Phenotypic Heterogeneity of Circulating Tumor Cells ...It has been hypothesized that the efﬁcacy of pathway-speciﬁc ... This is a retrospective analysis of a prospectively accrued

Clinical Studies

Phenotypic Heterogeneity of Circulating TumorCells Informs Clinical Decisions between ARSignaling Inhibitors and Taxanes in MetastaticProstate CancerHoward I. Scher1,2, RyonP.Graf3, NicoleA. Schreiber1, BrigitMcLaughlin1, AdamJendrisak3,YipengWang3, Jerry Lee3, StephanieGreene3, Rachel Krupa3,DavidLu3, Pascal Bamford3,Jessica E. Louw3, Lyndsey Dugan3, Hebert A. Vargas4, Martin Fleisher5, Mark Landers3,Glenn Heller6, and Ryan Dittamore3

Abstract

The heterogeneity of an individual patient's tumor has beenlinked to treatment resistance, but quantitative biomarkers torapidly and reproducibly evaluateheterogeneity ina clinical settingare currently lacking. Using established tools available in a Collegeof American Pathologists–accredited and Clinical LaboratoryImprovementAmendments–certified clinical laboratory,wequan-tified digital pathology features on 9,225 individual circulatingtumor cells (CTC) from 179 unique metastatic castration-resistantprostate cancer (mCRPC)patients todefinephenotypicallydistinctcell types.Heterogeneitywasquantifiedon thebasis of thediversity

of cell types in individual patient samples using the Shannon indexand associated with overall survival (OS) in the 145 specimenscollected prior to initiation of the second or later lines of therapy.Low CTC phenotypic heterogeneity was associated with better OSin patients treated with androgen receptor signaling inhibitors(ARSI), whereas high heterogeneity was associated with better OSin patients treated with taxane chemotherapy. Overall, the resultsshow that quantifying CTC phenotypic heterogeneity can helpinform the choice between ARSI and taxanes in mCRPC patients.Cancer Res; 77(20); 5687–98. �2017 AACR.

IntroductionThe heterogeneity of cancer has long been recognized by

phenotypic differences in cell morphology within a tumor spec-imen, and the observation that disease is rarely eliminated withany single systemic therapy. Recent multi-gene sequencing effortshave extended the concept of intrapatient cancer heterogeneity tothe single cell (1, 2), further revealing immensediversity in tumorsbetweenpatients (3, 4), individual lesionswithin the samepatient(5, 6), and different sites within each lesion (7, 8). It is thereforeunsurprising that reliance on solid tissue biopsies to guide "pre-cision medicine" is not so precise (9, 10). Recently reportedprofiling studies continue to elucidate the genomic complexityof cancer. Most are observational and are not reported as

biomarkers that can be associated with clinical outcomes or usedto impact medical decision making in patients (11, 12). Here, wereport the development of a quantitative biomarker of CTCheterogeneity fit-for-purpose (13) of informing treatment selec-tion at the time a change in therapy is needed, and explore therelationship between CTC phenotypic heterogeneity and patientsurvival following treatment with androgen receptor signalinginhibitors (ARSI) or taxane chemotherapies.

It has been hypothesized that the efficacy of pathway-specifictargeted agents would be inversely related to the degree of intra-patient heterogeneity present when a therapy is administered(Supplementary Fig. S1A–S1D; refs. 8, 14, 15). The currentlyapproved standard-of-care life-prolonging systemic therapiesfor progressive metastatic castration-resistant prostate cancer(mCRPC) include two therapeutic drug classes: pathway-specificARSI and non-pathway–specific taxane-based chemotherapy.ARSI, such as abiraterone acetate, enzalutamide, and apaluta-mide, target the androgen receptor (AR) signaling pathway, andinhibit the growth of cells dependent on AR signaling for survival,while taxane-based chemotherapies, docetaxel and cabazitaxel,stabilizemicrotubules, broadly inhibiting cell division in all cells,including those that harbor diverse drivers of resistance to targetedtherapy (16).

In current clinical practice, the choice between one therapy overanother at a decision point in mCRPC management is largelyempiric; the biology of an individual patient's cancer is typicallynot considered. In the first-line setting following progression onandrogen deprivation therapy (ADT), most patients are treatedwith an ARSI over a taxane based on the high response rates, morefavorable safety profile, and long survival times at this decision

1Genitourinary Oncology Service, Department of Medicine, Memorial SloanKettering Cancer Center, New York, New York. 2Department of Medicine, WeillCornell Medical College, NewYork, NewYork. 3Epic Sciences, La Jolla, California.4Department of Radiology, Memorial Sloan Kettering Cancer Center, New York,New York. 5Clinical Chemistry Service, Department of Laboratory Medicine,Memorial Sloan Kettering Cancer Center, New York, New York. 6BiostatisticsService, Department of Epidemiology and Biostatistics, Memorial Sloan Ketter-ing Cancer Center, New York, New York.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Corresponding Author: Howard I. Scher, Genitourinary Oncology Service,Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY10065. Phone: 646-422-4323; Fax: 212-988-0851; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-17-1353

�2017 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 5687

on January 22, 2020. © 2017 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst August 17, 2017; DOI: 10.1158/0008-5472.CAN-17-1353

http://crossmark.crossref.org/dialog/?doi=10.1158/0008-5472.CAN-17-1353&domain=pdf&date_stamp=2017-9-30

http://cancerres.aacrjournals.org/

point (17, 18). This contrasts with the second- or later line settingwhere there is no consensus onmanagement because the responseto sequential ARSI is lower and of shorter duration, and responseto the first does not predict response to the other (19). Everytreatment towhich a tumor has been exposed can alter the biologyof the cancer (20), necessitating the serial profiling of disease.Validated therapy-guiding biomarkers to better inform each treat-ment decision is a significant unmet medical need (21).

High intratumor and intrapatient heterogeneity has been asso-ciatedwith poor prognosis in small cohorts of patientswith breastand head and neck cancer (22, 23) and more recently in lungcancer (24). Expanded explorations of the relationship betweenheterogeneity and other clinical outcomes have been hindered bythe range of sampling methods to obtain tumor, the range ofdevices and analytic methods, and the lack of analytically andclinically validated assays. CTCs are a source of tumormaterial forbiologic profiling that can be obtained from routine, repeatedphlebotomy samples with minimal patient discomfort. CTCs canrepresent multiple metastatic lesions at once and are more likelyto provide information on the diversity of a patient's disease thana single site biopsy. However, existing single-cell genomic tech-niques (25) have yet to meet the regulatory requirements ofanalytic and clinical validity, turnaround time, and cost for morewidespread clinical utility.

In contrast, automated phenotypic (protein andmorphologic)assessment of single cells has established clinical utility in thearena of cytologic assessment of Pap smears for cervical cancerdetection. In this report, we captured CTCs utilizing an analyti-cally validated microscopy-based CTC detection technology (26)that allows for quantification of protein expression, and furthercharacterized individual cells by digital pathology features pre-viously validated and FDA approved for clinical use to identifyand characterize malignant cells in Pap smears (27–29).

Our central hypothesis is that the degree of pretherapy pheno-typic heterogeneity of CTCs inversely associates with patientsurvival following treatment with ARSI but not taxane therapy.To test this, individual features in each cell were used to definephenotypically distinct cell types. Heterogeneity was then quan-tified on the basis of the diversity of individual cell types in eachpatient sample using the Shannon index. The Shannon index iswidely utilized in ecology and biodiversity research to measurethe diversity or entropy of species, or in this case, CTCs. Previousapplications of the Shannon index include showing differences inthe genomic heterogeneity of separate and distinct metastaticlesions in an individual patient (30), and along with othermeasures, to predict the transition from Barrett esophagus toesophageal adenocarcinoma (31). Here, we associated this quan-titative heterogeneity biomarker with overall survival (OS) fol-lowing treatment with standard-of-care drug classes to demon-strate clinical utility: that patient outcomes can be improved byusing the heterogeneity biomarker test result to inform the choiceof treatment relative to nonuse of the test result.We also evaluatedwhether observed phenotypic heterogeneity was associated withthe presence ofmultiple genomic clones, elucidated by single-cellsequencing of CTCs in a subset of patient samples.

Patients and MethodsStudy design

This is a retrospective analysis of a prospectively accrued cross-sectional cohort of patients treated in a clinical practice setting at

Memorial Sloan Kettering Cancer Center (MSKCC, New York,NY). All patients provided signed informed consent for partici-pation on a MSKCC Institutional Review Board/Privacy Board–approved protocol prior to blood sampling and studies wereconducted in accordance with the Declaration of Helsinki, Bel-mont Report, and U.S. Common Rule. The choice of therapy wasat the discretionof the treating physician. All patients underwent ahistory that included details of the stage of disease at diagnosis,initial management, subsequent systemic therapies, physicalexamination, and laboratory studies including complete bloodcount, chemistry panel [albumin (ALB), alkaline phosphatase(ALK), lactate dehydrogenases (LDH), hemoglobin (Hgb), pros-tate specific antigen (PSA)], and serum testosterone to confirmcastration status (<50 ng/dL; Table 1).

Patient selectionBetween December 2012 and March 2015, 265 patients with

histologically confirmedmCRPCundergoing a change in systemictherapy for progressive disease were treated at MSKCC. Of these,179 unique patients starting an ARSI (abiraterone, enzalutamide,or apalutamide) or taxane (docetaxel, cabazitaxel, or paclitaxel)provided 319 samples (221 pretherapy and 98 on-therapy sam-ples), which formed two cohorts of samples (Table 1; Fig. 1). Thefirst, the CTC contributing cohort, included all 319 samples andwas used for automated feature extraction of protein biomarkerexpression and digital pathology features (Fig. 2A). The second,the clinical association cohort, is a subset of the CTC contributingcohort that consists only of samples obtained prior to the start ofthe second- or later line therapy (86 ARSI and 59 taxane, or 145total pretherapy samples). First-line samples were excludedbecause the majority of mCRPC patients progressing on standardhormones are treated with an ARSI due to the high response rate.Few receive taxanes in the first line, limiting the ability to explorethe association of heterogeneity with outcome to both drugclasses. There were no additional exclusions; all available sampleswere included in the analysis.

CTC collectionBlood samples (7.5 mL) were collected in Streck tubes and

processed at MSKCC or shipped to Epic Sciences and processedwithin 48 hours. Red blood cells were lysed, and approximately3 � 106 nucleated cells were dispensed onto 10–16 glass micro-scope slides and placed at �80�C for long-term storage asdescribed previously (26, 32, 33). Sample processing and testingwere conducted in laboratories following both Clinical Labora-tory Improvement Amendments (CLIA) and College of AmericanPathologists (CAP) regulations.

CTC immunofluorescent staining and detectionCTCs were characterized by automated immunofluorescent

staining for DAPI (a DNA stain), cytokeratins, CD45 (hemato-poietic lineage marker), and AR N-terminal domain (AR; CellSignaling Technology; D6F11) as described previously (26, 32,33). Up to two slides were evaluated per sample tested. CTCswereidentified using fluorescent scanners that imaged every nucleatedobject on the slides. All CTCs had an intact nucleus (DAPI), noCD45 signal, and morphologic features consistent with malig-nancy, as published previously (26). Clinical laboratory scientists(licensed in California) conducted final quality control of CTCidentification.

Scher et al.

Cancer Res; 77(20) October 15, 2017 Cancer Research5688




Individual CTC biomarker and digital pathologyWe built upon digital pathology software from FDA-cleared

devices used to characterize Pap smears for the context of use ofdiagnosing cervical cancer (27–29) to characterize CTCs. A sys-tematic analysis performed by key leaders in the field identifiedthe features that for decades were repeatedly found to be themostdiagnostically meaningful in cytopathology, histopathology, andhematopathology applications (34). A subset of these featureswas found to be suitable for the magnification of the microscopeobjectives utilized by the Epic Sciences platform. All featuresconsidered were measured, and part of the raw data used foranalyses, listed in Fig. 2A.

To allow for quantification of digital pathology features on acell-by-cell basis, individual CTC images of cytokeratin (CK),nuclear, and AR N-terminal staining were segmented from each

fluorescence channel image independently utilizing a library ofsingle-cell features optimized for fluorescence microscopy(Fig. 2A; ref. 34). Features utilized for this analysis includedmorphometric and texture patterns of nuclear and CK staining,as well as the densitometry (intensity) of the CK and AR signal,which were quantified as a ratio of signal intensity on the CTCrelative to surrounding white blood cells (WBC; Fig. 2A). Theidentification of a cluster of two or more CTCs was recorded as acategorical dichotomous variable, as was whether the level of CKand AR expression was above or below an analytic threshold ofintensity (Fig. 2A) based on cultured cell line control cells spikedinto healthy donor blood as described previously (26). ApoptoticCTCs were not included or reported, in part because nuclearfragmentation andmembrane blebbing (35) confound the digitalpathology features.

CTC mathematical phenotype identification and Shannonindex

Use of the Shannon index necessitated defining unique "spe-cies" within a given space. These "species" were defined using anunsupervised clustering analysis of the CTC protein, digitalpathology, and categorical features (Fig. 2B) into phenotypicallysimilar subtypes ofCTCs fromall samples in theCTC contributingcohort. To enable cross-feature comparisons, minimize digitalpathology feature redundancy, and distill the distinct sources ofvariance in the data, quantified biomarkers and features werez-score transformed prior to dimensionality reduction via prin-cipal component analysis, with the number of principal compo-nents selected via standard 85% variance inclusion (Supplemen-tary Fig. S2A).

K-means clustering was then used to group similar CTCs intodiscrete phenotypic cell subtypes based on the principal compo-nents of digital pathology features while minimizing the amountof variance of biomarker and feature ranges within each "cluster,"or group. The number of clusters ('k') was selected by the "elbowmethod" by which the within cluster sum of squares for k ¼ 1 tok ¼ 40 is plotted, and a human analyst looks for a bend in thecurve, signifying a point of diminishing returns where partition-ing the data into additional clusters (more 'k') only marginallyadds to the similarity within each cluster (Supplementary Fig.S2B). All CTCs in the cohort were then assigned a phenotypic cellsubtype ("A"–"O"; Fig. 3A and B) reflecting the mathematicalunsupervised cluster (not to be confused with a histologic cluster,or clump of CTCs traveling together within the bloodstream) andthen grouped by their individual protein, digital pathology, andcategorical features (Fig. 4A).

Next, the patient-level frequency of the defined CTC pheno-typic subtypes (categorized as "A"–"O"/mL) were determined ona per sample basis (Fig. 4B). These "species densities" were thenutilized to calculate a Shannon index for each sample in theclinical association cohort (Rpackage "vegan") that quantified theentropy of the individual CTC phenotypic subtypes present at thetime of the blood draw (Fig. 4B). A higher Shannon indexindicates both a higher number of different "species" (CTCphenotypic subtypes) present, as well as a more random distri-bution of those phenotypic subtypes, meaning that it would bemore difficult, compared with lower Shannon index samples, topredict which phenotypic subtypes would be seen next givenadditional observation. If 0 or 1CTCswere present, a value of zerois generated due to no sample present or the presence of only asingle-cell subtype.

Table 1. Demographics and sample characteristics of the CTC contributingcohort and clinical association cohort

CTCcontributing

cohort

Clinicalassociation

cohort

Patient characteristicsUnique patients 179 116Age, years: median (range) 68 (45–91) 69 (48–91)

Primary treatmentProstatectomy 84 (47%) 55 (48%)Radiation 34 (19%) 20 (17%)Brachytherapy 7 (4%) 7 (6%)None 54 (30%) 34 (29%)

Sample characteristicsTotal baseline (pretherapy) samples 221 145Follow-up (on-therapy) samples 98 0Total samples 319 145Samples with CTCs 264 131

Prior hormone therapies at baselinea

1–2 lines 82 (37%) 25 (17%)3 lines 50 (23%) 38 (26%)4 or more lines 89 (40%) 82 (57%)

Chemotherapy status at baselineChemo-na€�ve 136 (62%) 63 (43%)Chemo-exposed 85 (38%) 82 (57%)

Metastatic therapy initiated after baselineARS Inhibitor 150 86 (59%)Taxane 71 59 (41%)

Line of metastatic therapy at baseline1st line 76 02nd line 56 563rd line 40 40�4th line 49 49

Metastatic sites of disease at baselineBone onlyb 66 (30%) 46 (32%)Lymph node onlyb 24 (11%) 10 (7%)Bone and lymph nodeb 93 (42%) 63 (43%)Bone and visceral þ/� LNb 35 (16%) 25 (17%)Soft tissue only 3 (1%) 1 (1%)

Laboratory measures at baselinePSA, ng/mL: median (range) 37.7 (0.1–3,728.2) 62.7 (0.1–3,728.2)Hgb, g/dl: median (range) 12 (7–15) 11.6 (7–15)ALK, unit/L: median (range) 110 (25–2,170) 121 (42–1,816)LDH, unit/L: median (range)c 222.5 (123–1,293) 238.5 (123–1,004)ALB, g/dl: median (range) 4.2 (31.–4.9) 4.2 (3.1–4.9)

AR N-term CTC test at baselineTotal CTC/mL: median (range) 6.3 (0–991.3) 6.25 (0–991.3)

aIncludes GnRH agonists and antagonists, antiandrogens, and next-generationhormonal therapies (abiraterone acetate, enzalutamide, and apalutamide).bIncludes patients with other soft tissue disease.cThree samples did not have LDH available.

CTC Phenotypic Heterogeneity for mCRPC Treatment Selection

www.aacrjournals.org Cancer Res; 77(20) October 15, 2017 5689




Posttreatment outcomesOverall survival (OS) was calculated from initiation of therapy

to death from any cause. Patients still alive at the time of lastfollow-up were right-censored.

Statistical analysesPatient demographics and clinical characteristics at the time

of blood draw were evaluated by descriptive statistics. c2 andWilcoxon rank sum tests were used to compare treatmentgroups for categorical and continuous characteristics, respectively.Median survival time as a function of a patient's continuousheterogeneity measure was evaluated using nonparametric kernelestimates (R package "sm,"; ref. 36). Time-to-event outcomes,categorized by group, were evaluated with the Kaplan–Meier

method. Differences in time-to-event outcomes between catego-rized groups were evaluated using the maximum log-rank test,where the grouping was determined adaptively from the data (37).

The association of heterogeneity with time-to-event outcomeswas additionally evaluated with HRs estimated from univariableand multivariable Cox proportional hazards (PH) regressionmethods. Continuous variables with right-skewed distributionswere transformed by log2(x þ 1) to establish a normal distribu-tion. The pretherapy features evaluated for the multivariable CoxPH models included line of therapy (2, 3, or 4þ, as factors),visceralmetastases (present vs. not), pretherapy PSA (continuous,log2þ1), LDH(continuous, log2þ1), ALK (continuous, log2þ1),Hgb (continuous), ALB (continuous), patient age (continuous),CTC/mL (continuous, log2 þ1), Shannon index (continuous),

Figure 1.

The distribution of patient samples in the CTC contributing cohort and clinical association cohort. CONSORT diagram showing the breakdown of patientsamples analyzed for this study. Displayed from top to bottom is the total number of samples collected, samples included for CTC digital pathology forunsupervised clustering and phenotypic assessment (CTC Contributing Cohort), and the subset of samples used for the clinical association cohort by therapyclass administered and line of therapy.

Scher et al.





and therapy class (taxane vs. ARSI). These pretherapy featureswereindividually assessed for correlation with Shannon index (Sup-plementary Fig. S3).

Visceral metastases, patient age, and ALB were excluded fromthe final model by using a stepwise selection method based onunivariate proportional hazards (Schoenfeld residuals P > 0.05)and univariate significance with outcome (P < 0.05). Included inthemodels is the interaction between Shannon index and therapyadministered. CTC enumerationwas also tested for an interactionwith therapy and heterogeneity index because of its correlation toShannon index. All earlier line samples from patients who con-tributedmultiple pretherapy samples but ultimately expired wereright censored. In addition, the robust sandwich estimate for thecovariance matrix was implemented for all Cox PH models tocorrect for possible underestimation of variance (38). All statis-tical tests were two-sided and performed at the 5% significancelevel. Data consolidation was conducted using KNIME. Statisticalanalyses utilized theRpackages: "survival," "stats," "sm," "vegan,"and "maxstat." Graphical representations were generated withR packages: "ggplot2," "gridExtra," "scales," "survminer," and"ggthemes."

CTC isolation, genome amplification, and next-generationsequencing

Methods for CTC relocation, picking, and sequencing havebeen described previously (39, 40). In brief, individual CTCswere relocated and recovered from assayed slides using theEppendorf TransferMan NK4 micromanipulator. Single-cellwhole genome amplification (WGA) was performed using theSeqPlex enhancedDNAamplification kit (Sigma). Shotgun librar-ies were constructed from 100 ng of WGA material using the

NEBNext Ultra DNA Library Prep Kit and sequenced to approx-imately 0.3� depth by 2 � 150 bp PE sequencing.

Genome wide copy number variation (CNV) analysis wasperformed using the Epic Sciences single-cell CNV analysis pipe-line. FASTQ files were aligned to hg38 human reference genomefrom UCSC Genome database. BAM files were filtered for MAPQ30 reads, followed by two separate analyses for genome-wideprofiling (pipeline 1) and individual gene copy number changesdetermination (pipeline 2). Pipeline 1: Hg38 human genomewasdivided into approximately 3,000 1M bp bins and counted acrossbins for each cell. Read counts per bin were normalized againstWBC controls, and the circular binary segmentation algorithm(R Bioconductor package "DNAcopy") was used to segment DNAcopy number data (log2-normalized ratio, sample/reference) andidentify abnormal copy number. Pipeline 2: reads were countedfor each gene and for each sample, and normalized against thetotal sequencing reads for the particular sample. Normalizedreads were compared with reference WBCs and z-scores werecalculated for each gene. Z-score of > 3 and < �3 are used assignificant cutoff for calling gene gain or loss.

ResultsClinical characteristics of the patient population

A total of 179 unique patients contributed 319 samples prior tostarting (221 pretherapy) or while receiving (98 on-therapy)treatment with an ARSI or taxane therapy. Two cohorts of patientsamples were studied: a CTC phenotypic analysis cohort (CTCcontributing cohort), and a clinical association cohort that is asubset of the former (see Patients and Methods, Table 1; Fig. 1).The clinical association cohort included only the pretherapy

A CTC Iden�fica�on and digital pathology

CTCs Segmented: DAPI, CK &

AR channels

Protein Features

CK cRatio (protein expression)AR cRatio (protein expression)

Digital Pathology Features

Nuclear SolidityCytoplasmic Solidity

Nuclear EntropyNuclear / Cytoplasmic Ratio

NucleoliCK Speckles

Nuclear SpecklesNuclear Circularity

Cytoplasmic CircularityNuclear Area (um2)

Cytoplasmic Area (um2)Nuclear Convex Area (um2)

Cytoplasmic Convex Area (um2)Nuclear Major Axis (um)

Cytoplasmic Major Axis (um)Nuclear Minor Axis (um)

Cytoplasmic Minor Axis (um)

Categorical Features

CTC in cluster statusCK Positivity (analytical)AR Positivity (analytical)

Feature extrac�on

Example CTC feature: nuclear area

High nuclear area: 201.3 um2

Composite DAPI CK CD45 AR

Low nuclear area: 30.9 um2

Composite DAPI CK CD45 AR

Single-cell features

F1F2

F3

ShannonIndex:

Cell subtype heterogeneity

analysis

CTC Contribu�ng cohort: unsupervisedclustering of features into phenotypically

similar CTC subtypes

Shannon index calculated forClinical associa�on cohort

High heterogeneity example: high entropy of different

subtypes present

Low heterogeneity example: low entropy of different

subtypes present

Associa�on of CTC heterogeneity topa�ent outcomeB

Single-cell features

Cell

1Ce

ll 2

Cell

3Ce

ll n

Single-cell phenotypes

Figure 2.

CTC and clinical association analysis overview. Shown are schematics for CTC detection and digital pathology analysis on single cells (A), as well asgeneration of patient-level quantification of phenotypic heterogeneity by Shannon index (B).






samples from patients about to start a second- or later line ofsystemic therapy for mCRPC.

Mathematical phenotype identification in pretherapy andon-therapy samples

CTC phenotypic heterogeneity was evaluated using the Shan-non index, which measures the entropy of species or, in this case,the phenotypically defined CTC subtypes (Fig. 2B)

Unsupervised clustering of all of the single-cell features (Fig.4A) from the 9,225 CTCs identified from the 319 samples in theCTC contributing cohort revealed a k of 15 (Supplementary Fig.S2B), which was used to "define" or classify individual cells into15 phenotypic subtypes ("A"–"O") where the cells were moresimilar tomembers of the subtype to which they were categorizedrelative to the other 14 subtypes (Fig. 3). Each phenotypic subtyperepresents a unique signature of the single-cell features (Fig. 4A).Each patient sample was then evaluated for the diversity of thedefined CTC subtypes present in a blood draw, and quantifiedusing the Shannon index (see Patients and Methods, Fig. 4B).Examples of CTCs from low Shannon index (Fig. 4C) and highShannon index (Fig. 4D) samples are shown.

Intersample Shannon index is related to overall survival ofARSI, but not taxanes

The relationshipbetween Shannon index (phenotypic entropy)andOS following treatmentwith a pathway-specific targeted ARSI

or taxane chemotherapy was analyzed on a continuous basis withestimates of median survival by Shannon index (see Patientsand Methods) separated by therapy administered (ARSI ortaxanes; Fig. 5A). While median survival estimates remainedconstant for patients on taxanes with respect to Shannon index,the median survival of patients on ARSI was longer than ontaxanes with comparable Shannon index at the low end. Incontrast, patients with high Shannon index on ARSI had shortermedian survival than those on taxanes with comparable Shannonindex on the high end. This relationship was further visualizedwith Kaplan–Meier plots for each therapy class with patientpopulations dichotomized by the crossover point (Fig. 5A,arrow). Patients going onto ARSI (Fig. 5B) had a large differencein survival between "high" and "low" Shannon index samples(median OS: 8.8 months vs. 28.1 months, P ¼ 0.0015), whereaspatients going onto taxanes (Fig. 5C) did not (11.4 months vs.12.9 months, P ¼ 1).

Multivariate analysesTo correct for potential imbalances in the demographics of the

pre-ARSI and pretaxane patient samples due to greater use of theformer in the second line and the latter in the third- or later linesetting, we evaluated survival in the context of a multivariate CoxPH model utilizing known prognostic factors (see Patients andMethods). Integrated into the Cox PH model was the interactionbetween CTC phenotypic entropy (Shannon index) and therapy

Cell type Distinguishing phenotypic featuresA Low cytokeratin, no AR, huge cell sizeB High CK expression, AR overexpression, large cytoplasmC No AR expression, large cytoplasmD Very small cell and nuclear size, high CK expressionE Distinctly high AR expression, high nuclear entropyF High CK expression, AR expression, frequently in histological cluster of 2 CTCsG Very high CK expression and frequent high AR expressionH Low AR expression, spindly cell shapeI Low CK expression, Low AR expression, frequently found in histological CTC clustersJ Very small cell size, high n/c ratioK Huge nucleus, high n/c ratioL Very small cell, frequent AR overexpression, high n/c ratioM Huge nuclear size, high nuclear entropyN Non-circular cell shapeO Low AR expression, high nuclear entropy, large cell size

Cell type A B C D E F G H I J K L M N O

B

A

Figure 3.

Phenotypic cell types resulting from unsupervised clustering. A, Shorthand description of each cell type's distinguishing features. B, Example images ofeach cell type. Blue, DAPI (DNA); red, cytokeratins; green, CD45; white, AR. Note that AR signal indicates AR protein overexpression.

Scher et al.





class (taxanes vs. ARSI). CTC enumeration was also considered asan interaction term with therapy class (see Materials and Meth-ods), and while there was an interaction (data not shown), theinteraction between Shannon index and therapy class was stron-ger. Of note, there remained a significant interaction betweenincreasing Shannon index and higher risk of death on ARSIrelative to taxanes (HR, 2.48; 95% CI, 1.22–5.03; P ¼0.0119; Fig. 5D and E), even when adjusting for pretherapyprognostic factors. To assess the possibility that the observedtreatment interaction effects are being driven by low Shannonindex due to inclusion of samples with 0 or 1 CTCs, we created aseparate model in which we excluded these samples and keptthose from the clinical association cohort with a Shannon indexgreater than 0. The same trends as previously described wereupheld (Supplementary Fig. S4A and S4B).

The Pleomorphism indexThe relationship between phenotypic heterogeneity, drug class,

and OS was additionally evaluated with an alternate measure ofintrapatient phenotypic variance: the Pleomorphism index (seeSupplementary Methods). Similar to the Shannon index, patientswith low Pleomorphism index had longer survival on ARSI, andpatients with high Pleomorphism index had longer survival ontaxanes, observed as a continuous marker in univariate and

multivariate settings (Supplementary Figs. S5A and S5B andS6A–S6E).

Heterogeneous genomic profiles identify frequent subclonaldrivers of ARSI resistance observed in high phenotypicheterogeneity samples

Randomly selected CTCs from 17 patient samples, each with10 or more CTCs, were single-cell sequenced to assess intra-sample genomic heterogeneity with CNVs. From a time andresource perspective, it is currently impractical to individuallysequence all CTCs in each sample to inform clinical decisions.However, we still sought to explore genomic heterogeneity atthe single-cell level in a subset of samples using these researchtools.

Shown are two examples of patient samples with multipledistinct genomic profiles with unique CNV patterns (Fig. 6A andB). One patient (Fig. 6A) had 22 CTCs sequenced and appears tohave two major genomic patterns (I and II). Three sub-patternswere observed in pattern II with additional chromosome Y loss ora chr5q deletion, indicated by the red circles in Fig. 6A. The secondpatient (Fig. 6B) had 62 CTCs sequenced and fourmajor genomicpatterns identified (I–IV). Pattern II had many chromosomalbreakpoints and alterations, including chr8p loss and 8q gain,which commonly occur in prostate cancer, and III and IV had

Categorical

Pre-2nd Line Samples Pre-3rd Line Samples Pre-4th+ Line Samples

C DCTCs from Low Shannon index sample CTCs from High Shannon index sample

Nuclear and cytoplasmic size, as well as textures, are fairly consistent between CTCs

Nuclear and cytoplasmic size, shape, and textures are diverse between CTCs

A CTC Subtypes by phenotypic features

Protein Digital pathology

B

CTC Phenotypic

features

Shannon index of CTC phenotypic entropy

TypeA/mL

2.0

1.5

1.0

Sh

ann

on

Ind

ex (

no

un

its)

0.0

TypeB/mL

TypeC/mL

TypeD/mL

TypeE/mL

TypeF/mL

TypeG/mL

TypeH/mL

TypeI/mL

TypeJ/mL

TypeK/mL

TypeL/mL

TypeN/mL

TypeM/mL

TypeO/mL

Figure 4.

CTC phenotypic features, cell subtype classifications, and Shannon index of CTC phenotypic entropy in patient samples. A, CTC phenotypic features includedin this analysis and unsupervised clustering of the CTC phenotypic features identified across all CTCs in the CTC Contributing Cohort (n ¼ 9,225) wasused to categorize CTCs into 15 phenotypic subtypes ("A"–"O"). Shown is a heatmap of mean individual cell features per phenotypic subtype. High, red; low,blue. B, Heatmap of CTC phenotype densities detected per patient sample, organized by line of therapy. The bar plot above shows the resulting Shannonindex by samples from the observed intra-sample diversity of CTC phenotypes. C, An example of CTCs from a low Shannon index sample. D, An exampleof CTCs from a high Shannon index sample.






single chromosome loss or gain. This patient also had nineadditional CTCs that did not fit any of the four major genomicpatterns identified in Fig. 6B, each instead representing its ownunique pattern. These samples also had high Shannon indexscores of phenotypic heterogeneity, and are provided as examplesof the diverse genomic profiles observed in patient samples. Theyare not intended to offer correlations between phenotypic andgenotypic heterogeneity. Properly powered comparisons betweenphenotypic and genotypic heterogeneity are planned for futurework.

A limitation of this analysis is the reliance on CNVs, as thesingle-cell sequencing technique utilized does not measuremuta-tions or translocations that could reveal additional intercellulardiversity. For example, common driver mutations in prostatecancer include alterations like SPOP mutations, AR mutations,and ERG rearrangements, which were not seen with these CNVanalyses but very well could exist in the flat genomes shown inPattern I of Fig. 6A and B.

A broader view of CTCs sequenced from high phenotypicheterogeneity samples showed varying degrees of subclonal gainor loss of key alterations known to drive resistance to ARSI (Fig.6C). No patients were dominated by what is amore-or-less clonalgenotype (i.e., white or black, but not gray tiles). Instead, mostpatients hadmany "gray" areas representing subclonal alterations

of other drivers, suggesting that a single dominant clone was notidentified or not present across all CTCs.

Certain alterations appeared to be more clonal than others,such as MYC, which when amplified, appeared in high propor-tions of CTCs within a patient sample. Other alterations, such asAR amplification, appeared frequently as subclones. These obser-vations are consistent with intrapatient tumor clonality analyses(41), where MYC and AR amplifications identified in tissuebiopsies were primarily truncal and subclonal, respectively, in acohort of 10 autopsies. In contrast, the analyses reportedherewereconducted on cells from living patients, with the potential to useresults to inform clinical decision making.

DiscussionDeveloping biomarkers of heterogeneity to guide treatment

selection and improve patient outcomes is an unmet medicalneed. Our objective was to develop a quantitative CTC hetero-geneity biomarker assay that has achieved the level of methodvalidation to be fit-for-purpose (13) of exploring its relation-ship to survival in mCRPC patients treated with an ARSI ortaxane therapy. To do so, we retrospectively applied the Shan-non index to prospectively collected and clinically annotatedpatient samples about to start a second or later line of

B C

AMul�variable cox propor�onal hazard analysis of predictors of OS

Effect P-value HR (95% CI)Line of Therapy (factor) 0.488 1.11 (0.822 – 1.51)

PSA Pre-Therapy (con�nuous, log2 + 1) 0.603 1.03 (0.917 – 1.17)LDH Pre-Therapy (con�nuous, log2 + 1) 0.00780 1.74 (1.16 – 2.61)

Alkaline Phosphatase Pre-Therapy (con�nuous, log2 + 1) 0.138 1.21 (0.942 – 1.55)Hemoglobin Pre-Therapy (con�nuous) 0.00879 0.788 (0.660 – 0.942)

CTC/mL Pre-Therapy (con�nuous, log2 + 1) 0.0220 1.33 (1.04 – 1.70)Shannon Index Pre-Therapy (con�nuous) 0.0470 0.440 (0.195 – 0.989)

Therapy (ARSI vs. Taxane) 0.173 0.475 (0.162 – 1.38)Interac�on : Shannon index (con�nuous) &

Therapy (ARSI vs. Taxane)0.0119 2.48 (1.22 – 5.03)

D

E

0 .2 5 0 .5 1 2 4 8

Treatment-specific hazards of death (OS)

HR (95% CI)

Favors ARSI Favors Taxanes

Heterogeneity& therapy

All samples

Heterogeneity : therapy interac�on: Mul�variable Cox PH modelComparison HR (95% CI)

Heterogeneity Status & therapy

Therapy (ARSI vs. Taxane) 0.475 (0.162 – 1.38)Interac�on : Shannon index (con�nuous)

& Therapy (ARSI vs. Taxane)2.48 (1.22 – 5.03)

OS by Shannon index and therapy class

Shannon index (no units)

OS: 2nd+ Line Pre-ARSI Samples

Number at risk by time Number at risk by time

Su

rviv

al p

rob

abili

ty

Su

rviv

al p

rob

abili

ty

Months survival Months survival

Months survival

Median survival (Months): 11.4 vs. 12.9HR (Logrank): 1.38 (0.72 to 2.65)

P (Logrank, Maximal) = 1

Median survival (Months): 8.8 vs. 28.1HR (Logrank): 3.84 (2.11 to 7)

P (Logrank, Maximal) = 0.00151

Months survival

Med

ian

su

rviv

al (

mo

nth

s)

OS: 2nd+ Line Pre-Taxane Samples

Figure 5.

The degree of inter-sample Shannon index is related to overall survival of ARSI, but not taxanes. A, The relationship between degree of heterogeneity(Shannon index, x-axis) and overall survival (y-axis) is shown, along with nonparametric kernel estimates of median survival. Colors represent treatmentreceived after pretherapy draw. O, patient alive at last observation; X, patient died at time indicated. Overall survival is alternately visualized with Kaplan–Meier plotsfrom patients starting ARSI (B) and taxanes (C), with survival curves dichotomized with the survival crossover point from A, indicated with an arrow.Individual covariates were tested for additive power to predict overall survival using a Cox proportional hazards (PH) model. D, The resulting P values, HRs,and 95% confidence intervals. E, The interaction of therapy and heterogeneity integrated into the multivariate Cox PH model. The forest plot shows HRsand 95% confidence intervals.

Scher et al.





treatment. The analysis showed that patient survival on taxaneswas unrelated to the degree of heterogeneity, while lowerheterogeneity scores were associated with a longer mediansurvival on ARSI relative to taxanes, and higher heterogeneitywith shorter median survival on ARSI relative to taxanes. Asignificant interaction was also observed between the Shannonindex (Fig. 5D and E) and therapy class in multivariate modelscorrecting for potential intra-cohort imbalances. The resultssupport the hypothesis that CTC phenotypic heterogeneitymeasured prior to starting systemic therapy associates withdifferential outcomes on pathway-specific hormonal agents butnot on non-pathway–specific chemotherapy.

As noted by a recent heterogeneity focus group, there are nostandard methods to quantify heterogeneity, or consensus onwhether the method to evaluate heterogeneity should be genetic,transcriptomic, phenotypic, epigenetic, or a combination thereof(42). All may prove relevant. It is critical to develop and test theassociation of a reproducibly measurable heterogeneity biomark-er(s) with clinical outcomes: the context of use for which thebiomarker result will be used to inform treatment decisions in theclinic. The outcome explored here was OS.

In this analysis, we expanded upon previously validateddigital pathology features that were components of FDA-

cleared devices for single-cell characterization with provenclinical utility for detecting cervical cancers from cytology Papsmears and applied these same tools to characterize CTCsphenotypically. Relative to single-site biopsies, CTCs can orig-inate from multiple tumor sites within a patient, and provide amore global picture of an individual patient's cancer. Incontrast to single-site biopsies, the acquisition of CTCsthrough phlebotomy poses minimal risk to the patient, andis amenable to repeated sampling to monitor disease evolu-tion over time.

Critical to the assessment of intrapatient heterogeneity in CTCsis a CTC detection method that does not rely on pre-enrichmentby epitope, size, or shape. These positive selection methods oftenmiss critical CTC subtypes and do not consistently assess cellfeatures essential to the analysis of heterogeneity. The CAP-accre-dited and CLIA-certified Epic Sciences platform used here hasundergone the necessary analytic validation (accuracy, linearity,specificity, and intra/interassay precision) for CTC detection andenumeration (26) and has been previously demonstrated todetect a wide range of CTC phenotypes (32, 33, 40, 43). Theseinclude rare CTCs that are negative for epithelial markers withmalignant genomics (40, 43) and CTCs that are smaller thanWBCs (32, 43).

A

B PPattern I:39 CTCs

Pattern II:7 CTCs

Pattern III:4 CTCs

Pattern IV:3 CTCs

Pattern I:8 CTCs

Pattern II-a:5 CTCs

Pattern II-b:8 CTCs

Pattern II-c:1 CTC

C

1.00

TP53 loss

RB1 loss

PTEN loss

BRCA2 loss

BRCA1 loss

ATM loss

PTK2 gain

MYCN gain

MYC gain

CYP17A1 gain

AURKA gain

AR gain

AKT3 gain

0049

1200

4166

0057

81

0026

7500

8559

0109

63

0012

8800

7931

0054

9200

5575

0095

7400

8164

0056

1900

5229

0128

8200

5374

0064

92

Subclonality of genomic alterations present in high phenotpic heterogeneity samples

0.750.500.25

Samples IDs

0.00

Figure 6.

Heterogeneous genomic profiles observed in high phenotypic heterogeneity samples. Copy number variation plots are displayed in A and B, wherechromosomes are shown left to right, 1 to 22, X and Y, odd as red, even as blue. Log2-normalized copy number ratio (sample/reference) is indicated on they-axis for two patient samples that exhibited high CTC phenotypic heterogeneity (by both Shannon index and Pleomorphism index). A, Red circles indicatechromosome Y loss and a chr5q deletion that were observed in subpatterns of genomic pattern II. All cells identified for this patient are visualized here.B, The four dominant profiles found across 53 of the 62 individual cells sequenced are illustrated. C, Heatmap of degree of subclonality of genomicalterations present in high phenotypic heterogeneity samples, with tile darkness indicating the proportion of single-cell whole genome sequenced CTCs withthe indicated genomic alteration per sample. Black tiles, complete clonality of a given alteration; white tiles, complete absence; gray tiles, degrees ofsubclonality.






The approach used to analyze the digital pathology featuresrepresents a novel confluence of techniques adapted from diversefields. The z-score transformation and principal component anal-ysis is used in gene array applications to avoid redundancy anddistill distinct sources of variance within the data (44). K-meansclustering (45) is a machine learning technique with applicationsincluding, but not limited to, e-commerce, defense, ecology, andastronomy (46). Shannon index is a measure of species diversityfrequently utilized in ecology research (47), where quantificationandmonitoring of species biodiversity and evolutionary events iscommon. Shannon index increases with both the number ofunique species present and the evenness in species distribution.When applied to CTCs, it indicates the level of uncertainty topredictwhat phenotypic subtypes ofCTCs are going tobedetectednext: the higher the uncertainty (entropy), the higher the hetero-geneity. We additionally developed a simpler, alternate measureof CTC phenotypic heterogeneity within patient samples thatdoes not require unsupervised clustering, the PleomorphismIndex (see Supplementary Methods; Supplementary Fig. S5A andS5B), performed identical survival analyses, and observed similarresults with respect to phenotypic heterogeneity, drug classadministered, and OS (Supplementary Fig. S6A–S6E). Theseresults suggest that there are likely multiple ways to modelheterogeneity, but the underlying phenomena observed is con-sistent with our central hypothesis.

A limitation of this approach is the potential lack of utility inthe setting of low or no CTC counts. Recognizing this, as stated inPatients and Methods, our focus was the second or later line oftherapy setting where the frequency of detection and individualCTC counts are higher, and the decision to select one form oftherapy versus another is more critical. In our analyses, sampleswith 1 or 0 CTCs were assigned a heterogeneity score of 0 andincluded in our OS association analyses. However, despite thesepotential shortcomings, Shannon index does demonstrate theability to differentially predict OS by therapy class, even whentotal CTC count is included as a covariate in multivariate models.A supplemental sub-cohort analysis that excluded samples withheterogeneity scores of 0 yielded similar results as well (Supple-mentary Fig. S4A and S4B). Ultimately, it may be concluded thatthe determination of a heterogeneity score (for which there iscurrently no unified definition, units, or reference range deter-mined by the field; ref. 42) might have a requisite minimalnumber of CTCs for optimal quantification.

Another emerging technique that can infer heterogeneity on thegenomic level is cell-free DNA (cfDNA; ref. 48). cfDNA is severalsteps removed from the cells of origin and represents a homog-enizedmixture, or pool, of tumor DNA and nontumor DNA frommultiple cells, which poses additional technical hurdles whentrying to distill features of entropy or variance within a patientsample. cfDNA can also include fragments of apoptotic or deadcells that could represent parts of tumor that were sensitive to thetreatment and consequently eliminated (49). In part due to thesereasons, we excluded the analysis of apoptotic CTCs and onlyscored intact, whole CTCs.

In addition to clinical utility, the abilities to provide results intime to inform clinical decisions and at a reasonable cost areessential for the use of a biomarker in clinical practice. Image-based phenotypic measures of tumor heterogeneity meet thisrequirement. Previously, our group analytically validated a sin-gle-CTC whole genome sequencing technology (39) to potential-ly assess intrapatient genomic heterogeneity. While promising,

this approach is not scalable with available sequencing technol-ogies when considering reproducibility, turnaround time, andcost. Acknowledging these caveats, we performed whole-genomesequencing on CTCs from a subset of samples with at least 10evaluable CTCs. The results showed frequent subclonal altera-tions associatedwith ARSI resistance (Fig. 6C), consistent with thepoor outcome of patients on ARSI who had high phenotypicheterogeneity determined with the Shannon index reported here-in (Fig. 5). A greatly expanded cohort of single-cell sequencingresults would be required for meaningful statistical analyses ofgenomic heterogeneity. However, these data suggest that sampleswith high phenotypic heterogeneity have heterogeneous genomicprofiles as well.

In previous work, we studied an analytically validated IHCassay for AR-V7 as a treatment-selection biomarker for the samecontext (50), and found that 100% of patients with nuclear-localized AR-V7 protein in their CTCs had poor responses onARSI and lived longer on taxanes (50). The assay, however, did notidentify all patients who would have poor response on ARSI, asthe frequency of AR-V7 detection was only 20% among patientswhohad less than 50%PSAdecline by 12weeks. Thus,most of thepatientswhodid not respond toARSIwere AR-V7–negative. In thecurrent study, patients with an increasing Shannon index had anincreasing risk of death on ARSI relative to taxanes (Fig. 5). Thissuggests that patients with low heterogeneity scores would bemore likely to survive longer on an ARSI over a taxane, andpatients with high heterogeneity scores would likely survivelonger on a taxane than an ARSI. Specifically in the context ofAR-V7–negative patients, thismeasure could have the potential tomore reliably inform treatment selection at this clinical decisionpoint. Comparisons of AR-V7 to heterogeneity indices areplanned for future studies.

CTCheterogeneity biomarkersmay also be useful in the clinicaldevelopment of novel therapies, asmany trials evaluate cohorts ofheavily pretreated patients who might exhibit high levels ofheterogeneity that could portend resistance to targeted therapeu-tics. Such patients may be more suitably offered participation instudies of novel drug combinations, or combinations that includea cytotoxic drug more likely to affect diverse cell populations.

While the results reported here have the potential for broadimplications, this study was designed to test a hypothesis thatrepresents a first step in the clinical evaluation of an analyticallyvalid, quantitativemeasurement of heterogeneity that could serveas a therapy-guiding biomarker to inform the choice between anARSI or taxane for mCRPC patients starting a second- or later linetherapy. The gold standard for evaluating the clinical utility of apredictive biomarker is through randomized, interventionaltrials. A limitation of observational studies like this one is thepossibility of unobserved confounding factors. An additionallimitation of our study is the lack of an external validation cohort.Definitive clinical utility of these heterogeneity biomarkers willrequire a sequence of trials, analogous to the development of adrug focused on a context of use. Further clinical validation inseparate, larger cohorts is planned and assessment of longitudinaland kinetic changes over time in response to therapy are ongoing,both within the context of mCRPC clinical decisions and outsideof prostate cancer.

Disclosure of Potential Conflicts of InterestH.I. Scher reports receiving a commercial research grant from Janssen and is a

consultant/advisory board member for Astellas, Janssen Research and

Scher et al.





Development, LLC, andMedivation. Y.Wang has ownership interest (includingpatents) in Epic Sciences. D. Lu is a senior scientist at Epic Sciences. G. Hellerreports receiving a commercial research grant from Janssen Diagnostics. Nopotential conflicts of interest were disclosed by the other authors.

Authors' ContributionsConception and design: H.I. Scher, R.P. Graf, M. Landers, R.V. DittamoreDevelopment of methodology: H.I. Scher, R.P. Graf, A. Jendrisak, Y. Wang,S. Greene, M. Fleisher, M. Landers, R.V. DittamoreAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): N. Schreiber, B. McLaughlin, S. Greene, R. Krupa,D. Lu, H.A. Vargas, M. Fleisher, M. LandersAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis):H.I. Scher, R.P. Graf, A. Jendrisak, Y.Wang, S. Greene,H.A. Vargas, M. Fleisher, M. LandersWriting, review, and/or revision of the manuscript: H.I. Scher, R.P. Graf,N. Schreiber, B. McLaughlin, Y. Wang, P. Bamford, H.A. Vargas, R.V. DittamoreAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): R.P. Graf, N. Schreiber, B. McLaughlin, Y. Wang,J. Lee, R. Krupa, D. Lu, J. Louw, L. Dugan

Study supervision: H.I. Scher, R.P. Graf, J. Louw, L. Dugan, M. Landers,G. Heller, R.V. Dittamore

AcknowledgmentsWe would like to thank the patients and their families for taking part in this

study, and the clinical and laboratory staff at MSKCC and Epic Sciences.

Grant SupportThis work was supported by the NIH/NCI P50-CA92629 SPORE in Prostate

Cancer, NIH/NCI Cancer Center Support Grant P30-CA008748, and Depart-ment of Defense Prostate Cancer Research Program (PC121111 andPC131984). Prostate Cancer Foundation Challenge Award and David H. KochFund for Prostate Cancer Research were used to support the design and conductof the study.

The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received May 5, 2017; revised July 7, 2017; accepted August 9, 2017;published OnlineFirst August 17, 2017.

References1. Pestrin M, Salvianti F, Galardi F, De Luca F, Turner N, Malorni L, et al.

Heterogeneity of PIK3CA mutational status at the single cell level incirculating tumor cells from metastatic breast cancer patients. Mol Oncol2015;9:749–57.

2. Paguirigan AL, Smith J, Meshinchi S, Carroll M, Maley C, Radich JP. Single-cell genotyping demonstrates complex clonal diversity in acute myeloidleukemia. Sci Transl Med 2015;7:281re2.

3. Roberts NJ, Norris AL, Petersen GM, Bondy ML, Brand R, Gallinger S, et al.Whole genome sequencing defines the genetic heterogeneity of familialpancreatic cancer. Cancer Discov 2016;6:166–75.

4. Skoulidis F, Byers LA, Diao L, Papadimitrakopoulou VA, Tong P, Izzo J,et al. Co-occurring genomic alterations define major subsets of KRAS-mutant lung adenocarcinoma with distinct biology, immuneprofiles, and therapeutic vulnerabilities. Cancer Discov 2015;5:860–77.

5. Brastianos PK, Carter SL, Santagata S, Cahill DP, Taylor-Weiner A, Jones RT,et al. Genomic characterization of brain metastases reveals branchedevolution and potential therapeutic targets. Cancer Discov 2015;5:1164–77.

6. Russo M, Siravegna G, Blaszkowsky LS, Corti G, Crisafulli G, Ahronian LG,et al. Tumor heterogeneity and lesion-specific response to targeted therapyin colorectal cancer. Cancer Discov 2016;6:147–53.

7. Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E,et al. Intratumor heterogeneity and branched evolution revealed by multi-region sequencing. N Engl J Med 2012;366:883–92.

8. Burrell RA, McGranahan N, Bartek J, Swanton C. The causes and conse-quences of genetic heterogeneity in cancer evolution. Nature2013;501:338–45.

9. Hiley C, de Bruin EC,McGranahan N, Swanton C. Deciphering intratumorheterogeneity and temporal acquisition of driver events to refine precisionmedicine. Genome Biol 2014;15:453.

10. McGranahanN, Favero F, de Bruin EC, BirkbakNJ, Szallasi Z, SwantonC.Clonal status of actionable driver events and the timing ofmutational processes in cancer evolution. Sci Transl Med 2015;7:283ra54.

11. Alberter B, Klein CA, Polzer B. Single-cell analysis of CTCs with diagnosticprecision: opportunities and challenges for personalized medicine. ExpertRev Mol Diagn 2016;16:25–38.

12. Gorges TM, Kuske A, Rock K, Mauermann O, Muller V, Peine S, et al.Accession of tumor heterogeneity by multiplex transcriptome profiling ofsingle circulating tumor cells. Clin Chem 2016;62:1504–15.

13. Cummings J, Raynaud F, Jones L, Sugar R, Dive C. Fit-for-purpose bio-marker method validation for application in clinical trials of anticancerdrugs. Br J Cancer 2010;103:1313–7.

14. Swanton C. Intratumor heterogeneity: evolution through space and time.Cancer Res 2012;72:4875–82.

15. Crockford A, Jamal-Hanjani M, Hicks J, Swanton C. Implications ofintratumour heterogeneity for treatment stratification. J Pathol 2014;232:264–73.

16. Mohler JL, Armstrong AJ, Bahnson RR, D'Amico AV, Davis BJ, Eastham JA,et al. Prostate Cancer, Version 1.2016. J Natl Compr Canc Netw 2016;14:19–30.

17. Ryan CJ, Smith MR, de Bono JS, Molina A, Logothetis CJ, de Souza P, et al.Abiraterone in metastatic prostate cancer without previous chemotherapy.N Engl J Med 2013;368:138–48.

18. Beer TM, Armstrong AJ, Rathkopf DE, Loriot Y, Sternberg CN, Higano CS,et al. Enzalutamide in metastatic prostate cancer before chemotherapy. NEngl J Med 2014;371:424–33.

19. SchraderAJ, BoegemannM,OhlmannCH, Schnoeller TJ, Krabbe LM,HajiliT, et al. Enzalutamide in castration-resistant prostate cancer patientsprogressing after docetaxel and abiraterone. Eur Urol 2014;65:30–6.

20. Gillessen S, Omlin A, Attard G, de Bono JS, Efstathiou E, Fizazi K, et al.Management of patients with advanced prostate cancer: recommendationsof the St Gallen Advanced Prostate Cancer Consensus Conference(APCCC) 2015. Ann Oncol 2015;26:1589–604.

21. Scher HI, Morris MJ, Stadler WM, Higano C, Basch E, Fizazi K, et al. Trialdesign and objectives for castration-resistant prostate cancer: updatedrecommendations from the prostate cancer clinical trials working group3. J Clin Oncol 2016;34:1402–18.

22. Seol H, Lee HJ, Choi Y, Lee HE, Kim YJ, Kim JH, et al. Intratumoralheterogeneity of HER2 gene amplification in breast cancer: its clinico-pathological significance. Mod Pathol 2012;25:938–48.

23. Mroz EA, Tward AD, Pickering CR, Myers JN, Ferris RL, Rocco JW. Highintratumor genetic heterogeneity is related to worse outcome in patientswith head and neck squamous cell carcinoma. Cancer 2013;119:3034–42.

24. Jamal-Hanjani M, Wilson GA, McGranahan N, Birkbak NJ, Watkins TBK,Veeriah S, et al. Tracking the evolution of non–small-cell lung cancer. NEngl J Med 2017;376:2109–21.

25. Navin NE. The first five years of single-cell cancer genomics and beyond.Genome Res 2015;25:1499–507.

26. Werner SL, Graf RP, LandersML, Valenta DT, SchroederM, Greene SB, et al.Analytical validation and capabilities of the epic CTC platform: enrich-ment-free circulating tumour cell detection and characterization. J Circu-lating Biomarkers 2015;4:3. doi: 10.5772/60725.

27. Wied GL, Bahr GF, Oldfield DG, Bartels PH. Computer-assisted identifica-tion of cells from uterine adenocarcinoma. A clinical feasibility study withTICAS. I.Measurements atwavelength530nm.ActaCytol1968;12:357–70.

28. Zahniser DJ, Oud PS, Raaijmakers MC, Vooys GP, Van de Walle RT.BioPEPR: a system for the automatic prescreening of cervical smears. JHistochem Cytochem 1979;27:635–41.

29. Tucker JH, Shippey G. Basic performance tests on the CERVIFIP linear arrayprescreener. Anal Quant Cytol 1983;5:129–37.






30. Almendro V, Kim HJ, Cheng YK, Gonen M, Itzkovitz S, Argani P, et al.Genetic and phenotypic diversity in breast tumor metastases. Cancer Res2014;74:1338–48.

31. Maley CC, Galipeau PC, Finley JC,Wongsurawat VJ, Li X, Sanchez CA, et al.Genetic clonal diversity predicts progression to esophageal adenocarcino-ma. Nat Genet 2006;38:468–73.

32. Beltran H, Jendrisak A, Landers M, Mosquera JM, Kossai M, Louw J, et al.The initial detection and partial characterization of circulating tumorcells in neuroendocrine prostate cancer. Clin Cancer Res 2016;22:1510–9.

33. Punnoose EA, Ferraldeschi R, Szafer-GlusmanE, Tucker EK,Mohan S, FlohrP, et al. PTEN loss in circulating tumour cells correlates with PTEN loss infresh tumour tissue from castration-resistant prostate cancer patients. Br JCancer 2015;113:1225–33.

34. Rodenacker K, Bengtsson E. A feature set for cytometry on digitizedmicroscopic images. Anal Cell Pathol 2003;25:1–36.

35. Wickman G, Julian L, OlsonMF. How apoptotic cells aid in the removal oftheir own cold dead bodies. Cell Death Differ 2012;19:735–42.

36. Scher HI, Jia X, de Bono JS, Fleisher M, Pienta KJ, Raghavan D, et al.Circulating tumour cells as prognostic markers in progressive, castration-resistant prostate cancer: a reanalysis of IMMC38 trial data. Lancet Oncol2009;10:233–9.

37. MIller R, Siegmund D. Maximally selected chi square statistics. Biometrics1982;38:1011–6.

38. Lin DY, Wei LJ. The robust inference for the proportional hazards model. JAm Stat Assoc 1989;84:1074–8.

39. Greene SB, Dago AE, Leitz LJ, Wang Y, Lee J, Werner SL, et al. Chromosomalinstability estimation based on next generation sequencing and single cellgenomewide copynumber variation analysis. PLoSOne2016;11:e0165089.

40. Anantharaman A, Friedlander TW, Lu D, Krupa R, Premasekharan G,Hough J, et al. Programmed death-ligand 1 (PD-L1) characterization of

circulating tumor cells (CTCs) and white blood cells (WBCs) in muscleinvasive and metastatic bladder cancer patients. J Clin Oncol 34;2s, 2016(suppl; abstr 446).

41. Gundem G, Van Loo P, Kremeyer B, Alexandrov LB, Tubio JM, Papaem-manuil E, et al. The evolutionary history of lethal metastatic prostatecancer. Nature 2015;520:353–7.

42. Alizadeh AA, Aranda V, Bardelli A, Blanpain C, Bock C, Borowski C, et al.Toward understanding and exploiting tumor heterogeneity. Nat Med2015;21:846–53.

43. McDaniel AS, Ferraldeschi R, Krupa R, Landers M, Graf R, Louw J, et al.Phenotypic diversity of circulating tumour cells in patients with metastaticcastration-resistant prostate cancer. BJU Int 2016 Aug 18. [Epub ahead ofprint]. doi: 10.1111/bju.13631.

44. Cheadle C, Vawter MP, Freed WJ, Becker KG. Analysis of microarray datausing Z score transformation. J Mol Diagn 2003;5:73–81.

45. Jain AK. Data clustering: 50 years beyond K-means. Pattern Recognit Lett2010;31:651–66.

46. Honarkhah M, Caers J. Stochastic simulation of patterns using distance-based pattern modeling. Math Geosci 2010;42:487–517.

47. Brocchieri L. Phylogenetic diversity and the evolution of molecularsequences. J Phylogenet Evol Biol 2015;3:e109. doi: 10.4172/2329-9002.1000e109.

48. Wyatt AW, Azad AA, Volik SV, Annala M, Beja K, McConeghy B, et al.Genomic alterations in cell-free DNA and enzalutamide resistance incastration-resistant prostate cancer. JAMA Oncol 2016;2:1598–606.

49. Tannock IF, Hickman JA. Limits to personalized cancer medicine. N Engl JMed 2016;375:1289–94.

50. ScherHI, LuD, Schreiber NA, Louw J, Graf RP, VargasHA, et al. Associationof AR-V7 on circulating tumor cells as a treatment-specific biomarker withoutcomes and survival in castration-resistant prostate cancer. JAMAOncol2016;2:1441–9.


Scher et al.




2017;77:5687-5698. Published OnlineFirst August 17, 2017.Cancer Res Howard I. Scher, Ryon P. Graf, Nicole A. Schreiber, et al. Metastatic Prostate CancerClinical Decisions between AR Signaling Inhibitors and Taxanes in Phenotypic Heterogeneity of Circulating Tumor Cells Informs

Updated version

10.1158/0008-5472.CAN-17-1353doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2017/08/16/0008-5472.CAN-17-1353.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/77/20/5687.full#ref-list-1

This article cites 49 articles, 13 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/77/20/5687.full#related-urls

This article has been cited by 2 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/77/20/5687To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-17-1353

http://cancerres.aacrjournals.org/content/suppl/2017/08/16/0008-5472.CAN-17-1353.DC1

http://cancerres.aacrjournals.org/content/77/20/5687.full#ref-list-1

http://cancerres.aacrjournals.org/content/77/20/5687.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/77/20/5687


Documents

Phenotypic Heterogeneity of Circulating Tumor Cells ...It has been hypothesized that the efﬁcacy of pathway-speciﬁc ... This is a retrospective analysis of a prospectively accrued