Functional Characterization of Mobilized Tumor Cells
by
Xiaosai Yao
B.S. Biomedical Engineering
Johns Hopkins University, 2007
Submitted to the Department of Biological Engineering
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Biological Engineering
at the
Massachusetts Institute of Technology
February 2014
© 2014 Massachusetts Institute of Technology
All rights reserved
Signature of Author: ____________________________________________________________
Department of Biological Engineering
December 11, 2013
Certified by: ___________________________________________________________________
K. Dane Wittrup
C.P.Dubbs Professor of Chemical Engineering and Biological Engineering
Thesis Supervisor
Accepted by: __________________________________________________________________
Forest M. White
Associate Professor of Biological Engineering
Chair, Graduate Program Committee
1
Thesis committee members
John M. Essigmann, Ph.D. (Chair)
Besty P. Leitch Professor of Chemistry and Biological Engineering
Massachusetts Institute of Technology
Tyler E. Jacks, Ph.D.
Professor of Biology
Massachusetts Institute of Technology
J. Christopher Love, Ph.D.
Associate Professor of Chemical Engineering
Massachusetts Institute of Technology
Christina A. Williamson, M.D.
Senior Consulting Surgeon
Lahey Hospital and Medical Center
Assistant Clinical Professor of Surgery
Tufts University School of Medicine
2
Functional Characterization of Mobilized Tumor Cells
by
Xiaosai Yao
Submitted to the Department of Biological Engineering on December 11, 2013
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Biological Engineering
ABSTRACT
Despite being responsible for 90% of cancer mortality, metastasis is not well understood.
This thesis is focused on the circulation step of the metastatic cascade, examining three types of
mobilized tumor cells: circulating tumor cells (CTCs), intraoperatively shed tumor cells, and
malignant pleural effusions (MPE). We investigated the functional behavior of mobilized tumor
cells in order to explain the discrepancy between the number of tumor cells in circulation and the
number overt metastases.
The first part of this thesis examines the functional behavior of CTCs isolated from the
peripheral blood of metastatic castration-resistant prostate cancer patients. Individual CTCs were
compartmentalized using arrays of nanowells to enable clonal comparison and mapping of
heterogeneity. The viability, invasiveness and secretory profiles of CTCs were measured. Only a
subset of CTCs was found to possess malignant traits indicative of metastatic potential. These
CTCs were resistant to anoikis, were invasive or secreted proteolytic enzymes.
The second part of this thesis determines the presence of intraoperatively shed tumor
cells using blood samples withdrawn from the pulmonary vein after pulmonary lobectomy
procedures. Previous studies did not distinguish tumor cells from normal epithelial cells
specifically or sensitively. Single-cell genetic approaches were used to compare the genotype of
isolated single cells to matched tumor cells and normal adjacent tissue, thereby confirming the
malignancy of shed epithelial cells.
The third and last part of the thesis delineates the tumorigenic population with surface
markers using MPEs. A total of 35 surface antigens were screened from four categories: 1)
cancer stem cell 2) epithelial-mesenchymal transition 3) metastatic signature and 4) tyrosine
kinase receptors. Surface antigen CD24 was found to be specifically and abundantly expressed in
MPE, and was required for the colonization of the lung.
In conclusion, metastatic inefficiency is due to the presence of inactive cells and cellular
heterogeneity. Inactive cells are either normal epithelial cells or apoptotic tumor cells. Cellular
heterogeneity may arise from differences in surface marker expression or functional states.
Therefore, only a subset of mobilized tumor cells can give rise to metastases, and therapeutic
strategies should be focused on this subset.
Thesis Supervisor: K. Dane Wittrup
Title: C.P. Dubbs Professor of Chemical Engineering and Biological Engineering
3
Acknowledgments
The past five years at MIT have brought many mentors and friends into my life. They
have inspired me with their ingenuity and drive, helped me through moments of inadequacy and
allowed me to mature scientifically. I have learnt that many seemingly impossible things are in
fact possible, and there might be many ready solutions out there waiting to be utilized. The key
to success that I have gleaned from my mentors is an open mind to possibilities substantiated by
constant learning.
First of all, I would like to thank my advisor, Dane Wittrup. His scientific acumen has
charted the project to the right direction. He has also given me the freedom to pursue my own
interests and form collaborations beyond his lab, while reminding me the importance of stay
focused. His moments of excitement during discoveries have spurred me on to greater work
while his unwavering trust during my failures has kept me trying.
I would also like to thank Chris Love, who welcomed me into his lab of wonderful
people and state-of-the-art technology. He has taught me the importance of scientific rigor along
with the art of presentation. I thank him for his numerous comments and feedback during group
meetings and preparation of manuscripts.
Christina Williamson is committed as a collaborator, encouraging as a committee
member and motherly as a friend. She has provided me with over sixty invaluable patient
samples for the work described in Chapter 3 and has always responded to my enquiries promptly.
Despite her busy schedule, she and I have had hours of discussion at her office and over the
phone.
I thank my other committee members John Essigmann for his patience and enthusiasm
and Tyler Jacks not only for his criticism and feedback but also for creating the most conducive
environment for cancer research as the director of the Koch Institute.
My other mentors before MIT has laid the foundation of my graduate work. Dr. Lim Koh
Pang at the Institute of Molecular and Cell Biology Singapore patiently taught me all the
important molecular lab techniques and gave me a head start in my career. I was able to further
my research skills in the fascinating field of tumor immunology under the guidance of Dung Le
and Elizabeth Jaffee at Johns Hopkins.
My time at the Wittrup lab was very enjoyable, with special thanks to Tiffany Chen,
Nicole Yang, Cary Opel, John Rhoden, Alessandro Angelini, Jim van Deventer, Byron Kwan
and Michael Traxylmayr. Tiffany, Nicole and I spent almost every day of the past five years
together. I am grateful for their company during the otherwise lonesome nights, and for their
assistance during my major life events. Cary and I shared the “Awesome Lab” together, and he
always came to my rescue with his Doraemon collection of lab supplies. John always cheered me
up with his bag of pranks and jokes. The senior lab members, some of whom I didn’t overlap
very long, Ben Hackel, Kelly Orcutt, Annie Gai, Jamie Spangler, Jordi Mata-Fink, Eileen
Higham, nevertheless taught me the importance of hard work very early on.
The Love lab has given me tremendous scientific assistance. Yvonne Yamanaka
introduced me to the lab and helped me set up many experiments initially and throughout the five
years. I admire her for her scientific rigor, discipline and fun-loving nature outside the lab.
Viktor Adalsteinsson is also working on CTC and has offered invaluable scientific advice
especially on cancer genomics. Rachel Barry and Adebola Ogunniyi helped me pick many cells.
Todd Gierahn and Denis Loginov set up many software and network infrastructure instrumental
4
to data analysis. Li-Lun Ho, Qing Han and Rita Contento have helped me with bioinformatics
and microengraving analysis.
The help of Myriam Labelle and Shahinoor Begum in the lab of Richard Hynes has made
the animal work in the Chapter 4 possible. Myriam and Richard edited my manuscript very
meticulously and taught me important lessons as a writer.
My other collaborators at Lahey Clinic, Carla Lamb, Kimberly Christ and Donna Spencer
coordinated the delivery of pleural effusion samples for the work of Chapter 4. Dr. Atish
Choudhury from Dana Farber provided valuable blood samples from prostate cancer patients for
the work of Chapter 2.
The diligence and meticulousness of my UROPs, Robert William, Ryan Keating and
Ashty Karim, greatly increased my productivity.
The staff at Swanson Biotechnology Center, especially Glenn Paradis of the flow
cytometry core and staff of the sequencing facilities, were extremely patient and helpful to my
work.
My life at MIT is not complete without my friends from BE especially Amneet Gulati
and friends from home, Qunya Ong, Chyan Ying Ke, Teck Chuan Lim, Wang Jia.
Lastly, I’m most grateful to my supportive family. My husband, Shao Ning Pei, and I
vowed our lifetime commitment in the presence of friends at MIT. He is always there when I
need him the most. He kept me cool in moments of frustration and has brought me true
happiness. My mother-in-law gave up her job to take care of my son, Larry, so that I could sleep
through the night and focus on my work. Most importantly, I thank my parents for everything. I
did not appreciate the degree of their sacrifice and dedication until I have become a parent
myself. They have always put my needs before their own, and unburdened me with workload
unrelated to the pursuit of my academic ambitions. It is because of their upbringing that I am
able to come to MIT and do the work described in this thesis.
5
To my parents
For their unconditional love
&
In loving memory of my grandfathers
Who battled cancer
6
! !
!
!
!
Table of Contents
1! Chapter 1: Introduction and Motivation ............................................................................ 10!
1.1! Abstract ........................................................................................................................................ 10!
1.2! Cancer metastasis ........................................................................................................................ 10!
1.3! Types of mobilized tumor cells ................................................................................................... 12!
1.4! Single cell technology .................................................................................................................. 13!
1.5! Thesis overview ............................................................................................................................ 15!
2! Chapter 2: Functional Characterization of Circulating Tumor Cells in Metastatic
Castration-Resistant Prostate Cancer Using Nanowells ......................................................... 17!
2.1! Abstract ........................................................................................................................................ 17!
2.2! Introduction ................................................................................................................................. 17!
2.2.1! Isolation of circulating tumor cells ........................................................................................ 17!
2.2.2! Current understanding of circulating tumor cells .................................................................. 18!
2.3! Materials and methods ................................................................................................................ 19!
2.3.1! Patient recruitment ................................................................................................................. 19!
2.3.2! Fabrication of nanowell arrays ............................................................................................... 20!
2.3.3! CTC enrichment ..................................................................................................................... 21!
2.3.4! Staining and microscopy ........................................................................................................ 21!
2.3.5! Proliferation assay .................................................................................................................. 22!
2.3.6! Invasion assay ........................................................................................................................ 23!
2.3.7! Microengraving ...................................................................................................................... 24!
2.3.8! Proteolytic assay with FRET-based peptides ......................................................................... 25!
2.4! Results........................................................................................................................................... 26!
2.4.1! Isolation of circulating tumor cells with arrays of subnanoliter wells ................................... 26!
2.4.2! A subset of circulating tumor cells was viable at the time of isolation ................................. 30!
2.4.3! The majority of circulating tumor cells underwent rapid cell death ...................................... 31!
2.4.4! A subset of circulating tumor cells was invasive. .................................................................. 35!
2.4.5! A subset of circulating tumor cells secreted soluble factors .................................................. 37!
2.4.6! Some EpCAM- cell persisted in Matrigel .............................................................................. 43!
2.5! Discussion ..................................................................................................................................... 44!
2.5.1! Only a subset of circulating tumor cells exhibit malignant traits indicative of metastatic
potential .............................................................................................................................................. 44!
2.5.2! Nanowells allow functional characterization of circulating tumor cells................................ 47!
3! Chapter 3: Identification of Intraoperatively Dislodged Tumor Cells during Surgical
Resection of Lung Tumor........................................................................................................... 48!
3.1! Abstract ........................................................................................................................................ 48!
3.2! Introduction ................................................................................................................................. 48!
3.2.1! Putative tumor cells were shed during surgical manipulation of tumors ............................... 48!
3.2.2! Earlier studies did not distinguish normal epithelial cells from malignant cells ................... 49!
3.2.3! Isolation and retrieval of intraoperatively shed cells using nanowells .................................. 49!
3.3! Materials and methods ................................................................................................................ 50!
3.3.1! Patients and sample collection ............................................................................................... 50!
3.3.2! Enrichment of epithelial cells from blood samples ................................................................ 53!
3.3.3! Staining and microscopy ........................................................................................................ 53!
3.3.4! Single cell retrieval ................................................................................................................ 53!
7
3.3.5! Tumor disaggregation and flow sorting ................................................................................. 54!
3.3.6! Mutation analysis of single cells using nested PCR .............................................................. 54!
3.3.7! Whole genome amplification ................................................................................................. 55!
3.3.8! Copy number variation analysis ............................................................................................. 56!
3.3.9! Targeted sequencing .............................................................................................................. 56!
3.4! Results........................................................................................................................................... 56!
3.4.1! Surgery released EpCAM+ cells into the tumor-draining pulmonary vein. ........................... 56!
3.4.2! Copy number variation analysis revealed tumor signature in pulmonary vein blood after
surgery. ............................................................................................................................................... 58!
3.4.3! Targeted sequencing confirmed consistent mutations between primary tumor and pulmonary
vein blood after surgery. ..................................................................................................................... 61!
3.4.4! Mutation analysis of single cells identified tumor cells released into the pulmonary vein
during thoracotomy. ........................................................................................................................... 62!
3.5! Discussion ..................................................................................................................................... 66!
3.5.1! Surgical manipulation releases both normal and malignant epithelial cells. ......................... 66!
4! Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions....................... 68!
4.1! Abstract ........................................................................................................................................ 68!
4.2! Introduction ................................................................................................................................. 68!
4.2.1! Surface antigens for antibody targeting ................................................................................. 68!
4.2.2! Pleural effusions as a source of disseminated tumor cells ..................................................... 69!
4.3! Materials and methods ................................................................................................................ 70!
4.3.1! Collection of patient samples ................................................................................................. 70!
4.3.2! Surface marker expression analysis using flow cytometry .................................................... 71!
4.3.3! Quantification of surface receptors expression ...................................................................... 73!
4.3.4! Sorting and animal studies ..................................................................................................... 74!
4.3.5! Cell lines and shRNA knockdown ......................................................................................... 74!
4.3.6! Seeding study ......................................................................................................................... 74!
4.3.7! Lung colonization study ......................................................................................................... 75!
4.3.8! Platelet binding assay ............................................................................................................. 75!
4.3.9! Anoikis assay ......................................................................................................................... 75!
4.3.10! Statistical analysis ................................................................................................................ 75!
4.4! Results........................................................................................................................................... 76!
4.4.1! Only the EpCAM+ population of cells in pleural effusions was tumorigenic........................ 76!
4.4.2! CD24 was one of the several surface markers abundantly expressed in malignant pleural
effusions ............................................................................................................................................. 78!
4.4.3! CD24 predicted increased tumor growth in xenograft tumors of pleural effusions............... 83!
4.4.4! CD24 was critical for tumorigenic lung colonization ............................................................ 86!
4.4.5! CD24 was pro-survival .......................................................................................................... 89!
4.4.6! Higher expression of CD24 in malignant pleural effusions correlated with a worse patient
outcome .............................................................................................................................................. 91!
4.5! Discussion ..................................................................................................................................... 91!
4.5.1! CD24 is a cancer-specific, abundantly expressed and functionally important for lung
colonization of tumor cells. ................................................................................................................ 92!
4.5.2! CD24 promotes tumor growth ............................................................................................... 93!
4.5.3! Future improvements ............................................................................................................. 94!
5! Chapter 5: Discussions and Future Perspectives ............................................................... 96!
5.1! Discussions ................................................................................................................................... 96!
5.1.1! Circulating tumor cells and metastasis .................................................................................. 96!
5.1.2! Working with clinical specimens ........................................................................................... 97!
5.2! Future experiments ..................................................................................................................... 97!
8
5.2.1! Discovery of EpCAM- CTCs ................................................................................................. 97!
5.2.2! The switch from dormancy to proliferation ........................................................................... 99!
9
Chapter 1: Introduction and Motivation
1
1.1
Chapter 1: Introduction and Motivation
Abstract
This thesis examines the functional behaviors of mobilized tumor cells obtained directly
from cancer patients because these cells may advance into life-threatening metastases. In
particular, we wish to address the discrepancy between the number of tumor cells found in
circulation and the number of overt metastases (section 1.2). The three types of mobilized tumor
cells are circulating tumor cells, intraoperatively shed tumor cells and malignant pleural
effusions (section 1.3). The rarity of mobilized tumor cells and the low purity of tumor cells in
bodily fluids present a technical challenge to their characterization and require tumor enrichment
and single-cell technologies (section 1.4). We address specific questions for each type of clinical
specimens: functional characterization of circulating tumor cells (Chapter 2), malignancy of
intraoperatively shed epithelial cells (Chapter 3) and surface marker expression of malignant
pleural effusions (Chapter 4) (section 1.5).
1.2
Cancer metastasis
Metastasis is the lethal stage of cancer. Metastasis kills in a patient in two ways: 1) organ
failure and 2) cytokine overproduction [1] (Figure 1-1). As cancer spreads, tumor cells can
interfere with the vital functions of the organs they invade, such as brain, liver and lung. At the
same time, a systematic overproduction of pro-inflammatory cytokines can result in clinical
syndromes [1] including cachexia (wasting syndrome 30-87% patients), thrombosis
(hypercoagulation, 50% patients) and dyspnea (shortness of breath, 20-80% patients).
Lethality of Metastasis
Organ
Failure
- Brain
- Liver
- Lung
- Bone
Cytokine
Overproduction
- Cachexia (30-90% patients)
- Thrombosis (50% patients)
- Dyspnea (20-90% patients)
Figure 1-1: Metastasis is lethal because of organ involvement and clinical syndromes as a result of cytokine
overproduction
10
Chapter 1: Introduction and Motivation
The metastatic cascade (Figure 1-2) consists of five steps: 1) tumor invasion, 2) tumor
intravasation, 3) transport through the circulation, 4) extravasation and 5) colonization. Since the
process of metastasis may take years to complete, tracking metastasis in real time in patients has
not been achieved. However, through experiments that tried to model metastasis by intravenous
injection of mouse melanoma cells, Luzzi and colleagues estimated the efficiency of
extravasation and colonization to be 80% and 0.02% respectively [2], implying that the
colonization is the rate limiting step of metastasis.
Figure 1-2: Five steps of metastatic cascade
Understanding how cancer cells behave in circulation is important to the development of
prophylactic therapy against overt metastasis. The hope is to halt the proliferation of mobilized
tumor cells before they can progress to colonize new sites. In reality, by leaving their solid
milieu to enter a fluid environment, mobilized tumor cells have already subjected themselves to a
harsh selection pressure. Mobilized tumor cells have to severe their connection with their
neighboring tumor cells and lose anchorage from extracellular matrix where growth factors are
docked. In exchange for the harsh condition, however, mobilized tumor cells gain mobility, a
prerequisite for invasion and metastasis.
What happened to tumor cells after they became detached from the primary tumor and
while they get transported in the circulation? Two observations have provided important clues to
the above question (Figure 1-3A):
1. The number of shed tumor cells far exceeds the number of metastases.
Autopsy reports on cancer patients revealed that the number of metastases is 1-13 per
patient, in 2-3 organs [3]. This is far less than an average of 0 - 23,618 CTCs per 7.5 ml of blood
[4], or 10 - 7×106 cells shed in surgery [5].
2. Metastases are clonal.
Genetic comparison between single cells found in primary tumors, early disseminated
cells and overt metastases revealed that metastases are more homogeneous than primary tumors
11
Chapter 1: Introduction and Motivation
[6] and early disseminated cells [7], implying that metastases may have resulted from single
clones derived from the primary tumor [6].
Two models describe the mechanism of metastatic evolution - linear progression model,
and parallel progression model (Figure 1-3B) [3]. The linear progression model believes that
dissemination occurs late in tumor development and the more malignant clones break off late in
tumor progression. On the other hand, parallel progression model believes that dissemination
happens relatively early, and the disseminated tumor cells may evolve independently from the
primary tumors. However, the precise mechanism of metastatic evolution remains unclear, it is
unknown whether the linear progression model or the parallel progression model predominates.
A
B
Linear progression
Shed tumor cells >> Metastases
Metastases are clonal
Parallel progression
Figure 1-3: A) Current observations of metastasis, B) The linear progression and the parallel models to
describe the formation of metastases.
1.3
Types of mobilized tumor cells
In this thesis, I will examine three kinds of mobilized tumor cells: 1) circulating tumor
cells, 2) intraoperatively shed tumor cells and 3) malignant pleural effusions. Even though all
three types of cells are mobilized tumor cells, there are fundamental differences between them.
The first difference is the staging of the cancer patients from whom we obtained these
cells from. Pleural effusions and circulating tumor cells were retrieved from advanced, late stage
(stage IV) disease patients. These late stage patients had their primary tumors resected years
before the collection of the mobilized tumor cell specimens. In contrast, intraoperatively shed
tumor cells were retrieved from early stage (stage I-III) cancer patients whose tumors have not
spread and were still resectable.
The second difference between the three cell types is their abundance. Pleural metastases
are very abundant in number, ranging from 103-106 cells, allowing the use of flow cytometry as a
12
Chapter 1: Introduction and Motivation
suitable analysis method. On the other hand, both CTCs and intraoperatively shed cells are rare,
and are concealed amongst blood cells which are 6-9 orders of magnitude more abundant than
the tumor cells. Therefore, to retrieve and characterize these rare cells, we had to apply tumor
enrichment followed by single cell analysis (see section 1.4 for literature review).
The last difference between the three types of mobilized tumor cells is their
tumorigenicity. Tumor cells from pleural effusions form tumors with high efficiency (~75%,
chapter 4). On the other hand, the frequency of tumor formation by CTCs is very low (3/110
patients), and the minimal number of cells required is 1000 [8]. In reality, it is rather rare for
1000 CTCs to seed at the same spot, so the actual tumorigenicity of CTCs in patients might be
even lower. For intraoperatively shed tumor cells, there has been one report that showed tumor
formation from 105-106 tumor cells [9], which means tumor formation is even less likely.
A summary of the three types of cells and their properties is shown below in Table 1-1:
Mobilized cells
Location
Abundance
Staging
(cells/ available volume)
Circulating tumor cells
Intraoperatively shed
Blood
Blood
1- 100 cell/ 10 ml
2
3
3
6
10 -10 cells/ 2-10 ml
(minimal
cells required)
I - IV
3% (1000 cells)
I - IV
Not
tumor cells
Pleural effusions
Tumorigenicity
determined
(105-106
cells)
Pleural cavity
10 -10 cells/ 100 -1000 ml
IV
75% (100 cells)
Table 1-1: Types of mobilized tumor cells examined
1.4
Single cell technology
Because of the low abundance and high impurity of mobilized tumor cells in clinical
samples, sensitive techniques are necessary to analyze their properties. Understanding single
cells requires three technical components: 1) tumor enrichment, 2) single cell isolation and 3)
single cell characterization. Conventional single cell techniques, such as flow cytometry and
cytology, can be used to analyze pleural effusions but can become limited when applied to blood
specimens. Flow cytometry requires at least 50 cells to provide a high enough statistical
confidence in the results. Cytology enables examination of single cells, but is a rather low
throughput technology. Furthermore, cytology requires prior enrichment of tumor cells if the
13
Chapter 1: Introduction and Motivation
number of contaminating non-tumor cells exceeds one million/slide. In the case of circulating
tumor cells, the number of red blood cells is 109 cells/ml, so cytology alone cannot easily
identify tumor cells.
To circumvent the limitation of conventional single cell analysis, a plethora of tumor
enrichment techniques have been developed recently. They can be divided into three categories:
1) antibody-mediated, 2) size-based and 3) density-based. In the antibody-mediated approach, an
antibody against a surface antigen expressed on the tumor cells selectively binds to the tumor
cells. The most common surface antigen used is EpCAM as it is an epithelial marker expressed
in as high as 80% of various carcinomas such as colon, pancreas and prostate [10]. The antibody
could be immobilized on magnetic particles or devices [4, 11, 12]. In the size-based approach, a
microfiltration system retains tumor cells based on their bigger size [13]. Inertial focusing selects
cells based on their hydrodynamic radii [14, 15]. Lastly, the density of tumor cells could also be
used for their enrichment [16].
Tumor enrichment step only removes 3-4 logs of contaminating non-tumor cells,
therefore a second step – single-cell retrieval – is necessary to obtain a high purity of tumor cells.
Micromanipulation uses a glass capillary tube to manually aspirate the cells of interest, but it is
labor-intensive and requires training of the operator. Other microfluidics devices can perform
trapping of single cells [17], although they are not easily coupled to downstream molecular
analysis.
Lastly, a few single cell molecular analysis methods have emerged to enable the
characterization of rare cells. Two technologies, multiple displacement amplification (MDA)
using random priming and the strand-displacing
29 polymerase under isothermal conditions
[18], and multiple annealing and looping-based amplification cycle (MALBAC) [19], can
amplify the whole genome of single cells with improved coverage and decreased biases. The
amplified genomic products can be used for downstream assays such as copy number variation,
exome capture for mutation analysis [20]. Single cell transcriptomics methods sought to decrease
3’ positional bias and use principles including homopolymer tailing and template switching [20].
The major drawback with these single cell analysis techniques is that the tumor cells are lysed
and therefore dynamic information about their behaviors is lost. Our work hopes to expand upon
the existing single cell analysis to interrogate the secretion of soluble factors and cellular
behaviors including invasion and proliferation in viable tumor cells.
14
Chapter 1: Introduction and Motivation
1.5
Thesis overview
The overall goal of this thesis is to have a more in-depth understanding of the properties
of mobilized tumor cells since far less data exists on mobilized tumor cells than on solid tumors.
The primary hurdle in the study of mobilized tumor cells is their paucity, thus requiring
technological innovations to address the relevant biological questions. The biological question
this thesis hopes to address is why so few of the mobilized tumor cells can progress to overt
metastases – what cell types or functional properties are directly responsible for tumor
progression. The answer to this question is translatable to clinical benefit because pinpointing the
tumorigenic subset will allow us to design targeted therapy against the most relevant cells, and
better stratify patients and identify their risks.
In this thesis, I hope to characterize the three types of mobilized tumor cells described
above. Because of the differences in their properties, each clinical sample has a different focus in
the question we ask. In the characterization of circulating tumor cells, the goal is to optimize a
method of single cell analysis using nanowells and investigate the functional behaviors of single,
viable CTCs, specifically their viability, invasion and secretion of soluble factors.
Spatiotemporal tracking of single, viable CTCs can give insights to the function of CTCs as well
as the heterogeneity of CTCs. Interestingly, we identified that only a rare subset of CTCs
possessed malignant traits (anti-apoptotic, invasive or secreting proteolytic enzymes).
In the characterization of intraoperatively shed tumor cells, the goal is to distinguish
tumor cells from normal epithelial cells by using genetic mutations consistent between the
EpCAM+ cells found in the blood and the EpCAM+ cells found in primary tumor. This part of
the thesis also made use of the nanowell approach to isolate and retrieve single epithelial cells.
We found tumor cells mobilized during the course of the surgery along with normal epithelial
cells.
In the characterization of pleural effusions, the goal is to identify a surface marker that
can delineate the tumorigenic population. We screened a panel of 35 surface antigens implicated
in cancer metastasis and/or tumorigenicity using three criteria: 1) levels of expression, 2) cancer
specificity and 3) functional requirement for tumorigenicity. We narrowed down to CD24 as a
promising cancer target and showed that it was required for tumor formation in cancer cell line
and xenograft tumors.
15
Chapter 1: Introduction and Motivation
The underlying hypothesis for all three types of clinical specimens is that tumor cells are
heterogeneous. Indeed, I believe that the discrepancy between the number of mobilized tumor
cells and overt metastases is due to the inactivity of many mobilized tumor cells and the
heterogeneity of mobilized tumor cells. We observed normal epithelial cells shed
intraoperatively and many apoptotic circulating tumor cells even in advanced cancer patients,
implying that many of the mobilized epithelial cells have already lost the ability to form
metastases while in circulation. Differential expression of surface markers in pleural effusions
and the different degree of malignancy in circulating tumor cells affect the ability of these
mobilized tumor cells to colonize subsequently.
16
Chapter 2: Circulating Tumor Cells
2
Chapter 2: Functional Characterization of Circulating Tumor Cells in
Metastatic Castration-Resistant Prostate Cancer Using Nanowells
2.1
Abstract
Ample evidence supports genetic and functional heterogeneity in primary tumors, but it
remains unclear whether circulating tumor cells (CTCs) also exhibit the same hierarchical
organization. We examined the functional diversity of viable, single CTCs using an array of
subnanoliter wells (nanowells). The compartmentalization of single cells by nanowells allowed
clonal comparison and mapping of heterogeneity of single cells or preformed clusters of cells.
By measuring the short-term viability, invasiveness and secretory profiles of individual CTCs, it
was evident that only a rare subset of CTCs possessed malignant traits indicative of metastatic
potential in late-stage, progressing metastatic castration-resistant prostate cancer (mCRPC)
patients. These CTCs were resistant to anoikis after being in the circulation, were invasive in
their epithelial state, or secreted proteases capable of cleaving peptide substrates. Not every CTC
observed exhibited such metastatic potential, suggesting that enumeration of CTCs alone may be
insufficient to understand metastasis or stratify patients.
2.2
2.2.1
Introduction
Isolation of circulating tumor cells
Two broad categories of tumor enrichment – positive and negative selection – have been
used to isolate CTCs. Positive selection is affinity-based and usually involves the use of an
antibody against the epithelial lineage antigen, EpCAM, commonly expressed in most
carcinomas [10]. Devices using positive selection can vary depending on how the antibodies are
mobilized, and can come in the form of microfluidics devices such as the CTC chip [11, 21],
immunomagnetic particles such as the MagSweeper [12] and the FDA approved CellSearch
system [22] or magnetic nanoparticles and detection by photoacoustic signals [23]. The negative
selection method depletes the contaminating blood cells either by CD45 depletion or a debulking
process where the smaller erythrocytes and leukocytes were removed by inertial focusing [14,
15]. Other label-free depletion methods also include density gradient separation (OncoQuick)
[16], filtration (ScreenCell) [24], or dielectrophoresis [25].
17
Chapter 2: Circulating Tumor Cells
2.2.2
Current understanding of circulating tumor cells
1. CTC count is prognostic of patient outcome
Enumeration of CTCs by the CellSearch system, which defines CTC as
EpCAM+/CK+/DAPI+/CD45- cells, is prognostic of patient progression-free and overall survival
in a variety of cancer types. Using ≥5 CTCs as the threshold, overall survival was 21.7 vs. 11.5
months for castration-resistant prostate cancer patients before chemotherapy [26]. In stage III
and IV lung cancer patients, progression-free survival was 6.8 vs. 2.4 months (P < .001) and
overall survival was 8.1 vs. 4.3 months (P < .001) for patients with fewer than five CTCs
compared with five or more CTCs before chemotherapy [27].
2. CTCs evolve with treatment
Changes in CTC counts before and after chemotherapy can predict rates of relapses in
breast cancer patients [28]. Only 1/28 patients relapsed in the group with greater than 10-fold
decrease in the number of CTCs after the treatment, compared to 5/30 patients with less than 10
fold changes and 14/30 patients with greater than 10-fold increase in the number of CTCs after
treatment.
The mutation status of CTCs can also change with the course of treatment [29]. In lung
cancer patients treated with Gefitinib, CTCs continued to harbor the primary EGFG activating
mutation but the frequency of T790M drug-resistant mutation started to increase during the
course of the therapy. The CTCs of some patients also acquired new EGFR mutations that were
previously absent in the primary tumors. This indicates that CTC is a dynamic reflection of the
systemic tumor evolution in patients.
In another study with breast cancer patients, the number of CTCs and the percentage of
mesenchymal CTCs declined in patients who responded to therapy. In contrast, patients who had
progressive disease on therapy showed an increase in the number of mesenchymal CTCs [30].
3. CTCs are tumorigenic but the efficiency is low
Two recent studies established the ability of CTCs to form xenograft tumors in
immunocompromised mice [8, 31]. In the first study, CTCs were enriched by CD45 depletion
and injected directly into the femur of mice. Only three out of 113 patients had CTCs that
successfully engrafted, and all three patients had CTCs count > 1000. This indicates that even
though CTCs may have tumorigenic potential, this potential is low.
18
Chapter 2: Circulating Tumor Cells
The
second
study
isolated
EpCAM-
CTCs
with
surface
expression
of
HER2/EGFR/NSSE/NOTCH1. Not only could these CTCs propagate in vitro, they could also
establish xenograft tumors with spontaneous metastasis to the brain and the lung, closely
resembling the behavior of CTCs in patients. The frequency of engraftment is relatively high –
3/8 patients had CTCs that successfully established tumors in nude mice.
4. CTC vs. primary tumor
The relationship between CTCs and patient matched primary tumor has been explored by
copy number variation analysis using array CGH [32, 33]. DNA from single as well as pooled
CTCs was amplified and compared to the primary tumor. The degree of concordance between
primary tumor and the CTC ranges from very high to moderate (0.89-0.46) depending on the
patients [33]. CTCs and primary tumor share many common aberrations, but the CTCs also
acquired divergent genomic alternations. For example, the CTCs of one patient revealed a high
level of amplification of CDK8, which was absent in the primary tumor and the liver metastasis
34 and 24 months before the analysis the CTCs. The presence of CDK8 may change treatment
options since this patient could be eligible for CDK inhibitors.
2.3
2.3.1
Materials and methods
Patient recruitment
The patient cohort used in this study was generated from the Prostate Clinical Research
Information System (CRIS) at Dana-Farber Cancer Institute. The CRIS system consists of dataentry software, a central data repository, collection of patient data including comprehensive
follow-up of all patients, and tightly integrated security measures as previously described [34].
All patients provided written informed consent to allow the collection of tissue and blood and the
analysis of clinical and genetic data for research purposes. Patients with metastatic castrationresistant prostate cancer were identified for this trial based on 1) progression on a phase II study
of abiraterone in combination with dutasteride or 2) PSA >20 to enrich for patients likely to have
detectable circulating tumor cells. Patient status was assigned by changes in serum PSA levels,
with progression at the time of blood collection defined as a PSA increase of >5% per 30 days.
Refer to Table 2-1 for patient information. Blood was drawn into EDTA tubes and processed
within 4 hr. Whole blood from healthy donors was purchased from Research Blood Components.
19
Chapter 2: Circulating Tumor Cells
Patient
Age
ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
H1
H2
H3
H4
H5
H6
56
65
57
76
73
69
70
65
56
61
87
80
60
63
76
57
78
77
65
58
22
53
19
53
28
CTC&count&in&Fig&2
Staging
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
metastatic
-
PSA at time of
blood draw (ng/ml) CalceinAM+/ CalceinAM+ CalceinAMAnnexinV/AnnexinV+ /AnnexinV+
322.39
7336.98
248.68
5.57
140.58
191.41
125.8
75.93
110.13
97.49
20.3
109.2
48.99
1305.5
134.99
20.32
919.79
20.05
403.72
-
314
90
33
7
1
1
1
0
0
0
0
0
0
0
0
0
0
0
10
2
2
1
0
0
0
131
34
27
1
0
0
0
1
0
2
2
1
0
0
0
0
0
0
62
8
25
0
0
0
0
3
5
8
17
0
17
4
9
3
2
2
6
0
0
2
0
0
8
9
0
2
3
0
CalceinAM/AnnexinV39
26
315
3
5
0
1
15
4
2
24
3
2
0
6
1
7
1
2
20
31
2
4
3
1
Status at
blood draw
progressing
progressing
progressing
progressing
responding
progressing
responding
progressing
progressing
responding
responding
progressing
progressing
responding
progressing
stable
progressing
progressing
progressing
healthy
healthy
healthy
healthy
healthy
healthy
Table 2-1: Patient information
2.3.2
Fabrication of nanowell arrays
A silicon master [35] was microfabricated (Stanford foundry) and mounted in a metal
mold. Poly(dimethylsiloxane) (PDMS) (Dow Corning) (10:1 ratio of base to catalyst) was
injected through a port into the silicon mold, cured at 80°C for 4 hr, and then removed to
produce an array containing 84,672 cubic wells (65 µm). Before use, the PDMS array was
oxygen plasma treated for 2 min and immediately submerged in PBS to preserve the
hydrophilicity rendered by the plasma treatment. The array was then blocked in serumcontaining media for 15 min before cells were loaded
20
Chapter 2: Circulating Tumor Cells
2.3.3
CTC enrichment
Negative selection was performed using either the EasySep or RosetteSep CD45
depletion kit (StemCell Technologies). With the EasySep kit, 45 ml of red blood cell lysis buffer
(Biolegend) was added to 5 ml of whole blood and the mixture was incubated at room
temperature until the red blood cells were completely lysed (15 – 20 min). Blood was washed
once with wash buffer (2% Fetal Bovine Serum (FBS), 1% Bovine Serum Antigen (BSA), 5 mM
Ethylenediaminetetraacetic acid (EDTA) in Phosphate Buffered Saline (PBS)). CD45 depletion
was performed with the EasySep human CD45 depletion kit according to the manufacturer’s
instructions. The remaining cells were suspended in approximately 200 µl wash buffer and were
directly deposited onto the PDMS nanowells and allowed to settle for 5 min. With the
RosetteSep kit, 250 µl antibody cocktail was incubated with 5 ml of whole blood for 20 min. The
blood was then diluted with PBS at a 1:1 ratio and layered onto Ficoll-Paque Plus (GE
Healthcare) in a SepMate tube (StemCell Technologies) and centrifuged at 800x g for 10 min.
The upper layer containing the serum and buffy coat was removed and washed twice. Further red
blood cell lysis was sometimes necessary to remove residual red blood cells.
2.3.4
Staining and microscopy
Cells were either stained directly on the array of nanowells in a tube for 1 hr at room
temperature with EpCAM and a cocktail of lineage markers for leukocytes including CD3,
CD16, CD20, CD38 and CD45 (refer to the list of antibodies used in table 2-2.
For viability assays, the cells were rinsed with PBS and stained with Calcein AM violet
(Molecular Probes) and Annexin V FITC (BD Pharmingen) in Annexin V binding buffer (BD
Pharmingen) for 10 min at room temperature.
For viability assays, the cells were rinsed with PBS and stained with Calcein AM violet
(Molecular Probes) and Annexin V FITC (BD Pharmingen) in Annexin V binding buffer (BD
Pharmingen) for 10 min at room temperature.
The stamps were imaged with an epifluorescence microscope (Zeiss) with filter wheels at
the following wavelengths: Calcein AM violet (Ex: 390 nm, Em: 440/40 nm), FITC (Ex: 488
nm, Em: 525/36 nm), PerCP-eFluor710 (Ex: 488 nm, Em: 716/40nm), PE/CY7 (Ex: 570 nm,
Em: 809/81 nm)
21
Chapter 2: Circulating Tumor Cells
Compensation was performed with beads that were precoated with anti-Fc antibodies
(Bangs Laboratories). Each antibody was incubated with the beads and imaged in all the
fluorescent channels. The percentage of bleed over was computed by plotting the fluorescence
intensity of the signal channel versus the intensity of all other channels individually. The slope of
the linear plot gave the percentage bleed over of the signal channel into the second channel.
Cells were identified with Enumerator, a custom image analysis software developed in
house. For each nanowell array, we generated a list of cell information including the well IDs,
cell size and fluorescent intensities. The text file was converted into a FlowJo-readable text
format [36]. Gating and cell statistics were analyzed in FlowJo (Treestar).
Assay
Antigen
Clone
Fluorophore
Vendor
Surface staining
EpCAM
1B7
PerCP-eFluor710
eBiosciences
Surface staining
CD45
HI30
PE-CY7
Biolegend
Surface staining
CD3
UCHT1
PE-CY7
Biolegend
Surface staining
CD20
2H7
PE-CY7
Biolegend
Surface staining
CD16
3G8
PE-CY7
Biolegend
Surface staining
CD38
HIT2
PE-CY7
Biolegend
Microengraving
PSA (capture)
181811
N/A
R&D systems
Microengraving
PSA (detection)
Goat polyclonal
Alexa Fluor 488
R&D systems
N/A
Life Technologies
Alexa Fluor 700
Jackson
(Product Number:
BAF1344)
Microengraving
hIgG (capture)
Goat
polyclonal
(81-7100)
Microengraving
hIgG (detection)
09-175-098
ImmunoResearch
Table 2-2: List of antibodies used
2.3.5
Proliferation assay
CTCs were cultured directly in the nanowells. C4-2 cells were obtained from ATCC.
Matrigel (1 ml; BD Biosciences) was added directly onto the nanowells and allowed to solidify
for 1 hr at room temperature. Cells were maintained in a growth medium that was previously
22
Chapter 2: Circulating Tumor Cells
reported to enhance the proliferation of epithelial cells [37]. The growth medium consisted of 3:1
Ham’s F-12 Nutrient Mixture-Dulbecco’s modified Eagle’s medium (Cellgro), 5% FBS (SigmaAldrich), 0.4 µg/ml hydrocortisone (Sigma-Aldrich), insulin (5 µg/ml) - transferrin (5 µg/ml) sodium selenite (5 ng/ml) supplement (Roche), 8.4 ng/ml cholera toxin (Sigma-Aldrich), 100
U/ml Penicillin-Streptomycin (Cellgro), 10 ng/ml epidermal growth factor (Life Technologies),
24 µg/ml adenine (Sigma-Aldrich), 10 µM Y-27632 (Enzo Life Sciences) and 1 pM 5αAndrostan-17β-ol-3-one (Sigma-Aldrich). In addition to the growth conditions described above,
we applied a second culture condition that has previously been used to grow intestinal stem cells
and primary prostate cancer cells [38]. The growth medium consisted of 3:1 Ham’s F-12
Nutrient Mixture-Dulbecco’s modified Eagle’s medium (Cellgro), 1x N2 supplements (Life
Technologies), 1x B-27 supplements (Life Technologies), 1 mM N-acetylcysteine (SigmaAldrich), 1 µg/ml R-spondin (Life Technologies), 100 ng/ml Noggin (Life Technologies), 50
ng/ml epidermal growth factor (Life Technologies), 100 U/ml penicillin-streptomycin (Cellgro),
10 µM Rho-kinase (ROCK) inhibitor (Y-27632) (Enzo Life Sciences) and 1 pM 5α-Androstan17β-ol-3-one (Sigma-Aldrich). There was, however, no significant difference in the percentage
of CTC survival between the two culture conditions.
The predicted probability of the viability of a cluster was estimated based on the viability
of a single cell. A cluster was considered viable if it contained at least one viable cell. The
viabilities of clusters containing different numbers of cells were weighted by the frequencies of
each cluster according to the following formula:
!!"#$%&',!"#$%&
1
=!
!
!!"#
!×!! (1 − (1 − !!"#$%&,!"#$%& )! )
!!!
where ! = total number of cells found in clusters,!!= number of cells found in a cluster (ranging
from 2 to !!!"# ), !! = number of clusters containing i cells, and !!"#$%&,!"#$%& = viability of
single cells.
2.3.6
Invasion assay
After cell loading, 1 ml of Matrigel was pipetted directly onto the nanowell arrays and
allowed to solidify for 1 hr at room temperature. Cells embedded in Matrigel were stained with
antibodies (1:200) for 2 hr at 37 °C before imaging. The array was then imaged every three days.
23
Chapter 2: Circulating Tumor Cells
EpCAM+ cells were identified and tracked by their positive EpCAM staining. The coordinates of
the centroid of the individual cells or cell clusters and of their wells were read in AxioVision
(Zeiss). The relative coordinates of the cells to the wells were calculated as:
!"##!"# = (!"#! , !"#! ) ! = (!!"## − !!"## , !!"## − !!"## )
The distance that the cells had moved from their initial position was calculated as:
!"#$%&'( = ! (!"#!!"#!! − !"#!!"#!! )! + (!"#!!"#!! − !"#!!"#!! )!
2.3.7
Microengraving
Microengraving was performed as previously described [39]. Poly-lysine-coated glass
slides were coated with 1 µg of capture antibody in 80 µl sodium borate buffer (pH 9) for 1 hr at
room temperature or 4°C overnight and then blocked in PBS + 1% BSA for 30 min. The cellloaded array of nanowells was rinsed with basal media containing 0.04% human serum. Human
IgGs in the serum were used to mark the position of each well; every well should be positive for
human IgG because anti-human IgG capture and detection antibodies were included in the panel
of antibodies. The antibody-coated glass slide was then sealed on top of the nanowells in a
hybridization clamp for 4 hr at 37°C. The slide was then blocked with 5% milk + 0.5% TWEEN20 + PBS (blocking buffer) for 15 min and incubated with 0.3 µg fluorescently conjugated
detection antibodies in 150 µl of blocking buffer for 45 min at room temperature. The dried slide
was scanned with a GenePix 4400A scanner (Molecular Devices). The scanned image was
analyzed with Matlab programs developed in house, Crossword and Matchbox. The IgG
background channel was used to identify the position of each well. A positive event was defined
as a well that 1) had 50% of its pixels 2 standard deviations above the mean intensity of the
background and 2) 3 ≤ signal-to-noise ratio (SNR) ≤12. Each positive event was further
manually inspected for potential artifacts. To account for print-to-print variability, the
distribution of SNR of PSA was normalized to mean = 1, standard deviation = 0.2. The
normalization of the SNR was performed as follows:
< !"# >!= (!"# − !!)/! ∗ 0.2 + 1
24
Chapter 2: Circulating Tumor Cells
where < !"# > = normalized SNR value, !"#!= SNR of each well, !!= mean of the SNRs of
all the wells in a print, !!= standard deviation of the SNRs of all the wells in a print.
The calibration curve for determining rates of secreted PSA was constructed by spotting 1
µl of diluted PSA detection antibody (1 µg/µl) (5000x – 10,000,000x dilution) on a poly-lysine
slide. The slide was dried under vacuum for 5 min and then scanned using the same settings as
the clinical samples. The median intensity was quantified using GenePix Pro 6.0 (Molecular
Devices).
To estimate the probability of secretion of each cell type (Figure 2-10), we compiled the
percentage of secretion for each permutation of cell types from the print, and fit the observed
secretion rate to the following equation by non-linear regression in order to obtain the probability
!"##
of secretion for each cell type, !!"!!!"!
:
!"##!!(!"#$%"%$&,!"#$,!"#$,!"!#)
!"##
(!!"!!!"#
)!!"## !
!!"# = ! 1 − !!"!!!"# = 1 −
where !!"# = probability of secretion from the microengraving results,
!!"## = number of cells for that particular cell type, which was generated by tabulating the
manually gated cells.
2.3.8
Proteolytic assay with FRET-based peptides
Whole blood from a prostate cancer patient was first enriched for CTCs using the
RosetteSep CD45 depletion kit. The cells were then loaded into the nanowells and stained with
Calcein AM Violet, EpCAM and lineage antibody cocktail for 1 hr. Next, 500 µl of 5 µM FRET
polypeptides (BioZyme) mixture was added to the array of nanowells. A glass slide was sealed
over the array using a hybridization clamp for 3 hr in a humidifier chamber. The array of
nanowells was then imaged with the microscope using the FITC channel (Ex: 488 nm, Em:
525/36 nm) for the FRET peptides. The peptides are substrates of Matrix Metalloproteinases
(MMPs) 1, 2, 8, 9, 10, 12, 13 and 14 and A Disintegrin and Metalloproteinases (ADAMs) 8, 10,
17 as described previously [40]. The sequences of the peptides are shown in Table 2-3.
25
Chapter 2: Circulating Tumor Cells
After the initial time point, the arrays of nanowells containing CTCs were placed in the
lower chamber of a 0.2 µm transwell petri dish (Corning). C4-2 cells (1 x 104 cells, ATCC) and
MG-63 cells (1 x 106 cells, ATCC) were seeded on top of the transwell. The 0.2 µm transwell
allowed secreted factors from C4-2 cells and MG-63 cells to diffuse to the CTCs in the lower
chamber. The cells were maintained in growth medium as described in the proliferation assay for
one week, at which point the proteolytic activity was measured again.
PEPDAB005
Dabcyl-Leu-Ala-Gln-Ala-Homophenylalanine-Arg-Ser-Lys(5-FAM)-NH2
PEPDAB010
Dabcyl-Ser-Pro-Leu-Ala-Gln-Ala-Val-Arg-Ser-Ser-Lys(5-FAM)-NH2
PEPDAB022
Dabcyl-Leu-Arg-Ala-Glu-Gln-Gln-Arg-Leu-Lys-Ser-Lys(5-FAM)-NH2
Table 2-3: Sequences of FRET-peptides
2.4
2.4.1
Results
Isolation of circulating tumor cells with arrays of subnanoliter wells
In order to resolve the variance between individual CTC, we developed a system to
characterize single cells in a high-throughput manner. Our PDMS array comprises 84,762 cubic
wells of 275 pl each (65 µm x 65 µm x 65 µm). CTCs were enriched from whole blood by
negative selection against CD45 and loaded onto the array to settle into the nanowells by limiting
serial dilution (Figure 2-1A). Because CTCs are rare, the loading biased the occupancy of each
well to single CTCs or preformed clusters, allowing comparisons among individual CTCs.
To interrogate all the cells, we imaged the entire array and obtained the surface
fluorescence and position of every cell using a custom-designed image processing software (onchip cytometry). In addition to surface fluorescence, we were also interested in measuring the
secretion of individual cells. We placed glass slides functionalized with capture antibodies in
contact with the nanowells to capture soluble factors secreted by cells in the individual wells in a
process known as microengraving [41]. Because the unique well ID retains the spatial
information of each cell, we can visually inspect rare cells with their images, track the same cells
over time and map secretion events back to the respective cells (Figure 2-1B). Using this
approach, we made three types of measurements on viable CTCs isolated from patients with
26
Chapter 2: Circulating Tumor Cells
prostate cancer: (1) immediate and short-term viability of CTCs, (2) invasive potential of CTCs,
and (3) secretion of soluble factors.
Figure 2-1: Functional measurements of viability, invasion and secretory profiles of CTCs using arrays of
nanowells. (A) Scheme for the enrichment and functional characterization of CTCs using nanowells for viability,
invasion, or secretion of soluble factors. (B) Scatter plot (left) is generated from on-chip imaging cytometry of 50
C4-2 cells spiked into 5 ml of whole blood. A single cell can be mapped back to its original image (middle) or
secretion event determined by microengraving (right) based on its unique well ID.
We validated the performance of our method using spiked tumor cell lines. We
determined the overall yield using C4-2 prostate tumor cells and HT29 colon tumor cells to be
30% (Figure 2-2A). The depletion of CD45+ cells was the dominant source of loss (Figure 2-2B).
We have tried two commercially available CD45 depletion kits. RosetteSep depleted red blood
cells and leukocytes in a single step using a Ficoll-Paque gradient, and it had a yield of 50%.
EasySep kit required a prior red blood cell lysis step and depleted leukocytes by
immunomagnetic separation. It had a slightly lower yield of 45% because it involved one
additional step of red blood cell lysis. Cell loss also resulted from the loading of cells onto the
stamp. The loading efficiency was proportional to the total area of the wells. Removal of
channels from the original design also marginally improved the yield (Figure 2-2C) to close to
80%.
27
Chapter 2: Circulating Tumor Cells
A
B
C
Percentage yield
100
80
60
40
20
0
50 µm
channels
65 µm
65 µm
channels no channels
Figure 2-2: Nanowells can recover spiked tumor cells. (A) 5-118 C4-2 cells and 5-57 HT29 cells were spiked into
5 ml of whole blood. The recovery of tumor cells is about 30% with both C4-2 (top) and HT29 (bottom) cells. (B)
CTC enrichment from human whole blood using arrays of nanowells involves depleting the CD45 cells and loading
onto the arrays. We quantified the cell loss of each step. Shown is the recovery of spiked HT29 cells at each step.
CD45 depletion was performed using either of two commercially available kits (RosetteSep or EasySep) and each
kit usually results in 50% loss of spiked tumor cells. Cell loss from red blood cell lysis or loading onto the array is
less significant (10% and 20% respectively). (C) Yield of the cell loading step is affected by the design of the
device.
The efficiency of depletion could vary quite significantly from patient to patient. Here is
a typical distribution of the occupancy of the wells (Figure 2-3). But because CTCs are rare, the
loading biased the occupancy of the wells to single CTCs or preformed clusters per well,
allowing comparisons among individual CTCs.
28
Chapter 2: Circulating Tumor Cells
Figure 2-3: Occupancy of wells by cell type. Shown is a typical distribution of the number of cells by well after
the depletion of erythrocytes and leukocytes by RosetteSep. Black – white blood cells; blue – platelets; red – red
blood cells. The number of cells per well was tabulated by Enumerator, and plotted against the cumulative
distribution function of the corresponding well occupancy.
We have experimented with other tumor enrichment methods: 1) density separation
(Oncoquick) and 2) a positive EpCAM-based immunomagnetic enrichment kit from Veridex.
However, both methods gave lower yield than RosetteSep did (Figure 2-4).
Percentage yield
100
80
60
40
20
0
RosetteSep
OncoQuick
CellSearch
kit
Figure 2-4: Alternative CD45 depletion methods have a lower yield than RosetteSep. We compared the yield of
two other tumor enrichment kits using approximately 100 spiked C4-2 cells. OncoQuick uses a density gradient to
deplete CD45 and recovers tumor cells in the buffy coat layer. The CellSearch kit is a positive EpCAM-based
immunomagnetic enrichment kit. The yield of OncoQuick is 20% and CellSearch is only 10%.
The process maintained the viability of the isolated tumor cells (95% ± 10% for HT29
cells and 90% ± 7% for C4-2 cells). Therefore, viability of CTCs lower than 90% is a likely
result of previous apoptotic events, rather than death induced from processing.
29
Chapter 2: Circulating Tumor Cells
2.4.2
A subset of circulating tumor cells was viable at the time of isolation
Previously reported methods have not distinguished among viable and dead CTCs from
blood samples. Therefore, we determined the viability of primary CTCs isolated from the blood
of prostate cancer patients. We applied a combination of live (Calcein AM+) and apoptotic
(Annexin V+) markers in addition to lineage markers (EpCAM+, CD45-). The non-fluorescent
Calcein AM is hydrolyzed to a fluorescent form by intracellular esterases in live cells, while
Annexin V binds to phosphatidylserine on apoptotic cells. Using imaging cytometry, we
analyzed CTCs isolated from the blood of prostate cancer patients and categorized them as either
fully viable (Calcein AM+/ Annexin V-), apoptotic (Calcein AM+/ Annexin V+), dead (Calcein
AM-/ Annexin V+), or disintegrated (Calcein AM-/ Annexin V-) (Figure 2-5A).
Our results showed that a significant proportion of CTCs was already dead in the
circulation. Since most CTC enumeration methods count the total number of viable, apoptotic
and dead cells (Figure 2-5B, top panel), previous analysis might have overestimated the
abundance of biologically active CTCs. The range of viable CTCs observed in 18 prostate cancer
patients is 0 – 314 cells per 5 ml of blood (mean = 25, median = 0) (Figure 2-5B, bottom panel).
30
Chapter 2: Circulating Tumor Cells
TL
A
102
103
CD45
104
101
104
EpCAM
103
102
101
101
EpCAM Calcein Annexin
2
Calcein AM
102
103
104
1
10110 10210 10310 10410
Annexin V
1
65
3
4
2
3
4
+
viability
+
1000
100
10
1
Annexin V
in the count
1000
100
10
1
Calcein AM
B
+
+
-
+
+
-
+
1000
100
10
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 2 3 4 5 6
patient ID
Figure 2-5: Viability of CTCs at the time of isolation. (A) Assessment of viability of CTCs using on-chip
cytometry data for apoptotic (Annexin V) / viable (Calcein AM) markers. Scatter plot of EpCAM+ cells (left) yields
population of EpCAM+CD45- cells that are scored for both Annexin V and Calcein AM (middle) to classify viable
cells (1), apoptotic cells (2), dead cells (3) and disintegrated cells (4). Representative phase contrast and
epifluorescence micrographs (right) are shown for each classified state. The dimension of the side of each well is 65
µm. (B) Bargraphs of the number of CTCs by classified state of viability. The numbers of total intact EpCAM+ cells
(top), Calcein AM+/EpCAM+ cells (middle) and Calcein AM+/Annexin V-/EpCAM+ cells (bottom) are shown for 18
prostate cancer patients (red) and 6 healthy donors (black).
.
2.4.3
The majority of circulating tumor cells underwent rapid cell death
A subset of CTCs may represent “in transit” tumor cells with metastatic potential, but the
rate-limiting step of metastasis is thought to be colonization at the distant site (20). Next we
sought to determine whether viable CTCs isolated from blood maintained their viability and/or
possessed any proliferative potential.
We tracked the viability (Figure 2-6A) and surface area (Figure 2-6B) as a proxy for
proliferation (cells in a cluster fused and could not be accurately counted after a week in culture)
in 125 CTCs from patient ID2 and 48 spiked C4-2 prostate cancer cells for 16 days in Matrigel.
31
Chapter 2: Circulating Tumor Cells
Tracking was possible because CTCs maintained expression of EpCAM and other cell types
started to die as early as day 4. Both the CTCs and C4-2 cells had clusters of cells and single
cells. In the case of C4-2 cells, a subset of cells experienced substantial growth (Figure 2-6B),
forming colonies from single cells or clusters (Figure 2-6C). In contrast, none of the primary
CTCs exhibited sustained proliferation. Most surviving CTCs spread out after becoming
adherent to the Matrigel but remained dormant throughout the 16 days of culture (Figure 2-6
B,C), suggesting that CTCs have a low in vitro clonogenic potential compared with prostate
cancer cell lines. The absence of any proliferative cells might be a result of their low abundance
or the inability of our in vitro system to mimic the important stromal factors found in vivo. We
also cultured the CTCs using different conditions (see Methods) and with co-culture of
fibroblasts, but none of these culture conditions resulted in proliferation.
We observed that the overall viability of single cells declined at a faster rate than did cell
clusters for both CTCs and C4-2 cells (Figure 2-6A). We asked whether being in a cluster
imparted enhanced survival to cells. Since we considered the entire cluster to be viable if at least
one cell remained alive, we needed to correct for the higher number of cells found in a cluster
(see Methods for formula). We compared the observed and predicted viability of clusters
extrapolated from single cells (Figure 2-6A. See Methods for formula) and found that clusters of
CTCs did not show enhanced survival but clusters of C4-2 cells had an increased survival rate
compared with single cells.
We cultured CTCs from 10 more patients in Matrigel in nanowells for up to 7 days
(Figure 2-6D). Similar to that of patient ID2, most CTCs underwent cell death, but a small
number of cells remained viable after a week of culture in Matrigel (Figure 2-6D). We did not
observe outgrowth of CTCs in samples without any starting viable CTCs (in case some CTCs
could upregulate their EpCAM expression once becoming established in culture). The two
patients (ID1 and ID2) with the highest number of persistent CTCs were deceased within six
months of their blood draws. Therefore, it is suggestive that the presence of large numbers of
persistent CTCs indicates an aggressive clinical disease and poor patient outcome. These
persistent cells appeared resistant to anoikis after being in the circulation and thus may have the
potential to colonize when growth conditions become favorable.
None of the CTCs exhibited sustainable proliferation. There are two possibilities: 1) our
culture condition is not optimal, 2) CTCs have a low clonogenic potential. In addition to the
32
Chapter 2: Circulating Tumor Cells
culture condition described, we have tried two other alternative methods of culture. The first was
adapted from the culture conditions used for intestinal stem cells with CTCs seeded directly on
chip [38]. The second method involved seeding CTCs on a layer of support cells (irradiated
MG63 osteosarcoma cells) in a 6 well dish [42]. Neither method saw proliferation of CTCs after
one month. In contrast, single C4-2 cells proliferated well on chip, suggesting that CTCs
obtained from patients have a lower clonogenic potential than C4-2 cells (Figure 2-6C).
33
Chapter 2: Circulating Tumor Cells
D
A
patient ID2
Number of viable CTCs
per 5 ml blood
day 1
viability (%)
viability (%)
C4-2
days
day 7
100
10
1
days
1
2
3
4
5
6
7
8
9
15
17
patient ID
C4-2 clusters
days
patient ID2 clusters
fold change in area
patient ID2 single cells
fold change in area
fold change in area
C4-2 single cells
fold change in area
B
days
days
days
C
day 4
day 7
day 10
cluster (5 cells)
single cell
cluster (3 cells)
patient ID2
C4-2
single cell
day 1
34
day 13
day 16
Chapter 2: Circulating Tumor Cells
Figure 2-6: Short-term viability of CTCs is low compared to that of C4-2 cell line. (A) Left: viability of single
C4-2 cell line (black line, n = 34), clusters of C4-2 cells (red, solid line, n = 32) and predicted viability of clusters of
C4-2 (red, dashed line). Right: viability of single CTCs (black line, n = 23), clusters of CTCs red, solid line, n = 9)
and predicted viability of clusters of CTCs (red, dashed line) of patient ID2. (B) Fold changes in the surface area of
single cells or clusters of C4-2 cells or CTCs from patient ID2. C) Representative optical micrographs of the
proliferation of a single C4-2 cell, a cluster of 5 C4-2 cells, a single CTC from patient ID2, and a cluster of 3 CTCs
from patient ID2. EpCAM staining is in magenta. D) Bargraph of number of viable CTCs as a function of time in
culture. Persistent viable CTCs after one week in Matrigel culture (red bar) were only present in progressing patient
2.4.4
A subset of circulating tumor cells was invasive.
In addition to clonogenic potential, the invasiveness of tumor cells also correlates with
metastatic potential. When we cultured CTCs from patient ID2, we noticed that even though
CTCs did not proliferate, they could invade through the Matrigel layer. Using each CTC’s
original well as a reference point, we monitored cell movement by recording the coordinates of
their centroids over time (Figure 2-7A). Tracking was possible because CTCs maintained
expression of EpCAM and other cell types started to die as early as day 4. We quantified the
invasiveness of individual CTCs from patients ID1 and ID2. Both patients had single cells and
cell clusters in their blood samples (Figure 2-7B), although clusters from patient ID1 comprised
only doublets whereas clusters from patient ID2 ranged from two to nine cells.
Clusters from patient ID1 did not exhibit enhanced invasive behavior relative to single
cells (Figure 2-7B). On the other hand, cell clusters from patient ID2 exhibited a greater range of
migration compared to single cells (Figure 2-7B). Individual cells also differed significantly in
their invasive behavior, ranging from no movement to over 100 µm in two weeks.
35
Chapter 2: Circulating Tumor Cells
A
B
Figure 2-7: Some CTCs were invasive. (A) The movement of the CTCs can be tracked by their uniquely barcoded
wells. The centroids of the CTCs, cellref, are reported relative to the centroids of the wells which can be determined
by the custom-designed image analysis software, Enumerator. The distance moved by a CTC is the Euclidean
distance between the relative positions of the same CTC of two time points. Optical micrographs (top) of EpCAM+
36
Chapter 2: Circulating Tumor Cells
cluster of cells (red) in Matrigel and plot of spatial coordinate of the centroid of the cluster as a function of time
(bottom). (B) Heat maps of the cumulative distance moved by the CTCs isolated from two patients. Each row
represents a single cell or a cell cluster. The distribution of the number of single cells and number of cells in a
cluster is shown below the heatmap for each patient.
2.4.5
A subset of circulating tumor cells secreted soluble factors
One key advantage of the nanowells is the ability to use them to detect secreted factors
from single cells with high sensitivity. We used microengraving (4 hr) to detect the secretion of
prostate specific antigen (PSA) from CTCs. PSA is a secreted protein specific to cells from
prostate. While EpCAM staining can only ascribe the histologic classification of a cell to be
epithelial, secretion of a tissue-specific protein such as PSA can further reveal the origin of
CTCs. Surface staining could not differentiate PSA+ cell line (C4-2) from PSA- cell line (PC3).
Intracellular staining could but it also killed the cells. Microengraving could clearly distinguish
PSA secreting cell line from PSA non-secreting cell line and preserve cell viability at the same
time (Figure 2-8).
Surface
Intracellular Staining
Microengraving
Figure 2-8: Microengraving can detect secretion of soluble factors. PSA levels were measured in PSA+ cell line,
C4-2 (blue) and PSA- cell line (PC3) by surface staining (left), or first permeabilizing the cells and staining with
antibodies (middle), or microengraving (right). In the case of surface and intracellular staining, the single cells were
stained with anti-PSA 488 antibodies, imaged with microscope and the 488 fluorescence level was quantified by
image analysis software, Enumerator. In the case of microengraving, the print with captured PSA was scanned and
the fluorescence level quantified with Genepix.
We sought to determine whether CTCs in the blood produce serum PSA. Using a signalbackground-ratio of 3 to 12, we could distinguish spiked prostate cells (C4-2 and LNCaP) from
other blood cells and HT29 colon tumor cells with 100% specificity (Figure 2-9A). We found
that CTCs of prostate cancer patients could secrete PSA (Figure 2-9B), but surprisingly, only a
very small subset of them did during microengraving (4 hr). For one patient (ID3), only 2/31
CTCs secreted PSA above the detection level. The low percentage of cells actively secreting in a
37
Chapter 2: Circulating Tumor Cells
population of cells at one instance has been reported in both clinical and cell lines, indicating that
not all cells are constitutively secreting soluble factors.
Furthermore, the number of PSA-secreting events did not correlate with serum PSA
levels. The rate of detectable PSA secretion by CTCs was low, ranging from 0.05-0.2 pg/hr/cell
(Figure 2-9C), implying that PSA secreted by CTCs is a negligible source of total serum PSA
(typically in the ng/ml range). This result also suggests that the numbers of CTCs reflect only a
small fraction of the total tumor burden in a patient if serum PSA correlates with tumor volume.
38
Chapter 2: Circulating Tumor Cells
A
C
B
Figure 2-9: CTCs secrete PSA. (A) PSA-secreting prostate cancer cells (LNCaP and C4-2), and non-PSA-secreting
colon cancer cells (HT29) were spiked into whole blood, loaded onto separate arrays of nanowells and probed for
secretion of PSA by microengraving. The signal-to-noise ratio (SNR) of wells containing all the spiked prostate
tumor cells were plotted as a cumulative distribution function to determine the range of positive secretion events.
Similarly, wells containing HT29 cells or blood cells are used to estimate the range of negative secretion events. In
this example, a range of 3 to 12 of SNR can distinguish prostate cancer cells from HT29 cells and blood cells with
100% specificity. (B) Scatter plot of the signal-to-noise ratio of secreted PSA measured by microengraving for 5
39
Chapter 2: Circulating Tumor Cells
prostate cancer patients, 6 healthy donors and 2 prostate cancer cell lines. Each dot indicates one measured well.
Micrographs of the secreted PSA (green) and control measure (hIgG; red) are shown for a subset of events. Red line
indicates threshold used for positive events. (C) The rate of PSA secretion of CTCs was determined by constructing
a standard curve using different dilutions of PSA detection antibody (black dots). We assume that one PSA molecule
is bound to one PSA detection antibody, thus the signal intensity of PSA detection antibody can give an estimate of
the amount of PSA. The four positive CTC secretion events in B) (red triangles) are also superimposed onto the
calibration curve. The rates of PSA secretion of 4 events are between 0.05 – 0.2 pg/hr/cell.
We tried to detect other soluble factors in addition to PSA. We first screened a panel of
19 factors in five different prostate cancer cell lines (Figure 2-10A). We narrowed down to six
factors (PSA, VEGF, CXCL1, CXCL5, CXCL8 and MMP9) because these factors were most
abundantly expressed in cancer cell lines or were functionally important. We proceeded to screen
these factors in the blood of prostate cancer patients and healthy donors as a control (Figure
2-10C). Surprisingly, we found that CTCs were not a dominant source of angiogenic factors
(Figure 2-10C, D). In fact, leukocytes and platelets secreted these factors at a relatively high
frequency (Figure 2-10C).
40
Chapter 2: Circulating Tumor Cells
A
B
RBCs
WBCs
CTCs
calcein violet
platelets
CD41a
C
GlycoA
Lin
healthy donor 1
EpCAM
healthy donor 2
cancer pt 1
cancer pt 2
D
surface staining
microengraving
CalceinV CD41a
GlycoA
EpCAM
Lin
CXCL1
CalceinV CD41a
GlycoA
EpCAM
Lin
PSA
CXCL5
CXCL8
hIgG
VEGF
MMP9
hIgG
41
Chapter 2: Circulating Tumor Cells
Figure 2-10: CTCs did not secrete detectable level of angiogenic factors. (A) We screened a panel of 19 soluble
factors in five prostate cancer cells by microengraving. We chose five of the highest expressing factors (PSA,
MMP9, CXCL1, 5, 8) and a functionally important factor (MMP9) for further analysis in primary samples. (B)
Because whole blood contained various cell types even after tumor enrichment, we identified each cell type by
surface staining (CD41a – platelets, Glycophorin A- erythrocytes, lineage marker – leukocytes, EpCAM – CTCs).
Erythrocytes stained negative for Calcein Violet. Platelets alone stained slightly positive for Calcein Violet.
Leukocytes and CTC stained positive for Calcein Violets. Some platelets adhered onto leukocytes, accounting for
the Calcein Violet high/CD41a+ population. (C) The percentage of secretion in single cells was estimated by globally
fitting the microengraving results with the number of each cell type for all the wells. Platelets and leukocytes, not
CTCs, secreted angiogenic factors. (D) Shown are examples of two wells containing a single CTC (top) and a single
CTC with a leukocyte (bottom) from the blood of a prostate cancer patient. The single CTC did not secrete CXCL1,
5 or 8, but the CTC with the leukocyte secreted PSA and MMP9. Although we could pinpoint which cell secreted
which factor, we could deduce that the CTC secreted PSA based on PSA’s tissue specificity and that the leukocyte
secreted MMP9 based on the high probability of MMP9 secretion by leukocytes.
In addition to direct measurement of secreted factors, we also measured the activity of
enzymes secreted into the wells by observing the cleavage of their substrates. We incubated
enriched CTCs with a FRET-based polypeptide protease substrates (moderately specific for
matrix metalloproteinases and A disintegrin and metalloproteinases) in the nanowells. Only a
small proportion of CTCs cleaved the substrates immediately after isolation (Figure 2-11). To
mimic the paracrine interactions in the bone microenvironment since prostate cancer cells have a
tendency to metastasize to the bone tissue, we cultured the CTCs with conditioned media from
MG63 osteosarcoma cells and C4-2 prostate cancer cells. After a week in culture, two of the five
surviving CTCs cleaved the peptides to a greater degree than did freshly isolated CTCs, resulting
in an increase in the fluorescent signal within the well (Figure 2-11). These results indicate that
CTCs can secrete proteolytic enzymes necessary to break down extracellular matrix.
Number of CTCs
20
day 1
day 7
EpCAM / FRET peptide
15
10
5
0
1.10
1.20
1.30
1.40
signal to background ratio
Figure 2-11: CTCs secrete proteolytic enzymes. Histogram of proteolytic activity of CTCs was measured by the
cleavage of FRET-based polypeptide substrates immediately after isolation (gray) or one week in culture (red).
42
Chapter 2: Circulating Tumor Cells
Optical micrograph shows one example of positive signals generated by an EpCAM+ cell (magenta) after 7 days.
Each well is 65 µm in width.
2.4.6
Some EpCAM- cell persisted in Matrigel
When we cultured the EpCAM+ CTCs, we also observed that some EpCAM- cells
remained viable at the end of the seven days. Compared to the EpCAM+ CTCs, the majority of
the EpCAM- cells were smaller in size. The EpCAM- cells were also more abundant than the
EpCAM+ CTCs but their numbers did not correspond to the disease status of the cancer patients.
Therefore, it is likely that a large fraction of the EpCAM- cells were normal cells. We retrieved
some of these cells and performed whole-genome amplification, but we did not detect any p53
mutations in the cells that were successfully amplified.
43
Chapter 2: Circulating Tumor Cells
A
EpCAM+
100
B
10
EpCAM-
1
0.1
1
10
Day 1
Calcein AM AnnexinV EpCAM
1000
Day7
CD45
TL
Calcein AM AnnexinV EpCAM
CD45
EpCAM-
EpCAM+
TL
100
Figure 2-12: Some EpCAM- cells persisted for one week in Matrigel. (A) EpCAM+ cells (left) rapidly underwent
cell death after one week in Matrigel (white bar – starting number of cells, red bar – remaining number of cells after
one week). We also observed the persistence of some EpCAM- cells which were defined by Calcein
Violet+/Annexin V-/EpCAM-/Lin- staining. (B) Shown are examples of a surviving EpCAM+ cell (first row) and
EpCAM- cells (rest of the rows) from a cancer patient. Surviving cells are indicated by yellow arrows.
2.5
2.5.1
Discussion
Only a subset of circulating tumor cells exhibit malignant traits indicative of
metastatic potential
Heterogeneity among CTCs adds complexity to the understanding of cancer metastasis.
We demonstrated that CTCs found in mCRPC patients exhibit functional heterogeneity in terms
of viability, invasiveness and proteolytic activity, with as few as 2% of the total viable CTCs
44
Chapter 2: Circulating Tumor Cells
possessing malignant traits in progressing patients. For example, patient ID3 had 117 viable
CTCs at the time of isolation but only 2/117 remained viable on day 7. In a separate experiment,
only 2/86 CTCs of the same patient secreted proteolytic enzymes. The rarity of malignant CTCs
agrees with two pieces of evidence supporting metastatic inefficiency. First, the low engraftment
rate of CTCs in an immunocompromised mouse model, and the requirement for at least 103
CTCs to initiate tumor formation, suggest that metastatic initiating cells (MICs) comprise a rare
fraction of total CTCs [8]. Second, the discrepancy between the number of CTCs (0 – 23,618
cells per 7.5 ml of peripheral blood[4]) and the number of metastases (1-13 lesions per patient
[3]) implies that the majority of the shed tumor cells are incapable of progressing to overt
metastases.
Here we have provided two additional pieces of evidence to account for the low
frequency of tumorigenic CTCs: 1) significant apoptosis during and after circulation and 2)
dormancy of CTCs. By preserving the viability of isolated CTCs, we quantified the distribution
of viable and dead cells. The relatively high percentage of apoptotic cells we saw agree with
previous studies that detected a significant number of apoptotic CTCs by Terminal
deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay [43] or caspase cleavage at
cytokeratin 18 [44, 45]. Because these previous studies used an intracellular marker, however,
they could not directly quantify the presence of viable cells. Our data indicated that only 22% of
the patients have ≥ 1 viable cells in 1 ml of blood after isolation. The majority of these surviving
cells continued to undergo rapid cell death after leaving the circulation, and only a few cells
could persist in Matrigel culture for over two weeks.
Secondly, we found that these viable CTCs are in a quiescent state. The observation that
these CTCs are non-proliferative may seem surprising initially but corroborates an earlier study
that fails to detect any Ki-67 staining in the CTCs of 47 breast cancer patients [46]. A separate
genetic analysis of CTCs also showed that compared with established cell lines, CTCs had a
quiescent phenotype, with a decreased expression of growth factors such as VEGFA, MET,
ESR1, EGFR, and HER2 and the cell cycle genes downstream of these growth factors including
MYC, ATF3, TERT, RAC1, FOXA1, RRM1, CCNB1, and BIRC5 [47]. It is important to realize
that the proliferative potential of CTCs found in patients differs significantly from that of cell
lines. Conventional chemotherapy is ineffective in eliminating single dormant cells in the bone
marrow of cancer patients [48] because these dormant cells evade drugs that target proliferating
45
Chapter 2: Circulating Tumor Cells
cells. These dormant cells may have the potential to restart their proliferative program after a
number of years as minimal residual disease [49]. Accordingly, the presence of dormant
disseminated tumor cells correlates with a worse prognosis and a higher rate of relapse [50-52].
We demonstrated that clusters of CTCs could exhibit invasive phenotypic behaviors
while retaining their expression of EpCAM, an epithelial marker. This collective cell invasion, in
which multicellular units invade while maintaining their cell-cell junction molecules, was
thought to be the predominate form of invasion in highly differentiated tumors such as epithelial
prostate cancer [53, 54]. While experimental evidence has demonstrated the importance of
epithelial-mesenchymal transition for cell invasion to occur, and has thus raised the question
whether CTCs expressing EpCAM, an epithelial marker, can be invasive. This result suggests
that the loss of epithelial markers is not a requirement for invasion to occur. Furthermore,
consistent with our observation that the invasive CTCs were quiescent, an earlier study identified
primary tumor cells in the invasive margin of ductal carcinoma to be invasive yet dormant. These
cells were enriched for pro-migratory and anti-proliferative genes relative to intratumoral cells
[55]. An in vitro model using tumor cell lines encapsulated in 3D collagen matrices also
demonstrated that tumor cells could be highly invasive and dormant, but knockdown of p27
could reverse this dormancy [55].
We detected the secretion of PSA in 60% patients, which is lower than the reported
83.3% using ELISPOT [50]. The differences might be due to our higher limit of detection to
ensure specificity. But we showed that the number of PSA-secreting cells is very low, and CTCs
are not a significant source of PSA. It is possible that other sources of tumor cells, for example,
disseminated tumor cells in bone marrow, or other undetectable micrometastases also constitute
a significant reservoir of mobilized tumor cells.
Interrogating the functional behavior of individual CTCs allowed us to demonstrate that
CTCs from the same patient differ significantly in their viability, invasiveness and secretory
profiles. We identified a rare subset among isolated CTCs with phenotypes consistent with more
efficient metastasis in mCRPC patients. Cells in this subset can resist apoptosis and remain
dormant for the duration of our experiments. These dormant cells are, however, invasive, and
they can secrete proteolytic enzymes. Therapy to target this quiescent but potentially dangerous
subset of cells may be necessary to eliminate minimal residual disease. A larger cohort of
46
Chapter 2: Circulating Tumor Cells
patients and emergence of new CTC markers in the future can potentially bring insight into how
functional behavior of CTCs affects patient outcome.
2.5.2
Nanowells allow functional characterization of circulating tumor cells
The spatially-addressable array of nanowells offers three advantages: (1) the registration
of each cell allows the spatiotemporal tracking of individual cells; (2) the compartmentalization
of single cells allows clonal comparison and mapping of tumor heterogeneity; and (3) direct onchip assays measure native cell behaviors.
This approach can be applied to screen other rare cells such as cancer stem cells, and
elucidate the organization of the rare cells amongst the network of numerous other cell types.
47
Chapter 3: Intraoperatively Dislodged Tumor Cells
3
Chapter 3: Identification of Intraoperatively Dislodged Tumor Cells
during Surgical Resection of Lung Tumor
3.1
Abstract
Intraoperative tumor shedding is thought to be a potential cause of tumor dissemination.
Earlier studies defined tumor cells primarily by cytomorphological examination and/or epithelial
marker staining, and thus could not clearly distinguish normal epithelial cells from tumor cells. It
remains to be determined whether tumor cells are mobilized during surgery. In this study, we
used arrays of subnanoliter wells to analyze blood samples withdrawn from the tumor-draining
pulmonary vein at the end of pulmonary lobectomy procedures. Malignancy was determined by
comparing the genotype of shed EpCAM+ cells to matched tumor cells and normal adjacent
tissue. Genotyping was performed using three genetic approaches: 1) copy number variation of
pooled epithelial cells (10-20 cells), 2) targeted sequencing against a panel of frequently mutated
genes in lung cancer and 3) nested PCRs of single cells. We found that a majority of the
EpCAM+ cells shed were normal epithelial cells, implying that enumerating tumor cells by
EpCAM staining alone may lead to an overestimate of the amount of tumor cells shed
intraoperatively. The number of tumor cells detected varied depending on the type of surgeries
performed. VATS (no wedge) has the highest number of tumor cells due to post-surgery tissue
compression and VATS (wedge) has the lowest number due to the removal of tumor tissue prior
to the ligation of pulmonary vein. Thoracotomy procedures gave the most realistic estimate of
the number of tumor cells shed intraoperatively. Tumor cells were present in a subset of patients
who have undergone thoracotomy procedure.
3.2
3.2.1
Introduction
Putative tumor cells were shed during surgical manipulation of tumors
Surgical resection of the primary tumor is the first line of treatment in early stage non-
small cell lung cancer (NSCLC), but 30% of the patients relapse and succumb to distant
metastases [56]. Intraoperative tumor shedding can potentially contribute to tumor recurrence
[9]. A number of studies have reported incidences of tumor seeding during surgery [9, 57], or
local recurrences as a result of surgery [58, 59]. In particular, the study by Yamanaka et al. took
sampling of blood through a catheter inserted into the mesenteric vein, and found clusters of
48
Chapter 3: Intraoperatively Dislodged Tumor Cells
tumor cells released into the blood stream in patients with portal invasions [57]. In addition, notouch isolation technique, in which ligation of venous egress preceded the isolation of malignant
tissue, was developed to reduce intraoperative tumor shedding and shown to improve overall
survival by 15% in colon cancer [60, 61]. Therefore, it is of interest to quantify how much tumor
cells are dislodged during the physical manipulation of the tumor tissue.
3.2.2
Earlier studies did not distinguish normal epithelial cells from malignant cells
These earlier studies identified shed tumor cells primarily by cytomorphological
examination, immunohistochemical staining or indirect detection of epithelial cell markers such
as cytokeratin and EpCAM using RT-PCR [9, 62-64]. Using cytokeratin staining, it was
previously estimated that the number of tumor cells shed during surgery ranged from 10 to 7 x
106 [9]. Another study reported a high number of tumor cells found in the pulmonary vein (mean
1195, median 81) compared to the peripheral blood (mean 1, median 0 per 7.5 ml of blood) [63].
However, it remains to be determined whether the count of tumor cells is contaminated by
normal epithelial cells since none of the epithelial markers used is tumor-specific. The lack of
single cell isolation techniques when performing genetic analysis such as RT-PCR also limits the
sensitivity of detection to about ten cells [64, 65]. This sensitivity may be suboptimal when the
amount of tumor cells shed is extremely rare.
3.2.3
Isolation and retrieval of intraoperatively shed cells using nanowells
We made use of recent advances in single cell isolation techniques and genomic analysis
[20] to interrogate single epithelial cells shed intraoperatively. We obtained whole blood from
ligated tumor-draining pulmonary vein that was removed during pulmonary lobectomy
procedures. We isolated individual epithelial cells using arrays of subnanoliter wells previously
developed [41]. The array comprises 84,762 cubic wells of 65 um each. Because the shed cells
are rare, loading of cells biased the occupancy of the wells to single epithelial cells. We then
used a robotic micromanipulator to retrieve the single cells for targeted or whole genome
sequencing. By sequencing the matched tumor and adjacent normal tissue, we can pinpoint
whether the shed cells are tumor in origin.
49
Chapter 3: Intraoperatively Dislodged Tumor Cells
3.3
3.3.1
Materials and methods
Patients and sample collection
Patients were recruited according to Institutional Review Board-approved protocol at the
Lahey Hospital and Medical Center and Use of Humans as Experimental Subjects approved
study at MIT. Patients identified had biopsy proven lung cancer, or had lung tumors suspicious
for lung cancer by CT scan characteristics and/or PET scan findings. Lung cancer patients
underwent thoracotomy or video-assisted thoracoscopy. Blood were withdrawn from the
patients, 8 ml each, from their arterial line (referred to as peripheral blood) at the start of surgery
and placed in an EDTA tube. If a lobectomy, segmentectomy, or extended wedge resection was
performed for the treatment of the patient’s malignancy, then 8 ml of blood was obtained from
the draining pulmonary vein (referred to as pulmonary vein blood) once the ligature/stapler has
been placed around the pulmonary vein and was controlled. After the blood was drawn from the
pulmonary vein the vessel was ligated/stapled and divided. Once the lobe had been removed
from the patient, the few remaining blood (2 – 8 ml) in the pulmonary vein specimen (referred to
as specimen blood) was placed in a separate EDTA tube. If the tumor was at least 1.5 cm in size
then a 5 mm x 5 mm x 5 mm segment of tumor was removed and placed in saline and on ice. A 2
cm x 2 cm x 1 cm segment of the adjacent normal tissue was removed 4-5 cm outside the tumor
margin. The tissue specimens were transported to MIT within 2 hr. Table 3-1 shows the patient
information.
50
Chapter 3: Intraoperatively Dislodged Tumor Cells
ID
CW1
CW4
CW5
CW6
CW7
CW10
CW11
CW12
CW13
CW14
CW20
CW21
CW22
CW23
CW24
CW25
CW26
CW27
Procedure
VATS (wedge)
VATS(no wedge)
VATS(no wedge)
VATS(no wedge)
VATS (wedge)
VATS(no wedge)
VATS (wedge)
thoracotomy
VATS (wedge)
VATS (wedge)
VATS (wedge)
thoracotomy
thoracotomy
VATS(no wedge)
thoracotomy
thoracotomy
thoracotomy
VATS(no wedge)
Sex
M
M
M
F
F
M
M
F
F
F
F
F
F
M
F
F
M
M
Age
68
73
65
78
72
53
69
69
66
53
46
61
72
75
76
66
69
66
Smoker
1
1
1
0
1
1
1
1
0
1
1
0
1
1
1
1
1
1
Cl Stage
IA (T1aN0M0
IIA(T1aN1M0)
IB(T2aN0M0)
IA(T1aNoM0)
IB(T2aN0M0)
IB(T2aN0M0)
IB(T2aN0M0)
IIA(T1bN0M0)
IA(T1aN0M0)
IA(T1aN0M0)
IA(T1aN0M0)
IIB(T3N0M0)
IIB(T3N0M0)
IB(T2aN0M0)
IA(T1aN0M0)
IA(T1bN0M0)
IA(T1aN0M0)
IA(T1bN0M0)
Path Stage
IA (T1aN0M0)
IIA(T2aN1M0)
IIA(T2aN1M0)
IA (T1aN0M0)
IB(T2aN0M0)
IB(T2aN0M0)
IA(T1bN0M0)
IIB(T3N0M0)
IA(T1aN0M0)
IB(T2aN0M0)
IA(T1aN0M0)
IIB(T3N0M0)
IIB(T3N0M0)
IIIA(T2aN2M0)
IB(T2aN0M0)
IA (T1bN0M0)
IA(T1aN0M0)
IA (T1aN0M0)
Path
Adeno
Adeno
Squam
Adeno
Adeno
Adeno
Adeno
Adeno
Adeno
Adeno
Adeno
Squam
Squam
Adeno
Adeno
Adeno
Adeno
Adeno
Grade
1
3
2
2
2
3
2
1
1
3
1
3
3
3
2
2
3
3
# positive
nodes
0
1
1
0
0
0
0
0
0
0
0
0
0
3
0
0
0
0
CW28
thoracotomy
M
70
1
IIA(T2bN0M0)
IIA(T2bN0M0)
Squam
3
0
CW29
CW30
thoracotomy
VATS (wedge)
F
M
65
57
1
1
IIIA(T1aN2M0)
IA(T1aN0M0)
IA(T1aN0M0)
IA(T1aN0M0)
Adeno
Adeno
2
3
0
0
CW41
CW42
CW43
CW44
CW45
VATS (wedge)
VATS(no wedge)
VATS (wedge)
VATS(no wedge)
VATS(no wedge)
M
M
F
F
M
56
58
67
79
68
1
1
1
1
1
IA(T1aN0M0)
IA(T1bN0M0)
IA(T1bN0M0)
IIA(T1aN1M0)
IA(T1aN0M0)
IA (T1aN0M0)
IA(T1bN0M0)
IA(T1aN0M0)
IIIA(T1aN2M0)
IA(T1bN0M0)
Adeno
Adeno
Adeno
Adeno
Squam
2
1
3
2
3
0
0
0
5
0
CW46
CW47
VATS(no wedge)
VATS (wedge)
M
F
62
65
1
1
IA(T1aN0M0)
IA(T1aN0M0)
IIB(T2bN1M0)
IB(T2aN0M0)
Adeno
Adeno
3
3
6
0
CW48
CW49
CW50
VATS(no wedge)
VATS(no wedge)
thoracotomy
M
F
F
86
76
59
1
1
1
IIA(T1aN1M0)
IA (T1bN0M0)
IIB(T3N0M0)
IIA(T2aN1M0)
IA(T1bN0M0)
IIIA(T3N1M0)
3
3
3
2
0
2
CW51
CW53
CW54
thoracotomy
VATS(no wedge)
VATS(no wedge)
F
M
F
64
79
55
1
1
1
IIA(T1bN1M0)
IA(T1bN0M0)
IA(T1aN0M0)
IIIA(T1bN2M0)
IA(T1bN0M0)
IA(T1aN0M0)
Adeno
Squam
Adeno
15%
Squam,
70%
large
neuro,
10%
small
Squam
Adeno
4
2
2
8
0
0
51
LgV
Inv
LySm
V Inv
VPL
Inv
1
1
1
1
1
1
1
1
1
1
1
1
1
4.5
2.8
7
lobe
remove
d
LUL
RML
LLL
LLL
LUL
LLL
LUL
RLL
LUL
wedge
RUL
RUL
RUL
RLL
RUL
LUL
RUL
RUL
left
pneum
RLLR
ML
RUL
LUL
Tri
RUL
RUL
RUL
LLL
RMLR
LL
RUL
RMLR
LL
RLL
RLL
3
3
1.1
RMLR
LL
RUL
RLL
tumor
size
1.2
2.5
3
1.3
4
3.7
3
4
1
1
1
5
3.5
3.4
2.7
2.5
2
2
6
0.9
2
1
1.5
3.5
2
1.5
2.3
1
1
1
1
1
1
3
1.5
#
EpCAM+
cell
161
277
272
20
149
9
632
17
423
9406
1219
10
26
190
78
22
22
8
12
9
104
232
24
1015
99
41
509
53
131
21
17
31
140
469
Chapter 3: Intraoperatively Dislodged Tumor Cells
CW55
CW56
VATS(no wedge)
thoracotomy
F
M
86
62
1
1
IIB(T3N0M0)
IB(T2aN0M0)
IB(T2aN0M0)
IB(T2aN0M0)
Adeno
Squam
2
2
9
0
CW57
CW58
VATS(no wedge)
thoracotomy
F
F
59
45
1
0
IA(T1aN0M0)
IIIA(T1aN2M0)
IIB(T3N0M0)
IIIA(T1aN2M0)
Adeno
Adeno
2
4
0
10
CW59
CW60
CW61
CW62
thoracotomy
VATS (wedge)
VATS (wedge)
thoracotomy
F
F
F
F
69
63
57
63
1
0
1
1
IIB(T3N0M0)
IA(T1bN0M0)
IA(T1aN0M0)
IA(T1bN0M0)
IIB(T3N0M0)
IA(T1aN0M0)
IA(T1aN0M0)
IA(T1bN0M0)
Adeno
Adeno
Adeno
Adeno
1
1
2
3
0
0
0
0
1
1
1
4
4
1.8&7
mm
1.5
9
1.5
1.2
3
LUL
RLL
433
30
RUL
RML
RULR
ML
RLL
RUL
RLL
0
120
14
47
92
2920
Table 3-1: Patient characteristics: Cl- clinical; Path – pathology; LgV InV – large vein invasion; LySmV Inv – lymphatics and small vein invasion; VPL Inv –
visceral pleural invasion
52
Chapter 3: Intraoperatively Dislodged Tumor Cells
3.3.2
Enrichment of epithelial cells from blood samples
For each of blood sample, 5 ml of was enriched for epithelial cells using RosetteSep
CD45 depletion kits (StemCell Technologies). Subsequently, 250 µl antibody cocktail was
incubated with 5 ml of whole blood for 20 min at room temperature. The blood was then diluted
with PBS in 1:1 ratio and layered onto Ficoll-Paque Plus (GE healthcare) in a SepMate tube
(StemCell Technologies) and centrifuged at 800x g for 10 min. The upper layer containing the
serum and buffy coat was transferred to a new tube and washed twice. Further red blood lysis
was sometimes necessary to remove residual red blood cells.
3.3.3
Staining and microscopy
Following CD45 depletion, the residual cells from whole blood were stained with
EpCAM, a cocktail of lineage markers for leukocytes including CD3, CD16, CD20, CD38 and
CD45 in a dilution of 1:20 (refer to the list of antibodies used) and 1 µM Calcein AM violet
(Molecular Probes) at room temperature for 1 hr. For viability assay, the cells were rinsed with
PBS and stained with and Annexin V FITC (BD Pharmingen) in Annexin V binding buffer (BD
Pharmingen) for 10 min at room temperature.
3.3.4
Single cell retrieval
Enumerator generates a list of well positions containing the epithelial cells of interest. A
robot micromanipulator (CellCelector, AVISO GmbH) calibrates the position of each well of the
array and drives to the specified wells to retrieve the cells. The robot aspirates 1 µl of PBS
directly above each well and deposits the cells into a 96-well plate for downstream assays.
Borosilicate capillary tubings (internal diameter 0.86 mm, outside diameter 1.5 mm, length 10
cm, Sutter Instrument) were purchased and shaped with a micropipette puller (Sutter Instrument
P-97). The tip was manually scored with a ceramic tile (Sutter Instrument) to achieve an internal
diameter of 50 to 60 µm.
53
Chapter 3: Intraoperatively Dislodged Tumor Cells
3.3.5
Tumor disaggregation and flow sorting
Tumor and adjacent normal tissues were cross diced with a pair of scalpels, and
resuspended in 2 ml of digest media consisting of 1 mg/ml collagenase A (Roche) and 1 mg/ml
dispase (Stemcell Technologies). The diced tissue was incubated on a shaking platform at 37 °C
for 1 hr. The digested tissue was then rinsed twice in PBS and stained with the same antibodies
as the blood samples and flow sorted into a 96 well plate by FACSAria III (BD Biosciences).
The sorted cells were frozen until further genomic analysis.
3.3.6
Mutation analysis of single cells using nested PCR
Single cells from blood samples were deposited into 0.8% triton X-100 in 9 µl Ultrapure
water (Invitrogen) and 1 µl Proteinase K (600 mAU/ml, Qiagen) by the robot manipulator as
described in the cell retrieval section. Cell lysis was achieved by one freeze-thaw cycle and 1 hr
Proteinase K digestion at 55 °C. Two rounds of nested PCRs were performed to amplify the
exons of interest with multiplex PCR kit (Qiagen). The first round of PCR amplification was
performed for 25 cycles (each cycle consists of denaturation 95 °C 30 s, annealing 60 °C 4 min,
extension 72 °C 90 s). The second round of PCR amplification was performed for 35 cycles
(each cycle consists of denaturation 95 °C 30 s, annealing 60 °C 3 min, extension 72 °C 90 s).
Refer to Table 3-2 for primers used for p53 and kras sequencing.
Gene
Exon
PCR
Direction
Sequence (5'-3')
TP53
4
outer
forward
CTGAGGACCTGGTCCTCTGACT
TP53
4
outer
reverse
GGCCAGGCATTGAAGTCTCAT
TP53
4
inner
forward
ACCTGGTCCTCTGACTGCTCTT
TP53
4
inner
reverse
AAGCCAGCCCCTCAGGGCAA
TP53
4
sequencing
forward
CCTGGTCCTCTGACTGCTCTTTTCACCCA
TP53
5
outer
forward
GCTCGCTAGTGGGTTGCAGGAGGTGC
TP53
5
outer
reverse
TGTCGTCTCTCCAGCCCC
TP53
5
inner
forward
TGCTGCCGTGTTCCAGTTGCT
TP53
5
inner
reverse
TGTCGTCTCTCCAGCCCC
TP53
5
sequencing
forward
CAACTCTGTCTCCTTCCT
TP53
6
outer
forward
GGCTGGTTGCCCAGGGTCC
TP53
6
outer
reverse
GGTCAAATAAGCAGCAGGAGAAAGCCCCC
TP53
6
inner
forward
GGCTGGTTGCCCAGGGTCC
TP53
6
inner
reverse
CTTAACCCCTCCTCCCAGAG
54
Chapter 3: Intraoperatively Dislodged Tumor Cells
TP53
6
sequencing
forward
GGTCCCCAGGCCTCTGATTCC
TP53
7
outer
forward
GCCACAGGTCTCCCCAAGGCG
TP53
7
outer
reverse
AGCGGCAAGCAGAGGCTGGG
TP53
7
inner
forward
CGCACTGGCCTCATCTTGGGC
TP53
7
inner
reverse
AGTGTGCAGGGTGGCAAGTG
TP53
7
sequencing
forward
CCTCATCTTGGGCCTGTGTT
TP53
8
outer
forward
GGCTCCAGAAAGGACAAGGGTGG
TP53
8
outer
reverse
ATAACTGCACCCTTGGTCTC
TP53
8
inner
forward
TGGGAGTAGATGGAGCCTGGT
TP53
8
inner
reverse
CCCTTGGTCTCCTCCACCGCT
TP53
8
sequencing
forward
CCTTACTGCCTCTTGCTTCT
KRAS
2
outer
forward
CGTCTGCAGTCAACTGGAAT
KRAS
2
outer
reverse
TCATGAAAATGGTCAGAGAAACC
KRAS
2
inner
forward
GGTGGAGTATTTGATAGTGTATTAACC
KRAS
2
inner
reverse
GGTCCTGCACCAGTAATATGC
KRAS
2
sequencing
forward
TTAACCTTATGTGTGACATGTTCTAA
EGFR
18
outer
forward
GCGTACATTTGTCCTTCCAAATGAGCTGG
EGFR
18
outer
reverse
AGATGATGGAAATATACAGCTTGCAAGGAC
EGFR
18
inner
forward
CCGTGTCCTGGCACCCAAGC
EGFR
18
inner
reverse
TCTGGGCTCCCCACCAGACC
EGFR
18
sequencing
forward
TGGTGAGGGCTGAGGTGACCC
EGFR
19
outer
forward
GCTCCACAGCCCCAGTGTCC
EGFR
19
outer
reverse
CAGCATGGGAGAGGCCAGTGC
EGFR
19
inner
forward
CCTTCGGGGTGCATCGCTGG
EGFR
19
inner
reverse
GCCATGGACCCCCACACAGC
EGFR
19
sequencing
forward
GGGCAGCATGTGGCACCATCTC
EGFR
20
outer
forward
ACAGCCCTGCGTAAACGTCCC
EGFR
20
outer
reverse
GCTGCATGCACGCACACAC
EGFR
20
inner
forward
TGGCCACCATGCGAAGCCAC
EGFR
20
inner
reverse
GGAGCGCAGACCGCATGTGAG
EGFR
20
sequencing
forward
GCCACACTGACGTGCCTCTCC
EGFR
21
outer
forward
AGTCACTAACGTTCGCCAGCC
EGFR
21
outer
reverse
CAGCTGCTGCGAGCTCACCC
EGFR
21
inner
forward
TCCTCGACGTGGAGAGGCTCAG
EGFR
21
inner
reverse
GCAGCCTGGTCCCTGGTGTC
EGFR
21
sequencing
forward
ACCCTGAATTCGGATGCAGAGCTTC
Table 3-2: List of primers used for nested PCR of single cells to detect mutations in TP53, KRAS and EGFR
3.3.7
Whole genome amplification
Whole genome amplification (WGA) was performed using the REPLI-g single cell kit
(Qiagen). 10 – 20 single cells from blood samples were deposited into a single well containing 8
55
Chapter 3: Intraoperatively Dislodged Tumor Cells
µl Ultrapure water, 0.3 µl DLB buffer (reconstituted in 55 µl water) and 0.1 µl 1M DTT buffer
by the robot manipulator. The volume of aspiration was reduced to 0.2 µl per cell. The cells were
lysed at 65 °C for 10 min. 3 µl Stop solution was added to quench cell lysis, followed by 29 µl
Reaction Buffer and 2 µl DNA polymerase. Isothermal amplification was carried out at 30 °C for
8 hr.
3.3.8
Copy number variation analysis
WGA products were quantified by PicoGreen (Invitrogen), adjusted to 2.5 ng/ul for
library preparation using the Nextera DNA Sample Prep Kit (Illumina), and barcoded with the
Nextera Index Kit (Illumina). Sequencing of barcoded pools was performed with paired-end 150
reads using the Illumina MiSeq.
3.3.9
Targeted sequencing
Amplicon enrichment was performed using the Lung Cancer Panel within the GeneRead
DNAseq Targeted Exon Enrichment Panels for NGS (Qiagen). Library preparation was done
with the NEBNext DNA Library Prep Master Mix Set for Illumina (NEB) and barcoded with the
NEBNext Multiplex Oligos for Illumina (NEB). Sequencing of barcoded pools was performed
with paired-end 150 reads using the Illumina MiSeq and data were analyzed using a combination
of the Cloud-Based DNAseq Sequence Variant Analysis (Qiagen) and custom scripts.
3.4
3.4.1
Results
Surgery released EpCAM+ cells into the tumor-draining pulmonary vein.
The patients in our study underwent one of the three surgical procedures (Figure 3-1A).
In the video-assisted thoracic surgery (VATS) (no wedge) procedure, a lobe of lung
(approximately 20 cm) was removed through a 4-5 cm incision in the chest wall. In the VATS
(wedge) procedure, a diagnostic wedge containing the tumor was removed at the start of the
surgery. In a thoracotomy procedure, the tumor containing lobe was removed through a larger
incision of approximately 20 cm in the chest wall with minimal tissue compression.
In order to investigate whether tumor cells were shed intraoperatively, we removed blood
(1-10 ml) from the ligated tumor-draining pulmonary vein once the lobe was outside the patient’s
56
Chapter 3: Intraoperatively Dislodged Tumor Cells
body. Therefore, the cells we measured were shed from the time when the pulmonary vein was
ligated to the time when the lobe was removed.
Because EpCAM+ cells were rare in the specimen blood, we developed a platform to
enumerate and retrieve single cells. The 65 µm cubic nanowells partitioned the cells into units of
single cells or preformed clusters, and the robotic micromanipulator could accurately retrieve
only the cell(s) of interest but not the neighboring cells (Figure 3-1B). On the other hand,
epithelial cells were abundant in normal and tumor tissue, so we used flow cytometry to isolate
them. The Calcein AM+/Annexin V-/EpCAM+ /Lin- staining allowed us to isolate viable
epithelial cells.
We observed a significant number of EpCAM+ cells in the pulmonary vein (Figure 3-1C),
with the lowest in thoracotomy (mean 165, median 22, range 3 to 2920), followed by VATS (no
wedge) (mean 221, median 197, range 0 to 509) and the highest in VATS (wedge) (mean 1128,
median 115, range 47 to 9406). The high number of EpCAM+ cells in VATS (wedge) was
surprising initially because the tumor tissue was already removed prior to the ligation of
pulmonary vein in this procedure and tumor cells should not be detected in the pulmonary vein.
This implied that many normal EpCAM+ cells were released as a result of tissue compression of
the lobe as it was squeezed through a narrow incision during a VATS procedure. Therefore we
needed to distinguish between normal and malignant epithelial cells using methods other than
EpCAM staining in order to accurately quantify the tumor cells shed. In addition, since
thoracotomy does not involve tissue compression through a narrow incision, it might most
accurately reflect the degree of tumor cell mobilization during surgery.
57
Chapter 3: Intraoperatively Dislodged Tumor Cells
Figure 3-1: Isolation of intraoperatively shed tumor cells from three types of procedures. A) NSCLC patients
recruited in this study underwent either VATS or thoracotomy. VATS is a laparoscopic procedure whereas
thoracotomy involves opening of the chest wall. Within the VATS procedure, patients either have the entire lobe
removed with the tumor in place or first have the tumor removed through a diagnostic wedge. B) EpCAM+ cells
from tumor and normal tissue were isolated by flow cytometry whereas EpCAM+ cells from the tumor draining
pulmonary vein were isolated by nanowells. C) The number of EpCAM+ cells found in pulmonary vein was
enumerated using nanowells. ** p-value < 0.001
3.4.2
Copy number variation analysis revealed tumor signature in pulmonary vein blood
after surgery.
Copy number variation allowed us to quickly survey the genomic landscape of
intraoperatively shed tumor cells, without a priori knowledge of specific mutations. We
compared a low-pass copy number variation analysis between normal tissue, tumor and
specimen blood (10-20 pooled cells).
The profile of EpCAM+ cells in the blood of VATS (no wedge) procedures generally
resembled the profile of the tumor tissue in the two patients shown, suggesting a high tumor
content in the pooled epithelial cells (Figure 3-2A). On the other hand, the profile of VATS
(wedge) blood specimens was relatively uniform (Figure 3-2B), implying that no or few tumor
cells were released into the pulmonary vein since the tumor tissue was already removed from the
58
Chapter 3: Intraoperatively Dislodged Tumor Cells
lobe. Interestingly, the profiles of thoracotomy blood specimens were mixed (Figure 3-2C). Two
patients (CW56 and CW59) had a flat profile whereas two other patients (CW51 and CW62) had
gains and losses of chromosomes. The profile of CW51 blood specimen resembled that of the
matched tumor, but the degree of gains and losses was attenuated, suggesting that only a small
fraction of the EpCAM+ cells in the blood was malignant, and was being diluted by the presence
of normal cells. On the other hand, the degree of chromosomal aberrations of blood was greater
than that of the tumor tissue in patient CW62, implying that the blood specimen has an enriched
tumor content even more than the matched tumor.
59
Chapter 3: Intraoperatively Dislodged Tumor Cells
Figure 3-2: Copy number variation analysis. EpCAM+ cells from normal tissue, tumor and pulmonary vein were
whole-genome amplified and subjected to next generation sequencing. The copy number variation analysis was
performed with HMMCopy, a software that segments chromosomes using a Hidden Markov Model. Chromosomal
gains are colored in red; losses are colored in green; neutral copies are colored in blue. Patients underwent either
VATS (no wedge) in A), or VATS (wedge) in B) or thoracotomy in C).
60
Chapter 3: Intraoperatively Dislodged Tumor Cells
3.4.3
Targeted sequencing confirmed consistent mutations between primary tumor and
pulmonary vein blood after surgery.
The low purity of tumor cells in a pooled sample can attenuate the copy number changes,
making the analysis difficult to interpret. A low number of tumor cells could be present in
thoracotomy specimens and even VATS (wedge) specimens in case of residual tumor tissue. The
sensitivity of copy number variation analysis cannot resolve a low tumor content in a pooled
sample especially when the primary tumor has low copy number gains or losses. Therefore, we
performed targeted sequencing against a panel of 20 frequently mutated genes in NSCLCs.
Instead of sequencing across the entire genome at a low coverage as in the case of the copy
number variation analysis, targeted sequencing focuses only on 20 genes but yields a deep
coverage (50x to 5000x) for each gene. The deep coverage at common somatic mutations
increases the sensitivity of detection and can estimate the frequency of tumor cells in a pooled
population.
In the VATS (no wedge) samples we screened, we detected a high fraction of alternate
alleles (30%) in the blood specimens (Table 3-3), corroborating the earlier copy number
variation analysis (Figure 3-2A). The mutations detected in the blood matched that of the tumor,
with a nonsense mutation of TP53 (CW46) and activating mutation as KRAS codon 12 (CW48
and CW54). In the VATS (wedge) samples (Table 3-3), no somatic mutation was detected in the
blood specimens even though they were present in the primary tumor, implying that wedge
removed most if not all of the tumor tissue.
The thoracotomy specimens gave mixed results (Table 3-3) but were consistent with the
copy number variation analysis. CW56 had no detectable mutations in the blood specimens
whereas CW51 and CW59 has approximately 10% of the reads matching the TP53 mutation and
STK11 mutation found in the primary tumor respectively. The blood specimen of CW62 had a
high proportion of the reads (81%) harboring an activating EGFR mutation at codon 719,
matching that of the primary tumor. Interestingly, the fraction of reads mapped to the EGFR
mutation is rather low in in the primary tumor (17%), and could be due to impurity from normal
epithelial cells in the tumor microenvironment.
61
Chapter 3: Intraoperatively Dislodged Tumor Cells
Patient
ID
Procedure
CW46
VATS (no wedge)
CW48
Mutations
Tumor
Normal
Specimen Blood
Alt
Ref
Alt
Ref
Alt
Ref
TP53, W91*
250
27
4
810
104
244
VATS (no wedge)
KRAS, G12D
298
804
0
860
495
1382
CW54
VATS (no wedge)
KRAS, G12C
183
285
60
1426
2
115
CW47
VATS (wedge)
KRAS, G12C
296
7
0
524
0
34
CW61
VATS (wedge)
KRAS, G13C
303
9
3
483
0
67
CW51
Thoracotomy
TP53, R248Q
2337
129
8
5078
721
4370
CW56
Thoracotomy
TP53, K120E
42
168
1
431
2
1435
CW59
Thoracotomy
STK11,Y60*
2366
5
0
52
1
9
CW62
Thoracotomy
EGFR, G719A
195
936
0
1648
182
44
Table 3-3: Targeted sequencing. The whole-genome-amplification products of tumor, normal tissue or shed
epithelial cells found in the blood as described in Figure 3-2 were also used in targeted sequencing of 20 commonly
mutated genes in NSCLC. The table indicates the number of reads mapped to the reference alleles (indicating the
presence of normal cells) or alternate alleles (indicating the presence of tumor cells). Shown are three patients who
have undergone VATS (no wedge), two patients who have undergone VATS (wedge) and four patients who have
undergone thoracotomy.
3.4.4
Mutation analysis of single cells identified tumor cells released into the pulmonary
vein during thoracotomy.
While both copy number variation analysis and targeted sequencing gave convincing and
consistent evidence to the presence of shed tumor cells, these two methods of analysis were
conducted in a pooled sample of 10-15 cells (mainly due to cost limitations). We developed a
third method of detecting somatic mutations that is applicable to single cells. Multiplex nested
PCR followed by Sanger sequencing is cost effective and sensitive down to single cells. We
designed multiplex primers against TP53 exons 4-8, KRAS exon 2 and EGFR exons 18-21 as
these were the most common mutations in NSCLC. For each patient, we first confirmed the
mutations present in the bulk tumors, then screened the same mutation in single cells of the
blood specimens.
The results of the nested PCR were highly consistent with targeted sequencing, even
though the two methods were conducted independently. The former amplifies by PCR whereas
the latter utilizes multiple displacement amplification of the whole genome using
62
29
Chapter 3: Intraoperatively Dislodged Tumor Cells
polymerase under isothermal condition. Both methods detected the same mutations in the bulk
tumor cells, with comparable frequencies of alternate alleles (Figure 3-3, Table 3-2). Sanger
sequencing is less sensitive at detecting low abundance mutations. The EGFR mutation of CW62
would not have been easily detected by Sanger sequencing without the results from targeted
sequencing as this mutation has a low penetrance in the bulk tumor (Figure 3-3). Nevertheless,
the sensitivity of Sanger sequencing is sufficient for single cells.
We detected single and clusters of tumor cells in VATS (no wedge) procedures (Figure
3-3). Morphologically, the tumor cells were larger and more irregularly shaped. No tumor cells
were detected in the VATS (wedge) procedures, consistent with the copy number variation
analysis and targeted sequencing results. In the thoracotomy samples, we did not detect single
tumor cells in CW56, CW59 and CW50. However, we detected single tumor cells harboring the
activating EGFR G719A mutation in CW62 as observed in the matched tumor. Morphologically,
the tumor cells were indistinguishable from normal cells, further supporting the inadequacy of
cytomorphological means to quantify the number of tumor cells.
63
Chapter 3: Intraoperatively Dislodged Tumor Cells
A
CW48
kras G>A, G12D
CW46
p53 G>A, W91*
TL
TL
Calcein EpCAM
Calcein EpCAM
Lin
Lin
F28S
B
CW61
kras G>T, G13C
CW47
kras G>T, G12C
TL
Calcein EpCAM
Lin
TL
C
64
Calcein EpCAM
Lin
Chapter 3: Intraoperatively Dislodged Tumor Cells
C
CW51
p53 G>A, R248Q
CW50
p53 G>T, G244S
TL
Calcein EpCAM
Lin
CW56
p53 A>G, K120E
TL
Calcein EpCAM
TL
Calcein EpCAM
Lin
CW62
EGFR G>C, G719A
Lin
TL
65
Calcein EpCAM
Lin
Chapter 3: Intraoperatively Dislodged Tumor Cells
Figure 3-3: Nested PCR of single cells or single preformed clusters. Shown are 8 patients whose bulk tumor,
normal tissue and individual epithelial cells were subjected to two rounds of PCR amplification reactions against
specific somatic driver mutations in TP53, KRAS and EGFR. Shown are patients who have undergone VATS (no
wedge) in A), patients who have undergone VATS (wedge) in B) and patients who have undergone thoracotomy in
C). The sequences of the normal and tumor tissue are indicated next to the patient ID with the tumor tissue boxed in
red. The bright field and fluorescence images of single shed tumor cells are shown next to the sequences of codons
mutated in the bulk tumor. Mutations matching those of bulk tumor are boxed in red. For patient CW48, one cell
(the fifth cell down the column) did not habor the G12D mutation as in the primary tumor but had a F28S mutation
in KRAS.
3.5
3.5.1
Discussion
Surgical manipulation releases both normal and malignant epithelial cells.
We have shown that pulmonary lobectomy mobilizes viable tumor cells into the
pulmonary vein. EpCAM+ cells found in the blood harbored the same driver mutations as the
primary tumor, as confirmed by single-cell targeted sequencing. Along with the tumor cells, we
also detected normal epithelial cells mobilized during the surgery; these cells stained positive for
EpCAM but did not contain the same mutation as the primary tumor and have a uniform copy
number profile. This implies that enumerating the number of epithelial cells by
cytomorphological means alone can lead to an overestimation of the amount of tumor cells shed
intraoperatively.
The presence of mobilized tumor cells raises the question what is the best surgical
practice for minimizing local recurrence and tumor dissemination. For instance, should the tumor
be wedged out earlier or does the order of pulmonary vein ligation affect patient outcome? A
recent study compared the sequence of pulmonary vessel ligation in thoracotomy patients and its
impact on tumor cell shedding using CD44v6 and CK19 as the tumor marker, but the results
showed no appreciable differences in patients whether their pulmonary vein was ligated before
or after the ligation of the pulmonary artery [66]. It is possible that our genetic approaches may
give a different count of tumor cells and reveal a difference in the degree of tumor cell shedding
depending on the sequence of pulmonary vessel ligation.
In our limited sample size, only one out of the five thoracotomy samples (patient
CW62) had a significant number of tumor cells (~1000 tumor cells) mobilized into the
pulmonary vein. Other thoracotomy patients either have no detectable tumor cells (CW50, 56) or
a very low number of tumor cells based on the results of targeted sequencing (CW51, 59). In
addition to measuring the number of tumor cells, it remains to be determined whether the
mobilized tumor cells are capable of forming metastases. A more extensive, longitudinal study is
necessary to establish whether intraoperative tumor shedding affects patient outcome.
66
Chapter 3: Intraoperatively Dislodged Tumor Cells
In this study, we confirmed the presence of intraoperative tumor shedding by identifying
consistent mutations between EpCAM+ cells found in the pulmonary vein blood and patient
matched tumor. While it still remains to be determined whether the shed tumor cells are
tumorigenic, surgeons should take care to minimize tumor manipulation cell mobilization during
surgery.
67
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.1
Abstract
Many targets have been identified in solid tumors for antibody therapy but it is less clear
what surface antigens may be most commonly expressed on disseminated tumor cells. Using
malignant pleural effusions as a source of disseminated tumor cells, we compared a panel of 35
antigens for their cancer specificity, antigen abundance and functional significance. These
antigens have been previously implicated in cancer metastasis and fall into four categories: 1)
cancer stem cell 2) epithelial-mesenchymal transition 3) metastatic signature of in vivo selection
and 4) tyrosine kinase receptors. We determined the antigen density of all 35 antigens on the cell
surface, which ranges from 3x103 – 7x 106 copies per cell. Comparison between the malignant
and benign pleural effusions enables us to determine the antigens specific for cancer. We further
picked 6 antigens and examined the correlation between their expression levels with tumor
formation in immunocompromised mice. We concluded that CD24 is one of the few antigens
that could simultaneously meet all three criteria of an ideal target, as it is specifically and
abundantly expressed in malignant pleural effusions. CD24high tumor cells formed tumors in
mice at a faster rate than CD24low tumor cells, and shRNA-mediated knockdown of CD24 in
HT29 cells confirms a functional requirement for CD24 in the colonization of the lung.
Concomitant consideration of antigen abundance, specificity and functional importance can help
identify potentially useful markers for disseminated tumor cells.
4.2
4.2.1
Introduction
Surface antigens for antibody targeting
The discovery and validation of antigen targets is the first step in the development of an
efficacious antibody therapy [67]. Several successful targets have been identified in solid tumors,
but it is unclear whether the same targets are equally applicable to disseminated tumor cells,
which are cells detached from the primary site and often possess greater tumorigenic potential
than the primary tumor [68]. We wish to evaluate a panel of antigens for their suitability to target
disseminated tumor cells.
In terms of antigen characterization, an ideal antibody target should fulfill these three
criteria: 1) cancer specificity, 2) abundant expression and 3) functional importance [67]. One of
68
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
the most successful antibody targets, ErbB2, is an antigen that has met all three criteria. ErbB2 is
overexpressed in 82% of breast cancer samples [69]. However, antibody therapy against ErbB2
with Trastuzumab is only effective in patients expressing the highest levels of ErbB2 at around
5x105 receptors per cell but ineffective in patients expressing 2x105 receptors per cell, indicating
that abundant expression of the antigen is an essential determinant of therapeutic outcome [70].
Functionally, ErbB2 plays an important role in tumor growth as it can promote tumor growth by
recruiting survival and mitogenic pathways [71].
While many targets have been evaluated according to the above-mentioned guidelines in
solid tumors, few antigens have been assessed in disseminated tumor cells. Based on recent
evidence, we identified candidate antigens that are particularly involved in the molecular
pathways and cellular mechanisms of metastasis. They fall under four general categories. First,
cancer stem cell (CSC) markers have been shown to delineate subpopulations with markedly
increased tumor-initiating potential [72, 73]. Second, the loss of epithelial markers and the gain
of mesenchymal markers during epithelial-mesenchymal transition (EMT) are features of a more
invasive cell type [74]. Third, in vivo selection experiments have identified gene signatures
responsible for the organ tropism of metastasis [75] or metastasis in general [76]. The fourth
class is the family of receptor tyrosine kinases important for tumor growth, and monoclonal
antibodies targeting these are used to treat solid tumors [77].
4.2.2
Pleural effusions as a source of disseminated tumor cells
Acquisition of biological materials to evaluate the targets presents a challenge for target
discovery in disseminated tumor cells [67]. Circulating tumor cells are too rare (~1 cell/ml
blood) for direct screening of surface antigens [4]. On the other hand, malignant pleural effusions
are a large reservoir of highly malignant tumor cells (104- 106 cells/L), with suspended tumor
cells of increased tumorigenic potential than that of primary tumors [68]. In addition to cancer,
congestive heart and liver diseases can also cause the buildup of fluids in the pleural cavity,
forming benign effusions that can be used as negative controls in the measurement of cancer
specificity. Soluble factors have been evaluated as tumor markers for malignant pleural effusion
[78, 79] but to our knowledge there has not been a previous screening effort of cell-surface
antigens. Cell-surface antigens are desirable for antibody targeting because they are directly
accessible and localized to the cells of interest.
69
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.3
4.3.1
Materials and methods
Collection of patient samples
Pleural effusions were obtained from consenting patients as part of an Institutional
Review Board-approved protocol at the Lahey Hospital and Medical Center and Use of Humans
as Experimental Subjects approved study at MIT. Samples were collected from patients with
biopsy-proven cancer and patients with no cancer as determined by clinical presentation and
histology. 400 ml of pleural fluid was set aside for routine cytologic evaluation and the rest was
sent to MIT on ice within 24 hours after collection, and analyzed for surface expression within
48 hours after collection. 12 malignant pleural effusions and 8 benign effusions were used for
surface antigen analysis. Refer to Table 4-1 for detailed patient information.
70
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
Benign (n=8)
Malignant (n=12)
ID
Diagnosis
Survival
(months)
Site of Metastasis
LC28
lung cancer
7.2
brain
LC51
lung cancer
>12
brain
LC52
rectal cancer
2.7
pleura
LC74
endometrial cancer
5.3
ovaries, fallopian tubes, omentum
LC86
lung cancer
5.3
lymph nodes
LC90
lung cancer
2.4
rib, soft tissue
LC95
lung cancer
9.4
lymph nodes, pleura
LC101
lung cancer
>12
lymph nodes, bone
LC103
lung cancer
>12
lymph nodes
LC109
lung cancer
3.6
none reported
LC129
colon cancer
2.1
lungs, liver
LC131
lung cancer
1.4
lymph nodes, pleura
LC47
congestive heart failure
>12
-
LC48
congestive heart failure
>12
-
LC49
congestive heart failure
>12
-
LC55
renal disease
>12
-
LC77
liver cirrhosis
5.2
-
LC102
primary biliary cirrhosis
>12
-
LC122
primary biliary cirrhosis
>12
-
LC128
liver cirrhosis
>12
-
Table 4-1: Patient information
4.3.2
Surface marker expression analysis using flow cytometry
30 ml of pleural effusions were passed through a 40 µm cell strainer (BD Biosciences),
incubated with human FcR block (eBiosciences) in 1:20 dilution and then stained with antibodies
(Table 4-2) for 30 min on ice. DAPI is added immediately before flow cytometry analysis to gate
out the dead cells. Cells were analyzed on LSRII (BD Biosciences) with 4 different lasers
(355nm, 488 nm, 561nm and 635nm) in 5 different channels: DAPI, FITC (marker 1), PE
(marker 2), EpCAM-APC and CD45-PECY7 using HTS plate reader. Antibodies were
purchased directly conjugated, or labeled with Alexa488 dye (Invitrogen) or PE (Dojindo). The
fluorophores were chosen to be those with minimal bleed through. Compensations were
71
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
performed with beads (Bangs Laboratory) coated with individual clone of antibodies, and the
compensation matrix was applied to all the patient samples. Refer to Table 4-2 for the list of
antibodies.
Antigen
Clone
Fluorophore
Reactivity
IgG type
VendorCat #
488 isotype
MOPC-
488
hu
mouse IgG2A
Biolegend
400233
FITC
hu
mouse IgG2b
Biolegend
332014
PE
hu
goat polyclonal
R&D
FAB1790P
173
ABCG2
5D3
Cad-11
CD24
ML5
PE
hu
mouse IgG2A
Biolegend
311106
CD44
IM7
488
hu
rat IgG2b
Biolegend
103016
CD44v4
VFF-4
FITC
hu
mouse IgG1
Genetex
GTX75289
CD44v6
VFF-7
FITC
hu
mouse IgG1
Genetex
GTX75292
CD44v7
VFF-9
FITC
hu
mouse IgG1
Genetex
GTX75294
CD45
HI30
PE/Cy7
hu
mouse IgG1
Biolegend
304016
CD51/61
23C6
488
hu
mouse IgG1
Biolegend
304408
CEA
sm3e
488
hu
human
Developed in
sm3e
Wittrup lab
CLDN7
4D4
PE
hu
mouse IgG2a
Sigma
WH0001366M
1
CXCR4
12G5
PE
hu
mouse IgG2A
Biolegend
306506
CXCR7
8F11-M16
PE
hu
mouse IgG2b
Biolegend
331103
E-Cad
67A4
488
hu
mouse IgG1
Biolegend
324110
EGFR
EGFR.1
PE
hu
mouse IgG2b
BD
555997
EpCAM
9C4
488
hu
mouse IgG2b
Biolegend
324210
EpCAM
9C4
APC
hu
mouse IgG2b
Biolegend
324208
EPHB6
465327
488
hu
mouse IgG1
R&D
MAB33841
erbB2/HER2
24D2
PE
hu
mouse IgG1
Biolegend
324406
erbB3/HER3
1B4C3
PE
hu
mouse IgG2A
Biolegend
324706
FGFR1
M19B2
PE
hu
mouse IgG1
Genetex
GTX20823
(discont.)
ab823
AbCAM
FGFR2c
133730
PE
mouse
rat IgG2A
R&D
MAB7161
FGFR2b
98707)
488
hu
mouse IgG1
R&D
MAB665
FOLR1
548908
PE
hu
mouse IgG1
R&D
MAB5646
72
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
GLP-1R
197920
488
hu
mouse IgG2b
R&D
MAB2814
HLA
G46-2.6
APC
hu
mouse IgG1
BD
555555
ICAM1
BB1G-l1
FITC
hu
mouse IgG1
R&D
BBA20
IGFIRa
1H7
PE
hu
mouse IgG1k
BD
555999
N-cadherin
401408
PE
hu
rat IgG2A
R&D
FAB1388P
PDGFRa
PRa292
PE
hu
mouse IgG2
R&D
FAB1264P
PDGFRb
PR7212
PE
hu
mouse IgG3
R&D
FAB1283P
PE isotype
MOPC-21
PE
hu
mouse IgG1
BD
559320
SPARC
122511
488
hu
mouse IgG1
R&D
MAB941
TNFRSF11b
98A1071
488
hu
mouse IgG
AbCAM
ab79064
VCAMI
51-10C9
FITC
hu
mouse IgG1
BD
551146
VEGFc
193208
PE
hu
mouse IgG2b
R&D
MAB752
VEGFR1
49560
PE
hu
mouse IgG
R&D
FAB321P
Table 4-2: List of antibodies used
4.3.3
Quantification of surface receptors expression
The number of receptors per cell is determined by calibrating the fluorescence intensity
using microspheres with known Antibody Binding Capacities (Bangs Laboratory). The Quantum
Simply Cellular micropheres come with 1 blank and 4 standards with increasing levels of Fcspecific capture antibodies. The beads were labeled to saturation with the same antibodies used
to stain the cells and analyzed on the flow cytometer with the same settings used to analyze the
cells. The standards were then used to construct a calibration curve. The number of antigen was
determined by converting the mean fluorescence level to antibody binding capacities. Since we
only used monoclonal antibodies to stain the cells (except cadherin-11), each antibody only binds
to one antigen and the antibody binding capacities of the microspheres is equivalent to the
number of antigens. The mean fluorescence level is tabulated by Flowjo (Tree Star) and
compiled using a Matlab (Mathworks) program. Monoclonal antibody for Cadherin-11 was not
commercially available, so we used a goat polyclonal antibody. The calibration kit against goat
Fc was also not available, so we estimated the number of Cadherin-11 based on the fluorescence
intensity of N-cadherin antibody.
73
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.3.4
Sorting and animal studies
All experimental protocols were approved by the Animal Care and Use Committee of
MIT. Tumor from patient LC52’s pleural effusion was formed after injecting the cells pelleted
from 30 ml of pleural effusion subcutaneously into NOD/SCID/IL2γR-/- with 200 µl Matrigel
(BD Biosciences). LC52 was propagated in vivo by serial transplantation. The tumors were cut
up with razor blades, and digested in 1 mg/ml collagenase A (Roche) and 100 units/ml DNAse I
(Roche) in RMPI at 37°C for 20 min. Single cell suspensions were stained with antibodies for 30
min on ice and sorted on Aria (BD) by fluorescence-activated cell sorting (FACS). Sorted cells
were injected with Matrigel and observed for tumor formation for up to 1 year.
4.3.5
Cell lines and shRNA knockdown
HT29 cells (ATCC) were maintained in DMEM (Cellgro) and 10% FBS (PAA
Laboratories). shRNA lentiviral particles were purchased from Sigma-Aldrich (clone1 TRCN0000057675,
5'-CCGG-TCTTCTGCATCTCTACTCTTA-CTCGAG-
TAAGAGTAGAGATGCAGAAGA-TTTTTG-3' and clone 2- TRCN000007677, 5'-CCGGCGCAGATTTATTCCAGTGAAA-CTCGAG-TTTCACTGGAATAAATCTGCG-TTTTTG-3').
The non-targeted control uses an shRNA sequence targeting no known mammalian genes
(Sigma-Aldrich SHC002V). Cells were plated at 1 x 104/well in a 96-well plate, transduced with
lentiviral particles, and selected with puromycin for 10 days. After antibiotic selection, cells were
further FACS-sorted for CD24 low-expressing cells.
4.3.6
Seeding study
HT29 cells transduced with non-targeted shRNA were labeled with CellTracker Green
CMFDA (Invitrogen) and HT29 cells transduced with CD24 shRNA were labeled with
CellTracker Red CMPTX (Invitrogen). The cells were mixed at 1:1 ratio. Each mouse was
injected with a total of 1x106 cells through the tail vein. Their lungs were harvested 48 hours
later and examined under a fluorescence microscope (Zeiss). The number of cells in each field
was counted with ImageJ.
74
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.3.7
Lung colonization study
1x105 HT29 cells, either transduced with non-targeted shRNA or CD24 shRNA, were
injected into mice through the tail vein. The lung tissue was fixed with 4% paraformaldehyde for
15 min and rinsed with PBS. The lungs were embedded in paraffin, sectioned 100 µm apart and
stained by hematoxylin and eosin (H&E). The number of metastases observed was counted and
averaged over 5 sections per mouse.
4.3.8
Platelet binding assay
Platelets were isolated from 10 ml human whole blood (Blood Research Components) by
centrifuging once at 200 x g for 10 min, and centrifuging the supernatant a second time at 1600 x
g for 10mins. 1x104 HT29 cells were incubated with platelets at 37°C for 1 hour, washed twice
with PBS and stained with anti-CD41a antibody (BD Pharmingen).
4.3.9
Anoikis assay
Plates were coated with 500 µl 12mg/ml poly-HEMA dissolved in 95% ethanol (Sigma-
Aldrich) and dried overnight. HT29 cells were cultured in serum-free medium for 48 hours. Cells
were collected and stained with Annexin V-APC (Biolegend) and analyzed by flow cytometry.
4.3.10 Statistical analysis
ANOVA tests in Matlab were used to determine p-values of the heatmap. Kaplan-Meier
plots were created and p-value calculated by a log-rank test in GraphPad Prism. Sample size
determination for statistical powering of Type II error is computed by this formula:
where n = sample size,
d = effect or the difference between the mean of two population
s = standard deviation
C = 7.85 when significance level (α) = 0.05 and power (1-β) = 0.08
75
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.4
4.4.1
Results
Only the EpCAM+ population of cells in pleural effusions was tumorigenic.
The pleural effusion is a complex mixture of immune cells, mesothelial cells, fibroblasts,
and cancer cells. Since the proportion of stromal cells can be >90% of the total population,
taking the global average across different cell types would skew the result towards the more
abundant cell types and preclude antigen discovery in rarer tumor-forming cells. Therefore, we
used lineage markers – CD45 for hematopoietic cells and EpCAM for epithelial cells – to
provisionally separate the pleural effusions into different subsets. We observed three distinct
populations in malignant pleural effusions: 1) EpCAM+, 2) EpCAM-/CD45- and 3) CD45+
(Figure 4-1A). The subsets were morphologically distinct, with the EpCAM+ cells being round
and the EpCAM-/CD45- cells spindle-shaped (Figure 4-1A). The EpCAM+ subset was present in
50% of the malignant pleural effusions and was completely absent in the benign effusions
(Figure 4-1A).
Before we could evaluate the antigens for cancer specificity, we need to establish which
subset is tumorigenic. Therefore, we FACS sorted 7 malignant effusions into EpCAM+, EpCAM/CD45- and CD45+ subsets and injected the sorted cells subcutaneously into NOD/SCID/IL2γR-/mice, monitoring for tumor growth for up to a year (Table 4-3). All cells obtained from sorting
were injected without normalization to a common cell number, to maximize each subset’s
probability of tumor formation. The sorting results showed that only the EpCAM+ subset formed
tumors in mice, with a 75% efficiency (Table 4-4). Four of the seven effusions sorted contained
EpCAM+ cells and tumor formation was restricted to these EpCAM+ samples. No tumor
formation was observed with the EpCAM-/CD45- subset (Table 4-4).
When pleural effusions were injected into mice without prior sorting, the formed tumors
were entirely EpCAM+ (Figure 4-1B). For example, the percentage of EpCAM+ cells in the
LC52 sample (Figure 4-1B) increased from 0.4% to 99% during tumor growth.
76
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
Figure 4-1: The EpCAM+ subset is tumorigenic. (A) Cells from pleural effusion samples were separated into 5
different subsets based on their EpCAM and CD45 expression, and whether obtained from a cancer patient
(malignant) or otherwise (benign). The EpCAM+ subset was only present in the malignant effusions whereas the
EpCAM-/CD45- and CD45+ subsets were found in both the malignant and benign effusions. Cells were imaged
directly after sorting with bright field microscopy. (B) The pleural effusions and their corresponding xenograft
tumors were analyzed for their EpCAM and CD45 expression. The xenograft tumors were pre-gated with HLAABC antibody to remove contaminating mouse cells. Shown are the histology images of the xenograft tumors for
each pleural effusion.
77
Samples
Samples
EpCAM-
EpCAM+
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
ID
Diagnosis
EpCAM+
EpCAM-/CD45-
CD45+
LC28
lung cancer
2 x 105
3 x 105
8 x 105
LC42
breast cancer
4 x 105
1.4 x 105
8 x 104
LC60
lung cancer
1 x 104
1.5 x 105
2 x 105
LC74
endometrial cancer
2 x 103
2 x 103
Not injected
5
1 x 105
LC40
lung cancer
Absent
3 x 10
LC65
lung cancer
Absent
1 x 106
2 x 106
LC103
lung cancer
Absent
2 x 105
Not injected
Table 4-3: Malignant pleural effusions were sorted into the EpCAM+, EpCAM-/CD45- and CD45+ subsets by flow
cytometry and injected subcutaneously into NOD/SCID/IL2ϒR-/- mice and monitored for tumor formation for up to a
year. The table shows the number of cells injected for each subset of the patient.
Cancer
Non-cancer
1. EpCAM+
2. EpCAM-/CD45-
3.CD45+
unsorted
3/4
0/7
0/5
0/5
Table 4-4: This table shows the ratio of the number of tumors formed to the total number of clinical samples
injected. Only the EpCAM+ subset formed tumors.
4.4.2
CD24 was one of the several surface markers abundantly expressed in malignant
pleural effusions.
After establishing that the EpCAM+ subset is tumorigenic, we proceeded to screen
surface antigens for their abundance and cancer specificity. We analyzed 12 malignant and 8
benign pleural effusions using flow cytometry and stained the effusions with 35 directly
conjugated antibodies. Because the antibodies were labeled with different fluorophores and with
different labeling efficiency, we calibrated the raw fluorescence signal to the absolute number of
surface receptor molecules using beads with predetermined antibody binding capability (Figure
4-2A). We used monoclonal antibodies to calibrate the absolute number of receptors because
each monoclonal antibody should theoretically bind to a single receptor, allowing us to equate
the number of receptors to the number of monoclonal antibodies, with the assumption that all the
receptors are fully saturated with antibodies. Polyclonal antibodies are not suitable for this
application because multiple polyclonal antibodies may bind to a single receptor, thus
overestimating the number of receptors. We included carcinomas of various tissue types
78
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
including lung, endometrial or colorectal so as to identify ubiquitously expressed markers. The
benign effusions came from patients with congestive heart failure or liver diseases.
A
B
Figure 4-2: (A) This is an example calibration curve showing surface receptor quantification with calibration beads
of predetermined antibody binding capacity (ABC) values. Refer to methods section for details. (B) The pleural
effusions were analyzed by flow cytometry using the following gating: the bigger cells were gated out by the
forward and side scatters. The live cells were then gated on the DAPI- population. The live cells were stained for
CD45 and EpCAM. Within each of the subsets (EpCAM+, CD45-/EpCAM-, CD45+), we reported the mean
expression of each given marker (E-cadherin and N-cadherin are used as examples). The fluorophores were chosen
with minimal bleed through. Compensations were performed with beads bound to a single antibody respectively.
Expression levels of 35 surface markers for each of the five cell subsets (malignant
EpCAM+, EpCAM-/CD45- and CD45+; benign EpCAM-/CD45- and CD45+) (Refer Figure 4-2B
for gating strategy) is shown in the heatmap (Figure 4-3) (Refer to Table 4-5 for specific values).
The surface expression of receptors was obtained by averaging all the patient samples for each of
the subsets (Refer to Figure S2C for individual patient surface expression). An analysis of
variance (ANOVA) was applied to each marker and a small p-value indicates that the
distribution of any one subset (most commonly the cancer patient EpCAM+ subset) deviates
significantly from the other subsets. A preferred cancer marker was defined as one that is both
consistently overexpressed across multiple cancer patients and is found abundantly on the cell
surface (Figure 4-3).
79
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
!
A
B
Figure 4-3: Surface marker expression in malignant and benign pleural effusions. (A) Each of the 5 FACSdefined subsets was stained with a panel of 35 surface markers by flow cytometry. We determined the absolute
number of receptors on the cell surface with fluorescence calibration beads (exact values given in table S3). Shown
5
6
are the average numbers of receptors averaged across all patients from each subset. Range shown is 1x10 to 1x10
receptors/cell. Markers with * are differentially expressed in at least one subset with p-value< 0.05 (** p-value<
0.001) based on the ANOVA test. (B) For the same surface marker expression in (a), the p-values are plotted against
the absolute number of surface markers for the EpCAM+ subset to illustrate the point that an ideal marker is both
consistently overexpressed across different patients (p-value< 0.05) and highly expressed (>105 receptors/cell). The
dotted line separates the markers with p-value <0.05 from those with p-value> 0.05. Some example markers are
80
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
circled in red, representing the statistically significant markers, and in green representing the non-statistically
significant markers.
Marker
Cancer
EpCAM
Cancer
+
EpCAM
Cancer
-
CD45
+
/CD45-
Non-cancer
EpCAM
-
Non-cancer
CD45+
/CD45-
EpCAM
6.3E+05
2.8E+04
1.9E+04
1.4E+04
1.6E+04
CEA
8.9E+05
6.9E+04
1.3E+05
4.0E+04
1.0E+05
E-cad
9.7E+05
3.2E+04
3.9E+04
3.8E+04
3.4E+04
CLDN7
7.8E+04
6.6E+04
7.0E+04
6.4E+03
1.9E+04
FGFR2b
1.7E+05
2.6E+04
3.7E+04
2.0E+04
2.6E+04
N-cad
2.9E+03
1.0E+03
5.5E+02
1.6E+02
2.3E+02
Cad-11
4.5E+04
6.4E+02
2.9E+03
1.0E+03
1.6E+03
FGFR2c
8.8E+05
4.7E+04
3.2E+04
2.9E+03
6.0E+03
EGFR
1.4E+05
6.2E+04
8.3E+04
7.4E+04
4.3E+04
ERBB2
2.3E+05
5.7E+04
8.9E+04
6.6E+04
4.5E+04
ERBB3
3.1E+03
3.2E+02
1.2E+03
4.0E+02
4.0E+02
FGFR1
2.2E+05
4.9E+04
9.9E+04
3.0E+03
1.8E+04
FOLR1
4.8E+04
1.1E+03
5.2E+03
4.1E+02
1.7E+03
IGF-1 Ra
4.3E+04
4.0E+03
6.7E+03
3.2E+03
3.0E+03
GLP1R
4.8E+04
4.8E+04
3.9E+04
1.2E+04
1.4E+04
PDGFRa
1.2E+04
2.8E+03
1.1E+04
5.7E+03
8.8E+03
PDGFRb
2.2E+04
5.6E+03
2.6E+04
6.2E+03
3.5E+04
TNFRSF11b
3.4E+05
3.3E+04
2.7E+04
9.7E+03
8.9E+03
EPHB6
2.8E+04
2.8E+04
2.3E+04
2.7E+04
3.3E+04
VEGFR
3.7E+04
8.7E+03
2.5E+04
3.2E+03
1.5E+04
VEGFc
2.9E+05
3.5E+04
7.6E+04
3.2E+03
2.4E+04
ABCG2
8.5E+04
4.4E+04
3.6E+04
1.7E+04
2.0E+04
CXCR4
1.8E+05
1.5E+05
1.8E+05
5.0E+04
1.6E+05
CXCR7
1.7E+05
4.7E+04
8.9E+04
4.8E+04
5.0E+04
integrin-αvβ3
7.2E+04
4.9E+04
5.2E+04
4.8E+04
4.8E+04
SPARC
1.5E+05
4.4E+04
5.7E+04
3.6E+04
4.6E+04
VCAM
5.9E+04
3.6E+04
2.9E+04
8.2E+04
2.5E+04
ICAM
3.7E+05
6.1E+04
7.3E+04
1.0E+05
5.9E+04
CD44
8.6E+04
8.5E+04
2.3E+05
1.2E+05
2.2E+05
CD44v4
1.3E+06
2.1E+05
1.3E+05
2.0E+04
5.4E+04
CD44v6
4.9E+05
3.7E+04
3.0E+04
9.5E+03
1.1E+04
81
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
CD44v7
7.2E+06
9.4E+05
4.3E+05
4.3E+04
9.5E+04
CD24
6.8E+05
7.0E+03
4.0E+04
1.1E+03
1.1E+04
Table 4-5: Absolute number of surface markers expressed on cells found in pleural effusions (averaged across all
patients)
Epithelial markers such as EpCAM, CEA and E-cadherin (E-cad) were the most highly
expressed surface proteins, at close to 1 x 106 receptors/cell in the EpCAM+ subset. In contrast,
we did not observe high expression of mesenchymal markers such as N-cadherin (N-cad) and
cadherin-11 (Cad-11) in any of the subsets. These mesenchymal markers were expressed at 1 x
103 -1 x 104 receptors/cell, approaching background staining. The mesenchymal marker FGFR2c
was sometimes overexpressed, but its expression was inconsistent across different patient
samples. The expression of signaling receptors predominantly fell between 1x 104 to 2 x 105
receptors/cell. ErbB2, ErbB3, IGF-I Ra were significantly overexpressed in the EpCAM+ subset
although they were much less abundant than the other epithelial markers. Integrin αvβ3, SPARC
and ICAM1 were the adhesion molecules overexpressed in the EpCAM+ subset. Interestingly,
CD44 variants and CD24 but not CD44 were significantly overexpressed in the EpCAM+ subset.
The CD44 variants have an extra stem structure between the hyaluronan-binding amino terminal
domain and transmembrane region.
We also compared the marker expression of the EpCAM-/CD45- and CD45+ subsets
between the malignant and benign effusions by t-test. There was no significant difference
between cancer and non-cancer EpCAM-/CD45- cells. However, the CD45+ cells of cancer
samples showed a somewhat higher expression of signaling receptors, namely erbB2, erbB3 and
IGF-1Ra (Table 4-6).
82
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
Marker
EpCAM
CEA
E-cad
CLDN7
FGFR2b
N-cad
Cad-11
FGFR2c
EGFR
ERBB2
ERBB3
FGFR1
FOLR1
IGF-1 Ra
GLP1R
PDGFRa
PDGFRb
TNFRSF11b
EPHB6
VEGFR
VEGFc
ABCG2
CXCR4
CXCR7
integrin-αvβ3
SPARC
VCAM
ICAM
CD44
CD44v4
CD44v6
CD44v7
CD24
EpCAM-/CD45- p-value
4.3E-01
2.1E-01
7.9E-01
3.2E-01
5.8E-01
1.7E-01
1.0E+00
3.7E-01
9.7E-01
9.5E-01
7.1E-01
4.6E-01
1.1E-01
6.1E-01
2.2E-01
7.0E-01
7.5E-01
1.7E-01
9.9E-01
1.5E-01
3.5E-01
2.4E-01
2.1E-01
7.7E-01
7.9E-01
8.2E-01
3.1E-01
6.5E-01
7.1E-01
2.7E-01
4.9E-01
2.9E-01
8.5E-01
!
CD45+ p-value
7.4E-01
3.3E-01
4.2E-01
1.9E-01
2.1E-01
1.6E-02
2.9E-01
2.4E-01
2.1E-02
2.4E-02
3.3E-02
3.8E-01
8.5E-02
2.7E-02
9.5E-02
3.5E-01
9.1E-01
2.0E-01
8.6E-01
2.7E-01
2.1E-01
1.7E-01
5.0E-02
5.1E-02
2.7E-01
4.2E-01
3.2E-01
4.1E-01
5.8E-01
2.3E-01
2.7E-01
4.6E-01
1.9E-01
-
-
Table 4-6: t-test of surface marker expression between the malignant and benign samples in EpCAM /CD45 and
+
CD45 subsets
4.4.3
CD24 predicted increased tumor growth in xenograft tumors of pleural effusions
We further explored whether antigens overexpressed in the EpCAM+ population were
correlated with tumor growth. To address this question, tumor cells propagated in mice from the
83
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
pleural effusion of a colorectal cancer patient (LC52) were sorted based on surface marker
expression. This xenograft tumor was used because it retained the complex tissue morphology of
its origin (Figure 4-1B), and it could be propagated in the shortest amount of time (around 2
months for each passage) and could grow to a reasonable size yielding sufficient cells for the
subsequent experiment. A total of six antigens were examined for their ability to predict
tumorigenicity, in which four were overexpressed in malignant pleural effusions and they
represented the major categories of surface antigens we screened: CEA (epithelial marker),
ICAM1 (adhesion marker), CD44v7 and CD24 (stem cell markers), and two were mesenchymal
markers: Cadherin-11 and FGFR2c. After gating on the human cells with human leukocyte
antigen class I (HLA-A, B, C) staining, LC52 tumor cells were sorted into the top and bottom
10% expressing each of the six antigens (Figure 4-4A), reinjected subcutaneously into
NOD/SCID/IL2γR-/- mice, and monitored for tumor growth (Figure 4-4B). CD24high, CEAhigh,
CD44v7low, and Cadherin-11low populations form statistically significantly faster-growing
tumors. The faster tumor growth observed with cell populations expressing low levels of certain
cancer markers suggests that overexpression does not always correlate with faster tumor growth.
To investigate whether tumors retained the same level of surface expression as their
starting population, we compared the marker expression in tumors formed from high and low
populations by flow cytometry at the end of the tumor induction study (Figure 4-4C). Of note,
tumors from CD24low and CEAlow cells regained their CD24 and CEA expression, reaching the
same level of expression as CD24high and CEAhigh tumors. This implies that marker expression is
a dynamic process and that selected markers, in this case CD24 and CEA, are upregulated as
tumors form.
84
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
A
B
C
high
Figure 4-4: CD24 cells are tumorigenic. (A) To investigate which surface marker predicts tumorigenicity, we
used tumor formed from xenotransplants of the LC52 effusion, FACS-sorted the highest and lowest 10% of cells
expressing each of the 6 markers. (B) Sorted cells were injected subcutaneously into NOD/SCID/IL2ϒR-/- mice and
monitored for tumor growth (n = 4 per group). Red curves represent the average tumor growth from cells expressing
the highest 10% of the indicated markers whereas the green curves represent the tumor growth from cells expressing
the lowest 10% of the markers. Error bars shown are the standard errors of the mean. Overexpression of CD24 and
CEA correlates with faster tumor growth (*p-value <0.05 for each day). (C) Tumors formed from high (red bars)
and low (green bars) starting populations were analyzed for surface marker expression at the end of the tumor
growth studies shown in (b) by flow cytometry (*p-value < 0.05).
85
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.4.4
CD24 was critical for tumorigenic lung colonization
To investigate a possible causal role for CD24 in tumorigenicity, we knocked down CD24
expression by shRNA in the HT29 colorectal adenocarcinoma line (Figure 4-5A, B). A
colorectal line was chosen because LC52 was of rectal origin but LC52 tumor cells were
refractory to lentiviral transduction. HT29 was chosen because it has a marker expression profile
similar to LC52 and other EpCAM+ pleural effusions (Table 4-7); high in EpCAM, CEA and
CD24.
We found that HT29 cells transduced with CD24 shRNA grew at a similar rate in vitro as did
the non-targeted control (Figure 4-5C). When HT29 cells were injected subcutaneously, CD24
imparted a small but statistically significant growth advantage to tumors during the linear phase
(p<0.05) (Figure 4-5D). Both the in vitro and in vivo growth curves are consistent with previous
observations (15). Interestingly, whereas CD24high HT29 cells formed numerous tumor nodules
in the lungs of mice when injected intravenously, very few nodules were observed when HT29
cells in which CD24 was knocked down were injected. This result demonstrates that CD24
expression by tumor cells is required for efficient lung colonization (Figure 4-5E).
86
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
B
CD24 mRNA level
Percentage of maximum
A
NT
CD24-1 CD24-2
D
Mean absorbance
tumor size (mm2)
C
CD24 surface expression
Days
Days
Number of lung metastases
per section
E
NT
CD24-1
CD24-2
Figure 4-5: CD24 is critical for tumor colonization. To investigate the functional role of CD24, we knocked down
the expression of CD24 in HT29 cells. Both the mRNA levels (A) and the surface protein expression levels (B) of
CD24 were reduced by 90% in the shRNA clones. NT: non-targeted control; CD24-1: CD24 shRNA clone 1; CD242 shRNA clone 2. (C) HT29 cells were plated and monitored for in vitro proliferation. The knockdown of CD24 had
87
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
little effect on the in vitro proliferation of HT29. (D) HT29 cells were injected subcutaneously and monitored for
tumor formation (n = 4 per group). (*p-value < 0.05). Error bars shown are the standard errors of the mean. (E)
HT29 cells were injected intravenously and the numbers of lung nodules formed were counted one month post
injection (n = 5 per group). Scale bar = 100 µm. p-value < 0.001 for both shRNA clones. On the H&E section, black
arrows indicate some examples of lung nodules formed from HT29 cells with non-targeted shRNA control.
Pleural Effusion (EpCAM+ cells)
HT29
EpCAM
6.3E+05
2.6E+06
CEA
8.9E+05
2.5E+05
E-cad
9.7E+05
1.1E+05
CLDN7
7.8E+04
1.5E+04
FGFR2b
1.7E+05
2.5E+04
N-cad
2.9E+03
8.1E+02
Cad-11
4.5E+04
6.7E+02
FGFR2c
8.8E+05
8.4E+03
EGFR
1.4E+05
1.2E+05
ERBB2
2.3E+05
6.4E+05
ERBB3
3.1E+03
5.2E+03
FGFR1
2.2E+05
9.3E+04
FOLR1
4.8E+04
1.6E+04
IGF-I Ra
4.3E+04
1.6E+04
GLP1R
4.8E+04
1.3E+04
PDGFRa
1.2E+04
5.9E+03
PDGFRb
2.2E+04
6.3E+03
TNFRSF11b
3.4E+05
1.3E+04
EPHB6
2.8E+04
2.3E+04
VEGFR
3.7E+04
1.1E+04
VEGFc
2.9E+05
9.3E+03
ABCG2
8.5E+04
3.1E+04
CXCR4
1.8E+05
1.3E+05
CXCR7
1.7E+05
1.5E+05
integrin-αvβ3
7.2E+04
6.3E+04
SPARC
1.5E+05
8.4E+04
VCAM
5.9E+04
3.0E+04
ICAM
3.7E+05
4.4E+04
CD44
8.6E+04
1.7E+05
CD44v4
1.3E+06
1.7E+04
CD44v6
4.9E+05
6.1E+03
CD44v7
7.2E+06
6.5E+04
CD24
6.8E+05
1.8E+07
Marker
88
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
Table 4-7: Absolute number of surface markers on HT29 cell line
4.4.5
CD24 was pro-survival
To elucidate further the contributing functions of CD24 in tumorigenesis, we investigated
the effect of knocking down CD24 on the in vitro phenotypes of HT29 cells. First of all, there
was no significant loss of initial seeding efficiency of HT29 cells in the lungs after the
knockdown of CD24: an equal ratio of HT29 cells with non-targeted shRNA to CD24 shRNA
was observed in the lungs 48 hours after intravenous injection (Figure 4-6A). The majority of the
observed HT29 cells were physically trapped in the lung capillaries due to their large size.
We next tested whether CD24 could protect HT29 cells from apoptosis. The viability of
HT29 cells lacking CD24 was decreased after being maintained in both serum-free adherent and
suspension cultures (Figure 4-6C). Protection from anoikis is critical for tumor cells in
suspension and this might account for the enrichment of CD24+ cells in pleural effusions.
Since CD24 is a known ligand for P-selectin [80], one hypothesis to explain the
differences seen in the tumor colonization potential is that the binding of tumor cells to Pselectin on platelets protects them from immune-mediated clearance [81, 82]. A reduction of
platelet binding was observed upon CD24 knockdown (Figure 4-6D).
89
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
NT1 / CD24 shRNA clone 1
B
Non-targetde/CD24 shRNA
seeding ratio
A
1.2
1.0
0.8
0.6
0.4
0.2
0.0
CD24-1 CD24-2
D
80
60
40
* **
** *
20
re
su
nt
sp
en
si
on
0
*
**
80
60
40
20
0
no
ad
100
pl
at
el
et
C N
D T
C 24D 1
24
-2
100
No platelet
Non-targeted control
CD24 shRNA clone 1
CD24 shRNA clone 2
he
Percentage of Cells
Non-targeted control
CD24 shRNA clone 1
CD24 shRNA clone 2
CD41a levels normalized to
non-targeted control (%)
C
Figure 4-6: CD24 protects tumor cells from apoptosis and affects cell cycle progression. (A, B) To determine
the seeding efficiency in the lungs, pre-labeled non-targeted controls (green) and CD24 shRNA knockdowns (red)
were mixed at 1:1 ratio and injected intravenously through the tail vein. Lungs were harvested 48 hours later and the
ratio of green to red cells was counted. (C) HT29 cells were grown in adherent culture or in suspension in polyHEMA-coated plates. Cells were stained for Annexin V two days after seeding. Shown is the percentage of cells that
are dead after 2 days. (D) HT29 cells were incubated with platelets and stained for CD41a (a platelet marker). CD24
knocked-down cells bind platelets less effectively than do non-targeted controls.
90
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.4.6
Higher expression of CD24 in malignant pleural effusions correlated with a worse
patient outcome
From our dataset of 8 benign effusions and 15 pleural effusions (12 malignant pleural
effusions in the surface expression analysis and 3 additional lung cancer pleural effusion
samples), we observed that CD24 was overexpressed in cancer patients, and a cutoff of 1000
mean fluorescence intensity can exclude 7/8 benign effusions and divide the cancer patients into
CD24high and CD24low groups (Figure 4-7A). Using the survival data of 15 cancer patients, we
found that CD24 overexpression was correlated with worse patient outcomes in this study (pvalue = 0.008) (Figure 4-7B). We gated out the CD45+ cells and used the mean CD24 expression
of the EpCAM+ and the EpCAM-/CD45- cells combined. This is because in 8 out of the 15
pleural effusions, the EpCAM+ subset was absent. The worse prognosis of CD24 overexpression
is consistent with an earlier meta-analysis of multiple studies that found CD24 overexpression to
be correlated with metastasis and poor prognosis [83].
B
Fraction survival
CD24 expression
(mean fluorescence)
A
Benign (n=8)
Malignant (n=15)
Days
Figure 4-7: Patients with higher CD24 expression have reduced overall survival. (A) Cancer patients have a
higher expression of CD24 in their pleural effusions compared to that of benign patients. Shown is the mean
fluorescence intensity of CD24 of the CD45- cells. Patients with CD24 expression > 1000 mean fluorescence
intensity (MFI) are grouped as “CD24 high” and patients with CD24 expression < 1000 MFI are grouped as “CD24
low”. (B) Median survival of CD24 high group = 111 days, CD24 low group = 582 days. Hazard ratio = 6.471
(95% CI: 1.6 to 25.9). p-value = 0.008 (log-rank Mantel-Cox test). Patients who were still alive but not followed for
the entire duration of the study were considered censored.
4.5
Discussion
91
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.5.1
CD24 is a cancer-specific, abundantly expressed and functionally important for lung
colonization of tumor cells.
Identifying a surface antigen commonly expressed on disseminated tumor cells could help
improve therapeutic outcome in metastatic disease. With that objective, we evaluated 35
previously known cancer antigens based on three criteria: cancer specificity, abundance and
functional requirement for tumor growth. We used malignant pleural effusions as the source of
disseminated tumor cells because the tumor cells found in the effusions are highly tumorigenic
and much more abundant than other sources of disseminated tumor cells such as circulating
tumor cells in blood or cytokeratin positive cells in bone marrow aspirates.
CD24 emerged as a candidate antigen that fulfills all three criteria simultaneously – it is
abundantly and specifically expressed in malignant pleural effusions, and is required for the
colonization of the lung. In contrast, not all antigens can meet this tripartite assessment. The
CD44 splice variants were previously shown to mediate metastasis of rat carcinoma [84], but in
our screening, even though CD44v7 was specifically and abundantly expressed, it did not seem
to be required for tumor formation. Another example is erbB3, despite being cancer specific and
functionally important [85, 86], is expressed at a relatively low level of 3.1 x 103 copies per cell
on average. Targeting this low-abundance antigen can be challenging and may require highaffinity antibodies. Lastly, we examined the mesenchymal markers because tumor cells
undergoing EMT are highly invasive and anchorage-independent [74], so we expected the highly
malignant cells in pleural effusions to exhibit similar expression profile. Contrary to our
hypothesis, we found that N-cad and Cad-11, were neither abundant nor cancer specific. The
tumor formation experiment also did not suggest any requirement for Cad-11. The low
expression of mesenchymal markers in our screening indicates that tumor cells in pleural
effusions are not in the state of EMT but remain predominately epithelial.
We have applied a general guideline to evaluate the suitability of targets for disseminated
tumor cells: the concomitant consideration of antigen abundance, specificity and functional
importance, along with the use of tumor cells disseminated from the primary site. CD24 is an
antigen that has satisfied all three criteria, especially its functional importance in the colonization
of the lung and thus a potential relevance to metastasis. In the future, this screening method can
be extended to a greater panel of antibodies for the discovery of targets in disseminated tumor
cells.
92
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
4.5.2
CD24 promotes tumor growth
We demonstrated that CD24 is functionally important as a cancer antigen. It is interesting
to note that CD24 mediates the lung colonization of HT29 cells without contributing
significantly to tumor growth as seen in subcutaneous tumor growth experiment. The most
prominent effect is CD24’s protection against apoptosis, especially in suspension culture, which
mimics the effect of anoikis. Protection against anoikis plays a crucial role in metastasis [87]. It
is known that some genes important for metastasis do not actually have a direct effect on tumor
growth, such as tumor-derived endothelin-1[88]. The use of pleural effusion as a source of highly
malignant tumor cells can help identify targets unique to metastasis.
CD24 is a heavily glycosylated, GPI-anchored peptide expressed in naïve B cells and
cancer cells [89]. Antibody against CD24 has achieved therapeutic efficacy in preclinical models
[90, 91], especially in the prevention of metastatic formation in the lungs. Earlier studies
ascribed the role of CD24 in tumorigenesis and metastasis as two-fold: 1) proliferation and 2)
cell motility. CD24-siRNA knockdowns of HT29 cells showed a reduced exponential growth
and lower saturation density, with effects seen more prominently in serum-starved conditions
[90]. Network analysis using gene expression arrays revealed that downregulation of CD24 also
affected members in the mitogen-activated protein kinase (MAPK) pathway downstream of Ras.
Other pathways affected also included phospholipase C signaling, vascular endothelial growth
factor, hypoxia and angiogenesis [90]. Another group has shown that CD24 is regulated by Ral
GTPases, which mediate cell transformation through Ras [92]. Secondly, the involvement of
CD24 in cell motility was observed when CD24-siRNA clones of HT29 cells failed to transpass
transwell assays [90]. Indeed, studies have shown CD24 promotes binding to extracellular matrix
including fibronectin, laminin, collagen I and collagen V through α3β1 and α4β1 and is important
for cell spreading[93].
The importance of CD24 may seem contrary to the observation that CD44+/CD24- cells
were of increased tumorigenic potential [94] and potential cancer stem cells [95] in breast
cancer. However, the CD44+/CD24- phenotype is not always seen in other cancer types. For
example, it is the CD44+/CD24+ cells that take on the properties of cancer stem cells in
pancreatic cancer. Secondly, despite the evidence of CD24- cells being the cancer stem cells in
breast cancer, overexpression of CD24 has a poor prognosis in breast carcinomas including
increased staging, tumor grade and lymph node positivity [96]. In human bladder cancer, CD24
93
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
expression was observed in 75% of the primary tumors and 93% of the matched metastases,
indicating an increased CD24 expression with more advanced tumor staging [91].
4.5.3
Future improvements
One improvement that can be made to our study is to include a larger patient size with a
defined cancer type. In general, a sample size of 8 to 12 will only be powered to find effects of a
rather large size. This means that only highly overexpressed markers can be identified by this
sample size. A sample size of 8 is powered to find effects 1.5 times that of the standard deviation
of the population (see Methods). The difference seen for a few markers including CD24 between
the EpCAM+ population of the malignant effusions and the other benign populations exceeds this
magnitude, and thus our sample size is sufficiently powered to detect the large effects. But
increasing the sample size can detect subtler differences. With our small sample size, we
observed considerate patient heterogeneity, indicating the need for patient stratification. We did
not limit our patients to lung cancer because effusions from rectal (LC52) and endometrial
(LC74) cancers were both tumorigenic in mice.
Another limitation of our study is that we could not obtain matched primary tumors
because the resection of the primary tumor took place years before the development of pleural
effusions and not always performed in the same hospital. Therefore, it is unclear whether the
antigens overexpressed in pleural effusions were present in the primary tumor or in the adjacent
healthy tissue. Tumor tissue was formalin-fixed, paraffin-embedded (FFPE); membrane antigens
were sensitive to crosslinking, and not all the antibodies are developed for in FFPE slides. Future
study would involve the analysis of fresh tumor samples, and waiting 3-5 years for pleural
effusions to develop in a subset of samples. Nevertheless, an antigen expressed in normal tissue
may still be a useful cancer target when it was found in an unusual location. EpCAM, a marker
of both normal and tumor epithelia, indicates metastatic disease when present in pleural
effusions, blood or bone marrow. Antibody against EpCAM can completely eliminate EpCAM+
cells in the bone marrow aspirates of some patients [93].
Lastly, in the quantification of surface antigen abundance, we assume that all the
antibodies are equally accessible to the antigens and all antigens are labeled to saturation. This
may not always be the case because our results showed that although CD44 variants are
expressed at high levels, CD44 standard form is not. CD44 variants are only different from the
standard form by an extra stem structure [97], so their levels should be comparable. Therefore, it
94
Chapter 4: Analysis of Surface Markers in Malignant Pleural Effusions
might be more valid to verify the antigen expression by a different antibody targeting a separate
epitope. A more accurate quantification of surface antigen can be accomplished if the Kd of the
antibodies is known.
!!" =
!" !"
[!"]
!≈
!!
!" !" + !!
[!"] + !!
where yeq = fraction of saturation of surface antigens, [Ab] = concentration of antibody, and Kd
= affinity of the antibody. However, an estimate of Kd was not available for each of the 35
antigens in this study.
95
Chapter 5: Discussions and Future Perspectives
5
Chapter 5: Discussions and Future Perspectives
5.1
5.1.1
Discussions
Circulating tumor cells and metastasis
The finding in this thesis seems to suggest a rather dormant nature of circulating tumor
cells, in contrary to what might be the conventional belief – shed tumor cells are aggressive
players, actively invading the stroma. My data, however, seemed to suggest the inadvertent
release of CTCs into the blood stream perhaps as a result of 1) the proximity of tumor cells to the
blood supply and 2) leaky and damaged blood vessels due to inflammation and tumor necrosis.
CTCs are passive bystanders who got swept into the whirlpool of blood, which is a tremendous
source of selection pressure, and only the fittest tumor cells can survive through this selection. It
is conceivable that while inside the circulation, CTCs halt their proliferative program as
suggested by the absence of in vitro proliferation and secretion of soluble growth factors. They
have entered a hibernation state before landing on fertile soil again. However, it is not clear, how
upon landing, circulating tumor cells turn their proliferative program back on, and it is an
important topic for future studies.
Despite the passive nature of CTCs, the number of CTCs is prognostic of patient
outcomes. This is perhaps due to the correlation between the rate of tumor shedding to tumor
burden, the number of metastases, and how loosely the tumor cells are attached to the stroma –
all of which are indicators of poor prognosis. Therefore, the number of CTCs might be a good
predictor of clinical outcome.
However, the sensitivity of CTC as a prognostic method is limited by its rarity and the
lack of tumor-specific markers. We did encounter patients with progressive disease but no CTCs.
This is perhaps due to the fact that we are undersampling the amount of blood. We sampled 5 ml
of blood at one time point, while the total volume of blood is 5L, and the half-life of CTCs is
estimated to be 1-2 days. Since it takes years for metastasis to develop, the probability of
capturing a robust cancer cell capable of forming metastasis is extremely low.
On the other hand, circulating tumor DNA (ctDNA) is potentially a more sensitive source
of materials, and the stability of DNA may result in a longer half-life than CTCs. If the mutations
of the primary tumor are known (many drivers genes have high penetrance such as TP53 in most
cancers, KRAS in pancreatic cancer, TMPRSS2-ERG fusion in prostate cancer, BRCA in breast
96
Chapter 5: Discussions and Future Perspectives
cancer and APC in colon cancer), we can monitor the progress of the patients more easily with
ctDNA.
5.1.2
Working with clinical specimens
Working with clinical specimens has proven to be more challenging than working with cell
lines or animal models for the following reasons:
1. The paucity of CTCs and their failure to proliferate prevented me from performing more
in-depth mechanistic analysis. Therefore, the lack of sufficient materials is the bottleneck
of CTC characterization.
2. The pleural effusion samples were resistant to shRNA knockdown, and a cell line was
used instead to analyze the functional mechanism of CD24.
3. The large patient-to-patient variation makes statistically significant conclusion difficult to
obtain. For practicality concerns, it is best to perform mechanistic exploration in in vitro
models and then perform measurements of a limited set of parameters in a large cohort of
patients.
4. The clinical status of the patients may have a huge impact on the behavior of the cancer
cells, and this may partially account for the patient-to-patient variation observed. It is
plausible that shed tumor cells may become more invasive as cancer progresses.
Therefore, it is important to clearly define the patient characteristics at the start of the
trial so that convincing conclusions can be obtained with a more homogenous group of
subjects.
5. Some findings are not generalizable across cancer types. The requirement of CD24 for
colorectal cancer but not breast cancer is one such example.
5.2
5.2.1
Future experiments
Discovery of EpCAM- CTCs
We have shown convincing proof for the tumorigenicity of EpCAM+ cells, however,
many are speculating the presence of EpCAM- CTCs. Indeed, EpCAM- CTCs have been shown
to form tumors and spontaneous metastases in mice [31]. My preliminary data on EpCAM- cells
has revealed that there are ~103 EpCAM- cells in 5 ml of blood (Figure 2-12), much more than
the number of EpCAM+ cells. The first hypothesis to be tested is whether some of these cells are
malignant. Since it is likely that the tumor content is low, we need both a sensitive and a high
97
Chapter 5: Discussions and Future Perspectives
throughput method to determine the presence of tumor cells. This can be achieved by pooling the
EpCAM- cells and forming targeted amplification of mutation present in the primary tumor. In
this case, a single mutation is sufficient but it must be present in the tumor and absent in the
adjacent normal tissue. The coverage at the mutated site should be greater than 10 times the
number of cells.
Once it is confirmed that tumor cells are contained in the mixture, the next logical
question is which of the single cells are tumor cells. This can be obtained by performing
amplification of the mutation for each cell separately. This may be achieved with a mutation
specific primer (the 3’ end of the primer is complementary to the mutation) and performing onchip PCR or RT-PCR in the nanowell system. Since there are more copies of the mutation in the
RNA, we can consider probing RNA instead of DNA. Their cytosolic localization also renders
them more accessible to PCR amplification. Invitrogen has developed a Single Cell-to-Ct kit that
combines cell lysis, reverse transcription and PCR all in one step, thus eliminating the need to
open the nanowells repeatedly for several steps of the reaction. We should expect amplification
in only the wells containing the mutation. For background normalization, we should also include
a primer against the wildtype labeled with a second fluorescent.
Discovery of surface markers for EpCAM- cell is challenging because we have the
circular problem of not knowing what the phenotypes of tumor cells are and therefore not being
able to isolate them for surface marker screening. We can survey the entire genomic landscape of
all the EpCAM- cells but the cost is prohibitory given the number of cells we have to screen.
ENCODE recommends the number of reads for evaluating the similarity between two polyA+
samples as 30 million paired end [98]. One HiSeq run is ~ 150 million reads, which implies we
can only multiplex 5 cells in one lane. The cost can increase substantially if we are screening
1000 EpCAM- cells per sample. Therefore, limiting the number of transcripts may be a more cost
effective approach. This has been achieved in hematopoietic cells in which 280 surface
molecules were quantified during RT-PCR by custom designed primers [99]. The mutation
present in the primary tumor must be included in the panel of targeted genes to determine the
malignancy of the cells. Suppose we would like to screen 100 genes and these genes are
expressed at 100 transcripts per cell on average. The number of reads we need for a single cells
with 1x coverage is 100 genes x 100 copies/transcript/cell x 1 time coverage x 1000 bases
98
Chapter 5: Discussions and Future Perspectives
(average size of a gene) / 50 read lengths = 200,000. This means we can multiplex 750 cells in a
single HiSeq lane.
Ultimately, the circulating tumor cells that do progress to overt metastases will utilize the
same proliferative program that the primary tumor or the metastases thrive on. In fact, our data
seems to suggest that CTCs are just tumor cells inadvertently shed into the bloodstream and are
not more aggressive than the solid tumors. Therefore, genes important for the growth of primary
tumors should be transferrable to the circulating tumor cells. We could use primary tumors as a
source of materials for hypothesis generation because they represent a more complete repertoire
of cancer cells. Therefore, the 100 candidate genes mentioned in the preceding paragraph can be
based on previous research that identified genes important for the growth of primary tumors.
5.2.2
The switch from dormancy to proliferation
Since the majority of the CTCs are dormant, the central question remained to be explored
is how single quiescent cells turn on their proliferative programs, sometimes even years after the
removal of primary tumor. Finding out the genes that are important for the switch from an OFF
to ON state in tumor cells is the key problem that needs to be addressed.
As pointed out earlier, since CTCs are derived from solid tumors, the same “trigger”
genes that can cause activation of the solid tumor should exert similar functions on CTCs.
Working with solid tumors circumvents the problem of having inadequate materials for
exploratory studies. The only difference that exists between a solid tumor and CTCs is that CTCs
might have been deactivated by the circulation, that is, CTCs may be in the OFF state while
actively growing tumor cells are in the ON state. Therefore, we need to deactivate the tumor cells
before using them as the model system in the search for the trigger genes. That can be
accomplished by disaggregating a solid tumor into single cells with a protease digestion,
followed by culturing them in a suspension culture. Subsequently, disaggregated tumor cells can
serve as a good proxy to circulating tumor cells. It is plausible that even a panel of cell lines can
be used as relevant model system. We would then compare the genetic information between the
clones that grow with the clones that stagnate. The growth of the cells can be marked by labeling
the cells with a dye that is diluted every replication cycle. For initial analysis, we can sort the
cells into non-proliferative and proliferative groups and analyzed the gene expression of the
pooled sample by RNAseq analysis. RNAseq analysis can generate a list of genes that might be
involved in turning on the proliferative program. The causal relationship of the putative genes to
99
Chapter 5: Discussions and Future Perspectives
tumorigenesis needs to be evaluated by removing the genes individually and evaluating their
effect on tumor initiation.
Alternatively, an shRNA library can be used to infect a cell line [100] which is then
followed for the rate of tumor initiation after culturing in suspension culture. The shRNA clones
can then be amplified from both the proliferative cells and the dormant cells. Comparison
between the proliferative and non-proliferative cells can potentially lead to the discovery of
genes or network important for tumor initiation, which can then be targeted using inhibitors
against the genes or network.
We hope that by discovering the genes important for the reactivation and colonization of
mobilized tumor cells, we can target cancer cells before they become full-fledged metastases.
Recolonization of a new site is the most vulnerable step in the spread of cancer, and we should
make use of this latent window to treat cancer, thereby preventing the onset of metastasis.
100
Chapter 5: Discussions and Future Perspectives
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
Loberg, R.D., et al., The lethal phenotype of cancer: The molecular basis of death due to
malignancy. Ca-a Cancer Journal for Clinicians, 2007. 57(4): p. 225-241.
Luzzi, K.J., et al., Multistep nature of metastatic inefficiency: dormancy of solitary cells
after successful extravasation and limited survival of early micrometastases. Am J
Pathol, 1998. 153(3): p. 865-73.
Klein, C.A., Parallel progression of primary tumours and metastases. Nature Reviews
Cancer, 2009. 9(4): p. 302-312.
Allard, W.J., et al., Tumor cells circulate in the peripheral blood of all major carcinomas
but not in healthy subjects or patients with nonmalignant diseases. Clinical Cancer
Research, 2004. 10(20): p. 6897-6904.
Hansen, E., et al., Tumor-Cells in Blood Shed from the Surgical Field. Archives of
Surgery, 1995. 130(4): p. 387-393.
Navin, N., et al., Tumour evolution inferred by single-cell sequencing. Nature, 2011.
472(7341): p. 90-U119.
Klein, C.A., et al., Genetic heterogeneity of single disseminated tumour cells in minimal
residual cancer. Lancet, 2002. 360(9334): p. 683-689.
Baccelli, I., et al., Identification of a population of blood circulating tumor cells from
breast cancer patients that initiates metastasis in a xenograft assay. Nature
Biotechnology, 2013. 31(6): p. 539-U143.
Hansen, E., et al., Tumor cells in blood shed from the surgical field. Arch Surg, 1995.
130(4): p. 387-93.
Went, P.T., et al., Frequent EpCam protein expression in human carcinomas. Human
Pathology, 2004. 35(1): p. 122-128.
Nagrath, S., et al., Isolation of rare circulating tumour cells in cancer patients by
microchip technology. Nature, 2007. 450(7173): p. 1235-U10.
Talasaz, A.H., et al., Isolating highly enriched populations of circulating epithelial cells
and other rare cells from blood using a magnetic sweeper device. Proceedings of the
National Academy of Sciences of the United States of America, 2009. 106(10): p. 39703975.
Williams, A., et al., Size-based enrichment technologies for CTC detection and
characterization. Recent Results Cancer Res. 195: p. 87-95.
Ozkumur, E., et al., Inertial focusing for tumor antigen-dependent and -independent
sorting of rare circulating tumor cells. Sci Transl Med, 2013. 5(179): p. 179ra47.
Hou, H.W., et al., Isolation and retrieval of circulating tumor cells using centrifugal
forces. Scientific Reports, 2013. 3.
Gertler, R., et al., Detection of circulating tumor cells in blood using an optimized density
gradient centrifugation. Molecular Staging of Cancer, 2003. 162: p. 149-155.
Weng, Y.C., et al., Mass sensors with mechanical traps for weighing single cells in
different fluids. Lab on a Chip, 2011. 11(24): p. 4174-4180.
Dean, F.B., et al., Rapid amplification of plasmid and phage DNA using phi29 DNA
polymerase and multiply-primed rolling circle amplification. Genome Research, 2001.
11(6): p. 1095-1099.
Zong, C.H., et al., Genome-Wide Detection of Single-Nucleotide and Copy-Number
Variations of a Single Human Cell. Science, 2012. 338(6114): p. 1622-1626.
101
Chapter 5: Discussions and Future Perspectives
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
Shapiro, E., T. Biezuner, and S. Linnarsson, Single-cell sequencing-based technologies
will revolutionize whole-organism science. Nat Rev Genet, 2013. 14(9): p. 618-30.
Stott, S.L., et al., Isolation of circulating tumor cells using a microvortex-generating
herringbone-chip. Proceedings of the National Academy of Sciences of the United States
of America, 2010. 107(43): p. 18392-18397.
Riethdorf, S., et al., Detection of circulating tumor cells in peripheral blood of patients
with metastatic breast cancer: A validation study of the CellSearch system. Clinical
Cancer Research, 2007. 13(3): p. 920-928.
Galanzha, E.I., et al., In vivo magnetic enrichment and multiplex photoacoustic detection
of circulating tumour cells. Nature Nanotechnology, 2009. 4(12): p. 855-860.
Desitter, I., et al., A New Device for Rapid Isolation by Size and Characterization of Rare
Circulating Tumor Cells. Anticancer Research, 2011. 31(2): p. 427-441.
Shim, S., et al., Antibody-independent isolation of circulating tumor cells by continuousflow dielectrophoresis. Biomicrofluidics, 2013. 7(1).
de Bono, J.S., et al., Circulating Tumor Cells Predict Survival Benefit From Treatment in
Metastatic Castration-Resistant Prostate Cancer Editorial Comment. Journal of Urology,
2009. 181(6): p. 2535-2536.
Krebs, M.G., et al., Evaluation and prognostic significance of circulating tumor cells in
patients with non-small-cell lung cancer. J Clin Oncol. 29(12): p. 1556-63.
Pachmann, K., et al., Monitoring the response of circulating epithelial tumor cells to
adjuvant chemotherapy in breast cancer allows detection of patients at risk of early
relapse. Journal of Clinical Oncology, 2008. 26(8): p. 1208-1215.
Maheswaran, S., et al., Detection of mutations in EGFR in circulating lung-cancer cells.
New England Journal of Medicine, 2008. 359(4): p. 366-377.
Yu, M., et al., Circulating Breast Tumor Cells Exhibit Dynamic Changes in Epithelial
and Mesenchymal Composition. Science, 2013. 339(6119): p. 580-584.
Zhang, L.X., et al., The Identification and Characterization of Breast Cancer CTCs
Competent for Brain Metastasis. Science Translational Medicine, 2013. 5(180).
Heitzer, E., et al., Complex Tumor Genomes Inferred from Single Circulating Tumor
Cells by Array-CGH and Next-Generation Sequencing. Cancer Research, 2013. 73(10):
p. 2965-2975.
Magbanua, M.J.M., et al., Genomic Profiling of Isolated Circulating Tumor Cells from
Metastatic Breast Cancer Patients. Cancer Research, 2013. 73(1): p. 30-40.
Oh, W.K., et al., Development of an integrated prostate cancer research information
system. Clin Genitourin Cancer, 2006. 5(1): p. 61-6.
Ogunniyi, A.O., et al., Screening individual hybridomas by microengraving to discover
monoclonal antibodies. Nat Protoc, 2009. 4(5): p. 767-82.
Simm, M.
Liu, X.F., et al., ROCK Inhibitor and Feeder Cells Induce the Conditional
Reprogramming of Epithelial Cells. American Journal of Pathology, 2012. 180(2): p.
599-607.
Sato, T. and H. Clevers, Primary mouse small intestinal epithelial cell cultures. Methods
Mol Biol. 945: p. 319-28.
Han, Q., et al., Multidimensional analysis of the frequencies and rates of cytokine
secretion from single cells by quantitative microengraving. Lab on a Chip, 2010. 10(11):
p. 1391-1400.
102
Chapter 5: Discussions and Future Perspectives
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
Miller, M.A., et al., Proteolytic Activity Matrix Analysis (PrAMA) for simultaneous
determination of multiple protease activities. Integr Biol (Camb). 3(4): p. 422-38.
Love, J.C., et al., A microengraving method for rapid selection of single cells producing
antigen-specific antibodies. Nat Biotechnol, 2006. 24(6): p. 703-7.
Li, Y.M., et al., Osteoblasts can stimulate prostate cancer growth and transcriptionally
down-regulate PSA expression in cell line models. Urologic Oncology-Seminars and
Original Investigations, 2011. 29(6): p. 802-808.
Mehes, G., et al., Circulating breast cancer cells are frequently apoptotic. American
Journal of Pathology, 2001. 159(1): p. 17-20.
Rossi, E., et al., Dynamic changes of live/apoptotic circulating tumour cells as predictive
marker of response to Sunitinib in metastatic renal cancer. British Journal of Cancer,
2012. 107(8): p. 1286-1294.
Larson, C.J., et al., Apoptosis of circulating tumor cells in prostate cancer patients.
Cytometry Part A, 2004. 62A(1): p. 46-53.
Muller, V., et al., Circulating tumor cells in breast cancer: Correlation to bone marrow
micrometastases, heterogeneous response to systemic therapy and low proliferative
activity. Clinical Cancer Research, 2005. 11(10): p. 3678-3685.
Powell, A.A., et al., Single Cell Profiling of Circulating Tumor Cells: Transcriptional
Heterogeneity and Diversity from Breast Cancer Cell Lines. Plos One, 2012. 7(5).
Murray, N.P., et al., Redefining micrometastasis in prostate cancer - a comparison of
circulating prostate cells, bone marrow disseminated tumor cells and micrometastasis:
Implications in determining local or systemic treatment for biochemical failure after
radical prostatectomy. International Journal of Molecular Medicine, 2012. 30(4): p. 896904.
Pantel, K., C. Alix-Panabieres, and S. Riethdorf, Cancer micrometastases. Nature
Reviews Clinical Oncology, 2009. 6(6): p. 339-351.
Alix-Panabieres, C., et al., Detection and characterization of putative metastatic
precursor cells in cancer patients. Clinical Chemistry, 2007. 53(3): p. 537-539.
Aguirre-Ghiso, J.A., Models, mechanisms and clinical evidence for cancer dormancy.
Nat Rev Cancer, 2007. 7(11): p. 834-46.
Wiedswang, G., et al., Isolated tumor cells in bone marrow three years after diagnosis in
disease-free breast cancer patients predict unfavorable clinical outcome. Clin Cancer
Res, 2004. 10(16): p. 5342-8.
Friedl, P. and K. Wolf, Tumour-cell invasion and migration: diversity and escape
mechanisms. Nat Rev Cancer, 2003. 3(5): p. 362-74.
Friedl, P., et al., Classifying collective cancer cell invasion. Nature Cell Biology, 2012.
14(8): p. 777-783.
Raviraj, V., et al., Dormant but migratory tumour cells in desmoplastic stroma of
invasive ductal carcinomas. Clinical & Experimental Metastasis, 2012. 29(3): p. 273292.
Whitson, B.A., et al., Surgery for Early-Stage Non-Small Cell Lung Cancer: A Systematic
Review of the Video-Assisted Thoracoscopic Surgery Versus Thoracotomy Approaches to
Lobectomy. Annals of Thoracic Surgery, 2008. 86(6): p. 2008-2018.
Yamanaka, N., et al., Do the tumor cells of hepatocellular carcinomas dislodge into the
portal venous stream during hepatic resection? Cancer, 1992. 70(9): p. 2263-7.
103
Chapter 5: Discussions and Future Perspectives
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
Paolucci, V., et al., Tumor seeding following laparoscopy: international survey. World J
Surg, 1999. 23(10): p. 989-95; discussion 996-7.
Wang, P.H., et al., Risk factors contributing to early occurrence of port site metastases of
laparoscopic surgery for malignancy. Gynecol Oncol, 1999. 72(1): p. 38-44.
Turnbull, R.B., Jr., Cancer of the colon. The five- and ten-year survival rates following
resection utilizing the isolation technique. Ann R Coll Surg Engl, 1970. 46(5): p. 243-50.
Turnbull, R.B., Jr., et al., Cancer of the colon: the influence of the no-touch isolation
technic on survival rates. Ann Surg, 1967. 166(3): p. 420-7.
Sawabata, N., et al., Circulating tumor cells in peripheral blood caused by surgical
manipulation of non-small-cell lung cancer: pilot study using an immunocytology
method. Gen Thorac Cardiovasc Surg, 2007. 55(5): p. 189-92.
Okumura, Y., et al., Circulating Tumor Cells in Pulmonary Venous Blood of Primary
Lung Cancer Patients. Annals of Thoracic Surgery, 2009. 87(6): p. 1669-1675.
Brown, D.C., et al., Detection of Intraoperative Tumor-Cell Dissemination in Patients
with Breast-Cancer by Use of Reverse Transcription and Polymerase Chain-Reaction.
Surgery, 1995. 117(1): p. 96-101.
Koch, M., et al., Detection of hematogenous tumor cell dissemination predicts tumor
relapse in patients undergoing surgical resection of colorectal liver metastases. Ann
Surg, 2005. 241(2): p. 199-205.
Song, P.P., et al., Effects of different sequences of pulmonary artery and vein ligations
during pulmonary lobectomy on blood micrometastasis of non-small cell lung cancer.
Oncology Letters, 2013. 5(2): p. 463-468.
Carter, P., L. Smith, and M. Ryan, Identification and validation of cell surface antigens
for antibody targeting in oncology. Endocrine-Related Cancer, 2004. 11(4): p. 659-687.
Verschraegen, C.F., et al., Establishment and characterization of cancer cell cultures and
xenografts derived from primary or metastatic Mullerian cancers. Clin Cancer Res, 2003.
9(2): p. 845-52.
Camp, R.L., et al., Quantitative analysis of breast cancer tissue microarrays shows that
both high and normal levels of HER2 expression are associated with poor outcome.
Cancer Research, 2003. 63(7): p. 1445-1448.
Vogel, C.L., et al., Efficacy and safety of trastuzumab as a single agent in first-line
treatment of HER2-overexpressing metastatic breast cancer. Journal of Clinical
Oncology, 2002. 20(3): p. 719-726.
Harari, D. and Y. Yarden, Molecular mechanisms underlying ErbB2/HER2 action in
breast cancer. Oncogene, 2000. 19(53): p. 6102-6114.
Visvader, J.E. and G.J. Lindeman, Cancer stem cells in solid tumours: accumulating
evidence and unresolved questions. Nat Rev Cancer, 2008. 8(10): p. 755-68.
Hermann, P.C., et al., Distinct populations of cancer stem cells determine tumor growth
and metastatic activity in human pancreatic cancer. Cell Stem Cell, 2007. 1(3): p. 31323.
Yang, J., et al., Twist, a master regulator of morphogenesis, plays an essential role in
tumor metastasis. Cell, 2004. 117(7): p. 927-939.
Minn, A.J., et al., Genes that mediate breast cancer metastasis to lung. Nature, 2005.
436(7050): p. 518-24.
Winslow, M.M., et al., Suppression of lung adenocarcinoma progression by Nkx2-1.
Nature, 2011. 473(7345): p. 101-U120.
104
Chapter 5: Discussions and Future Perspectives
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
Adams, G.P. and L.M. Weiner, Monoclonal antibody therapy of cancer. Nature
Biotechnology, 2005. 23(9): p. 1147-1157.
Liang, Q.L., et al., Diagnostic accuracy of tumour markers for malignant pleural
effusion: a meta-analysis. Thorax, 2008. 63(1): p. 35-41.
Lombardi, G., et al., Diagnosis and treatment of malignant pleural effusion: a systematic
literature review and new approaches. Am J Clin Oncol. 33(4): p. 420-3.
Aigner, S., et al., CD24, a mucin-type glycoprotein, is a ligand for P-selectin on human
tumor cells. Blood, 1997. 89(9): p. 3385-3395.
Gay, L.J. and B. Felding-Habermann, Contribution of platelets to tumour metastasis.
Nature Reviews Cancer, 2011. 11(2): p. 123-134.
Labelle, M., S. Begum, and R.O. Hynes, Direct Signaling between Platelets and Cancer
Cells Induces an Epithelial-Mesenchymal-Like Transition and Promotes Metastasis.
Cancer Cell, 2011. 20(5): p. 576-590.
Lee, J.H., et al., CD24 overexpression in cancer development and progression: a metaanalysis. Oncol Rep, 2009. 22(5): p. 1149-56.
Hofmann, M., et al., CD44 splice variants confer metastatic behavior in rats:
homologous sequences are expressed in human tumor cell lines. Cancer Res, 1991.
51(19): p. 5292-7.
Alimandi, M., et al., Cooperative Signaling of Erbb3 and Erbb2 in Neoplastic
Transformation and Human Mammary Carcinomas. Oncogene, 1995. 10(9): p. 18131821.
Schoeberl, B., et al., Therapeutically Targeting ErbB3: A Key Node in Ligand-Induced
Activation of the ErbB Receptor-PI3K Axis. Science Signaling, 2009. 2(77).
Simpson, C.D., K. Anyiwe, and A.D. Schimmer, Anoikis resistance and tumor
metastasis. Cancer Letters, 2008. 272(2): p. 177-185.
Said, N., et al., Tumor endothelin-1 enhances metastatic colonization of the lung in
mouse xenograft models of bladder cancer. Journal of Clinical Investigation, 2011.
121(1): p. 132-147.
Kristiansen, G., M. Sammar, and P. Altevogt, Tumour biological aspects of CD24, a
mucin-like adhesion molecule. J Mol Histol, 2004. 35(3): p. 255-62.
Sagiv, E., et al., Targeting CD24 for treatment of colorectal and pancreatic cancer by
monoclonal antibodies or small interfering RNA. Cancer Res, 2008. 68(8): p. 2803-12.
Overdevest, J.B., et al., CD24 offers a therapeutic target for control of bladder cancer
metastasis based on a requirement for lung colonization. Cancer Res. 71(11): p. 3802-11.
Smith, S.C., et al., The metastasis-associated gene CD24 is regulated by Ral GTPase and
is a mediator of cell proliferation and survival in human cancer. Cancer Research, 2006.
66(4): p. 1917-1922.
Braun, S., et al., Monoclonal antibody therapy with Edrecolomab in breast cancer
patients: Monitoring of elimination of disseminated cytokeratin-positive tumor cells in
bone marrow. Clinical Cancer Research, 1999. 5(12): p. 3999-4004.
Al-Hajj, M., et al., Prospective identification of tumorigenic breast cancer cells.
Proceedings of the National Academy of Sciences of the United States of America, 2003.
100(7): p. 3983-3988.
Mani, S.A., et al., The epithelial-mesenchymal transition generates cells with properties
of stem cells. Cell, 2008. 133(4): p. 704-15.
105
Chapter 5: Discussions and Future Perspectives
96.
97.
98.
99.
100.
Athanassiadou, P., et al., CD24 expression has a prognostic impact in breast carcinoma.
Pathology Research and Practice, 2009. 205(8): p. 524-533.
Ponta, H., L. Sherman, and P.A. Herrlich, CD44: from adhesion molecules to signalling
regulators. Nat Rev Mol Cell Biol, 2003. 4(1): p. 33-45.
ENCODE. Available from: http://www.rna-seqblog.com/best-practices-for-rna-seq/.
Guo, G., et al., Mapping cellular hierarchy by single-cell analysis of the cell surface
repertoire. Cell Stem Cell, 2013. 13(4): p. 492-505.
Meacham, C.E., et al., In vivo RNAi screening identifies regulators of actin dynamics as
key determinants of lymphoma progression. Nature Genetics, 2009. 41(10): p. 1133U118.
106