stm.sciencemag.org/cgi/content/full/12/548/eabc4220/DC1

Supplementary Materials for

ARID1A mutation plus CXCL13 expression act as combinatorial biomarkers to

predict responses to immune checkpoint therapy in mUCC

Sangeeta Goswami, Yulong Chen, Swetha Anandhan, Peter M. Szabo, Sreyashi Basu, Jorge M. Blando, Wenbin Liu, Jan Zhang, Seanu Meena Natarajan, Liangwen Xiong, Baoxiang Guan, Shalini Singh Yadav, Abdel Saci, James P. Allison,

Matthew D. Galsky, Padmanee Sharma*

*Corresponding author. Email: padsharma@mdanderson.org

Published 17 June 2020, Sci. Transl. Med. 12, eabc4220 (2020)

DOI: 10.1126/scitranslmed.abc4220

The PDF file includes:

Fig. S1. CONSORT diagram for CheckMate275 and IMvigor210. Fig. S2. Loss of ARID1A enhances the immunogenicity of bladder tumor cells. Fig. S3. Patients harboring ARID1A gene mutation have higher TMB. Fig. S4. Analysis of TGFβ1 and PD-L1 expression in patients with and without ARID1A mutations. Fig. S5. TLS in baseline tumor tissues correlated with improved response in the discovery cohort. Fig. S6. CXCL13–/– mice are resistant to anti–PD-1 therapy. Fig. S7. Combination of ARID1A mutation and baseline expression of CXCL13 in the tumor tissue predicts performance of ICT.

Other Supplementary Material for this manuscript includes the following: (available at stm.sciencemag.org/cgi/content/full/12/548/eabc4220/DC1)

Data file S1 (Microsoft Excel format). Patient characteristics with correlative assays performed. Data file S2 (Microsoft Excel format). Expression data of NanoString Gene Panel. Data file S3 (Microsoft Excel format). Antibodies used for murine CyTOF analysis.

Fig. S1. CONSORT diagram for CheckMate275 and IMvigor210.

CONSORT diagram describing the numbers of samples used for specific studies in the CheckMate275

cohort (A) and IMvigor210 (B).

Checkmate275 Trial

cohort

n=270(100%)

WES- data available cohort

(QC- passed samples from both

tumor tissues and peripheral blood)

n=139(51.5%)

GEP (EdgeSeq)-data

available cohort (QC-passed

samples from tumor tissues)

n=217 (80.4%)

WES and GEP-data available cohort

n=120(44.4%)

IMvigor210 Trial cohort

Cohort I: n = 119

Cohort II: n = 310

Total n=429

Targeted next-generation

sequencing (NGS)-based genomic

FoundationOne®panel

Profiling n=293(68.3%)

GEP (TruSeq RNA Access

technology)-data available

cohort (QC-passed samples

from tumor tissues)

n=348 (81.1%)

FoundationOne® panel and GEP-

data available cohort n=275(64.1%)

A

B

Fig. S1

Fig. S2. Loss of ARID1A enhances the immunogenicity of bladder tumor cells.

(A) Graph representing weight (mg) of ARID1A knockdown and scrambled control MB49 tumors with

and without anti-PD-1 treatment (n=5 in each group,*p<0.05, ** p<0.01, ***p<0.001 determined by

Student’s t test). (B) Representative heat map indicating transcriptomic profiles of ARID1A knockdown

and scrambled control MB49 cells. (C) Gene Set Enrichment (GSEA) plot of the upregulated and

A B

KD ScrambledTXNIPCASP4

TNFAIP3

CXCL10VCAM1

CD40FAS

ISG15

CD274IFI27

RSAD2IFIT3

IRF1

CCL7PDE4B

CCL2CDKN1A

TNFAIP6

TAP1PTGS2

DDX58IRF9

NMI

PLA2G4AIRF5

NFKBIAARID5B

BTG1

PIM1STAT2

LATS2IRF2

UBE2L6

RNF213TRIM26

WARSB2M

PTPN2

ICAM1PARP14

NCOA3EPSTI1

BATF2

MTHFD2IFI35

RIPK2IFITM3

PSME2

BST2JAK2

ISG20TRIM21

CD74

XAF1VAMP8

PARP12PLSCR1

TRAFD1

ADARISOC1

CASP1TRIM25

SRI

DHX58CASP8

MVPIFNAR2

SAMD9L

SLC25A28IL15RA

SSPNOGFR

SAMHD1

GCH1STAT1

PSMB2PSMB10

ARL4A

HIF1AIFIT2

NOD1TOR1B

TRIM14

UPP1TAPBP

ZNFX1NAMPT

VAMP5

PELI1IL7

PNPT1SPPL2A

BPGM

PSMA3RNF31

NLRC5STAT3

IFITM2

LGALS3BPSOD2

EIF2AK2SOCS3

PSME1

TDRD7PSMA2

RIPK1LAP3

LY6E

RAPGEF6RBCK1

CASP3PNP

TNFAIP2

IL15PTPN1

CASP7OAS3

PML

NFKB1PFKP

HERC6NUP93

CMPK2

SOCS1IFI30

ARID1APTPN6

ITGB7

C

p < 0.001 p < 0.001

p < 0.001

En

rich

me

ntScore(ES)

En

ric

hm

en

t S

co

re(E

S)

p < 0.001

Fig. S2

-1 0 1

MB49

-Scr

ambled

MB49

-Scr

ambled

+ant

i-PD-1

MB49

-ARID

1A K

D

MB49

-ARID

1A K

D+a

nti-P

D-1

0

50

100

150

Tu

mo

r w

eig

ht (m

g)

✱

✱✱

✱✱✱

downregulated pathways in ARID1A knockdown and scrambled control MB49 tumor cell line

(p<0.001).

Fig. S3. Patients harboring ARID1A gene mutation have higher TMB.

(A) Representative plot of relative TMB in patients with no ARID1A mutation (ARID1A-WT) (n=100)

and with ARID1A mutation (ARID1A mutant) (n=39) in the CheckMate275 trial; TMB was log

transformed (***p<0.001). (B) Representative plot of relative TMB in patients with no ARID1A

mutation (ARID1A-WT) (n=213) and with ARID1A mutation (ARID1A mutant) (n=62) in the

IMvigor210 trial; TMB was log transformed (****p<0.0001) (C) Representative plot of relative TMB

A B C

D

Fig. S3

Hazard Curv es f or TMB by ARID1A mutation status

Estim

ate

d lo

g h

aza

rd

TMB

WT mutant

****

0.0

0.5

1.0

1.5

WT mutant

ARID1ATM

B

IMv igor210 (n = 275)CheckMate275(n = 139)

TM

B

WT mutant

***

1.0

1.5

2.0

2.5

3.0

3.5

ARID1A

***

0

1

2

3

4

WT mutant

ARID1A

TM

B

TCGA (n = 412)

in patients with no ARID1A mutation (ARID1A-WT) (n=311) and with ARID1A mutation (ARID1A

mutant) (n=101) in the TCGA BLCA cohort; TMB was log transformed (***p<0.001). (D) Predicted

log hazard curves demonstrating associations between TMB at various ARID1A mutation status with

PFS and OS. Hazard curve estimates are from the Cox PH model with linear predictors TMB and

ARID1A mutation status and TMB: ARID1A mutation status interaction. Shaded areas give 95%

pointwise confidence intervals for the hazard curves. Hazard curve estimates were scaled to be zero at

the median of TMB.

Fig. S4. Analysis of TGFβ1 and PD-L1 expression in patients with and without ARID1A

mutations.

(A) Representative plot of relative TGFβ1 expression in patients with no ARID1A mutation (ARID1A-

WT) (n=213) and with ARID1A mutation (ARID1A mutant) (n=62) in the IMvigor210 trial (*p<0.05).

(B) Representative plot of TGFβ1 expression in patients with no ARID1A mutation (ARID1A-WT)

(n=97) and with ARID1A mutation (ARID1A mutant) (n=37) in patients with stage IV cancer in the

TCGA BLCA cohort (** p<0.01). (C) Representative box plot of PD-L1 gene expression in patients in

the ARID1-WT group and in the ARID1A-mutant group in the discovery cohort (p= 0.45). (D)

Percentages of PD-L1 positive tumor cells (p=0.32) and immune cells (p=0.43) in patients with no

ARID1A mutation (ARID1A-WT) and with ARID1A mutation (ARID1A mutant) in the CheckMate275

biomarker cohort. (E) Stack bar plot showing the frequencies of patients with different PD-L1

*

4

6

8

WT mutant

ARID1A

TG

FB

1

CheckMate275

PD

-L1

positiv

e im

mune c

ell

(%)

0

20

40

60

0

25

50

75

100

WT mutant

ARID1A

Fre

quency

(%

)

IC_Lev el IC0 IC1 IC2+

IMv igor210 (n = 275, p = 0.4022)

IMv igor210 (n = 275)

A B

D

WT mutant

ARID1A

ns

Fig. S4

C

E

**

2

4

6

8

WT mutant

ARID1A

TG

FB

1

TCGA (n = 134)

ns

0

2

4

6

WT mutant

ARID1A

PD

-L1

Discov ery (n = 21)

PD

-L1

positiv

e tum

or cell

(%)

0

25

50

75

WT mutant

ARID1A

ns

expression in the immune cells (IC) in the group with no ARID1A mutation (WT) and group with

ARID1A mutation (mutant) in the IMvigor210 trial. P value from Pearson's Chi-squared test.

Fig. S5. TLS in baseline tumor tissues correlated with improved response in the discovery cohort.

(A) Composite image showing positive expression for CD20 cells (white), CD8 (red), and CD4

A

B C

STAT1CXCL9

CCL5

IKBKBJAK3

CXCL13CCL19

STAT2

IKBKGSTAT3

CCL24CXCR6

NFKBIA

CXCL11ITK

LYNCXCR4

CCR7

CCL14CXCR3

CCR5JAK2

RELA

CCL21CXCL1

CXCL10CXCL12

CXCL16

CCL4PRKCD

PIK3CDCXCR5

XCR1

CCL22MAPK3

CCL2PPBP

CCL7

CCL28XCL2

CCR6NFKB1

CHUK

IL8CCL27

CCR2MAP2K1

HCK

CCR4CXCL2

CCL8PIK3CG

CCL17

CCL25CXCL14

CCL18CX3CL1

CCR3

CXCL3CX3CR1

CCL3STAT5B

CCR9

CCL20CCL13

CCL16HRAS

CCL3L1

CCR1CXCR2

CXCR1CCL26

CXCL5

CCL11CXCL6

CCL15AKT3

CCL23

CCL1MAPK1

RNR

1.0

0.0

-1.0

NR R

0.0

0.5

1.0

TL

S d

en

sity

(# o

f T

LS

/ m

m2

of

tum

or

are

a) ****

NR R

0.00

0.02

0.04

0.06

Ra

tio

TL

S a

rea

/ t

ota

l a

rea

****

a

b c d

CD20CD8CD4

CD20 CD8 CD4

D

RNRRNR

CD20

CD8

CD4

CD20 CD8 CD4

Fig. S5

(yellow). (B-C) Quantification of ratio of TLS/ total area (B) and TLS density (C) in both non-

responders (blue) and responders (red). Each dot represents an individual sample. ****p<0.0001. (D)

Representative heat map of transcriptome profiling using GEP of baseline mUCC tumor specimens

(n=21; R=8, NR=13) using a customized 739-gene Nanostring panel.

Fig. S6. CXCL13–/– mice are resistant to anti–PD-1 therapy.

(A) Graph representing the tumor weight (mg) of WT and CXCL13-/- mice bearing MB49 tumors with and

without anti-PD-1 treatment. (B) Graph depicting frequency of intra-tumoral

CD3+CD8+ICOS+GzmB+PD-1+ T cells (n=5 in each group, *p<0.05, **p<0.01, ***p<0.001 determined

by Student’s t test). Data are representative of three independent experiments.

A B

Fig. S6

WT u

ntre

ated

WT+ a

nti-P

D-1

trea

ted

CXCL1

3-/- un

treat

ed

CXCL1

3-/- + a

nti-P

D-1

trea

ted

0

100

200

300

Tum

or

weig

ht

(mg)

✱✱

✱

ns

WT

untre

ated

WT+

ant

i-PD-1

trea

ted

CXCL1

3-/- un

treat

ed

CXCL1

3-/- +

anti-

PD-1

trea

ted

0

1

2

3

4

5

Fre

qu

en

cy o

f C

luste

r 1

3 (

%)

Cluster 13: CD3+CD8+ICOS+GzmB+PD-1+

✱✱

✱✱✱

ns

Fig. S7. Combination of ARID1A mutation and baseline expression of CXCL13 in the tumor tissue

predicts performance of ICT.

Predicted log hazard curves demonstrating associations of CXCL13 expression at various ARID1A

mutation statuses with OS in the IMvigor210 trial. Hazard curve estimates are from the Cox PH model

with linear predictors CXCL13 and ARID1A mutation status and CXCL13:ARID1A mutation status

interaction. Shaded areas give 95% pointwise confidence intervals for the hazard curves. Hazard curve

estimates were scaled to be zero at the median of CXCL13 values.

Fig. S7

WT Mutant

-5 0 5 -5 0 5

-3

-2

-1

0

1

2

3

CXCL13

Estim

ate

d lo

g H

aza

rdHazard Curv es f or CXCL13 by ARID1A mutation status

OS