Molecular profiling of small cell bladder cancer (SCBC) to reveal gene expression determinants of an aggressive phenotype

Vadim S Koshkin1, Jordan Reynolds1, Paul Elson1, Cristina Magi-Galluzzi1, Jesse McKenney1, Karen S Smith1, Bonnie Shadrach1, Kumiko Isse2, Laura R Saunders2, Ming Hu1, Rahul D Tendulkar,1 Andrew J Stephenson1, Amr F Fergany1, Mohamed Abazeed1, Brian I Rini1,

Jorge A Garcia1, Byron Lee1, Omar Y Mian1, Petros Grivas1

1. Cleveland Clinic, Cleveland, OH 2. AbbVie Stemcentrx, San Francisco, CA

Background Methods• A subset of 53 patients had tissue assessed via IHC for DLL3 expression

with anti-DLL3 (SC16.65) antibody (validated Ventana assay) and PD-L1 (using both SP263 and SP142 antibodies)

• Multivariable analyses (MVA) were used to identify patient/tumor characteristics and tumor biomarkers predictive of overall survival (OS), progression-free survival (PFS) and time to progression (TTP) (p ≤ .05)

Results

Conclusions

Small Cell Bladder Cancer• Small Cell Bladder Cancer (SCBC) is a rare subtype representing

about 1% of bladder cancers which has a more aggressive clinical course and worse outcomes compared to urothelial carcinoma

• The data on treatment and outcomes of SCBC is limited and treatment patterns are often extrapolated from small cell lung cancer (SCLC) and urothelial carcinoma

• Biology of SCBC is poorly understood. There have been limited reports on tumor markers and genomic profiling in SCBC, such as somatic alterations linked to treatment response (Teo et al. J Clin Oncol 35, 2017 (suppl 6S; abstract 294))

• Molecular profiling can shed light on biology of SCBC and help identify biomarkers and treatment targets

Tumor Markers and Gene Expression Analysis in SCBC• DLL3 is a Notch pathway protein that is overexpressed on the

surface of SCLC tumor cells and other neuroendocrine tumors • DLL3 represents a potential therapeutic target since it is targeted by

an antibody-drug conjugate, Rovalpituzumab tesirine (Rova-T) which has shown significant anti-tumor efficacy in a Phase I trial in SCLC

• Programmed death-ligand 1 (PD-L1) is an important immune checkpoint targeted by an expanding class of agents, whose expression has not previously been reported in SCBC

• Gene expression profiling of a diverse cohort of SCBC patients allows the classification of these tumors into clusters that correlate with clinical phenotypes

• Differential gene expression analyses compare gene expression of tumor to normal tissue and also among tumor subtypes

Figure 2A: DLL3 Expression: Negative control (top) and tissue with 95% of tumor

cells expressing DLL3 (bottom)

All Patients N = 63

Median Age 71 (39-90)

Male/Female 52 (83%) / 11 (17%)

Current or Former Smoker 44 (77%)

Hematuria at Presentation 56 (89%)

Hydronephrosis at Presentation 9 (15%)

Metastatic Disease at Presentation 6 (9.5%)

TURBT only / Cystectomy only / Both 22 (35%) / 15 (24%) / 26 (41%)

Patients With Cystectomies N = 41

Tumor Location 37(90%) Bladder / 3(7%) Bladder+Urethra/Ureter / 1(2%) No tumor (T0)

Surgical Margins 8 (22%) Positive / 29 (78%) Negative

Carcinoma in Situ (CIS) 24 (60%) Present / 16 (40%) Absent

Lymphovascular Invasion (LVI) 10 (37%) Present / 17 (63%) Absent

Pathologic T-stage 15 (37%) T0-T2 / 26 (63%) T3-T4

Pathologic N-stage 23 (56%) N0 / 18 (44%) N1-3

Factor HR (95% CI) p Factor HR (95% CI) pOS From Diagnosis OS From Cystectomy

% cells (DLL3+)

(>10% vs ≤10%)2.8 (1.1-6.7) .03

% cells (DLL3+) (>10% vs ≤10%) 2.5 (1.0-6.3) .05

pT Stage (T3/T4 vs T0/T1/T2) 2.5 (1.0-6.2) .05

PFS From Diagnosis PFS From Cystectomy

% cells (DLL3+)

(>10% vs ≤10%)2.5 (1.1-5.8) .04

Margins (positive vs negative) 2.3 (0.97-5.3) .06

% small cell (>50% vs ≤50%) 3.6 (1.2-10.9) .02

TTP From Cystectomy% small cell (>50% vs

≤50%) 4.0 (1.1-14.4) .03

• Unsupervised hierarchical clustering of gene expression patterns from a heterogeneous cohort of small cell bladder cancer patients produced 4 distinct gene expression clusters that correlated with clinical phenotypes

• This is the first study to reveal distinct gene expression patterns in SCBC that define aggressive behavior and are associated with worse clinical outcomes including shorter overall survival

• DLL3 gene expression had a strong correlation with DLL3 protein expression suggesting its regulation at the transcriptional level

• The majority of SCBC tumors (68%) had DLL3 protein expression, and 30% had PD-L1 expression

• Higher DLL3 expression and increased small cell component were prognostic of worse clinical outcomes in SCBC

• Prognostic value of differential gene expression networks and the presence of underlying genomic and epigenetic alterations is the subject of ongoing investigation in this patient cohort

Table 1: Clinical/Pathological Characteristics and Survival of SCBC Patients

Results

Table 2: Independent Prognostic Factors of Survival

Outcomes From Diagnosis, N=63 (Median Follow-Up 16.6 months)

Median Overall Survival (OS) 22.8 months (95% CI: 11.9-42.4)

Median Progression Free Survival (PFS) 13.7 months (95% CI: 11.2-19.4)

Figure 9: Survival Outcomes According to DLL3 Expression and Small Cell %

Tumor Markers• Majority of tumors (59%, 37/63) had pure small cell histology (100% small cell) and 79% (50/63)

were at least 50% small cell • DLL3 protein expression (≥1% cells in tumor sample) was noted in 68% (36/53) of patients, with 58%

(31/53) having expression in >10% of cells • Moderate positive correlation was observed between small cell % of tumor and % of tumor cells

expressing DLL3 (Spearman r = 0.33, p= 0.01)• PD-L1 positive staining with either antibody (≥1% of cells) was noted in 16/53 (30%) of patients. All

PD-L1 staining was seen on tumor infiltrating immune cells and no significant correlation was noted between PD-L1 protein expression and SC%, DLL3 protein expression or PD-L1 gene expression

• DLL3 protein expression correlated with DLL3 gene expression (Spearman r = 0.70, p < 0.01)

OS Based on DLL3 Expression

PFS Based on DLL3 Expression

OS from Cystectomy Based on DLL3 Expression

PFS from Cystectomy Based on Small Cell %

Gene Expression Profiling and Classification• A subset of 39 patients within the 53 patient cohort had gene expression

profiling using HTG EdgeSeq Oncology Biomarker Panel, a commercially available platform with probes for 2568 genes

• Gene expression analysis was done on 39 primary SCBC tumor samples, 6 samples of adjacent normal urothelial tissue from the same patient cohort and 1 metastatic SCBC sample from same cohort (46 total)

• Unsupervised hierarchical clustering analysis was done using 46 samples

Methods• A retrospective review of clinical and pathological characteristics of

63 patients with pathology-confirmed SCBC seen at Cleveland Clinic from 1993 to 2016 was done following IRB approval

• Small cell histology was confirmed and percentage of small cell component (SC%) was defined in all 63 patient tissues at the time of this analysis by an experienced GU pathologist

• Tumor marker analysis (DLL3 and PD-L1) via IHC and gene expression profiling of a subset of these patients with available tissue specimens was undertaken (Figure 1)

Please send correspondence to Vadim Koshkin (koshkiv@ccf.org)

Figure 2B: PD-L1 Expression: Negative control (top) and tissue with 5-10% tumor

infiltrating cells expressing PD-L1 (bottom)

Figure 1: Analysis Schema

Results

• Unsupervised hierarchical clustering of gene expression from 46 samples produced 4 clusters• Patients with tumors that were in the same cluster as most normal samples (Cluster 2: “normal-

like”) had a more favorable clinical phenotype as they did not have metastases at diagnosis or distant recurrence later in the disease course

• Patients with tumors that clustered with the metastatic sample (Cluster 3: “metastasis-like”) had shorter OS (median OS 6.0 months) compared to the other 3 clusters (log rank p value 0.046)

• Gene expression variability among tumor samples from different patients was lower than variability of gene expression between tumor and normal tissue samples from the same patient

Figure 7: Differential gene expression analysis based on tumor SC% (>50% vs ≤50% )

Figure 8: Differential gene expression analysis based on DLL3% of tumor (>10% vs ≤10%)

Figure 5: Differential gene expression analysis among the 4 clusters

Gene ANOVA_P valueTNFAIP3 1.26E-15DUSP1 3.54E-14

ADAMTS1 1.13E-13GADD45B 1.92E-13

IER3 3.03E-13SOCS3 3.10E-13DUSP5 1.82E-12PARP1 3.11E-12SGK1 4.58E-12CD44 6.37E-12SOD2 7.96E-12

UBE2T 8.08E-12LIF 1.23E-11

ICAM1 1.38E-11EMP1 1.41E-11BIRC3 1.67E-11CCT2 1.74E-11

KPNA2 2.70E-11IL1R1 3.13E-11FOS 3.14E-11

Figure 3: Dendrogram of hierarchical clustering analysis

Gene Mean_Tumor Mean_Normal P value FDRCXCL16 10.2 12.02 7.40E-07 0.0019

CTSB 10.48 12.24 4.46E-06 0.0057MDC1 11.32 10.21 6.71E-06 0.0057

CYP3A5 10.39 13.64 1.69E-05 0.0103SFN 10.82 14.6 2.00E-05 0.0103

TOP3B 8.71 7.61 2.53E-05 0.0108RFC3 13.09 10.39 4.54E-05 0.0146

DUSP5 10.84 14.53 4.57E-05 0.0146SMC3 13.27 11.62 6.65E-05 0.0148RFC4 12.86 10.02 6.85E-05 0.0148

CASP4 9.53 11.43 7.46E-05 0.0148CAV2 10.32 12.58 7.48E-05 0.0148UNG 12.89 11.31 7.76E-05 0.0148

FANCB 12.05 9.5 8.65E-05 0.0148IL18 9.25 11.17 9.61E-05 0.0148

CBX5 14.28 11.67 9.92E-05 0.0148CAPN7 11.59 10.69 0.000100113 0.0148PPARG 9.82 13.02 0.000105597 0.0148RAD51C 11.26 9.4 0.000113639 0.0148

REL 10.57 12.53 0.000117258 0.0148

Figure 6: Differential gene expression analysis between matched tumor and normal tissue samples

Average EZH2 gene expression in tumor sample = 13.66Average EZH2 gene expression in matched control sample = 10.41Paired t-test p-value = 0.000371EZH2 mutation or overexpression associated with numerous malignancies (PMID: 26845405)

Figure 4: Heat map of hierarchical clustering analysis and Kaplan-Meier Curve

A. Top 20 differentially expressed genes among 4 clusters B. One way ANOVA for selected genes with associated p values

A. Top 20 differentially expressed genes comparing tumor and normal B. Representative example of EZH2 differential expression

C3 vs C1,2,4: log rank p value = 0.046

63 patients with SCBC

53 patients with DLL3 and PD-L1 IHC

39 patients with gene expression profiling of SCBC tissue

6 patients with gene expression profiling of SCBC tissue AND adjacent normal urothelial tissue

4 clusters

Gene expression in normalized counts per million

p = 0.03

p = 0.04 p = 0.05

p = 0.02