Supplementary Figure S1

B40 candidate genes based on single cell prostate CTC RNA-seq data and public databases:

High expression in CTCs and/or prostate cancer and low expression in WBCs (107 candidate ddPCR probes)

Testing of probes with qRT-PCR:Robust signal in prostate cancer cells and absence of signal in WBCs (22 genes (55%); 34 probes (32%))

Testing of probes based on ddPCR evaluation of blood processed with CTC-iChip :Absence of signal in healthy donor blood (11 probes (10%))

D

CTMPRSS2100% FAM

FAT165% FAM35% HEX

KLK240% FAM60% HEX

STEAP2100% HEX

Reaction 1KLK3100% FAM

HOXB1370% FAM30% HEX

AGR250% FAM50% HEX

FOLH1100% HEX

Reaction 2

FAM

Inte

nsity

FAM

Inte

nsity

HEX Intensity HEX Intensity

1 cell 3 cells 5 cells 10 cells 50 cells

Final selection of probes based on ddPCR evaluation of blood processed with CTC-iChip :Robust signal in blood with spiked cancer cells and mCRPC patient blood (8 probes (10%))

A

FAM

Inte

nsity

HEX Intensity

Blood CTC-iChip Intact CTCsMultiplex Droplet

Digital PCR

Bulk LNCaP cells

Supplementary Figure S1. (A) Schematic of d-CTC assay. Whole blood is processed using the CTC-iChip, resultingin a bulk product highly enriched for intact CTCs (labeled green) but also contains WBCs (labeled red) and RBCs(unlabeled). Multiplex droplet digital PCR detects lineage-specific and cancer-specific genes expressed in CTCs.(B) Schematic of procedure for identification and selection of candidate gene targets and droplet digital PCR primers/probes for d-CTC assay. (C) Top panels: theoretical modeling of multiplex d-CTC assay based on FAM/HEXfluorophore ratios used in probes. Bottom panels: multiplex d-CTC assay results inthe LNCaP prostate cancer cell line.(D) Pie charts showing relative distribution of gene signal detected using the d-CTC assay after CTC- iChip processingof varying numbers of LNCaP cells micromanipulated into healthy donor blood, as well as bulk LNCaP cell cDNA.

Supplementary Figure S2

Supplementary Figure S2. Graphs showing droplet digital PCR signal for each gene in 12 metastatic prostatecancer patients compared to 34 healthy male control subjects.

Supplementary Figure S3

A BSerum PSA vs CTCM Score

R =0.16P=0.049

2

Serum PSA vs KLK3 in CTCs

R =0.18P=0.038

2

C D

0 5 10 15 20 25 30

0.0

0.2

0.4

0.6

0.8

1.0

≥5 CTC/7.5mL<5 CTC/7.5mL

Months From Start of Therapy

OS

0 5 10 15 20 25 30

0.0

0.2

0.4

0.6

0.8

1.0

P<0.001

CTC enumeration (OS) CTC enumeration (R-PFS)

≥5 CTC/7.5mL<5 CTC/7.5mL

P<0.001

Months From Start of Therapy

R-P

FS

EOverall Survival (OS) Radiographic Progression-Free Survival (R-PFS)

F

Supplementary Figure S3. (A, B) Graphs of relationships between Prostate CTCM Score and serum PSA,and CTC droplet digital PCR KLK3 signal and serum PSA at the pre-treatment time point.(C, D) Kaplan-Meier curves for overall survival (OS) and radiographic progression-free survival (R-PFS) byCTC enumeration at pretreatment, determined through immunofluorescence staining and microscopy.(E, F) Tables showing univariate analysis of CTCM Score and individual genes, according to associationwith overall survival and radiographic progression-free survival in prospective study of mCRCPC patientsreceiving first-line abiraterone.

Supplementary Figure S4

E F

AR-V7High Low

Alive

Dead 2 4

0 16P = 0.071

Alive >

12mo (N

=16)

Dead at

12mo (N

=6)0.1

1

10

100

1000

AR

-V7

Tran

scrip

ts/m

L

AR-V7 - Overall Survival

P = 0.051

Alive >

12mo (N

=21)

Dead at

12mo (N

=6)

0.1

1

10

100

1000

HO

XB13

Tra

nscr

ipts

/mL

HOXB13 - Overall SurvivalP = 0.014

HOXB13High Low

Alive

Dead 6 0

7 14P = 0.006

A B AR-V7 (R-PFS)

C

+

+ + ++ +

+ + +

0.00

0.25

0.50

0.75

1.00

0 6 12 18 24Time (months)

PFS

Pro

babi

lity

20 11 5 1 02 0 0 0 0>14.73

<=14.73

0 6 12 18 24Time (months)

ARV7

Number at risk by time

+

+ + ++ +

+ + +

0.00

0.25

0.50

0.75

1.00

0 6 12 18 24Time (months)

PFS

Prob

abilit

y

21 11 5 1 01 0 0 0 0>850.32

<=850.32

0 6 12 18 24Time (months)TMPR

SS2.

ERG Number at risk by time

D TMPRSS2:ERG (R-PFS)+

++ +

++ + + +

+ +

0.00

0.25

0.50

0.75

1.00

0 6 12 18 24Time (months)

OS

Prob

abilit

y

21 16 7 3 01 1 1 0 0>850.32

<=850.32

0 6 12 18 24Time (months)TM

PRSS

2.ER

G Number at risk by time

TMPRSS2:ERG (OS)

+

+ ++ +

+ + + ++

+ +

0.00

0.25

0.50

0.75

1.00

0 6 12 18 24Time (months)

PFS

Pro

babi

lity

14 10 5 2 113 4 2 0 0>2.39

<=2.39

0 6 12 18 24Time (months)

HO

XB13

Number at risk by time

HOXB13 (R-PFS)

P=0.004HR=10.4

P=0.011HR=3.9

P=0.38 P=0.10

Supplementary Figure S4. (A) Kaplan-Meier curves for radiographic progression-free survival (R-PFS) by CTCHOXB13 status at pretreatment. (B) Kaplan-Meier curves for radiographic progression-free survival (R-PFS) byCTC AR-V7 status at pretreatment. (C, D) Kaplan-Meier curves for overall survival (OS) and radiographicprogression-free survival (R-PFS) by CTC TMPRSS2:ERG status at pretreatment. (E, F) Graphs showingquantitation of HOXB13 and AR-V7 CTC transcripts in mCRPC patients, according to survival status at 12 months after initiation of abiraterone therapy. 5 patients had pre-treatment data available for HOXB13 but not for AR-V7.

Supplementary Figure S5

B

C

Seminal Vesicle Invasion

Lymph Node Involvement

P=0.481 P=0.005 P=0.009 P=0.014

P=0.014 P=0.004 P=0.243 P=0.041

P=0.193 P=0.001 P=0.002 P=0.004

P=0.004 P=0.003 P=0.135 P=0.016

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000

AG

R2

sign

al in

CTC

s

AGR2

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

FAT1

sig

nal i

n C

TCs

FAT1

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000

FOLH

1 si

gnal

in C

TCs

FOLH1

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000

HO

XB13

sig

nal i

n C

TCs

HOXB13

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000

KLK

2 si

gnal

in C

TCs

KLK2

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000

KLK

3 si

gnal

in C

TCs

KLK3

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000ST

EAP2

sig

nal i

n C

TCs

STEAP2

No SVI (N=2

9)

SVI (N=5

)0.1

1

10

100

1000

10000

TMPR

SS2

sign

al in

CTC

s

TMPRSS2

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000

10000

AG

R2

sign

al in

CTC

s

AGR2

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

FAT1

sig

nal i

n C

TCs

FAT1

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000

10000

FOLH

1 si

gnal

in C

TCs

FOLH1

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000H

OXB

13 s

igna

l in

CTC

sHOXB13

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000

KLK

2 si

gnal

in C

TCs

KLK2

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000

10000

KLK

3 si

gnal

in C

TCs

KLK3

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000

STEA

P2 s

igna

l in

CTC

s

STEAP2

LN(-) (N

=30)

LN(+) (N

=4)

0.1

1

10

100

1000

TMPR

SS2

sign

al in

CTC

s

TMPRSS2

A

AGR2FAT1

FOLH1HOXB13

KLK2KLK3

STEAP2TMPRSS2

CTC L Score

HD

F1H

DF2

HD

F3H

DF4

HD

F5H

DF6

HD

F7

HD

M8

HD

M9

HD

M10

HD

M11

HD

M12

HD

M13

HD

M14

HD

M15

HD

M16

HD

M17

HD

M22

HD

M23

HD

M24

HD

M1

HD

M2

HD

M3

HD

M4

HD

M5

HD

M6

HD

M7

HD

M18

HD

M19

HD

M20

HD

M21

HD

M25

HD

M26

HD

M27

HD

M28

HD

M29

HD

M30

PrR

P1Pr

RP2

PrR

P3Pr

RP4

PrR

P5Pr

RP6

PrR

P7Pr

RP8

PrR

P9Pr

RP1

0

PrR

P11

PrR

P12

PrR

P13

PrR

P14

PrR

P15

PrR

P16

PrR

P17

PrR

P18

PrR

P19

PrR

P20

PrR

P21

PrR

P22

PrR

P23

PrR

P24

PrR

P25

PrR

P26

PrR

P27

PrR

P28

PrR

P29

PrR

P30

PrR

P31

PrR

P32

PrR

P33

PrR

P34

Female Male <50yo Male >50yo Low Risk Intermediate Risk High Risk

Healthy Donor Localized Prostate Cancer

log10(transcripts + 1)

0 2 4 6 80 2 4 6 8

Supplementary Figure S5. (A) Heatmap of droplet digital PCR CTC signal after whole transcriptome amplificationfor blood samples from healthy donor controls arranged by sex and age, and from patients with clinically localizedprostate cancer, arranged by D’Amico Risk Group. (B) Graphs showing droplet digital PCR signal for each gene inlocalized prostate cancer patients with and without pathologic seminal vesicle invasion identified at radicalprostatectomy. (C) Graphs showing droplet digital PCR signal for each gene in localized prostate cancer patients withand without pathologic lymph node involvement identified at radical prostatectomy.

Supplementary Figure S6

C

D

Grade Group 1 (Gleason 6)

Grade Group 2 (Gleason 3+4)

Grade Group 3 (Gleason 4+3)

Grade Group 4 (Gleason 4+4)

Grade Group 5 (≥Gleason 4+5)

0 2 4 6

CTCL Score (LOOCV) vs. Grade Group

Log10(CTCL Score + 1)

SVI(+) or LN(+)SVI(-) and LN(-)

0.5

1.0

1.5

0 2 4 6

log1

0(P

SA

+1)

SVI(+) or LN(+)SVI(-) and LN(-)

CTCL Score (LOOCV) vs. PSA

Log10(CTCL Score + 1)

3/33/31

0/0

3/3

1/3

2/28

ACTCL Score (LOOCV) vs. Gleason Score

B

Gleason Score

Log1

0(C

TCL S

core

+ 1

)

0

2

4

6

CTCL Score (LOOCV) and InvasionLo

g10(

CTC

L Sco

re +

1)

SVI(-) and LN(-) SVI(+) or LN(+)

0

2

4

6

6 7 ≤8

Supplementary Figure S6. (A) Box plots showing pre-operative leave-one-out cross validated (LOOCV) CTCL Scorein clinically localized prostate cancer patients according to microscopic SVI or pelvic LN involvement identified at thetime of radical prostatectomy. Dashed line represents a threshold of 2 standard deviations above the average CTCL

Score signal in healthy donor controls. (B) Box plots showing pre-operative LOOCV CTCL Score in clinically localizedprostate cancer patients according to Gleason score of diagnostic prostate biopsies. (C, D) Graphs of relationshipbetween pre-operative LOOCV CTCL Score and Serum PSA (upper panel) or Gleason score (lower panel). Red dotsrepresent patients who had SVI or pelvic LN involvement identified on pathologic examination of the radicalprostatectomy specimen. Blue dots represent patients who did not have SVI or LN involvement. Perpendicular dashedline represents a threshold of 2 standard deviations above the average CTCL Score signal in healthy donor controls.