR-N terminal (AR_N), AR-C terminal (AR_C), PSMA,
EpCAM, E-Cadherin, N-Cadherin, Vimentin (VIM), PCNA, and β-catenin. B.
Immunofluorescence and image mass cytometry images of representative CTC
subtypes. C. Comparisons of proteomic expressions among CTC subtypes. Center line
37

of box: median; upper/lower hinges of box: 75% or 25% quartile; upper/lower whiskers
of box: hinge +/- 1.5*IQR (inter-quartile range). *p<0.05, **p<0.01, ***p<0.001.

Discussion
Phenotypic heterogeneity of tumor cells along disease progression arises from
somatic mutations, adaptation to new microenvironments, or resistance against
treatments. The purpose of this study was to distinguish different phenotypic states,
particularly epithelial-like and mesenchymal-like states, of CTCs with clonal identity and
along clonal lineage to analyze genotype-phenotype association and understand CTC
biology in mPC. Previous studies from our group (Chai et al., 2021; Malihi et al., 2018)
and others (Rangel-Pozzo et al., 2020; Soler et al., 2018) have shown the feasibility of
single cell multi-omic approach, e.g., immunostaining, whole genome/exome
sequencing, transcriptomic profiling, to simultaneously dissect phenotypic and genotypic
heterogeneity of detected or in vitro cultured CTCs. HDSCA3.0, as our most upgraded
liquid biopsy workflow, utilizes immunofluorescence (DAPI, CK, VIM, CD45/CD31) to
characterize mesenchymal feature and platelet attachment status of CTCs followed with
either single cell copy number profiling or targeted proteomics. Here, we applied this
approach to paired PB and BMA samples from one prostate cancer patient with bone
metastasis and fast progression following multiple lines of treatments including
hormonal therapies and chemotherapies. In this index patient, we have identified a
subgroup of genetically clonal CTCs that acquired a new phenotypic state, i.e.,
mesenchymal-like state with the additional expression of VIM, which shows the potential
of molecular characterization of EMT on genetically confirmed CTCs in different
compartments of fluid microenvironment.
38

Most importantly, the presence of the mesenchymal-like state in CTCs was
significantly associated with a genetic subclone emerged along the clonal lineage. Prior
studies have revealed that not only epigenetic events and environmental factors could
drive EMT trajectories through transcriptional changes, but also genomic alterations
could impact this cell state transition(Genna et al., 2020). For example, CAMK1D
(localized at chr.10p13) was highly expressed in basal-like breast cancer due to its gene
amplification and exogeneous overexpression could induce EMT(Bergamaschi et al.,
2008). Similarly, high expression of PLS3 was significantly related to copy number gain
of chr.Xq23 which is its genetic locus and promoted EMT through transforming growth
factor (TGF)-β signaling in colorectal cancer cells(Sugimachi et al., 2014; Yokobori et
al., 2013). Oppositely, SLC38A3 (localized at chr.3p21) expression was lower in the
tumor tissue than the adjacent normal tissue, linked to its high frequency of gene
deletion and further experiments revealed that deletion of SLC38A3 could stimulate
EMT in esophageal squamous cell carcinoma (Liu et al., 2020). However, those
potential associations were inferred from the observations in bulk tumor samples or the
validations in cell lines, while intra-patient genotype-phenotype relationship at single cell
level is still unknown, especially in CTCs. Here, our clonal lineage tracing of CTCs has
shown that there were at least 3 distinctive subclones emerged under the main clone,
include clade2, clade3, and clade4/5, based on the optimal number of clades from
hierarchical clustering determined in a mathematical way. More interestingly, the VIM+
group had significantly higher percentage of the clade4/5 subclone while the VIM- group
was enriched with the clade2 subclone. Further comparison analysis between those two
main subclones (clade2 vs clade4/5) had demonstrated a variety of different
39

chromosome or gene level CNAs across the whole genome. We hypothesized the
reasons to be that genetic defects of EMT-related genes could cause mesenchymal
transformation in clade4/5 subclone or loss of cell state transition ability in clade2
subclone, e.g., copy number gains of PDGFRA, PIK3CA, etc in the clade4/5 subclone
while copy number loss of TWIST2, JAK2, etc in the clade2 subclone. In addition to 4
subtypes of CTCs, we also included “VIM+ only” classification of cells to represent the
full spectrum of rare cells detected and to further investigate the potential of CTCs that
completed the epithelial mesenchymal transition and lost epithelial biomarker, e.g.,
cytokeratin (Yang et al., 2021; Zhang et al., 2017). Single cell copy number profiling
showed those cells did not carry the cancer cell genomic architecture (the main clone)
which is consistent with our observation in the previous publication (Chai et al., 2021),
demonstrating clonality in CK+ cell groups in metastatic prostate cancer. Meanwhile,
recent studies have shown hybrid and dynamic phenotypes generated from EMT of
CTCs which enhance adaptation ability, treatment resistance, and metastatic potential
(Balcik-Ercin et al., 2021; Genna et al., 2020; Hassan et al., 2021). We acknowledge
that the mesenchymal phenotype characterized in our immunostaining has yet to be
sufficient to depict the heterogeneity of EMT. We are developing new IF-based assays
which will include additional biomarkers to further characterize the plasticity of EMT on
CTCs, as well as transcriptomic profiling using single cell RNA sequencing.
Spatial heterogeneity of CTCs during metastasis across multiple liquid
compartments have been reported at the transcriptomic and proteomic level perhaps as
an adaptation to different circulation microenvironments (Sun et al., 2021). Here, we
investigated the phenotypic and genotypic differences of CTCs from two different
40

compartments of the fluid microenvironment during tumor metastasis. Consistent with
phenotypic differences in our previous studies (Carlsson et al., 2017; Malihi et al.,
2018), the abundance and phenotypic diversity of CTCs in the BMA are significantly
higher than in the PB which indicates site-specific enrichment and enhanced
heterogeneity of CTCs adapting to the bone marrow microenvironment. In terms of
genotypic differences, the clade4/5 was uniquely detected in the 66.2% (45/68) of clonal
CTCs from BMA as the dominant subclone, while the clade1 minor clone with
chromosome 13 and 22 losses was only detected in PB. Interestingly, the cells in the
minor clone are all VIM positive with or without CK expression, likely to be circulating
tumor microenvironment (TME) cells with clonal alterations while the presence of RB1
and BRCA2 losses in the minor clone indicated the probability of the generation of a
new cancer clone. To determine if they are TME cells with clonal alteration or CTCs with
the new cancer clone, additional biological characterization is warranted. Due to high
CTC abundance and additional observation of phenotypic and genotypic heterogeneity
in BMA, despite its invasiveness, we suggest that BMA could be a robust supplemental
approach, especially when CTC is not detected in PB. Meanwhile, BMA is still less
invasive than prostate biopsy and also provides a unique perspective of metastatic
cancer cells which usually present aggressive behavior.
CTC clusters or aggregates represent another unique phenotype and studies has
associated their presence with worse prognosis (Carlsson et al., 2017) as well as the
epigenetic regulations during its formation (Gkountela et al., 2019). Here, we explored
the genotypic and phenotypic heterogeneity of CTCs from the same clusters. Out of 14
CTC clusters from the BMA, 4 of them had heterogeneous subclones and the rest of
41

them possessed the homogeneous subclones. This observation relates to the
hypothesis of CTC cluster formation: 1) the cluster with the same subclone could be
formed by cell division from one parental cell or cell aggregation of tumor cells with the
same genetic background; 2) the cluster with different subclones could be formed by
physical aggregation between 2 or more cells with various genetic backgrounds. It will
be interesting to investigate different mechanisms of CTC cluster formation as well as
their potential clinical significance.

2.5 Conclusion
Overall, this case report demonstrated a robust single cell multi-omic approach to
simultaneously deconvolute genotype and phenotype of CTCs in paired liquid biopsy
samples from bone metastatic prostate cancers and to sophisticatedly delineate the
potential mechanisms of hematogenous dissemination and bone colonization in the
aggressive disease. The uniqueness of this case is the identification of epithelial-like
and mesenchymal-like CTCs within the clonal lineage and it avoids the common
misclassification of mesenchymal cells as mesenchymal state of CTCs with protein
marker only. Most importantly, association analysis connects a subclone genotype with
the mesenchymal phenotype at the single cell level. Besides, we hypothesize that the
presence of clonal CTCs with mesenchymal state could be a potential prognostic
biomarker that requires a cohort of patients for further validation, as well as a candidate
target that can be leveraged for new drug development.

42

Chapter 3
Identifying the Combination of Platelet-coated CTC and Aggressive Variant Prostate
Cancer Molecular Signature as a Predictive Biomarker

This chapter is published as:
Shoujie Chai, Nicholas Matsumoto, Ryan Storgard, Chen-Ching Peng, Ana Aparicio,
Benjamin Ormseth, Kate Rappard, Katherine Cunningham, Anand Kolatkar, Rafael
Nevarez, Kai-Han Tu, Ching-Ju Hsu, Paymaneh Malihi, Paul Corn, Amado Zurita,
James Hicks, Peter Kuhn, Carmen Ruiz-Velasco. Platelet-Coated Circulating Tumor
Cells Are a Predictive Biomarker in Patients with Metastatic Castrate-Resistant Prostate
Cancer. Mol Cancer Res. 2021 Dec;19(12):2036-2045. doi: 10.1158/1541-7786.

3.1 Abstract
Metastatic Castration-Resistant Prostate Cancer (mCRPC) includes a subset of
patients with particularly unfavorable prognosis characterized by combined defects in
at least two of three tumor suppressors PTEN, RB1, and TP53 as aggressive variant
prostate cancer molecular signature (AVPC-MS). We aimed to identify CTC signatures
that could inform treatment decisions of mCRPC patients with cabazitaxel-carboplatin
combination therapy versus cabazitaxel alone. Liquid biopsy samples were collected
prospectively from 79 patients for retrospective analysis. CTCs were detected,
classified, enumerated through a computational pipeline followed by manual curation,
and subjected to single cell genome-wide copy number profiling for AVPC-MS
detection. Based on immunofluorescence intensities, detected rare cells were
classified into 8 rare cell groups. Further morphological characterization categorized
CTC subtypes from 4 cytokeratin positive rare cell groups, utilizing presence of
mesenchymal features and platelet attachment. Of 79 cases, 77 (97.5%) had CTCs,
43

24 (30.4%) were positive for platelet-coated CTCs (pc.CTCs) and 25 (38.5%) of 65
sequenced patients exhibited AVPC-MS in CTCs. Survival analysis indicated that the
presence of pc.CTCs identified the subset of AVPC-MS positive patients with the worst
prognosis and minimal benefit from combination therapy. In AVPC-MS negative
patients, its presence showed significant survival improvement from combination
therapy. Our findings suggest the presence of pc.CTCs as a predictive biomarker to
further stratify AVPC subsets with the worst prognosis and the most significant benefit
of additional platinum therapy.

3.2 Introduction
Advanced prostate cancer presents in certain cases with clinical characteristics of a
distinct subtype of the disease known as aggressive-variant prostate cancer (AVPC)
(Aparicio et al., 2013; Aparicio et al., 2016; Manucha & Henegan, 2020). This subtype is
characterized by an aggressive clinical course with poor prognosis and more frequently
arises following hormonal therapy in patients with metastatic castration-resistant prostate
cancer (mCRPC) (Aparicio et al., 2013; Aparicio et al., 2016; Manucha & Henegan, 2020).
Studies have shown that while AVPC responds poorly to hormonal therapies, it is
vulnerable to chemotherapy, although the response is short-lived (Aparicio et al., 2013;
Corn et al., 2019). Our current knowledge about this aggressive variant is incomplete and
guidelines on specific treatment recommendations have not been established. To
overcome this challenge, diagnostic and predictive biomarkers are urgently needed to
stratify mCRPC patients for this subtype who could benefit from alternative therapeutic
interventions.
44

Although several clinicopathological criteria of AVPC (AVPC-C) were initially described
to facilitate its recognition (Aparicio et al., 2013), clinical manifestations are difficult to
identify, necessitating molecular biomarkers. So far, genomic characterization of solid
biopsies from primary and metastatic tumor tissues has revealed that AVPC can be
characterized by a molecular signature (AVPC-MS) composed of combined defects in at
least two of the three tumor suppressor genes PTEN, RB1, and TP53 (Aparicio et al.,
2016; Hamid et al., 2019). Importantly, the presence of this molecular signature may
predict AVPC vulnerability to taxane-platinum combination therapies relative to taxane
alone (Corn et al., 2019). However, a missing aspect in the clinical management of
mCRPC is the minimally invasive assessment of changes in the tumor to identify patients
with AVPC in a timely manner at the correct time point for the appropriate therapeutic
decision-making. This aspect is ideally suited for a liquid biopsy approach.
Liquid biopsy approaches offer a potential source of circulating tumor cells (CTCs)
amongst a spectrum of analytes that can be serially obtained during the course of the
disease and treatment interventions (Dago et al., 2014; Shishido et al., 2019; Welter et
al., 2020). This may overcome critical limitations of solid biopsies in dissecting the
intratumor heterogeneity and its evolution in an individual patient over time (Russano et
al., 2020; Russo & Bardelli, 2017). Successfully used in clinical management of late-stage
prostate cancer (Scher et al., 2018; Scher et al., 2016), the direct imaging liquid biopsy
HDSCA (High Definition Single Cell Assay) (Marrinucci et al., 2012) is used here in an
advanced implementation for the quantitative separation of multiple CTC subtypes. In
addition to describing phenotypic, morphometric, and organizational features, the HDSCA
allows for parallel downstream single cell genomic and targeted proteomic
45

characterization via low-pass whole-genome sequencing (6) and imaging mass cytometry
(Gerdtsson et al., 2018; Malihi et al., 2018), respectively. It has been used in both
peripheral blood (PB) and bone marrow aspirates (BMA) to discern features most
significant to the dynamics of metastatic progression in advanced PC (Carlsson et al.,
2017; Malihi et al., 2018) and to portray genomic instability as a distinctive feature in
patients with AVPC in comparison to patients with mCRPC that are AVPC-negative (Malihi
et al., 2020).
Herein, we report on the use of the advanced capabilities of the third generation HDSCA
(HDSCA3.0) in separating a wider spectrum of CTC subtypes and other disease-related
cells from patients with mCRPC. Its application in mCRPC identifies platelet-coated CTCs
(pc.CTCs) as a predictive biomarker for improved response to combination versus single
chemotherapy.

3.3 Methods
Clinical trial and patient selection
PB and BMA samples were collected from participants immediately starting treatment on
trial NCT01505868 entitled “Study of cabazitaxel with or without carboplatin in patients
with metastatic castration-resistant prostate cancer”. The trial aimed to test the hypothesis
that carboplatin improves the efficacy of cabazitaxel in men with advanced PC with the
additional intention of evaluating the effect of aggressive variant features on response
and outcome. This was a prospective, randomized, open-label, phase 1 (n=9 patients)
and 2 (n=160 patients) study at the University of Texas MD Anderson Cancer Center and
Barbara Ann Karmanos Cancer Institute (Corn et al., 2019). All patients were required to
46

have castration-resistant disease progression, an Eastern Cooperative Oncology Group
(ECOG) performance status between 0 and 2, and adequate organ function. For phase
2, patients were stratified by factors including ECOG performance status, previous
docetaxel treatment, response to docetaxel among those who received it, and the
presence of at least one of the seven AVPC-C criteria (Aparicio et al., 2013). The study
was approved by the corresponding institutional review boards and was conducted in
accordance with ethical principles founded in the Declaration of Helsinki. All patients gave
written informed consent.
For this study, we evaluated a subset of 79 patients based on the availability of liquid
biopsy samples (PB and BMA) for analysis. All patients had at least one sample type
collected and 48 had paired samples. A total of 68 PB and 60 BMA samples were
analyzed. PB samples from 11 normal blood donors (NBD) from our repository, and
defined as individuals with no known malignancy, were used as negative controls.
Biospecimen collection, preparation, and imaging
PB (7.5 mL) and BMA (7.5 mL) samples were collected in 10-mL collection tubes (Cell-
free DNA BCT, Streck) at the clinical sites and sent to the Convergent Science Institute
in Cancer (CSI-Cancer) at the University of Southern California, for sample processing
within 24 hours as previously described (Carlsson et al., 2017; Malihi et al., 2020;
Marrinucci et al., 2012). In brief, upon receipt, samples were subjected to erythrocyte lysis
in isotonic ammonium chloride solution and the entire nucleated cell population was
plated as a monolayer onto custom cell adhesion glass slides (Marienfeld, Lauda,
Germany) with approximately 3 million cells per slide. The cells were then incubated in
7% BSA, dried, and stored at −80°C for long-term storage before use.
47

Slides were stained with the use of an IntelliPATH FLX™ autostainer (Biocare Medical
LLC, Irvine, CA, USA) in batches of 50 (46 patient slides [2 per patient], plus 2 NBD slides
and 2 NBD slides spiked with LnCAP cells (ATCC® CRL-1740™) as negative and
positive quality controls, respectively). All steps were performed at room temperature.
Slides were thawed for 1 hour and cells were then fixed with 2% neutral buffered formalin
solution (VWR, San Dimas, CA) for 20 minutes. Non-specific binding sites were blocked
with 10% goat serum (Millipore, Billerica, MA) for 20 minutes. Slides were subsequently
incubated with a conjugate containing 2.5 ug/mL of a mouse IgG1 anti-human
CD31:Alexa Fluor® 647 mAb (clone: WM59, MCA1738A647, BioRad, Hercules, CA) and
100 ug/mL of a goat anti-mouse IgG monoclonal Fab fragments (115-007-003, Jackson
ImmunoResearch, West Grove, PA) for 4 hours. After incubation with CD31-Fabs, cells
were permeabilized using 100% cold methanol for 5 minutes. They were then incubated
with an antibody cocktail consisting of mouse IgG1/IgG2a anti-human cytokeratin (CK) 1,
4, 5, 6, 8, 10, 13, 18, and 19 (clones: C-11, PCK-26, CY-90, KS-1A3, M20, A53-B/A2,
C2562, Sigma, St. Louis, MO), mouse IgG1 anti-human CK 19 (clone: RCK108,
GA61561-2, Dako, Carpinteria, CA), mouse anti-human CD45:Alexa Fluor® 647 (clone:
F10-89-4, MCA87A647, AbD Serotec, Raleigh, NC), and rabbit IgG anti-human vimentin
(VIM): Alexa Fluor® 488 (clone: D21H3, 9854BC, Cell Signaling, Danvers, MA) for 2
hours. Slides were then incubated with Alexa Fluor® 555 goat anti-mouse IgG1 antibody
(A21127, Invitrogen, Carlsbad, CA) and counterstained with 4′,6-diamidino-2-
phenylindole (DAPI; D1306, Thermo Fisher Scientific, Waltham, MA) for 1 hour. Finally,
slides were mounted with a glycerol-based aqueous mounting media before adding
coverslips to maintain cell integrity. Meanwhile, platelet confirmation was performed with
48

manual staining with VIM replaced by rabbit IgG anti-human CD61 (clone: SJ19-09; MA5-
32077, Thermo Fisher Scientific, Waltham, MA) on the NBD sample spiked with SKRB3
cells (ATCC® HTB-30™) followed by Alexa Fluor® 488 goat anti-rabbit IgG antibody
(A11034, Invitrogen, Carlsbad, CA).
The cells on the slides were imaged as previously reported (Marrinucci et al., 2012).
Briefly, the slides were scanned using automated high-throughput fluorescence scanning
microscopy at 10x objective magnification and producing 2304 frame images per
fluorescence channel per slide. Exposure times and gain for Alexa Fluor® 555 (CK),
Alexa Fluor® 488 (VIM or CD61), Alexa Fluor® 647 (CD45/CD31), and DAPI (DNA)
fluorescence channels were automatically set up by the scanning microscope set to yield
the same background intensity level across slides for normalization purposes.
Rare cell detection and CTC subtype classification
The HDSCA methodology is based on cell imaging data analysis to identify several
rare cells in the blood as opposed to cell enrichment approaches that can only identify a
specific cell phenotype. For this reason, we first detect and classify rare cells amongst all
nucleated cells to then identify CTC subtypes within them based on morphologic features.
Rare cell candidates are detected using OCULAR (Outlier Clustering Unsupervised
Learning Automated Report), a custom pipeline based on image processing,
dimensionality reduction and unsupervised clustering methods. The algorithm first uses
the “EBImage” R package (Pau et al., 2010) (EBImage_4.12.2) to segment the images of
every event on the slide, separating DAPI
+
(cells) and DAPI
-
events. OCULAR then
performs feature extraction for each cell with the “computeFeatures” function from the
“EBImage” package, collecting 761 quantitative cellular and nuclear parameters. Principal
49

component analysis (PCA) of these parameters followed by a hierarchical clustering on
principal components (using the top 350 components) are then applied to identify rare
cells and common cells from all DAPI
+
events. Since PCA and hierarchical clustering are
static methods and introduce no stochasticity, all results are repeatable.
Rare cells are further classified into eight distinct groups based on their fluorescence
signal intensities for Alexa Fluor® 555 (CK), Alexa Fluor® 488 (VIM), and Alexa Fluor®
647 (CD45/CD31) channels. Importantly, OCULAR is independent of specific biomarkers
in the fluorescence channels and is compatible with other fluorescence protocols
including previous versions of the HDSCA (Marrinucci et al., 2012). Further manual
curation of the classified rare cells is performed to remove artifacts and non-interesting
cells providing the cleaned rare cell groups and their cell counts. Subsequently,
morphologic parameters are used to distinguish 4 CTC subtypes amongst the 4 CK
+
rare
cell groups and details were described in the results.
Single cell CNA analysis
We followed previously described methods for rare cell relocation, re-imaging, isolation,
next-generation sequencing, and copy number alteration (CNA) analysis (Baslan et al.,
2012; Dago et al., 2014; Malihi et al., 2018). In brief, slides were transferred to a Nikon
80i microscope, cells of interest were relocated using registered coordinates, and 40x
images were captured. Subsequently, individual cells were extracted from slides using a
robotic micromanipulator system followed by single cell whole genome amplification
(WGA; Sigma-Aldrich; Cat# WGA4). Libraries were constructed using the DNA Ultra
Library Prep Kit (New England Biolabs; Cat# E7370) and sequenced using Illumina HiSeq
at USC Genomics Core or Fulgent Genetics (Temple City, CA). The reads uniquely
50

mapped to the human genome were used to reconstruct the copy number profile of each
individual cell. Only cells with total reads above 30,000 per cell, total alignment rate above
50%, non-significant noise, and no apoptosis-induced alterations were included in the
analysis. The bin-based copy number gains or losses was used to estimate gene-level
CNAs in PTEN, RB1 and TP53 for AVPC-MS assessment.
Survival and statistical analysis
The clinical characteristics and outcomes, including progression free survival (PFS)
and overall survival (OS), were previously reported (Corn et al., 2019) and subsequently
updated for further correlation analysis. PFS was calculated from the date of specimen
collection to the date of first occurrence of progression or last follow-up used for
censorship. OS was calculated from the date of specimen collection to the date of death
or last follow-up used for censorship. For the Kaplan–Meier analyses, via the “survival”
and “survminer” R package (versions 3.1.7 and 0.4.8), two-sample and multiple
comparison log-rank tests and univariate models were used to measure observed survival
differences between groups. The “Complex Heatmap” R package (version 2.1.0) was
used to generate summary heatmaps of CTC subgroup presence and AVPC-MS status.
In the tables, categorical features were listed as numbers of cases and their percentages
and numeric features were presented as median values and their ranges. Association
analyses were performed via a student’s t-test (two groups) for parametric data, Mann–
Whitney U test (two groups) for non-parametric data, and chi-squared test for categorical
data.

51

3.4 Results and Discussion
Identification of CTC subtypes from CK
+
rare cell groups
In this study 68 PB and 60 BMA samples from 79 mCRPC patients and PB samples
from 11 NBDs were evaluated. Patients’ clinical characteristics at time of draw are
summarized in Table 3.

52

Table 3. Patient demographic and clinical characteristics

We utilized immunofluorescence intensity and morphological features to perform the
biological categorization of rare cell groups. A trained analyst further curated the detected
rare cells from those four CK
+
groups and identified CTC subtypes from them (Fig. 9A):
53

epithelial-like CTCs (epi.CTCs) were classified as cells that are CK positive and
CD45/CD31 negative identified from the “CK
+
only” group, with distinct appearing nucleus
by DAPI morphology as previously described (Marrinucci et al., 2012). Epi.CTCs
expressing VIM were classified as mesenchymal-like CTCs (mes.CTCs) from the “double
positive CK|VIM” group. Additionally, epi.CTCs and mes.CTCs presenting a punctuated
pattern in the Alexa Fluor® 647 (CD45/CD31) corresponding to CD31 signal from
platelets were tracked separately as platelet-coated (pc) epi.CTC (pc.epi.CTC) and
mes.CTC (pc.mes.CTC), respectively from the “double positive CK|(CD45/CD31)” and
“triple positive CK|VIM|(CD45/CD31)” groups. These punctuated particles were also
observed by the immunofluorescence staining with the platelet-specific biomarker CD61
in the spiked NBD sample (Supplementary Fig. 8).
54

Figure 9. Identification of CTC subtypes. A. Representative cell images of CTC
subtypes; Blue: DAPI; Red: CK; White: VIM; Green: CD45/CD31 in merged image. B.
Enumeration table of CTCs including subtypes in PB and/or BMA samples from 79
mCRPC patients and PB samples from 11 NBDs. C. Presence of CTC and its subtypes
in PB and/or BMA samples from 79 mCRPC patients. Epi: epithelial; mes:
mesenchymal; pc: platelet coated.

Next, we applied the newly identified CTC subtypes into the same group of 79 mCRPC
patients and 11 NBDs in Fig. 9B-C & Supplementary Fig. 9. Overall, CTCs were detected
in 77 of 79 (97.5%) patients with median enumeration 6.0 cells per mL in the PB and
231.2 cells per mL in the BMA. Regarding the CTC subtypes, epi.CTC was the dominant
subtype, detected in 77 of 79 (97.5%) patients, while a significant reduction of frequency
55

was observed in other three subtypes, 50/79 (63.3%) with mes.CTC, 23/79 (29.1%) with
pc.epi.CTC and 11/79 (13.9%) with pc.mes.CTC. As for the difference between PB and
BMAs, the presence frequency of mes.CTC (71.7% vs 26.5%), pc.epi.CTC (35.0% vs
2.9%) and pc.mes.CTC (18.3% vs 0.0%) were higher in the BMA than the PB samples.
Also, we combined pc.epi.CTC and pc.mes.CTC and found 24 of 79 (30.4%) patients that
had at least 1 pc.CTC in the PB and/or BMA. In the PB samples from 11 NBDs, only two
of them were detected with low numbers of CTCs, one with 1 epi.CTC and the other one
with 2 mes.CTCs and none of them had pc.epi.CTC or pc.mes.CTC.
Molecular signature of AVPC in CTCs
In addition to clonality assessment, single cell CNA analysis was also utilized to
determine copy number changes for individual gene regions, particularly the three tumor
suppressor genes, PTEN, RB1, and TP53 (Fig. 10A). The molecular signature of AVPC
in single CTC (AVPC-MS) is defined as losses of at least two of those three tumor
suppressor genes as previously described (Malihi et al., 2020). CTCs from PB and/or
BMA samples were individually sequenced across the 65 patients. A patient qualified as
AVPC-MS if at least one CTC matched the AVPC-MS criteria. Of these patients, 25
(38.5%) were AVPC-MS positive (Fig. 10B): 12 of them harbored all three gene losses
while the other 13 patients had two gene losses within which the combination of PTEN
and RB1 (7/13, 53.8%) was more frequent than the other two combinations. As for
individual gene analysis, the RB1 frequency (28/65, 43.1%) was higher than PTEN (22/65,
33.8%) and TP53 (20/65, 30.8%) and a high percentage of homozygous loss was
observed in PTEN (12/22, 54.5%) while only a few or none existed in RB1 (4/28, 14.3%)
or TP53 (0/20, 0.0%). To further assess the sensitivity in different sample types, we
56

compared 46 PB with 48 BMA samples and did not observe significant differences either
for the AVPC-MS detection or single gene CNA analysis.

Figure 10. AVPC-MS detection in CTCs. A. Copy number profiling of CTC with PTEN,
RB1 and TP53 losses B. AVPC-MS and its individual gene CNA status in PB and/or
BMA samples. Homo: homozygous loss; hetero: heterozygous loss.

Next we analyzed the concordance with previously published AVPC-C, AVPC-MS in
primary tumor by immunohistochemistry (AVPC-MS-IHC) and AVPC-MS in circulating
tumor DNA by next-generation sequencing (AVPC-MS-ctDNA) (Corn et al., 2019) and
found that AVPC-MS in CTCs and ctDNA shared the highest concordance (21/33, 63.6%)
compared to the concordances with AVPC-C (35/65, 53.8%) and AVPC-MS-IHC (15/24,
62.5%) (Supplementary Table 1). Consistent with previous clinical correlation analysis of
AVPC-C, AVPC-MS-IHC and AVPC-MS-ctDNA (Corn et al., 2019), patients with AVPC-
MS in CTCs showed more aggressive phenotypes, including significantly higher prostate
57

specific antigen (PSA), lactate dehydrogenase (LDH) and bone specific alkaline
phosphatase (BAP), compared with negative counterparts (Supplementary Table 1) as
well as shorter PFS (4.7 vs 6.0 months), OS (14.6 vs 21.2 months) (Supplementary Fig.
10A) and greater treatment response from combination vs cabazitaxel (6.4 vs 3.0 months),
despite absence of statistical significance (Supplementary Fig. 10B-C).
Survival analysis using pc.CTC and AVPC-MS
To establish the clinical utility of AVPC-MS in CTCs, we examined the value of pc.CTC
in combination with the AVPC-MS as prognostic and predictive biomarkers in mCRPC.
Patients with pc.CTC and AVPC-MS positive status had a median PFS of 1.7 months
versus 5.8 months in patients with pc.CTC but AVPC-MS negative (p=0.08) and 6.0
months in those without pc.CTC (p<0.01). Patients with pc.CTC and AVPC-MS positive
had a median OS of 8.2 months versus 27.3 months (p<0.01) and 19.9 months (p<0.001)
for the same groups (Fig. 11A). Patients with pc.CTC and AVPC-MS positive had a
median PFS of 2.3 months when treated with cabazitaxel versus 1.8 months when treated
with the combination (HR = 0.85, 95%CI = 0.19-3.87, p = 0.84) and a median OS of 10.6
months versus 8.1 months (HR = 1.83, 95%CI = 0.32-10.38, p = 0.49) (Fig. 11B). Patients
with pc.CTC but AVPC-MS negative had a median PFS of 3.7 months when treated with
cabazitaxel versus 7.8 months when treated with the combination (HR = 0.36, 95%CI =
0.11-1.22, p = 0.10) and a median OS of 20.7 months versus 41.3 months (HR = 0.17,
95%CI = 0.03-0.88, p = 0.03) (Fig. 11C). Survival analysis of patients without pc.CTCs
are shown in (Supplementary Fig. 11).
58

Figure 11. Survival analysis using pc.CTC and AVPC-MS status. A. Survival analysis by
the integration of pcCTC and AVPC-MS for PFS and OS. B-C. Subgroup analysis by
single or combination therapy for PFS and OS. C: Cabazitaxel only; CC: Cabazitaxel and
Carboplatin.

Discussion
The purpose of this study was to expand the available predictive biomarkers for
directing treatment of mCRPC using an enhanced version of the HDSCA liquid biopsy.
Non-invasive methods for patient stratification are critical for molecular characterization
in clinical scenarios such as AVPC where the disease subtype emerges, advances or
changes over the course of treatment. The commercialized version of the HDSCA has
previously demonstrated clinical utility and been recently approved for reimbursement for
the evaluation of patients with advanced prostate cancer as the Oncotype DX AR-V7
Nucleus Detect ® test (Scher et al., 2018; Scher et al., 2016). We have now extended the
capabilities of the HDSCA to identify a broader spectrum of rare cells in PB and BMA
samples from patients with progressive mCRPC. This version of the assay combines an
59

automated rare cell detection system (OCULAR) that sorts rare cell populations based on
different combinations of CK, VIM, and CD45/CD31 into eight distinct rare cell groups.
Additional morphologic characterization was then used to subdivide the eight staining
groups and to interpret their clinical and biological importance. This detailed analysis led
to the identification, among others, of platelet coated cells that are the focus of the clinical
results we report here.
The combination of CD31 with CD45 into the same channel technically solves the
problem of 5 markers into the HDSCA 4-channel system and utilizes the imaging features
to separate different phenotypes. This enables the observation of the platelet-CTC
interaction due to the robust expression of CD31 in the platelets, megakaryocytes or
endothelial cells (DeLisser et al., 1993). The further separation between CD31 and CD45
relies on subcellular location (e.g., membrane vs cytoplasm), combination with other
markers (e.g. VIM) and morphological features such as cell shape. Of particular note for
this study were cells in which the CD31 signal appeared to be localized to the perimeter
of the cell with punctuated pattern, and which we identify as putative pc.CTCs. This
identification, echoed by CD61 immunofluorescence staining in the spiked NBD sample,
requires future multiple platelet biomarkers confirmation with patient samples using the
Imaging Mass Cytometry. Although the role of platelet association with CTCs has not
been established clinically (Brady et al., 2020; Jiang et al., 2017), studies in model
systems have suggested multiple roles for platelet-coated CTCs in cancer progression,
from stimulating the epithelial-mesenchymal transition (EMT) and extravasation from
blood vessels, to protection from immune surveillance as well as physical shear stress
and extended survival in circulation (Heeke et al., 2019; Labelle et al., 2011; Lou et al.,
60

2015; Placke et al., 2012; Wang et al., 2018; Ward et al., 2021). Several molecular
mechanisms were postulated from in vitro and in vivo experiments, including the activity
of CD97, an adhesion G protein-coupled receptor, as the adaptor protein to bridge platelet
and tumor cells to stimulate bidirectional signaling including ATP-release induced
endothelial disruption and Rho-activation induced cell migration, eventually leading to
cancer metastasis (Ward et al., 2018). Another study discovered that Hsp47, a chaperone
facilitating collagen secretion and deposition, could be genetically amplified and highly
expressed in circulating tumor cells to enhance the platelet attachment (Xiong et al.,
2020). Further, the potential impact of platelets has been inferred mainly from studies on
the hematogenous dissemination of CTCs. In this study, the first to observe them in a
clinical liquid biopsy, we found that the pc.CTCs were more frequently detected in BMA
rather than PB samples, suggesting a potential role in bone metastases of mCRPC. Given
these pivotal roles of platelets as a potential prognostic and predictive biomarker, we
further evaluated this CTC subtype in combination with the AVPC-MS status and
compared it with the AVPC-MS biomarker alone.
Our group has previously reported that the presence of AVPC-C and AVPC-MS in
primary tumor and ctDNA is associated with clinically meaningful improvements in both
PFS and OS when treated with the carboplatin/cabazitaxel combination therapy (Corn et
al., 2019), and that AVPC-MS in CTCs is associated with poor prognosis (Malihi et al.,
2020). We therefore sought to evaluate whether AVPC-MS discerned by CTCs from PB
and/or BMA could predict improvements in survival with the combination
carboplatin/cabazitaxel relative to cabazitaxel alone. We therefore investigated whether
the presence of pc.CTC, in combination with AVPC-MS in CTCs could predict improved
61

performance be associated with improvements in PFS and OS with the
carboplatin/cabazitaxel combination therapy relative to cabazitaxel alone. Strikingly, our
results show that patients with pc.CTC and AVPC-MS had the shortest OS (8.2 months).
Contrary to our previously reported data using AVPC-MS in IHC and/or ctDNA alone that
showed improvements in PFS and OS with the combination therapy, we found no benefit
in PFS or OS in the patients with this combined signature. Interestingly, we showed that
patients with pc.CTC but without AVPC-MS had the longest OS (27.3 months) and
present significant improvement in OS when treated with the combination therapy (41.3
months). Regardless of the AVPC-MS status, patients with no detectable pc.CTCs
showed no significant differences in survival and a modest improvement of PFS and OS
was observed with the combination. Thus, to our knowledge, we are the first to include
the presence of pc.CTCs and AVPC-MS in CTCs as a combined predictive and
prognostic biomarker for the stratification of patients that could benefit from the addition
of carboplatin to cabazitaxel regime or present unfavorable prognosis in advanced
prostate cancer. Taken together, for any given patient a potential decision tree hypothesis
can be generated: the addition of platinum to taxane would be suggested to the patients
with pc.CTC
+
/AVPC-MS
-
, while taxane only would be applied for the pc.CTC
+
/AVPC-MS
+

and the pc.CTC
-
groups (Supplementary Fig. 12).
Meanwhile, there are several caveats which concern the validity of results. The original
trial had enrolled 169 patients while liquid biopsy samples were only collected from 79
patients for retrospective analysis. Among those patients, 24 of them (30.4%) were
detected with pc.CTCs, 25 of 65 (38.5%) sequenced patients were positive for AVPC-MS,
and 8 of them (12.3%) were observed with the dual presence of pc.CTCs and AVPC-MS.
62

Thus, considering the small cohort size, low incidence of positivity, and retrospective
analysis, the validation of initial observation and its generated hypothesis will, of course,
require additional validation and expansion studies. We additionally did further research
into the incidence of pc.CTCs and found one study that reported 18 out of 61(29.5%)
participants with platelets on CTCs (Brady et al., 2020), similar to our results, which
echoes the necessity of larger cohort studies.

3.5 Conclusion
In conclusion, we have utilized a next generation of the HDSCA liquid biopsy for the
detection, categorization, and genomic characterization of circulating rare cell populations.
We identified the presence of a specific CTC subtype, platelet-coated CTCs and the
tumor suppressor gene molecular signature related to AVPC, as a combined prognostic
and predictive biomarker to taxane-platinum combination therapy. Clinically, the presence
of this signature could extend the therapeutic stratification of advanced prostate cancer
patients in addition to previously identified AVPC biomarkers. Biologically, the platelet
attachment and the acquisition of AVPC-MS in CTCs might decipher the mechanisms of
aggressiveness in the subset of mCPRC patients, which could be new therapeutic targets
for drug developments.

63

Chapter 4
Early Detection of ‘Aggressive Variant’ Genotype in CTC and ctDNA from Men with
Treatment-naïve De Novo Metastatic Prostate Cancer

This chapter is prepared as a manuscript reviewed by all co-authors and ready for
submission:

Shoujie Chai, Paymaneh Malihi, Brian Chapin, Ana Apricio, Lisa Welter, Nikki Higa,
Amin Naghdloo, Anand Kolatkar, Rafael Nevarez, James Hicks, Peter Kuhn. Early
Detection of Aggressive Genotype in Circulating Tumor Cells and DNA from Men with
Treatment-Naïve de Novo Metastatic Prostate Cancer. Pre-submission

4.1 Abstract
The absence of biomarkers that identify clinically meaningful biological subsets of
prostate cancer (PCa) results in the homogeneous application of therapies to a
heterogeneous disease, both in clinical practice and clinical research, hampering the
development of novel and effective treatments. The aggressive variant PCa molecular
signature (AVPC-MS), composed of combined alterations in TP53, RB1 and PTEN,
stratifies for a subset of metastatic castration-resistant prostate cancer (mCRPC)
associated with androgen-indifference and a dismal prognosis. Whether the AVPC-MS
can be detected in the pre-treatment liquid biopsies of men with hormone-naïve
metastatic (M1) PCa is unknown. In this study, 17 of 29 men with hormone-naïve de
novo M1 PCa harbored CTCs and their enumerations were highly correlated with
clinically defined tumor volume and ctDNA tumor fraction (TFx). Four patients were
positive for AVPC-MS in CTCs and/or ctDNA and presented significantly and
independently shorter progression-free survivals than their signature negative
64

counterparts. Single cell and bulk sequencing showed inter- and intra-patient AVPC-MS
heterogeneity within CTCs, and between CTCs and ctDNA. Moreover, using our novel
immunofluorescence-based High Definition Single Cell Assay (HDSCA), we observed a
strong correlation between CTC phenotypic heterogeneity and genomic alterations. In
conclusion, pre-treatment liquid biopsies can serve to stratify men with advanced
hormone-naïve PCa into clinically-meaningful subsets with the potential to inform
therapy selection. In addition, multi-modal analyses of liquid biopsies provide insight into
tumor heterogeneity and an opportunity to further refine biomarker signatures of
therapeutic relevance.

4.2 Introduction
Prostate cancer (PCa) is a heterogeneous disease with a wide range of therapy
responses and outcomes. While the androgen receptor (AR) is considered the central
driver of the disease, approximately 20% of men with metastases have a poor response
to AR signaling inhibitors and succumb to their disease shortly after their diagnosis
(Fizazi et al., 2017; Hussain et al., 2006; James et al., 2017). The development of
therapies for this ‘androgen-indifferent’ subset has been hampered by the absence of
markers that can identify them accurately. Efforts to develop such markers led to the
definition of the ‘Aggressive Variant PCa’ (AVPC), a subset of the disease that shares
clinical features with small cell or poorly differentiated neuroendocrine carcinomas of the
prostate, a rare, morphologically defined subset of the disease that has an atypical and
virulent clinical behavior, is resistant to androgen signaling inhibitors but shows a high
65

response rate (albeit short-lived) to platinum-based chemotherapies (Aparicio et al.,
2013).
The clinically defined AVPCs are characterized by a molecular signature composed
of combined defects in at least two of the tumor suppressors PTEN, RB1, and Tp53
(Aparicio et al., 2016). The significance of this AVPC Molecular Signature (AVPC-MS) is
supported by preclinical studies, in which loss of two or more of these tumor
suppressors recapitulates metastatic (M1) PCa that is poorly responsive to androgen
signaling inhibition, while loss of one alone does not (Ku et al., 2017; Mu et al., 2017;
Zou et al., 2017). The AVPC-MS identified by immunohistochemistry in solid tumor
biopsies and/or next-generation sequencing of circulating tumor DNA (ctDNA) predicted
benefit from the addition of carboplatin to cabazitaxel in men with metastatic castration-
resistant PCa (mCRPC) participating in a randomized phase II trial (Corn et al., 2019).
Our previous studies showed that AVPC-MS detected in circulating tumor cells (CTCs)
from the same trial, defined an aggressive subpopulation with poor chemotherapy
response and survival and indicated a predictive biomarker for additional platin therapy
in combination with platelet-coated CTCs (Chai et al., 2021; Malihi et al., 2020). In
addition, targeted sequencing of formalin-fixed, paraffin-embedded biopsies from men
with localized or metastatic castration-naïve and castration-resistant PCa showed that
combined alterations in these tumor suppressors were associated with worse outcomes
(Hamid et al., 2019).
In addition to ctDNA characterization in plasma, the High Definition Single Cell Assay
(HDSCA) workflow has been developed as an enrichment-free, immunofluorescence-
based liquid biopsy platform for the enumeration and morpho-genomic characterization
66

of CTCs in multiple cancer types (Gerdtsson et al., 2019; Shishido et al., 2019; Shishido
et al., 2022; Shishido et al., 2020). Our previous publications have established the
HDSCA AR assay for PCa, utilizing DAPI to identify nuclei and antibodies against
epithelial cytokeratin (CK), and CD45 (leukocytes) to identify CTCs (DAPI+, CK+,
CD45-), plus a fourth antibody to measure AR expression levels. Individual stained cells
are then picked by micromanipulation for low-pass single cell DNA sequencing and
copy number profiling to directly confirm the tumor lineage of each cell. This assay has
been applied in localized PCa, as well as biochemically recurrent, castration-sensitive
and castration-resistant PCa (Carlsson et al., 2017; Chalfin et al., 2018; Dago et al.,
2014; Lazar et al., 2012; Malihi et al., 2018), however its potential clinical utility in
advanced treatment-naïve PCa has not been explored. We therefore investigated
whether the AVPC-MS and other molecular signatures could be detected in the CTCs
and ctDNA of men with treatment-naïve de novo M1 PCa, and further, whether these
signatures were predictive of outcome.

4.3 Methods
Liquid Biopsy Cohort, Clinal Trial, and Specimen Preparation
This liquid biopsy cohort consists of blood samples from 29 consecutively enrolled,
treatment-naïve de novo M1 PCa patients in the “M1” trial (“Randomized, Phase II Trial
of Best Systemic Therapy or Best Systemic Therapy (BST) Plus Definitive Treatment
(DT) of the Primary Tumor in M1 PCa”, clinicaltrials.gov/ct2/show/NCT01751438)
conducted at the University of Texas MD Anderson Cancer Center between May 2014
and October 2017. This study sought to gain support for the hypothesis that adding
67

treatment of the primary tumor with surgery or radiation after 6 months of treatment with
standard systemic therapies, would improve the progression free survival (PFS) of men
presenting with de novo M1 PCa. In the liquid biopsy cohort, 6 of 29 patients provided
consent for initial registration but not for randomization at the 6-month timepoint, and
thus survival information is not available. Clinical characteristics, including age, TNM
stage, local symptoms, tumor volume, PSA concentration, and treatment arm were
collected from all 29 patients at the time of diagnosis (Table 4). High tumor volume was
defined per the CHAARTED trial, as visceral metastasis and/or four or more bone
metastasis with at least one outside the vertebral column and pelvis (Kyriakopoulos et
al., 2018). All patients provided written informed consent.
All samples were collected and processed via HDSCA workflow as previously
published (Carlsson et al., 2017; Dago et al., 2014; Malihi et al., 2018). Briefly, samples
were collected in the Cell-Free DNA BCT tubes (Streck; Cat#62790315) at the time of
diagnosis of de novo M1 PCa and shipped overnight to the University of Southern
California. Upon receipt, plasma was extracted for ctDNA analysis, red blood cells were
lysed, and the remaining nucleated cells plated on a custom-made glass slide
(Marienfeld) as a monolayer of ~3.0 × 10
6
per slide. Slides and plasma were cryo-
banked at −80°C for future experiments.

68

Table 4 Demographics of 23 treatment-naïve de novo M1 PCa patients with clinical
information

HDSCA2.0 AR Staining Assay and Tumor Cell Characterization
The protocol of HDSCA2.0 androgen receptor (AR) staining assay has been
previously established (Carlsson et al., 2017; Dago et al., 2014; Malihi et al., 2018). In
brief, two slides from each sample were stained with four markers, DAPI for nuclear
DNA (Thermo Fisher Scientific; Cat# D1306), Pan-Cytokeratin cocktail (pan-CK: Sigma;
Cat# C2562; Clone: C-11, PCK-26, CY-90, KS-1A3, M20, A53-B/A2 and CK19: Dako;
69

Cat# GA61561–2; Clone: RCK108), AR (Cell Signaling Technology; Cat# 5153; Clone:
D6F11), and CD45 (AbD Serotec; Cat# MCA87A647; Clone: F10–89–4) in four
separate fluorescence channels (DAPI, Alexa Fluor® 555, Alexa Fluor® 488, Alexa
Fluor® 647) as shown in Fig. 12A. The stained slides were automatedly imaged using a
computerized fluorescence microscope at 10X magnification and normalized by setting
the exposure to yield the same background intensity level enabling a semi-quantitative
analysis and comparison across slides and samples. Candidate tumor cells were
computationally screened and manually confirmed based on fluorescence intensities
and morphological features generated by the EBImage Bioconductor R Package (Pau
et al., 2010). For fluorescence intensities of candidate cells, standard deviations over
the mean (SDOM) signal value derived from the surroundings cells was calculated for
each channel. Nuclear area, cell area, and ratio of cell/nuclear area were measured for
morphometric analysis.

Sequencing and Copy Number Alteration (CNA) Analysis
Single cells and cell free DNA were isolated and sequenced as previously described
(Carlsson et al., 2017; Dago et al., 2014; Malihi et al., 2018). Briefly, tumor cells were
picked from the slides using a robotic micromanipulator system for subsequent single
cell whole genome amplification (Sigma-Aldrich; Cat# WGA4). Libraries were
constructed using the NEBNext® Ultra™ II FS DNA Library Prep Kit (New England
Biolabs; Cat# E7805) and sequenced using Illumina NextSeq 500 at the USC Genomics
Core. Cell free DNA was extracted from plasma by QIAamp Circulating Nucleic Acid Kit
(QIAGEN; Cat# 55114). Libraries were constructed using NEBNext® Ultra™ II DNA
70

Library Prep Kit (New England Biolabs; Cat# E7645) and sequenced using Illumina
HiSeq 3000/4000 at Fulgent Genetics. The number of total reads was separated into
each of 5000 bins followed by the normalization and segmentation of the bin counts
across each chromosome to generate a genome wide CNV profile (Fig. 12B&C) (Baslan
et al., 2012). The quality control of copy number profile was defined as follows: total
reads above 30,000 per cell, total alignment rate above 50%, and no significant noises.
The bin-based copy-number gains or losses were used to estimate gene-level CNAs for
AVPC-MS assessment (PTEN, RB1, and TP53). Large scale transition (LST) was a
count of large-scale CNAs (> 10Mb) across the genome. Clonal alteration was defined
as two or more CTCs shared two or more same CNAs (same breakpoints) in the patient
or genomic profile similar to published PCa profile if only 1 CTC was available. Tumor
fraction (TFx) in ctDNA was estimated using “ichorCNA” package (Adalsteinsson et al.,
2017).

71

Figure 12. Multi-modal liquid biopsy. A. Immunofluorescence staining for CTC detection,
including DAPI, AR, Pan-CK, and CD45. B. Copy number profile of single CTC with
further genomic characterization of AVPC-MS, LST, and clonality. C. Copy number
profile of ctDNA with further genomic characterization of AVPC-MS and TFx.

72

Statistical Methods
The association analyses were performed by Mann–Whitney U/ Wilcoxon test for non-
parametric data and correlation analyses were conducted by Spearman Correlation test.
Primary outcome measurement is progression-free survival (PFS), defined as the time
interval from the start of initial BST treatment to the date of disease progression or
death, whichever occurred first (or last follow-up if censored). Log-rank test was used
for univariate survival analysis and Cox regression test was used for multivariate
survival analysis. Heatmap of CTC/ctDNA characteristics were generated by the
“Complex Heatmap” R Package (Gu et al., 2016).

4.4 Results and Discussion
PSA Concentration, CTC Enumeration, ctDNA Tumor Fraction, and Tumor Volume
Concentrations of multiple cell-free and cell-based analytes from the blood samples,
including PSA, CTC, and ctDNA, were separately quantified for association analyses in
23 treatment-naïve de novo M1 PCa with clinically defined tumor volume information.
Results showed (Fig. 13A) that the high-volume group had a significantly higher CTC
enumeration compared to the low-volume group (median: 1.6 vs 0.0 cells/mL; p =
0.048) and a similar trend was observed for ctDNA TFx (median:10.0% vs 0.0%; p =
0.062; only 16 of 23 samples available for ctDNA analysis). There was no association
between PSA concentration and tumor volume (median 23.0 vs 24.7 ng/mL; p = 0.801).
Additional, Spearman correlation analyses (Fig. 13B) were performed among those
three blood analytes and we found that CTC enumeration was significantly correlated
73

with ctDNA TFx (R = 0.54; p = 0.03) while no correlations were observed between PSA
and CTC or ctDNA levels.

Figure 13. Concentrations of multi-modal blood analytes and tumor volume. A.
Association analyses between concentration of different blood analytes (PSA, CTC,
ctDNA) and tumor volume. Among 23 treatment-naïve de novo M1 PCa patients with
tumor volume data, all of them were available for PSA and CTC analysis while 16 of
them had ctDNA analysis. B. Spearman correlation analyses among three different
blood analytes.

Molecular Signatures of CTCs and ctDNA
With the inclusion of the 6 patients who did not finish the study and for which we have
no outcome information, 17 of total 29 de novo M1 PCa patients (17/29, 58.6%) had
detectable CTCs, defined as one or more CTCs per test, with a median count of 2.5
cells/mL (Interquartile range (IQR) 1.5-11.2 cells/mL). Seven of 17 CTC positive
patients (7/17, 41.1%) were also AR positive (Fig. 14), defined as one or more AR
positive CTC per test, with a median count of 2.6 cells/mL (IQR 1.6-14.5 cells/mL).
74

Single cell copy number profiling was successfully performed on 14 CTC positive
patients (Fig. 14). Results indicated that 3 of them (3/14, 21.4%) were detected with
AVPC-MS in at least one CTC per patient and 5 of them (5/14, 35.7%) presented high
LST (above 10 events) and clonal alterations in CTCs. Meanwhile, copy number
profiling was also conducted on ctDNA extracted from 22 paired plasma samples (Fig.
14) and 4 of them (4/22, 18.2%) were detected with ctDNA TFx above 10%. Among
them, 3 patients (3/22,13.6%) were positive for AVPC-MS in ctDNA and 2 of them were
concordant with their molecular signatures in CTCs while the third one failed on single
cell genomics of CTC. Overall, the majority of genomic alterations in CTC or ctDNA
were mainly observed in the high-volume group.
75

Figure 14. Molecular signature derived from both CTC and ctDNA profiles. Complex
heatmap of values for PSA, CTC enumeration and ctDNA TFx combined with
components of the molecular signature from CTC and ctDNA in 29 treatment-naïve de
novo M1 PCa patients. Red arrow denotes patients carrying the molecular signature; *
denotes patients discontinued in the trial. Red line indicates 10% ctDNA TFx.

Progression Free Survival Analysis by Molecular Signatures
To evaluate the prognostic value of those molecular signatures in advanced
treatment-naïve PCa, we further investigated their associations with progression free
survival (PFS) among 23 patients with survival information. We first performed the log-
rank test for univariate analysis (Fig. 15A) and found that patients who were positive for
CTCs (7.0 vs 38.5months, p = 0.042), AR expression in CTCs (6.0 vs 37.0months, p =
76

0.021), AVPC-MS (3.5 vs 35.0months, p < 0.001) in CTCs/ctDNA, or high tumor volume
(6.0 vs 45.0 months, p = 0.005) presented significantly shorter PFS compared to their
counterpart groups while opposite trend was observed in patients with above median
PSA (43.5 vs 8.0months, p = 0.095), despite no statistical significance. However, there
was no significant difference of PFS found between two treatment arms (BST+DT vs
BST: 17.0 vs 11.5 months, p = 0.383). Furthermore, we conducted Cox regression test
for multivariate analysis within 4 significant signatures from univariate analysis (Fig.
15B) and results showed that AVPC-MS had an independent association with worse
PFS from other signatures (undetected vs detected: HR 0.17, 95 CI 0.03-1.00, p =
0.049).
77

Figure 15. Univariate and multivariate progression-free survival analysis. A. Univariate
survival analysis by molecular signatures in CTC and/or ctDNA, PSA, tumor volume,
and treatment arm. BST: best systemic therapy; DT: definitive therapy. B. Multivariate
survival analysis by signatures selected from univariate analysis. HR: hazard ratio; CI:
confidence interval.

Genomic Profiling and Heterogeneity in Four AVPC-MS Positive Patients
To further understand genomic heterogeneity of 4 AVPC-MS positive patients, we
analyzed copy number profiles of paired CTCs and ctDNA from those patients (Fig.
78

16A&B). Results indicated that 2 of them (#142 and #148) had AVPC-MS in both CTCs
and ctDNA, 1 patient (#91) was detected with AVPC-MS in CTCs rather than ctDNA,
while the other patient (#87) was positive for AVPC-MS in ctDNA but failed in CTC
sequencing. Inter-patient heterogeneity of AVPC-MS gene composition was observed
among those patients: all 3 genes losses (PTEN, RB1, and TP53) in patients #87 and
#148, losses of PTEN and RB1 in patient #91, and losses of RB1 and TP53 in patient
#142 (Fig. 14). Further single cell analysis illustrated the intra-patient heterogeneity of
AVPC-MS status (Fig. 16A): 5 of 8 (62.5%) sequenced CTCs in patient #148 and 2 of 2
(100%) sequenced CTCs in patient #142 were AVPC-MS positive, while only 2 of 22
(9.1%) sequenced CTCs in patient #91 harbored AVPC-MS which might explain its
negative detection in paired ctDNA. Similar to previous results in breast cancer (Hicks et
al., 2006), a unique “firestorm” pattern on chromosome 8, characterized by multiple
closely spaced alterations, was observed in patient #148, which included 5 gains and 5
losses and covered a number of key genes, such as MYC, FGFR1, and HEY1 gains,
and NCOA2, NDRG1, and RAD21 losses (Supplementary Fig. 13).
79

Figure 16. Genomic profiling of CTC and ctDNA in four AVPC-MS positive patients. A.
Heatmaps of copy number profiles of CTCs by singe cell sequencing. Red arrow
indicates cell with AVPC-MS. Red, copy number gain; Blue, copy number loss; White,
copy number neutral. B. Copy number profiles of ctDNA by bulk sequencing.
80

Phenotypic Heterogeneity and Its Correlation with Genomic Alterations
In order to link the phenotype with the genotype at the single cell level, we
investigated phenotypic features among CTCs with distinctive genomic profiles as well
as their correlation with LST. Out of 69 sequenced CTCs from 14 patients, 29 cells
carried clonal alterations and 9 of them presented AVPC-MS (Supplementary Fig. 14)
while 40 cells had either no (31 cells) or nonclonal (9 cells) alterations. Phenotypic
features of CTCs with different genotypes across patients were visualized in Fig. 17A,
including CK intensity, AR intensity, CD45 intensity, cell size, nucleus size, and
cell/nucleus size ratio. Phenotypic heterogeneity of CTCs was observed at inter- and
intra- patient levels. Meanwhile, there were visible phenotypic variations and LST
differences between different genotypes, although, no statistical analysis was performed
due to the unequal number of CTCs sequenced across patients (Fig. 17A and
Supplementary Fig.15). Furthermore, we used Spearman test to perform correlation
analysis among phenotypic features and LST and found that LST was highly correlated
with CK intensity (p < 0.001) and negatively related to CD45 intensity (p = 0.003) (Fig.
17B).
81

Figure 17. Association between genomic alterations and phenotypic features. A.
Phenotypic features of 69 CTCs with different genotypes across 14 patients. c_AVPC:
clonal alterations with AVPC-MS; c_nonAVPC: clonal alterations without AVPC-MS.
Three cells in patients #22, #85, #95 were missing AR staining. B. Spearman correlation
analyses among phenotypic features and LST. Color indicates correlation coefficient
and size indicates p value.

Discussion
Liquid biopsies are minimally invasive tools to characterize CTCs, cell-free
DNA/RNA/protein, extracellular vesicles, or tumor-educated platelets in the fluid
82

specimen (Siravegna et al., 2017) that complement solid tumor biopsies in early cancer
screening, minimal residue detection, therapeutic selection, and the study of tumor
evolution. The HDSCA workflow is a comprehensive liquid biopsy platform to
characterize both cell-based and cell-free analytes. Here, we applied this workflow to
identify and enumerate CTCs in immunofluorescence images, to determine their AR
expression based on fluorescence intensity, to assess AVPC-MS in CTC by single cell
sequencing, to estimate ctDNA TFx from copy number profile, and to evaluate AVPC-
MS in ctDNA by bulk sequencing, using blood samples from men with de novo M1 PCa
at the treatment-naïve state. In this multi-modal liquid biopsy cohort, CTC and ctDNA
were utilized to detect the AVPC-MS, deconvolute tumor heterogeneity, and explore
phenotype-genotype associations.
The CHAARTED clinical trial demonstrated that tumor volume is a key stratification
factor to identify a subset of hormone-sensitive M1 PCa patients with overall survival
benefit from chemo-hormonal combination therapy over hormonal therapy alone
(Kyriakopoulos et al., 2018). Its definition relies on the assessment of the location and
number of metastatic tumor sites using clinical imaging (see Methods). Here, we
investigated if the concentrations of three blood components (CTC, ctDNA and PSA)
are related to clinically defined tumor volume. CTC enumeration was significantly
associated with tumor volume and similar trend was observed in ctDNA TFx while there
was no association for PSA. This indicates that blood analytes have the potential to
confirm the tumor volume for therapeutic prediction.
Importantly, the presence of CTC and ctDNA in treatment-naïve de novo M1 PCa
enabled the further characterization of molecular signatures to identify the AVPC, a
83

subset with poor prognosis. Overall, 17 of 29 patients (58.4%) had at least one CTC per
test and, of these, 7 (7/17, 41.1%) harbored at least one AR positive CTC per test in the
blood samples. Meanwhile, 4 of 22 patients had ctDNA tumor fractions above 10%.
Previous preclinical studies supported the biological significance of the combined
genomic defects in AVPC-MS genes (PTEN, RB1 and TP53) in transgenic mouse
models, showing that they can initiate PCa and result in antiandrogen resistance (Ku et
al., 2017; Mu et al., 2017; Zou et al., 2017). In this study, we found that AVPC-MS could
be detected in CTCs and/or ctDNA among treatment-naïve M1 PCa patients ahead of
disease progression or drug resistance. Although the small sample size and the
absence of a head-to-head comparison limit the interpretation, our data suggest that the
AVPC-MS is detectable at a lower frequency in the pretreatment, hormone-naive setting
(4/29, 13.7%) than in castrate-resistant disease observed in our previous study (25/65,
38.5%) (Chai et al., 2021). This could be due to an enrichment of the AVPC-MS positive
patient population due to failure of systemic therapy or an evolutionary acquirement of
AVPC-MS related alterations under therapeutic pressure. Consistent with the prognostic
value of AVPC-MS in mCRPC (Chai et al., 2021; Malihi et al., 2020), the identified
subgroup of treatment-naïve de novo M1 PCa with AVPC-MS presented an accelerated
disease progression or hormonal resistance independent from other potential
biomarkers, e.g., CTC presence, AR expression in CTC, or tumor volume. This result
not only supports the prognostic value of the AVPC-MS in the hormone-naïve setting,
but also calls for new trials to include additional therapy early-on for this androgen-
indifferent subset.
84

Similar to previous observation in mCRPC (where 12 of 25 patients had 3-gene
losses while the remaining harbored 2-gene losses) (Chai et al., 2021), this study also
revealed inter-patient AVPC-MS heterogeneity in that 2 of 4 AVPC-MS positive patients
were detected with all three genes losses. Besides, single cell genomics further
illustrated the intra-patient heterogeneity that not all the CTCs from a given patient
harbor the AVPC-MS and the percentage of CTCs positive for the AVPC-MS varies
across patients. Based on the result that patient #148 had only 2 of 22 (9.1%) CTCs
with AVPC-MS and negative signature in ctDNA, we speculated that the low proportion
of specific genomic alteration in CTC might affect its limit of detection in paired ctDNA.
Single cell multi-omic characterization provides an opportunity to examine phenotype-
genotype associations. We observed inter- and intra- patient phenotypic heterogeneity
as well as phenotypic variations between CTCs with different genotypes. Besides,
CTCs with high LST were likely to have higher CK and lower CD45 levels, which needs
larger cohort to further validate the observation. These observations suggest that
phenotypic features from immunostaining may be able to predict genomic alterations in
CTCs.

4.5 Conclusion
In conclusion, AVPC-MS can be detected in CTCs and ctDNA from treatment-naïve
de novo M1 PCa and identifies an aggressive subgroup of patients who presented
accelerated progression following initiation of standard systemic therapies. We show the
feasibility of using multi-modal and multi-omic liquid biopsies for the discovery of
comprehensive biomarkers and the investigation of tumor heterogeneity. This feasibility
85

encourages us to further validate additional biomarkers at the treatment-naïve stage or
even in patients with localized prostate cancer for early detection of aggressive disease.
86

Summary
HDSCA3.0 has widened the horizons of cell-based liquid biopsy applications.
Different from enrichment-based technologies, HDSCA3.0 enables simultaneous
characterization of a variety of cell types and states for many systemic and circulating
diseases. Here, we utilized the Landscape immunostaining protocol and OCULAR
computational pipeline to detect and classify 8 rare cell groups in liquid biopsy samples
from patients with prostate cancer. Furthermore, single cell copy number profiling
identified rare cells with clonal alterations mainly in 4 CK+ groups, and morphology
analysis categorized a spectrum of epithelial, mesenchymal, endothelial, platelet-
related, or immune cell types or states. The development of HDSCA3.0 workflow has
transformed the cell-based liquid biopsy from ‘targeted cell detection’ to ‘whole rare cell
profiling’ as the next generation technology.
CTCs are the key components to connect the primary and metastatic tumors and
represent a variety of biological heterogeneity. Here, HDSCA3.0 provides the detailed
insights into CTC phenotypes and genotypes at the single cell level. Importantly, 4 CTC
subtypes, i.e., epi.CTC, mes.CTC, pc.epi.CTC, and pc.mes.CTC, were identified in the
clonal lineage of an index patient with aggressive prostate cancer. Association analysis
further indicated that the presence of the mesenchymal phenotype was associated with
a subclone genotype that carried copy number alterations of EMT-related genes.
Besides, clonal heterogeneity was detected within CTC clusters and across sample
types. All observations from this patient with HDSCA3.0 analysis have laid a foundation
for the delineation of mechanisms and biomarker discovery of aggressive disease.
87

‘Aggressive variant’ subset or AVPC, constitutes 10~20% of advanced prostate
cancer patients who presented with an unfavorable prognosis but who may benefit from
additional platinum therapy. Both CTC subtypes as a phenotype and AVPC-MS as a
genotype were detected in our cohort. Furthermore, we found that the presence of
pc.CTCs identified a subset of AVPC-MS positive patients with the worst prognosis and
minimal therapeutic benefit from the combination therapy, while also recognizing the
subset of AVPC-MS negative patients with the most benefit from the combination
therapy. Most interestingly, AVPC-MS can be early detected in the CTC and/or ctDNA
in men with treatment-naïve prostate cancer and those positive patients showed worse
PFS compared to their negative counterparts. The applications in those two cohorts
highlight the robustness of a comprehensive liquid biopsy for treatment decision making
and early detection of aggressive disease.
In conclusion, HDSCA3.0 provides a workflow for rare cell profiling to detect a
spectrum of cell types and states, an approach with single cell multi-omic analysis to
deconvolute CTC heterogeneity, a platform to identify liquid biopsy biomarkers for
treatment decision making and early detection of AVPC, and a method to delineate the
biological mechanisms of aggressive disease for new drug developments.
88

Future Directions
In this thesis, we highlighted the deconvolution of CTC heterogeneity using the
developed HDSCA3.0, and its implications to identify prognostic and predictive
biomarkers for treatment decision and early detection of AVPC. As for future directions,
there are several aspects that need to be further explored and we also speculate a
future comprehensive liquid biopsy test to delineate ‘AVPC’.
This first aspect is how to further deconvolute the heterogeneity of CTCs and other
rare cells. EMT is a dynamic process and adding more EMT-related markers in
technologies like IMC could further characterize and quantify its heterogeneity. CTCs
‘never travel alone’ and the investigation of various cellular interactions will be able to
decode the crosstalk between CTCs and the circulatory microenvironment. Aside from
CTC heterogeneity, there is additional heterogeneity of other rare cells and DAPI-
events, e.g., endothelial-like cells, fibroblast-like cells, megakaryocytes, and large
extracellular vesicles. Their biological features, correlation with CTCs, and clinical
utilities need to be studied as the next step.
The second aspect is how to further decipher the mechanisms of aggressive disease.
The mechanistic studies require a series of basic experiments to identify and validate
the key signal pathways and molecules that are involved in various biological processes
and behaviors. For examples, are the TWIST2 loss and PDGFRA gain the key
regulators of EMT in CTCs? Why are the pc.CTCs highly enriched in the BMA
compared with the PB while AVPC-MS is equally distributed in both sample types? Do
CD97 and HSP47 play important roles in the platelet attachment? Can anti-coagulation
drugs reduce the incidence of bone metastasis and improve outcomes? Would the
89

restoration of tumor suppressor genes in AVPC-MS alter the aggressiveness?
Investigating those questions not only provides the detailed insights into the biology of
aggressive disease, but also reveals valuable targets for new drug developments.
The third aspect is how to further validate the implications for AVPC. There are three
potential biomarkers identified in the study, i.e., 1) clonal CTC with mesenchymal state
as a prognostic biomarker in mCRPC, 2) the combination of pc.CTC and AVPC-MS in
CTC as a predictive biomarker for taxane-platinum therapy in mCRPC, 3) early
detection of AVPC-MS in CTC and/or ctDNA as a prognostic biomarker in treatment-
naïve metastatic prostate cancer. Those implications were initially observed in either
single case or small cohort, necessitating larger and ideally prospective clinical trials for
further validation of their clinical value. Meanwhile, we have established a ‘Realtime
AVPC’ project to test the turnaround time and validate the feasibility of AVPC-MS
evaluation across IHC, CTC, and ctDNA for the realtime clinical decision making.
The fourth aspect is how to expand the utility of HDSCA3.0. Beyond AVPC, is it
feasible to detect prostate cancer early and then separate different stages using this
comprehensive liquid biopsy assay? Would it be possible to identify post-surgery
recurrence with rare cell evidence in the liquid samples? How does the heterogeneity
change in longitudinal analysis during disease progression? Those are clinical
questions that we could provide unique perspectives with the HDSCA3.0 technology.
Lastly, based on what we learned here and what others reported previously, we have
pictured a future comprehensive liquid biopsy test to delineate the ‘AVPC’ for patients
with mCPRC. The ‘AVPC’ here is a broader concept that contains, not only the AVPC-
MS defined subset, but also other aggressive subsets that can be detected by various
90

biomarkers in liquid biopsy and achieve benefits from existing treatments beyond the
AR signaling inhibitors or new therapies that need to be developed. With just one tube
of blood sample, it is feasible to stratify patients with AV-V7 nuclear expression in CTC
for taxane therapy, with PSMA expression in CTC for targeted radioligand therapy, with
AVPC-MS in CTC/ctDNA plus pc.CTC for additional platinum therapy, or with DDR
gene alterations in CTC/ctDNA for PPAR inhibitors (Fig. 18). Meanwhile, novel
therapies are highly demanded for targeting newly identified biomarkers, e.g., clonal
CTC with mesenchymal state, and replacing non-selective therapies such as taxane
and platinum (Fig. 18). Most importantly, future studies on early detection of ‘AVPC’ at
treatment-naïve stage would have a significant impact on optimizing the time window of
many therapies that are currently available for treatment-resistant stage and eventually
improve their outcomes.

Figure 18. A future comprehensive liquid biopsy test to delineate ‘AVPC’
91

References
Adalsteinsson, V. A., Ha, G., Freeman, S. S., Choudhury, A. D., Stover, D. G., Parsons,
H. A., . . . Meyerson, M. (2017). Scalable whole-exome sequencing of cell-free
DNA reveals high concordance with metastatic tumors. Nat Commun, 8(1), 1324.
https://doi.org/10.1038/s41467-017-00965-y
Aparicio, A. M., Harzstark, A. L., Corn, P. G., Wen, S., Araujo, J. C., Tu, S. M., . . .
Logothetis, C. J. (2013). Platinum-based chemotherapy for variant castrate-
resistant prostate cancer. Clin Cancer Res, 19(13), 3621-3630.
https://doi.org/10.1158/1078-0432.CCR-12-3791
Aparicio, A. M., Shen, L., Tapia, E. L., Lu, J. F., Chen, H. C., Zhang, J., . . . Logothetis,
C. J. (2016). Combined Tumor Suppressor Defects Characterize Clinically
Defined Aggressive Variant Prostate Cancers. Clin Cancer Res, 22(6), 1520-
1530. https://doi.org/10.1158/1078-0432.CCR-15-1259
Ashworth, T. R. (1869). A Case of Cancer in Which Cells Similar to Those in the
Tumours Were Seen in the Blood after Death. In (Vol. 14, pp. 146-147): The
Medical Journal of Australia.
Balcik-Ercin, P., Cayrefourcq, L., Soundararajan, R., Mani, S. A., & Alix-Panabières, C.
(2021). Epithelial-to-Mesenchymal Plasticity in Circulating Tumor Cell Lines
Sequentially Derived from a Patient with Colorectal Cancer. Cancers (Basel),
13(21). https://doi.org/10.3390/cancers13215408
Baslan, T., Kendall, J., Rodgers, L., Cox, H., Riggs, M., Stepansky, A., . . . Hicks, J.
(2012). Genome-wide copy number analysis of single cells. Nat Protoc, 7(6),
1024-1041. https://doi.org/10.1038/nprot.2012.039
Batth, I. S., Dao, L., Satelli, A., Mitra, A., Yi, S., Noh, H., . . . Li, S. (2020). Cell surface
vimentin-positive circulating tumor cell-based relapse prediction in a long-term
longitudinal study of postremission neuroblastoma patients. Int J Cancer,
147(12), 3550-3559. https://doi.org/10.1002/ijc.33140
Berg, S., Kutra, D., Kroeger, T., Straehle, C. N., Kausler, B. X., Haubold, C., . . .
Kreshuk, A. (2019). ilastik: interactive machine learning for (bio)image analysis.
Nat Methods, 16(12), 1226-1232. https://doi.org/10.1038/s41592-019-0582-9
Bergamaschi, A., Kim, Y. H., Kwei, K. A., La Choi, Y., Bocanegra, M., Langerød, A., . . .
Pollack, J. R. (2008). CAMK1D amplification implicated in epithelial-
mesenchymal transition in basal-like breast cancer. Mol Oncol, 2(4), 327-339.
https://doi.org/10.1016/j.molonc.2008.09.004
Bertolini, F., Shaked, Y., Mancuso, P., & Kerbel, R. S. (2006). The multifaceted
circulating endothelial cell in cancer: towards marker and target identification. Nat
Rev Cancer, 6(11), 835-845. https://doi.org/10.1038/nrc1971
92

Brady, L., Hayes, B., Sheill, G., Baird, A. M., Guinan, E., Stanfill, B., . . . Finn, S. (2020).
Platelet cloaking of circulating tumour cells in patients with metastatic prostate
cancer: Results from ExPeCT, a randomised controlled trial. PLoS One, 15(12),
e0243928. https://doi.org/10.1371/journal.pone.0243928
Carlsson, A., Kuhn, P., Luttgen, M. S., Dizon, K. K., Troncoso, P., Corn, P. G., . . .
Zurita, A. J. (2017). Paired High-Content Analysis of Prostate Cancer Cells in
Bone Marrow and Blood Characterizes Increased Androgen Receptor
Expression in Tumor Cell Clusters. Clin Cancer Res, 23(7), 1722-1732.
https://doi.org/10.1158/1078-0432.CCR-16-1355
Chai, S., Matsumoto, N., Storgard, R., Peng, C. C., Aparicio, A., Ormseth, B., . . . Ruiz-
Velasco, C. (2021). Platelet-Coated Circulating Tumor Cells Are a Predictive
Biomarker in Patients with Metastatic Castrate-Resistant Prostate Cancer. Mol
Cancer Res, 19(12), 2036-2045. https://doi.org/10.1158/1541-7786.MCR-21-
0383
Chalfin, H. J., Glavaris, S. A., Malihi, P. D., Sperger, J. M., Gorin, M. A., Lu, C., . . .
Pienta, K. J. (2018). Prostate Cancer Disseminated Tumor Cells are Rarely
Detected in the Bone Marrow of Patients with Localized Disease Undergoing
Radical Prostatectomy across Multiple Rare Cell Detection Platforms. J Urol,
199(6), 1494-1501. https://doi.org/10.1016/j.juro.2018.01.033
Chen, Y., Li, S., Li, W., Yang, R., Zhang, X., Ye, Y., . . . Tang, W. (2019). Circulating
tumor cells undergoing EMT are poorly correlated with clinical stages or
predictive of recurrence in hepatocellular carcinoma. Sci Rep, 9(1), 7084.
https://doi.org/10.1038/s41598-019-43572-1
Conteduca, V., Ku, S. Y., Fernandez, L., Dago-Rodriquez, A., Lee, J., Jendrisak, A., . . .
Beltran, H. (2021). Circulating tumor cell heterogeneity in neuroendocrine
prostate cancer by single cell copy number analysis. NPJ Precis Oncol, 5(1), 76.
https://doi.org/10.1038/s41698-021-00211-1
Corn, P. G., Heath, E. I., Zurita, A., Ramesh, N., Xiao, L., Sei, E., . . . Aparicio, A. M.
(2019). Cabazitaxel plus carboplatin for the treatment of men with metastatic
castration-resistant prostate cancers: a randomised, open-label, phase 1-2 trial.
Lancet Oncol, 20(10), 1432-1443. https://doi.org/10.1016/S1470-2045(19)30408-
5
Cristofanilli, M., Budd, G. T., Ellis, M. J., Stopeck, A., Matera, J., Miller, M. C., . . .
Hayes, D. F. (2004). Circulating tumor cells, disease progression, and survival in
metastatic breast cancer. N Engl J Med, 351(8), 781-791.
https://doi.org/10.1056/NEJMoa040766
Cristofanilli, M., Hayes, D. F., Budd, G. T., Ellis, M. J., Stopeck, A., Reuben, J. M., . . .
Terstappen, L. W. (2005). Circulating tumor cells: a novel prognostic factor for
newly diagnosed metastatic breast cancer. J Clin Oncol, 23(7), 1420-1430.
https://doi.org/10.1200/JCO.2005.08.140
93

Dago, A. E., Stepansky, A., Carlsson, A., Luttgen, M., Kendall, J., Baslan, T., . . . Kuhn,
P. (2014). Rapid phenotypic and genomic change in response to therapeutic
pressure in prostate cancer inferred by high content analysis of single circulating
tumor cells. PLoS One, 9(8), e101777.
https://doi.org/10.1371/journal.pone.0101777
Danila, D. C., Heller, G., Gignac, G. A., Gonzalez-Espinoza, R., Anand, A., Tanaka,
E., . . . Scher, H. I. (2007). Circulating tumor cell number and prognosis in
progressive castration-resistant prostate cancer. Clin Cancer Res, 13(23), 7053-
7058. https://doi.org/10.1158/1078-0432.CCR-07-1506
de Bono, J. S., Scher, H. I., Montgomery, R. B., Parker, C., Miller, M. C., Tissing, H., . . .
Raghavan, D. (2008). Circulating tumor cells predict survival benefit from
treatment in metastatic castration-resistant prostate cancer. Clin Cancer Res,
14(19), 6302-6309. https://doi.org/10.1158/1078-0432.CCR-08-0872
DeLisser, H. M., Newman, P. J., & Albelda, S. M. (1993). Platelet endothelial cell
adhesion molecule (CD31). Curr Top Microbiol Immunol, 184, 37-45.
https://doi.org/10.1007/978-3-642-78253-4_3
Fizazi, K., Tran, N., Fein, L., Matsubara, N., Rodriguez-Antolin, A., Alekseev, B. Y., . . .
Investigators, L. (2017). Abiraterone plus Prednisone in Metastatic, Castration-
Sensitive Prostate Cancer. N Engl J Med, 377(4), 352-360.
https://doi.org/10.1056/NEJMoa1704174
Genna, A., Vanwynsberghe, A. M., Villard, A. V., Pottier, C., Ancel, J., Polette, M., &
Gilles, C. (2020). EMT-Associated Heterogeneity in Circulating Tumor Cells:
Sticky Friends on the Road to Metastasis. Cancers (Basel), 12(6).
https://doi.org/10.3390/cancers12061632
Gerdtsson, A. S., Thiele, J. A., Shishido, S. N., Zheng, S., Schaffer, R., Bethel, K., . . .
Kuhn, P. (2019). Single cell correlation analysis of liquid and solid biopsies in
metastatic colorectal cancer. Oncotarget, 10(66), 7016-7030.
https://doi.org/10.18632/oncotarget.27271
Gerdtsson, E., Pore, M., Thiele, J. A., Gerdtsson, A. S., Malihi, P. D., Nevarez, R., . . .
Kuhn, P. (2018). Multiplex protein detection on circulating tumor cells from liquid
biopsies using imaging mass cytometry. Converg Sci Phys Oncol, 4(1).
https://doi.org/10.1088/2057-1739/aaa013
Gkountela, S., Castro-Giner, F., Szczerba, B. M., Vetter, M., Landin, J., Scherrer,
R., . . . Aceto, N. (2019). Circulating Tumor Cell Clustering Shapes DNA
Methylation to Enable Metastasis Seeding. Cell, 176(1-2), 98-112.e114.
https://doi.org/10.1016/j.cell.2018.11.046
Gu, Z., Eils, R., & Schlesner, M. (2016). Complex heatmaps reveal patterns and
correlations in multidimensional genomic data. Bioinformatics, 32(18), 2847-
2849. https://doi.org/10.1093/bioinformatics/btw313
94

Hamid, A. A., Gray, K. P., Shaw, G., MacConaill, L. E., Evan, C., Bernard, B., . . .
Sweeney, C. J. (2019). Compound Genomic Alterations of TP53, PTEN, and
RB1 Tumor Suppressors in Localized and Metastatic Prostate Cancer. Eur Urol,
76(1), 89-97. https://doi.org/10.1016/j.eururo.2018.11.045
Hanahan, D., & Weinberg, R. A. (2011). Hallmarks of cancer: the next generation. Cell,
144(5), 646-674. https://doi.org/10.1016/j.cell.2011.02.013
Hassan, S., Blick, T., Thompson, E. W., & Williams, E. D. (2021). Diversity of Epithelial-
Mesenchymal Phenotypes in Circulating Tumour Cells from Prostate Cancer
Patient-Derived Xenograft Models. Cancers (Basel), 13(11).
https://doi.org/10.3390/cancers13112750
Heeke, S., Mograbi, B., Alix-Panabières, C., & Hofman, P. (2019). Never Travel Alone:
The Crosstalk of Circulating Tumor Cells and the Blood Microenvironment. Cells,
8(7). https://doi.org/10.3390/cells8070714
Heitmeir, B., Deniz, M., Janni, W., Rack, B., Schochter, F., & Wiesmüller, L. (2022).
Circulating Tumor Cells in Breast Cancer Patients: A Balancing Act between
Stemness, EMT Features and DNA Damage Responses. Cancers (Basel), 14(4).
https://doi.org/10.3390/cancers14040997
Hicks, J., Krasnitz, A., Lakshmi, B., Navin, N. E., Riggs, M., Leibu, E., . . . Zetterberg, A.
(2006). Novel patterns of genome rearrangement and their association with
survival in breast cancer. Genome Res, 16(12), 1465-1479.
https://doi.org/10.1101/gr.5460106
Horimoto, Y., Tokuda, E., Murakami, F., Uomori, T., Himuro, T., Nakai, K., . . . Saito, M.
(2018). Analysis of circulating tumour cell and the epithelial mesenchymal
transition (EMT) status during eribulin-based treatment in 22 patients with
metastatic breast cancer: a pilot study. J Transl Med, 16(1), 287.
https://doi.org/10.1186/s12967-018-1663-8
Hussain, M., Tangen, C. M., Higano, C., Schelhammer, P. F., Faulkner, J., Crawford, E.
D., . . . (INT-0162), S. O. G. T. (2006). Absolute prostate-specific antigen value
after androgen deprivation is a strong independent predictor of survival in new
metastatic prostate cancer: data from Southwest Oncology Group Trial 9346
(INT-0162). J Clin Oncol, 24(24), 3984-3990.
https://doi.org/10.1200/JCO.2006.06.4246
Iacobuzio-Donahue, C. A., Litchfield, K., & Swanton, C. (2020). Intratumor
heterogeneity reflects clinical disease course. Nat Cancer, 1(1), 3-6.
https://doi.org/10.1038/s43018-019-0002-1
Ignatiadis, M., Sledge, G. W., & Jeffrey, S. S. (2021). Liquid biopsy enters the clinic -
implementation issues and future challenges. Nat Rev Clin Oncol, 18(5), 297-
312. https://doi.org/10.1038/s41571-020-00457-x
95

James, N. D., de Bono, J. S., Spears, M. R., Clarke, N. W., Mason, M. D., Dearnaley, D.
P., . . . Investigators, S. (2017). Abiraterone for Prostate Cancer Not Previously
Treated with Hormone Therapy. N Engl J Med, 377(4), 338-351.
https://doi.org/10.1056/NEJMoa1702900
Jiang, X., Wong, K. H. K., Khankhel, A. H., Zeinali, M., Reategui, E., Phillips, M. J., . . .
Toner, M. (2017). Microfluidic isolation of platelet-covered circulating tumor cells.
Lab Chip, 17(20), 3498-3503. https://doi.org/10.1039/c7lc00654c
Karabacak, N. M., Spuhler, P. S., Fachin, F., Lim, E. J., Pai, V., Ozkumur, E., . . . Toner,
M. (2014). Microfluidic, marker-free isolation of circulating tumor cells from blood
samples. Nat Protoc, 9(3), 694-710. https://doi.org/10.1038/nprot.2014.044
Keller, L., & Pantel, K. (2019). Unravelling tumour heterogeneity by single-cell profiling
of circulating tumour cells. Nat Rev Cancer, 19(10), 553-567.
https://doi.org/10.1038/s41568-019-0180-2
Kilgour, E., Rothwell, D. G., Brady, G., & Dive, C. (2020). Liquid Biopsy-Based
Biomarkers of Treatment Response and Resistance. Cancer Cell, 37(4), 485-
495. https://doi.org/10.1016/j.ccell.2020.03.012
Krijthe, J. (2015). Rtsne: T-Distributed Stochastic Neighbor Embedding using a Barnes-
Hut Implementation. In https://github.com/jkrijthe/Rtsne
Ku, S. Y., Rosario, S., Wang, Y., Mu, P., Seshadri, M., Goodrich, Z. W., . . . Goodrich,
D. W. (2017). Rb1 and Trp53 cooperate to suppress prostate cancer lineage
plasticity, metastasis, and antiandrogen resistance. Science, 355(6320), 78-83.
https://doi.org/10.1126/science.aah4199
Kyriakopoulos, C. E., Chen, Y. H., Carducci, M. A., Liu, G., Jarrard, D. F., Hahn, N.
M., . . . Sweeney, C. J. (2018). Chemohormonal Therapy in Metastatic Hormone-
Sensitive Prostate Cancer: Long-Term Survival Analysis of the Randomized
Phase III E3805 CHAARTED Trial. J Clin Oncol, 36(11), 1080-1087.
https://doi.org/10.1200/JCO.2017.75.3657
Labelle, M., Begum, S., & Hynes, R. O. (2011). Direct signaling between platelets and
cancer cells induces an epithelial-mesenchymal-like transition and promotes
metastasis. Cancer Cell, 20(5), 576-590.
https://doi.org/10.1016/j.ccr.2011.09.009
Lazar, D. C., Cho, E. H., Luttgen, M. S., Metzner, T. J., Uson, M. L., Torrey, M., . . .
Kuhn, P. (2012). Cytometric comparisons between circulating tumor cells from
prostate cancer patients and the prostate-tumor-derived LNCaP cell line. Phys
Biol, 9(1), 016002. https://doi.org/10.1088/1478-3975/9/1/016002
Li, T. T., Liu, H., Li, F. P., Hu, Y. F., Mou, T. Y., Lin, T., . . . Li, G. X. (2015). Evaluation
of epithelial-mesenchymal transitioned circulating tumor cells in patients with
96

resectable gastric cancer: Relevance to therapy response. World J
Gastroenterol, 21(47), 13259-13267. https://doi.org/10.3748/wjg.v21.i47.13259
Lin, D., Shen, L., Luo, M., Zhang, K., Li, J., Yang, Q., . . . Zhou, J. (2021). Circulating
tumor cells: biology and clinical significance. Signal Transduct Target Ther, 6(1),
404. https://doi.org/10.1038/s41392-021-00817-8
Lindsay, C. R., Le Moulec, S., Billiot, F., Loriot, Y., Ngo-Camus, M., Vielh, P., . . .
Farace, F. (2016). Vimentin and Ki67 expression in circulating tumour cells
derived from castrate-resistant prostate cancer. BMC Cancer, 16, 168.
https://doi.org/10.1186/s12885-016-2192-6
Liu, R., Hong, R., Wang, Y., Gong, Y., Yeerken, D., Yang, D., . . . Zhan, Q. (2020).
Defect of SLC38A3 promotes epithelial-mesenchymal transition and predicts
poor prognosis in esophageal squamous cell carcinoma. Chin J Cancer Res,
32(5), 547-563. https://doi.org/10.21147/j.issn.1000-9604.2020.05.01
Lou, X. L., Sun, J., Gong, S. Q., Yu, X. F., Gong, R., & Deng, H. (2015). Interaction
between circulating cancer cells and platelets: clinical implication. Chin J Cancer
Res, 27(5), 450-460. https://doi.org/10.3978/j.issn.1000-9604.2015.04.10
Malihi, P. D., Graf, R. P., Rodriguez, A., Ramesh, N., Lee, J., Sutton, R., . . . Zurita, A.
J. (2020). Single-Cell Circulating Tumor Cell Analysis Reveals Genomic
Instability as a Distinctive Feature of Aggressive Prostate Cancer. Clin Cancer
Res, 26(15), 4143-4153. https://doi.org/10.1158/1078-0432.CCR-19-4100
Malihi, P. D., Morikado, M., Welter, L., Liu, S. T., Miller, E. T., Cadaneanu, R. M., . . .
Kuhn, P. (2018). Clonal diversity revealed by morphoproteomic and copy number
profiles of single prostate cancer cells at diagnosis. Converg Sci Phys Oncol,
4(1). https://doi.org/10.1088/2057-1739/aaa00b
Manucha, V., & Henegan, J. (2020). Clinicopathologic Diagnostic Approach to
Aggressive Variant Prostate Cancer. Arch Pathol Lab Med, 144(1), 18-23.
https://doi.org/10.5858/arpa.2019-0124-RA
Marrinucci, D., Bethel, K., Kolatkar, A., Luttgen, M. S., Malchiodi, M., Baehring, F., . . .
Kuhn, P. (2012). Fluid biopsy in patients with metastatic prostate, pancreatic and
breast cancers. Phys Biol, 9(1), 016003. https://doi.org/10.1088/1478-
3975/9/1/016003
McQuin, C., Goodman, A., Chernyshev, V., Kamentsky, L., Cimini, B. A., Karhohs, K.
W., . . . Carpenter, A. E. (2018). CellProfiler 3.0: Next-generation image
processing for biology. PLoS Biol, 16(7), e2005970.
https://doi.org/10.1371/journal.pbio.2005970
Mego, M., Gao, H., Lee, B. N., Cohen, E. N., Tin, S., Giordano, A., . . . Reuben, J. M.
(2012). Prognostic Value of EMT-Circulating Tumor Cells in Metastatic Breast
Cancer Patients Undergoing High-Dose Chemotherapy with Autologous
97

Hematopoietic Stem Cell Transplantation. J Cancer, 3, 369-380.
https://doi.org/10.7150/jca.5111
Mego, M., Karaba, M., Minarik, G., Benca, J., Silvia, J., Sedlackova, T., . . . Mardiak, J.
(2019). Circulating Tumor Cells With Epithelial-to-mesenchymal Transition
Phenotypes Associated With Inferior Outcomes in Primary Breast Cancer.
Anticancer Res, 39(4), 1829-1837. https://doi.org/10.21873/anticanres.13290
Miyamoto, D. T., Ting, D. T., Toner, M., Maheswaran, S., & Haber, D. A. (2016). Single-
Cell Analysis of Circulating Tumor Cells as a Window into Tumor Heterogeneity.
Cold Spring Harb Symp Quant Biol, 81, 269-274.
https://doi.org/10.1101/sqb.2016.81.031120
Mohme, M., Riethdorf, S., & Pantel, K. (2017). Circulating and disseminated tumour
cells - mechanisms of immune surveillance and escape. Nat Rev Clin Oncol,
14(3), 155-167. https://doi.org/10.1038/nrclinonc.2016.144
Mu, P., Zhang, Z., Benelli, M., Karthaus, W. R., Hoover, E., Chen, C. C., . . . Sawyers,
C. L. (2017). SOX2 promotes lineage plasticity and antiandrogen resistance in
TP53- and RB1-deficient prostate cancer. Science, 355(6320), 84-88.
https://doi.org/10.1126/science.aah4307
Ortiz-Otero, N., Clinch, A. B., Hope, J., Wang, W., Reinhart-King, C. A., & King, M. R.
(2020). Cancer associated fibroblasts confer shear resistance to circulating tumor
cells during prostate cancer metastatic progression. Oncotarget, 11(12), 1037-
1050. https://doi.org/10.18632/oncotarget.27510
Papadaki, M. A., Stoupis, G., Theodoropoulos, P. A., Mavroudis, D., Georgoulias, V., &
Agelaki, S. (2019). Circulating Tumor Cells with Stemness and Epithelial-to-
Mesenchymal Transition Features Are Chemoresistant and Predictive of Poor
Outcome in Metastatic Breast Cancer. Mol Cancer Ther, 18(2), 437-447.
https://doi.org/10.1158/1535-7163.MCT-18-0584
Paradis, E., & Schliep, K. (2019). ape 5.0: an environment for modern phylogenetics
and evolutionary analyses in R. Bioinformatics, 35(3), 526-528.
https://doi.org/10.1093/bioinformatics/bty633
Pau, G., Fuchs, F., Sklyar, O., Boutros, M., & Huber, W. (2010). EBImage--an R
package for image processing with applications to cellular phenotypes.
Bioinformatics, 26(7), 979-981. https://doi.org/10.1093/bioinformatics/btq046
Placke, T., Örgel, M., Schaller, M., Jung, G., Rammensee, H. G., Kopp, H. G., & Salih,
H. R. (2012). Platelet-derived MHC class I confers a pseudonormal phenotype to
cancer cells that subverts the antitumor reactivity of natural killer immune cells.
Cancer Res, 72(2), 440-448. https://doi.org/10.1158/0008-5472.CAN-11-1872
Rangel-Pozzo, A., Liu, S., Wajnberg, G., Wang, X., Ouellette, R. J., Hicks, G. G., . . .
Mai, S. (2020). Genomic Analysis of Localized High-Risk Prostate Cancer
98

Circulating Tumor Cells at the Single-Cell Level. Cells, 9(8).
https://doi.org/10.3390/cells9081863
Russano, M., Napolitano, A., Ribelli, G., Iuliani, M., Simonetti, S., Citarella, F., . . .
Santini, D. (2020). Liquid biopsy and tumor heterogeneity in metastatic solid
tumors: the potentiality of blood samples. J Exp Clin Cancer Res, 39(1), 95.
https://doi.org/10.1186/s13046-020-01601-2
Russo, M., & Bardelli, A. (2017). Lesion-Directed Therapies and Monitoring Tumor
Evolution Using Liquid Biopsies. Cold Spring Harb Perspect Med, 7(2).
https://doi.org/10.1101/cshperspect.a029587
Rzhetsky, A., & Nei, M. (1993). Theoretical foundation of the minimum-evolution method
of phylogenetic inference. Mol Biol Evol, 10(5), 1073-1095.
https://doi.org/10.1093/oxfordjournals.molbev.a040056
Satelli, A., Mitra, A., Brownlee, Z., Xia, X., Bellister, S., Overman, M. J., . . . Li, S.
(2015). Epithelial-mesenchymal transitioned circulating tumor cells capture for
detecting tumor progression. Clin Cancer Res, 21(4), 899-906.
https://doi.org/10.1158/1078-0432.CCR-14-0894
Schapiro, D., Jackson, H. W., Raghuraman, S., Fischer, J. R., Zanotelli, V. R. T.,
Schulz, D., . . . Bodenmiller, B. (2017). histoCAT: analysis of cell phenotypes and
interactions in multiplex image cytometry data. Nat Methods, 14(9), 873-876.
https://doi.org/10.1038/nmeth.4391
Scher, H. I., Graf, R. P., Schreiber, N. A., Jayaram, A., Winquist, E., McLaughlin, B., . . .
Heller, G. (2018). Assessment of the Validity of Nuclear-Localized Androgen
Receptor Splice Variant 7 in Circulating Tumor Cells as a Predictive Biomarker
for Castration-Resistant Prostate Cancer. JAMA Oncol, 4(9), 1179-1186.
https://doi.org/10.1001/jamaoncol.2018.1621
Scher, H. I., Graf, R. P., Schreiber, N. A., McLaughlin, B., Jendrisak, A., Wang, Y., . . .
Dittamore, R. (2017). Phenotypic Heterogeneity of Circulating Tumor Cells
Informs Clinical Decisions between AR Signaling Inhibitors and Taxanes in
Metastatic Prostate Cancer. Cancer Res, 77(20), 5687-5698.
https://doi.org/10.1158/0008-5472.CAN-17-1353
Scher, H. I., Lu, D., Schreiber, N. A., Louw, J., Graf, R. P., Vargas, H. A., . . . Dittamore,
R. (2016). Association of AR-V7 on Circulating Tumor Cells as a Treatment-
Specific Biomarker With Outcomes and Survival in Castration-Resistant Prostate
Cancer. JAMA Oncol, 2(11), 1441-1449.
https://doi.org/10.1001/jamaoncol.2016.1828
Shishido, S. N., Carlsson, A., Nieva, J., Bethel, K., Hicks, J. B., Bazhenova, L., & Kuhn,
P. (2019). Circulating tumor cells as a response monitor in stage IV non-small
cell lung cancer. J Transl Med, 17(1), 294. https://doi.org/10.1186/s12967-019-
2035-8
99

Shishido, S. N., Masson, R., Xu, L., Welter, L., Prabakar, R. K., D' Souza, A., . . . Kuhn,
P. (2022). Disease characterization in liquid biopsy from HER2-mutated, non-
amplified metastatic breast cancer patients treated with neratinib. NPJ Breast
Cancer, 8(1), 22. https://doi.org/10.1038/s41523-022-00390-5
Shishido, S. N., Welter, L., Rodriguez-Lee, M., Kolatkar, A., Xu, L., Ruiz, C., . . . Kuhn,
P. (2020). Preanalytical Variables for the Genomic Assessment of the Cellular
and Acellular Fractions of the Liquid Biopsy in a Cohort of Breast Cancer
Patients. J Mol Diagn, 22(3), 319-337.
https://doi.org/10.1016/j.jmoldx.2019.11.006
Siravegna, G., Marsoni, S., Siena, S., & Bardelli, A. (2017). Integrating liquid biopsies
into the management of cancer. Nat Rev Clin Oncol, 14(9), 531-548.
https://doi.org/10.1038/nrclinonc.2017.14
Soler, A., Cayrefourcq, L., Mazard, T., Babayan, A., Lamy, P. J., Assou, S., . . . Alix-
Panabières, C. (2018). Autologous cell lines from circulating colon cancer cells
captured from sequential liquid biopsies as model to study therapy-driven tumor
changes. Sci Rep, 8(1), 15931. https://doi.org/10.1038/s41598-018-34365-z
Sugimachi, K., Yokobori, T., Iinuma, H., Ueda, M., Ueo, H., Shinden, Y., . . . Mimori, K.
(2014). Aberrant expression of plastin-3 via copy number gain induces the
epithelial-mesenchymal transition in circulating colorectal cancer cells. Ann Surg
Oncol, 21(11), 3680-3690. https://doi.org/10.1245/s10434-013-3366-y
Sun, Y. F., Wu, L., Liu, S. P., Jiang, M. M., Hu, B., Zhou, K. Q., . . . Yang, X. R. (2021).
Dissecting spatial heterogeneity and the immune-evasion mechanism of CTCs
by single-cell RNA-seq in hepatocellular carcinoma. Nat Commun, 12(1), 4091.
https://doi.org/10.1038/s41467-021-24386-0
Swanton, C. (2012). Intratumor heterogeneity: evolution through space and time.
Cancer Res, 72(19), 4875-4882. https://doi.org/10.1158/0008-5472.CAN-12-
2217
Taylor, B. S., Schultz, N., Hieronymus, H., Gopalan, A., Xiao, Y., Carver, B. S., . . .
Gerald, W. L. (2010). Integrative genomic profiling of human prostate cancer.
Cancer Cell, 18(1), 11-22. https://doi.org/10.1016/j.ccr.2010.05.026
Thiele, J. A., Pitule, P., Hicks, J., & Kuhn, P. (2019). Single-Cell Analysis of Circulating
Tumor Cells. Methods Mol Biol, 1908, 243-264. https://doi.org/10.1007/978-1-
4939-9004-7_17
Wang, S., Li, Z., & Xu, R. (2018). Human Cancer and Platelet Interaction, a Potential
Therapeutic Target. Int J Mol Sci, 19(4). https://doi.org/10.3390/ijms19041246
Ward, M. P., E Kane, L., A Norris, L., Mohamed, B. M., Kelly, T., Bates, M., . . . J
O'Leary, J. (2021). Platelets, immune cells and the coagulation cascade; friend or
100

foe of the circulating tumour cell? Mol Cancer, 20(1), 59.
https://doi.org/10.1186/s12943-021-01347-1
Ward, Y., Lake, R., Faraji, F., Sperger, J., Martin, P., Gilliard, C., . . . Kelly, K. (2018).
Platelets Promote Metastasis via Binding Tumor CD97 Leading to Bidirectional
Signaling that Coordinates Transendothelial Migration. Cell Rep, 23(3), 808-822.
https://doi.org/10.1016/j.celrep.2018.03.092
Welter, L., Xu, L., McKinley, D., Dago, A. E., Prabakar, R. K., Restrepo-Vassalli, S., . . .
Hicks, J. (2020). Treatment response and tumor evolution: lessons from an
extended series of multianalyte liquid biopsies in a metastatic breast cancer
patient. Cold Spring Harb Mol Case Stud, 6(6).
https://doi.org/10.1101/mcs.a005819
Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis. In
https://ggplot2.tidyverse.org
Wu, F., Zhu, J., Mao, Y., Li, X., Hu, B., & Zhang, D. (2017). Associations between the
Epithelial-Mesenchymal Transition Phenotypes of Circulating Tumor Cells and
the Clinicopathological Features of Patients with Colorectal Cancer. Dis Markers,
2017, 9474532. https://doi.org/10.1155/2017/9474532
Xiong, G., Chen, J., Zhang, G., Wang, S., Kawasaki, K., Zhu, J., . . . Xu, R. (2020).
Hsp47 promotes cancer metastasis by enhancing collagen-dependent cancer
cell-platelet interaction. Proc Natl Acad Sci U S A, 117(7), 3748-3758.
https://doi.org/10.1073/pnas.1911951117
Xu, L., Mao, X., Guo, T., Chan, P. Y., Shaw, G., Hines, J., . . . Lu, Y. J. (2017). The
Novel Association of Circulating Tumor Cells and Circulating Megakaryocytes
with Prostate Cancer Prognosis. Clin Cancer Res, 23(17), 5112-5122.
https://doi.org/10.1158/1078-0432.CCR-16-3081
Yang, G., Xie, J., Zhang, S., Gu, W., Yuan, J., Wang, R., . . . Yang, B. (2021). Clinical
Significance of Mesenchymal Circulating Tumor Cells in Patients With
Oligometastatic Hormone-Sensitive Prostate Cancer Who Underwent
Cytoreductive Radical Prostatectomy. Front Oncol, 11, 812549.
https://doi.org/10.3389/fonc.2021.812549
Yokobori, T., Iinuma, H., Shimamura, T., Imoto, S., Sugimachi, K., Ishii, H., . . . Mori, M.
(2013). Plastin3 is a novel marker for circulating tumor cells undergoing the
epithelial-mesenchymal transition and is associated with colorectal cancer
prognosis. Cancer Res, 73(7), 2059-2069. https://doi.org/10.1158/0008-
5472.CAN-12-0326
Yu, G. (2020). Using ggtree to Visualize Data on Tree-Like Structures. Curr Protoc
Bioinformatics, 69(1), e96. https://doi.org/10.1002/cpbi.96
101

Zanotelli, V., & Bodenmiller, B. (2019). A flexible image segmentation pipeline for
heterogeneous multiplexed tissue images based on pixel classification. In.
Zhang, S., Wu, T., Peng, X., Liu, J., Liu, F., Wu, S., . . . Ma, S. (2017). Mesenchymal
phenotype of circulating tumor cells is associated with distant metastasis in
breast cancer patients. Cancer Manag Res, 9, 691-700.
https://doi.org/10.2147/CMAR.S149801
Zou, M., Toivanen, R., Mitrofanova, A., Floch, N., Hayati, S., Sun, Y., . . . Abate-Shen,
C. (2017). Transdifferentiation as a Mechanism of Treatment Resistance in a
Mouse Model of Castration-Resistant Prostate Cancer. Cancer Discov, 7(7), 736-
749. https://doi.org/10.1158/2159-8290.CD-16-1174
102

Appendix1: Supplementary Information

103

Supplementary Figure 1. Immunofluorescence composite and all-channel images of
CTCs from PB sample and BMA sample. Color coding: DAPI (blue); CK (red); VIM
(white); CD45/CD31 (green). Scale bar: 10μm.
104

Supplementary Figure 2. Feature selection pipeline and correlation analysis. a. Image
features selection pipeline including 1) non-relevant features removal – 2) Pearson
105

correlation analysis – 3) Representative feature selection within each group – 4) 10
selected features. b. Ten groups of highly correlated features identified through Pearson
correlation analysis. 1’, 2’, 3’, 4’ are negatively correlated features with features in group
1, 2, 3, 4. Arrows are the selected features for each group.

106

Supplementary Figure 3. Intensity and morphological features comparison across CTC
subtypes.

107

Supplementary Figure 4. Genomic comparisons among subclones. a. Chromosomal
and gene copy number alteration differences among genomic subclones. Y: positive for
108

CNA; N: negative for CNA. b. Comparison of LST among genomic subclones. c.
Breakpoint difference of copy number alterations on chromosome 13 between main
clone cells and minor clone cells.

Supplementary Figure 5. Single cell and clonal lineage trees. a. Minimum evolution tree
of single cell copy number profiles (cells with diluted profiles were excluded) with
Manhattan distance. The first number is the clade ID, the second number is the sample
ID (BMA: 5151; PB: 5152), and the third number is the sequencing cell ID. b. Minimum
evolution tree of median copy number profiles of clades with Manhattan distance.

109

Supplementary Figure 6. Association between genomic subclones and phenotypes. a.
Fluorescence intensity distributions in three different subclone groups. b. t-SNE analysis
for dimensionality reduction of copy number ratios of 5k bins labeled with hierarchical
clustering based subclones, morpho-proteomic features and sample types

110

Supplementary Figure 7. Intra-cluster characterization of genomic heterogeneity. a.
Table represents the clonality of single cells from 14 CTC clusters in the BMA sample.
Y: heterozygous; N: homozygous. The cell ID is the sequencing cell ID. b.
Immunofluorescence images and single cell copy number profiles of representative
homogeneous CTC clusters. Images order: composite, DAPI, CK, VIM, and
CD45/CD31.

111

Supplementary Figure 8. Platelet confirmation with CD61 staining in the SKBR3-spiked
NBD sample

Supplementary Figure 9. Enumeration of CTC subtypes

Cells/mL PB
750
500
250
0
20
15
10
5
0
2000
1500
1000
500
0
Cells/mL BMA
6000
4000
2000
0
% cells % cells 100
75
50
25
0
100
75
50
25
0
Patients Patients Patients
epi.CTC
mes.CTC
pc.mes.CTC
pc.epi.CTC
112

Supplementary Figure. 10. Survival analysis using AVPC-MS-CTC status

113

Supplementary Figure. 11. Survival analysis in patients without pc.CTC

Supplementary Figure 12. Treatment decision tree

114

Supplementary Figure 13. “Fire storm” genomic pattern on chromosome 8 in patients
#148

115

Supplementary Figure 14. Images and single cell copy number profiles of AVPC-MS
positive CTCs and matched white blood cells (WBCs) as the negative control from the
same patients

116

Supplementary Figure 15. LST of 69 CTCs with different genotypes across 14 patients.
c_AVPC: clonal alterations with AVPC-MS; c_nonAVPC: clonal alterations without
AVPC-MS.

117

Supplementary Table 1. Correlation between CTC and clinical characteristics
118

Appendix2: Data Availability
Chapter 2
The single cell sequencing data is available through the Sequence Read Archive with
BioProject accession number PRJNA827940. The immunofluorescence image data is
available in figshare at https://doi.org/10.6084/m9.figshare.19617717.v1. The image
mass cytometry data is available in figshare at
https://doi.org/10.6084/m9.figshare.19619007.v1.

Chapter 3
The single cell sequencing data and associated 10x and 40X single cell fluorescence
images and enumeration of cell types and evaluation of patient samples is available in
BloodPac Data Commons at https://data.bloodpac.org/discovery/BPDC000121/.

Abstract (if available)

Abstract Tumor heterogeneity is a consequence of the mutation, selection, and adaptation of cells along the pathway of disease progression and metastasis. It is considered the major cause of treatment failure and its decoding may provide insights into the development of new targeted therapies and their predictive biomarkers. Circulating tumor cells (CTCs), as important analytes for liquid biopsy, have the potential to offer a minimally invasive and realtime assessment of the heterogeneous and evolving landscape of cancer. However, current approaches to detect CTCs mostly rely on one identifier, such as epithelial markers or the cell’s biophysical traits, which can result in poor sensitivity or misclassification. This warrants the need for method developments that allow for the identification of various types of rare events, and then deconvolution of the heterogeneity. Here, we developed a third generation of liquid biopsy technology to detect and classify a spectrum of rare events followed with molecular characterization (Chapter 1). This technology was applied to an index patient with prostate cancer in which numerous CTCs were detected in the liquid biopsy. Results showed the phenotypic and genotypic heterogeneity, including epithelial/mesenchymal state and platelet attachment in CTCs with clonal lineage, which laid the foundation for biomarker discovery (Chapter 2).

Prostate cancer is becoming recognized not as a single disease, but as many, ranging from slow-growing tumors to highly aggressive and lethal lesions. A goal of this work was to improve patient outcomes for lethal prostate cancer, especially the ‘Aggressive Variant’ prostate cancer (AVPC) that poorly responds to androgen receptor inhibitors and has fast progression. This requires approaches to detect the variant early and inform the treatment decisions in a noninvasive and timely manner. Applying our technology, we identified the presence of platelet-coated CTCs in combination to the AVPC molecular signature (AVPC-MS), i.e., at least 2 defects in PTEN, RB1, TP53, as a predictive biomarker of treatment selection for AVPC (Chapter 3). Furthermore, AVPC-MS was detected in CTCs and other liquid biopsy analytes at the treatment-naïve stage, demonstrating the clinical utility of early detection for AVPC (Chapter 4). Overall, the development and application of our technology deciphered the CTC heterogeneity for biomarkers identification and validated their clinical implications for AVPC.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Molecular signature of aggressive disease and clonal diversity revealed by single-cell copy number analysis of prostate cancer cells across multiple disease states

PDF

Heterogeneity and plasticity of malignant and non-malignant circulating analytes in breast carcinomas

PDF

Applying multi-omics in cancer liquid biopsy for improved patient monitoring and biomarker discovery

PDF

Exploring the effects of CXCR4 inhibition on circulating tumor cell populations in metastatic prostate cancer

PDF

Early detection of lung cancer by characterizing circulating rare cells using peripheral blood liquid biopsy

PDF

Exploration of the roles of cancer stem cells and survivin in the pathogenesis and progression of prostate cancer

PDF

Ancestral inference and cancer stem cell dynamics in colorectal tumors

PDF

Optimization of circulating tumor cells isolation for gene expression analysis

PDF

An image-based identification of aggressive circulating tumor cell subtypes

PDF

The effect of tumor-mediated immune suppression on prostate cancer immunotherapy

PDF

Studies of murine prostate cancer stem / progenitor cells

PDF

Developing a robust single cell whole genome bisulfite sequencing protocol to analyse circulating tumor cells

PDF

Circulating tumor cells as biomarkers in lung cancer

PDF

Study of bone morphogenetic protein-2 and stromal cell derived factor-1 in prostate cancer

PDF

Multimodal single-cell biology and machine learning to characterize plasma cell neoplasms

PDF

CpG methylation profiling in lung cancer cell lines, tumors and non-tumors

PDF

Homologous cell systems for the study of progression of androgen-dependent prostate cancer to castration-resistant prostate cancer

PDF

LIGHT-ing up prostate cancer: remodeling the tumor microenvironment and enhancing therapeutic vaccine efficacy through forced LIGHT expression in prostate cancer

PDF

RNA methylation in cancer plasticity and drug resistance

PDF

Contribution of cancer associated fibroblasts to cancer progression

Asset Metadata

Creator Chai, Shoujie (author)

Core Title Deconvolution of circulating tumor cell heterogeneity and implications for aggressive variant prostate cancer

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Molecular Biology

Degree Conferral Date 2022-08

Publication Date 07/22/2024

Defense Date 05/19/2022

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

Tag aggressive variant,circulating tumor cell,heterogeneity,OAI-PMH Harvest,prostate cancer,single cell multiomics

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kuhn, Peter (committee chair), Chen, Lin (committee member), Chen, Xiaojiang (committee member), Hicks, James (committee member), Nieva, Jorge (committee member)

Creator Email chaishoujie@gmail.com,shoujiec@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC111375162

Unique identifier UC111375162

Legacy Identifier etd-ChaiShouji-10933

Document Type Dissertation

Format application/pdf (imt)

Rights Chai, Shoujie

Type texts

Source 20220728-usctheses-batch-962 (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