Purpose:

In an exploratory analysis, we investigated the association between programmed death ligand 1 (PD-L1), tumor mutational burden (TMB), T-cell–inflamed gene expression profile (TcellinfGEP), and stromal signature with outcomes of pembrolizumab in urothelial carcinoma (UC).

Patients and Methods:

Patients with advanced UC received first-line pembrolizumab 200 mg every 3 weeks in the single-arm phase II KEYNOTE-052 trial (NCT02335424) and salvage pembrolizumab 200 mg every 3 weeks or chemotherapy (paclitaxel/docetaxel/vinflunine) in the randomized phase III KEYNOTE-045 trial (NCT02256436). The association of each biomarker (continuous variable) with objective response rate (ORR), progression-free survival (PFS), and overall survival (OS) was evaluated using logistic regression (ORR) and Cox PH (PFS, OS), adjusted for ECOG PS; nominal P values were calculated without multiplicity adjustment (one-sided, pembrolizumab; two-sided, chemotherapy). Significance was prespecified at α = 0.05.

Results:

In KEYNOTE-052, PD-L1, TMB, and TcellinfGEP were significantly associated with improved outcomes; stromal signature was significantly associated with worse outcomes. In KEYNOTE-045, although findings for TMB and TcellinfGEP with pembrolizumab were consistent with those of KEYNOTE-052, PD-L1 was not significantly associated with improved outcomes, nor was stromal signature associated with worse outcomes with pembrolizumab; chemotherapy was not associated with outcomes in a consistent manner for any of the biomarkers. Hazard ratio (HR) estimates at prespecified cutoffs showed an advantage for pembrolizumab versus chemotherapy regardless of PD-L1 or TMB, with a trend toward lower HRs in the combined positive score ≥10 and the TMB ≥175 mutation/exome subgroup. For TcellinfGEP, PFS and OS HRs were lower in the TcellinfGEP-nonlow subgroup regardless of treatment.

Conclusions:

Multiple biomarkers characterizing the tumor microenvironment may help predict response to pembrolizumab monotherapy in UC, and potential clinical utility of these biomarkers may be context-dependent.

Translational Relevance

Exploratory analyses were conducted to investigate the association between programmed death ligand 1 (PD-L1) expression, tumor mutational burden (TMB), T-cell–inflamed gene expression profile (TcellinfGEP), and the stromal signature and outcomes of pembrolizumab monotherapy in advanced urothelial carcinoma. In addition to PD-L1 and TMB—two regulatory-approved biomarkers—exploratory biomarkers, TcellinfGEP and the stromal signature are predictive in the first-line but not the salvage setting in pembrolizumab-treated patients with advanced urothelial carcinoma. Our results suggest that PD-L1, TMB, TcellinfGEP, and potentially several axes of gene expression may play independent or complementary roles in predicting clinical outcomes to anti–PD-1/PD-L1 therapy. Further study of these biomarkers in the monotherapy setting may provide insights that support the potential to combine PD-1/PD-L1 inhibitor therapy with other novel agents to overcome the effects of immunosuppressive microenvironmental features that can be tested in clinical trials. Innovative clinical trial designs may evaluate the clinical utility of integral/integrated biomarkers.

Recommended first-line therapy for locally advanced/metastatic urothelial cancer (UC) includes gemcitabine plus cisplatin with a subsequent switch to maintenance with the programmed death ligand 1 (PD-L1) inhibitor avelumab (1, 2). However, many patients with comorbidities are ineligible to receive cisplatin-based chemotherapy (2). The PD-L1 inhibitor atezolizumab is a first-line therapy option for cisplatin-ineligible patients whose tumors express PD-L1 (PD-L1–stained tumor-infiltrating immune cells covering ≥5% of the tumor area), and programmed death 1 (PD-1) inhibitor pembrolizumab and atezolizumab are first-line therapy options for patients who are ineligible for any platinum-containing chemotherapy regardless of PD-L1 expression (2).

In patients with platinum-refractory UC, pembrolizumab demonstrated superior overall survival (OS) versus chemotherapy, regardless of PD-L1 status (HR, 0.73; 95% confidence interval [CI], 0.59–0.91); the HR was smaller in the PD-L1 combined positive score (CPS) ≥10 population (HR, 0.57; 95% CI, 0.37–0.88; ref. 3). Although PD-L1 is overexpressed in 12% to 20% of UC (irrespective of assay; refs. 4, 5) and may correlate with response to PD-1/PD-L1 inhibitors, a number of patients with PD-L1–negative tumors have responded and others with PD-L1–positive tumors do not; finding additional useful biomarkers may help identify patients most likely to benefit from PD-1/PD-L1 inhibitors, and improve the risk–benefit ratio.

Pembrolizumab is approved by the FDA for patients with unresectable/metastatic tumor mutational burden-high (TMB-H; ≥10 mut/Mb) solid tumors who experience disease progression after previous treatment but have no satisfactory alternative treatment options (6, 7). Although therapeutic standards are traditionally dictated by tissue or organ origin, this is a tissue-agnostic approval.

Several exploratory biomarkers are possibly associated with response/resistance to PD-1/PD-L1 inhibitors. In a pan-tumor analysis including advanced UC tumor samples, the 18-gene T-cell–inflamed gene expression profile (TcellinfGEP) containing IFNγ-responsive genes capturing key features of the tumor microenvironment (TME) was associated with clinical response (8, 9). Although IFNγ can associate with favorable response to PD-1/PD-L1 inhibitors, TGFβ signaling pathway has been associated with lack of response and reduced OS with atezolizumab in UC (10). Similarly, epithelial–mesenchymal transitional (EMT)/stroma-related gene expression has been associated with lower response rates and shorter progression-free survival (PFS) in nivolumab-treated UC (11). In addition, in a pan-tumor setting, stroma/EMT/TGFβ (i.e., stromal) signature was significantly associated with a lower response rate to pembrolizumab (12).

We performed exploratory analyses using data from the single-arm phase II KEYNOTE-052 and randomized phase III KEYNOTE-045 trials to test our hypothesis that PD-L1, TMB, and TcellinfGEP were positively associated, and stromal signature was negatively associated, with outcomes of pembrolizumab monotherapy in advanced UC.

Study design and patients

Study design details for KEYNOTE-052 (NCT02335424; ref. 13) and KEYNOTE-045 (NCT02256436; ref. 3) have been published. KEYNOTE-052 included histologically/cytologically confirmed, locally advanced unresectable or metastatic UC and no previous systemic chemotherapy for advanced disease (13). KEYNOTE-045 included histologically/cytologically confirmed UC and progressive disease (PD) after platinum-based chemotherapy for advanced disease or recurrence <12 months after platinum-based (neo)adjuvant therapy (3). Archival specimens, but not new biopsy specimens, were mandated by the protocols.

In both studies, patients received intravenous pembrolizumab 200 mg every 3 weeks (3, 13). KEYNOTE-045 also randomly assigned patients (1:1) to receive pembrolizumab or chemotherapy [investigator's choice of paclitaxel (175 mg/m2), docetaxel (75 mg/m2), or vinflunine (320 mg/m2) every 3 weeks; randomization was stratified by Eastern Cooperative Oncology Group performance status (ECOG PS; 0/1 vs. 2), liver metastases (yes vs. no), hemoglobin concentration (<10 g/dL vs. ≥10 g/dL), and time since last chemotherapy dose (<3 months vs. ≥3 months; ref. 3).

The study protocols and all amendments were approved by the institutional review board or ethics committee at each institution. The studies were conducted in accordance with the protocol and its amendments, with Good Clinical Practice guidelines, and with the provisions of the Declaration of Helsinki. All patients provided written informed consent before enrollment (3, 13).

Outcomes and assessments

The primary objective of this prespecified exploratory analysis was to test whether TMB and TcellinfGEP (as continuous variables) were associated with improved outcomes and stromal signature with worse outcomes of pembrolizumab (KEYNOTE-052 and KEYNOTE-045) and chemotherapy (KEYNOTE-045).

Secondary objectives were to explore the clinical utility of each biomarker (dichotomized using clinically relevant cutoffs) by estimating HRs for PFS and OS by biomarker-defined subgroups in KEYNOTE-045. Cutoffs were: TMB (≥175 mut/exome vs. <175 mut/exome; previously shown to be concordant with 10 mut/Mb with FoundationOne®CDx; ref. 14), TcellinfGEP [<first tertile (low) vs. ≥first tertile (nonlow); ref. 8], and stromal signature (<median vs. ≥median; ref. 12).

The association between PD-L1 expression as a continuous variable and clinical outcomes as well as the clinical utility using a CPS cutoff of 10 (≥10 vs. <10) was evaluated in a parallel analysis for context.

TcellinfGEP and stromal signature were evaluated as robust hypotheses with significant prior evidence of potential to differentiate responder populations: TcellinfGEP continues to serve as a robust exploratory gene signature to support the mechanism of pembrolizumab in highly inflamed tumors and is associated with response to pembrolizumab across several tumor indications (8, 9), whereas stromal signature is negatively associated with PD-1/PD-L1 inhibitors in UC (10–12). The genes included in each signature are listed in Supplementary Table S1.

Clinical outcomes were ORR per RECIST v1.1 by blinded independent central radiological review, wherein a responder was defined as achieving a complete response (CR) or partial response (PR); PFS per RECIST v1.1, for which an event was defined as having PD or death; and OS, for which an event was defined as death.

PD-L1 expression was categorized using PD-L1 CPS, defined as number of PD-L1–staining cells (tumor cells, lymphocytes, and macrophages) divided by total number of viable tumor cells, multiplied by 100, and was assessed in formalin-fixed tumor samples at a central laboratory using PD-L1 IHC 22C3 pharmDx (Agilent). TMB (mutations/exome), defined as the sum of somatic nonsynonymous single-nucleotide variants that met all criteria described, was assessed using whole exome sequencing (WES) of tumor samples and matched normal DNA (9). TcellinfGEP score was calculated as the weighted sum of normalized expression value of 18 genes (8). TcellinfGEP and stromal signatures were profiled by RNA-sequencing using HiSeq4000 (Illumina) from the transcriptome of archival formalin-fixed paraffin-embedded tissue from primary tumors.

Statistical analysis

The association between continuous scores for PD-L1, TMB, TcellinfGEP, and stromal signature was evaluated using logistic regression (ORR) and Cox proportional hazards regression (PFS, OS) models, adjusted for ECOG PS (as a covariate); HRs were not applicable to this analysis. Nominal P values were calculated without multiplicity adjustment using a one-sided test for pembrolizumab (hypothesized positive association) and two-sided test for chemotherapy (no assumed direction hypothesized); significance was prespecified at α = 0.05. Because the primary objective was to evaluate whether each biomarker was generally associated with a specific clinical outcome (ORR, PFS, or OS), multiplicity-adjusted P values were not relevant. Clinical data cutoff dates were September 26, 2018 (KEYNOTE-052) and November 30, 2018 (KEYNOTE-045).

Data availability

Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA (MSD), is committed to providing qualified scientific researchers access to anonymized data and clinical study reports from the company's clinical trials for the purpose of conducting legitimate scientific research. MSD is also obligated to protect the rights and privacy of trial participants and, as such, has a procedure in place for evaluating and fulfilling requests for sharing company clinical trial data with qualified external scientific researchers. The MSD data sharing website (available at: http://engagezone.msd.com/ds_documentation.php) outlines the process and requirements for submitting a data request. Applications will be promptly assessed for completeness and policy compliance. Feasible requests will be reviewed by a committee of MSD subject matter experts to assess the scientific validity of the request and the qualifications of the requestors. In line with data privacy legislation, submitters of approved requests must enter into a standard data-sharing agreement with MSD before data access is granted. Data will be made available for request after product approval in the US and EU or after product development is discontinued. There are circumstances that may prevent MSD from sharing requested data, including country or region-specific regulations. If the request is declined, it will be communicated to the investigator. Access to genetic or exploratory biomarker data requires a detailed, hypothesis-driven statistical analysis plan that is collaboratively developed by the requestor and MSD subject matter experts; after approval of the statistical analysis plan and execution of a data-sharing agreement, MSD will either perform the proposed analyses and share the results with the requestor or will construct biomarker covariates and add them to a file with clinical data that is uploaded to an analysis portal so that the requestor can perform the proposed analyses.

KEYNOTE-052

Of 370 pembrolizumab-treated patients in KEYNOTE-052, 361 (97.6%), 231 (62.4%), 194 (52.4%), and 187 (50.5%) had available PD-L1, TMB, TcellinfGEP, and stromal signature data, respectively (Fig. 1A). Key baseline characteristics were similar between the populations with biomarker data available (Supplementary Table S2). The pairwise Spearman correlation of these four biomarkers is reported in Supplementary Fig. S1A.

Figure 1.

CONSORT Diagram. A, KEYNOTE-052. B, KEYNOTE-045.

Figure 1.

CONSORT Diagram. A, KEYNOTE-052. B, KEYNOTE-045.

Close modal

PD-L1 CPS was higher in responders than nonresponders (Fig. 2A). When correlating with best response, PD-L1 was highest in patients with CR or PR versus stable disease (SD) or PD (Supplementary Fig. 2A). Area under the receiver operating characteristic (AUROC) was 0.67 (95% CI, 0.61–0.73). PD-L1 as a continuous variable was also significantly associated with improved outcomes (Table 1). We also evaluated OS using a CPS 10 cutoff, demonstrating a positive trend in OS among patients with CPS ≥10 versus CPS <10 (Fig. 2A).

Figure 2.

Clinical outcomes in KEYNOTE-052 based on biomarker. A, Response and (B) OS based on PD-L1 status. C, Response and (D) OS based on TMB score. E, Response and (F) OS based on T-cell–inflamed GEP score. G, Response and (H) OS based on stromal signature score. CR, complete response; NR, nonresponder; PR, partial response; R, responder.

Figure 2.

Clinical outcomes in KEYNOTE-052 based on biomarker. A, Response and (B) OS based on PD-L1 status. C, Response and (D) OS based on TMB score. E, Response and (F) OS based on T-cell–inflamed GEP score. G, Response and (H) OS based on stromal signature score. CR, complete response; NR, nonresponder; PR, partial response; R, responder.

Close modal
Table 1.

Association between biomarkers as continuous variables and clinical outcomes by treatment group in KEYNOTE-052 and KEYNOTE-045.a

KEYNOTE-052
ORR per RECIST v1.1PFS per RECIST v1.1OS
BiomarkerNNominal one-sided PbNominal one-sided PbNominal one-sided Pb
PD-L1 CPS  361 0.002 0.006 0.021 
TMB (log10 scale)  231 <0.001 0.001 0.012 
T-cell–inflamed GEP  194 0.002 <0.001 <0.001 
Stromal signature  187 0.001 0.006 0.029 
KEYNOTE-052
ORR per RECIST v1.1PFS per RECIST v1.1OS
BiomarkerNNominal one-sided PbNominal one-sided PbNominal one-sided Pb
PD-L1 CPS  361 0.002 0.006 0.021 
TMB (log10 scale)  231 <0.001 0.001 0.012 
T-cell–inflamed GEP  194 0.002 <0.001 <0.001 
Stromal signature  187 0.001 0.006 0.029 
KEYNOTE-045
PembrolizumabChemotherapy
ORR per RECIST v1.1PFS per RECIST v1.1OSORR per RECIST v1.1PFS per RECIST v1.1OS
BiomarkerNNominal one-sided PbNominal one-sided PbNominal one-sided PbNNominal two-sided PcNominal two-sided PcNominal two-sided Pc
PD-L1 CPS (square root scale) 252 0.503 0.811 0.469 246 0.277 0.389 0.100 
TMB (log10 scale) 182 0.007 0.002 0.015 165 0.645 0.625 0.326 
T-cell–inflamed GEP 144 0.051 0.115 0.118 125 0.903 0.605 0.810 
Stromal signature 144 0.156 0.317 0.455 125 0.557 0.070 0.117 
KEYNOTE-045
PembrolizumabChemotherapy
ORR per RECIST v1.1PFS per RECIST v1.1OSORR per RECIST v1.1PFS per RECIST v1.1OS
BiomarkerNNominal one-sided PbNominal one-sided PbNominal one-sided PbNNominal two-sided PcNominal two-sided PcNominal two-sided Pc
PD-L1 CPS (square root scale) 252 0.503 0.811 0.469 246 0.277 0.389 0.100 
TMB (log10 scale) 182 0.007 0.002 0.015 165 0.645 0.625 0.326 
T-cell–inflamed GEP 144 0.051 0.115 0.118 125 0.903 0.605 0.810 
Stromal signature 144 0.156 0.317 0.455 125 0.557 0.070 0.117 

aAssociation was assessed with univariate analysis performed using logistic regression (ORR) and Cox proportional hazards (PFS, OS) modeling, adjusted for ECOG PS. Nominal P values were calculated without multiplicity adjustment, and significance was prespecified at α = 0.05.

bOne-sided because the a priori hypothesis was that PD-L1 CPS, TMB, and 18-gene T-cell–inflamed GEP were positively associated with outcomes and that the stromal signature was negatively associated with outcomes.

cTwo-sided because there was no a priori hypothesis regarding the direction of the association between PD-L1 CPS, TMB, 18-gene T-cell–inflamed GEP, or stromal signature and outcomes.

TMB was higher in responders than nonresponders (Fig. 2B). When correlating with best response, TMB was highest in patients with CR or PR versus SD or PD (Supplementary Fig. S2B). AUROC was 0.67 (95% CI, 0.59–0.75). TMB as a continuous variable was also significantly associated with improved outcomes (Table 1). When using a cutoff of 175 mut/exome, there was no trend for OS between the TMB subgroups (Fig. 2B).

TcellinfGEP was higher in responders than nonresponders (Fig. 2C). When correlating with best response, TcellinfGEP was highest in patients with CR, PR, or SD versus PD (Supplementary Fig. S2C). AUROC was 0.63 (95% CI, 0.54–0.73). TcellinfGEP was significantly associated with improved outcomes (Table 1). Using the first tertile as a cutoff, there was a positive OS trend in patients with TcellinfGEP-nonlow versus TcellinfGEP-low (Fig. 2C).

In contrast to the positive associations observed with PD-L1, TMB, and TcellinfGEP, the opposite was observed with stromal signature score, which was higher in nonresponders than responders (Fig. 2D). When correlating with best response, stromal signature was highest in patients with SD or PD versus CR or PR (Supplementary Fig. S2D). AUROC was 0.64 (95% CI, 0.55–0.72). Higher stromal signature score was significantly associated with worse outcomes (Table 1). When using the median as a cutoff, there were trends toward longer OS in the <median versus ≥median subgroup (Fig. 2D).

When comparing Kaplan–Meier curve patterns between the biomarkers, it is important to not draw overly broad conclusions, since all biomarkers were not obtained on the same set of patients with different prevalence above and below the prespecified cutoffs. Patients may vary in their membership in the biomarker low or high group across the figures.

KEYNOTE-045

Of 266 pembrolizumab-treated patients in KEYNOTE-045, 257 (96.6%), 185 (69.5%), 147 (55.3%), and 147 (55.3%) had available PD-L1, TMB, TcellinfGEP, and stromal signature data, respectively (Fig. 1B). Of 255 chemotherapy-treated patients, 250 (98.0%), 168 (65.9%), 127 (49.8%), and 127 (49.8%) had available PD-L1, TMB, TcellinfGEP, and stromal signature data, respectively. Key baseline characteristics were similar between the biomarker populations (Supplementary Table S3). The pairwise Spearman correlation of these four biomarkers is reported in Supplementary Fig. S1B. PD-L1 CPS was similar in responders and nonresponders, regardless of treatment group (Fig. 3A). When evaluated based on best response (Supplementary Fig. S3A), PD-L1 was not significantly associated with outcomes of pembrolizumab or chemotherapy (Table 1) with an AUROC of 0.53 (95% CI, 0.46–0.61) for the pembrolizumab treatment group (Supplementary Table S4). Using a CPS cutoff of 10, PFS was comparable between treatment groups, regardless of CPS cutoff (Table 2). A positive OS trend was observed in pembrolizumab-treated versus chemotherapy-treated patients regardless of CPS cutoff, although there was a trend toward lower HR in the CPS ≥10 subgroup (Table 2; Fig. 3A).

Figure 3.

Clinical outcomes in KEYNOTE-045 based on biomarker and treatment group. A, Response and (B) OS based on PD-L1 status. C, Response and (D) OS based on TMB score. E, Response and (F) OS based on T-cell–inflamed GEP score. G, Response and (H) OS based on stromal signature score. CR, complete response; NR, nonresponder; PR, partial response; R, responder.

Figure 3.

Clinical outcomes in KEYNOTE-045 based on biomarker and treatment group. A, Response and (B) OS based on PD-L1 status. C, Response and (D) OS based on TMB score. E, Response and (F) OS based on T-cell–inflamed GEP score. G, Response and (H) OS based on stromal signature score. CR, complete response; NR, nonresponder; PR, partial response; R, responder.

Close modal
Table 2.

Clinical outcomes by biomarker cutoff and treatment group in KEYNOTE-045.

PFSOS
BiomarkerCutoffTreatment groupEvents,anMedian (95% CI)HRb (95% CI)Events,cnMedian (95% CI)HRb (95% CI)
PD-L1 CPS ≥10 Pembrolizumab n = 56 49 2.1 (1.9–3.4) 0.95 (0.64–1.41) 39 11.1 (6.2–25.3) 0.64 (0.42–0.98) 
  Chemotherapy n = 67 60 2.8 (2.2–3.6)  54 4.9 (3.9–8.7)  
 CPS <10 Pembrolizumab n = 196 175 2.1 (2.0–2.3) 0.97 (0.78–1.21) 160 9.9 (8.0–12.6) 0.76 (0.61–0.95) 
  Chemotherapy n = 179 162 3.3 (2.2–4.6)  154 7.7 (6.9–9.8)  
TMB TMB ≥175 mut/exome Pembrolizumab n = 53 38 4.8 (2.1–10.2) 0.64 (0.41–1.00) 36 11.6 (8.0–33.3) 0.62 (0.39–0.97) 
  Chemotherapy n = 52 47 3.3 (2.3–6.1)  43 8.8 (6.1–12.5)  
 TMB <175 mut/exome Pembrolizumab n = 129 121 2.0 (2.0–2.1) 1.18 (0.91–1.54) 107 8.6 (6.9–12.9) 0.75 (0.57–0.98) 
  Chemotherapy n = 113 105 3.2 (2.2–3.8)  101 7.4 (6.0–9.4)  
T-cell–inflamed GEP Nonlow Pembrolizumab n = 97 82 2.2 (2.1–3.9) 0.82 (0.59–1.13) 72 11.3 (8.6–17.0) 0.61 (0.44–0.85) 
  Chemotherapy n = 82 77 3.1 (2.2–3.7)  73 7.4 (5.2–9.5)  
 Low Pembrolizumab n = 47 44 2.1 (1.9–3.3) 1.49 (0.94–2.35) 41 6.7 (5.0–15.2) 1.04 (0.66–1.65) 
  Chemotherapy n = 43 36 5.9 (3.3–7.6)  35 10.1 (7.0–12.2)  
Stroma/EMT/TGF-β ≥Median Pembrolizumab n = 70 61 2.1 (2.0–3.6) 0.86 (0.58–1.27) 53 11.3 (8.3–17.6) 0.54 (0.36–0.80) 
  Chemotherapy n = 65 61 2.4 (2.1–3.7)  59 6.9 (4.0–10.2)  
 <Median Pembrolizumab n = 74 65 2.1 (2.1–5.0) 1.12 (0.77–1.62) 60 8.1 (6.2–14.4) 0.94 (0.64–1.38) 
  Chemotherapy n = 60 52 4.7 (3.3–6.3)  49 8.4 (7.3–12.5)  
PFSOS
BiomarkerCutoffTreatment groupEvents,anMedian (95% CI)HRb (95% CI)Events,cnMedian (95% CI)HRb (95% CI)
PD-L1 CPS ≥10 Pembrolizumab n = 56 49 2.1 (1.9–3.4) 0.95 (0.64–1.41) 39 11.1 (6.2–25.3) 0.64 (0.42–0.98) 
  Chemotherapy n = 67 60 2.8 (2.2–3.6)  54 4.9 (3.9–8.7)  
 CPS <10 Pembrolizumab n = 196 175 2.1 (2.0–2.3) 0.97 (0.78–1.21) 160 9.9 (8.0–12.6) 0.76 (0.61–0.95) 
  Chemotherapy n = 179 162 3.3 (2.2–4.6)  154 7.7 (6.9–9.8)  
TMB TMB ≥175 mut/exome Pembrolizumab n = 53 38 4.8 (2.1–10.2) 0.64 (0.41–1.00) 36 11.6 (8.0–33.3) 0.62 (0.39–0.97) 
  Chemotherapy n = 52 47 3.3 (2.3–6.1)  43 8.8 (6.1–12.5)  
 TMB <175 mut/exome Pembrolizumab n = 129 121 2.0 (2.0–2.1) 1.18 (0.91–1.54) 107 8.6 (6.9–12.9) 0.75 (0.57–0.98) 
  Chemotherapy n = 113 105 3.2 (2.2–3.8)  101 7.4 (6.0–9.4)  
T-cell–inflamed GEP Nonlow Pembrolizumab n = 97 82 2.2 (2.1–3.9) 0.82 (0.59–1.13) 72 11.3 (8.6–17.0) 0.61 (0.44–0.85) 
  Chemotherapy n = 82 77 3.1 (2.2–3.7)  73 7.4 (5.2–9.5)  
 Low Pembrolizumab n = 47 44 2.1 (1.9–3.3) 1.49 (0.94–2.35) 41 6.7 (5.0–15.2) 1.04 (0.66–1.65) 
  Chemotherapy n = 43 36 5.9 (3.3–7.6)  35 10.1 (7.0–12.2)  
Stroma/EMT/TGF-β ≥Median Pembrolizumab n = 70 61 2.1 (2.0–3.6) 0.86 (0.58–1.27) 53 11.3 (8.3–17.6) 0.54 (0.36–0.80) 
  Chemotherapy n = 65 61 2.4 (2.1–3.7)  59 6.9 (4.0–10.2)  
 <Median Pembrolizumab n = 74 65 2.1 (2.1–5.0) 1.12 (0.77–1.62) 60 8.1 (6.2–14.4) 0.94 (0.64–1.38) 
  Chemotherapy n = 60 52 4.7 (3.3–6.3)  49 8.4 (7.3–12.5)  

aPFS events are defined as patients who experienced disease progression or died.

bHRs were estimated using Cox proportional hazards modeling, adjusted for ECOG PS.

cOS events are defined as patients who died.

TMB was higher in pembrolizumab-treated responders than pembrolizumab-treated nonresponders; this effect was not observed in chemotherapy-treated patients (Fig. 3B). When evaluated based on best response (Supplementary Fig. S3B), TMB was significantly associated with improved outcomes of pembrolizumab but not chemotherapy (Table 1) with an AUROC of 0.65 (95% CI, 0.54–0.75) for the pembrolizumab treatment group (Supplementary Table S4). Using a cutoff of 175 mut/exome, ORR (pembrolizumab vs. chemotherapy) was 35.2% versus 15.1% in patients with TMB ≥175 mut/exome and 16.0% versus 14.8% in patients with TMB <175 mut/exome. A positive PFS trend was observed in pembrolizumab-treated versus chemotherapy-treated patients with TMB ≥175 mut/exome; PFS was similar between treatment groups in patients with TMB <175 mut/exome (Table 2). A positive OS trend was observed in pembrolizumab-treated versus chemotherapy-treated patients regardless of TMB cutoff, although the trend was toward lower HR in the TMB ≥175 mut/exome subgroup (Table 2; Fig. 3B).

TcellinfGEP was higher in pembrolizumab-treated responders than pembrolizumab-treated nonresponders; this effect was not observed in chemotherapy-treated patients (Fig. 3C). When evaluated based on best response (Supplementary Fig. S3C), TcellinfGEP was positively associated with ORR with pembrolizumab but not chemotherapy (Table 1). The AUROC of 0.62 (95% CI, 0.54–0.75) for the pembrolizumab treatment group (Supplementary Table S4) was very similar to the 0.63 cited above for KEYNOTE-052. The TcellinfGEP P values did not reach nominal significance levels for testing associations with PFS or OS for pembrolizumab or chemotherapy (Table 1); however, the role of sample size when conducting hypothesis testing should be respected when comparing across KEYNOTE-052 and KEYNOTE-045. TcellinfGEP was not associated with PFS or OS for pembrolizumab or chemotherapy (Table 1). Using the first tertile as a cutoff, ORR (pembrolizumab vs. chemotherapy) was 27.0% versus 14.5% in patients with TcellinfGEP-nonlow tumors and 14.9% versus 15.9% in patients with TcellinfGEP-low tumors. HRs were smaller for both PFS and OS in the TcellinfGEP-nonlow versus TcellinfGEP-low subgroup (Table 2).

Stromal signature was higher in chemotherapy-treated nonresponders than chemotherapy-treated responders; no difference was observed between pembrolizumab-treated patients (Fig. 3D). When evaluated based on best response (Supplementary Fig. S3D), stromal signature was not associated with worse outcomes with pembrolizumab or chemotherapy (Table 1). Using the median as a cutoff, ORR (pembrolizumab vs. chemotherapy) was 22.2% versus 12.3% in patients with ≥median and 24.0% versus 17.7% in patients with <median. HRs estimated for both PFS and OS were smaller in the ≥median subgroup versus the <median subgroup (Table 2). The most notable trend appeared to be toward poor OS for chemotherapy-treated patients with a stromal signature score ≥median (Fig. 3D).

In this exploratory analysis of biomarkers in advanced UC, higher PD-L1, TcellinfGEP, and TMB scores were associated with improved clinical outcomes whereas higher stromal signature was associated with worse clinical outcomes with first-line pembrolizumab monotherapy in KEYNOTE-052. In KEYNOTE-045, both TMB and TcellinfGEP showed associations with improved clinical outcomes with salvage pembrolizumab monotherapy but not chemotherapy. OS benefit for pembrolizumab versus chemotherapy was observed regardless of TMB status; among the smaller subset of patients with available gene expression data, the HR estimate in the TcellinfGEP-nonlow subgroup was lower than that in the TcellinfGEP low subgroup, respectively. The stromal signature score did not show trends in the anticipated direction in KEYNOTE-045. These findings demonstrate that biomarkers beyond PD-L1 may play a role in the first-line and salvage settings and merit ongoing investigation. Because this work on these research biomarkers is exploratory and the intention-treat-population contains a large fraction of missing data, we believe these findings offer continued support to the current FDA approval of pembrolizumab for patients with locally advanced or metastatic UC who are ineligible for any platinum-containing chemotherapy (regardless of PD-L1 status) and patients who experience PD during or after platinum-containing chemotherapy (6).

PD-L1 is overexpressed in 12% to 20% of UC tumors (irrespective of assay; refs. 4, 5) and is a common biomarker in trials investigating PD-1/PD-L1 blockade. However, PD-L1 assays, cutoffs, and scoring algorithms vary (15). PD-L1 has not been consistently predictive of response in prospective phase II and phase III trials investigating PD-1/PD-L1 inhibitors in metastatic UC. In KEYNOTE-052, which used PD-L1 IHC 22C3 pharmDx, ORR increased with increasing PD-L1 enrichment with first-line pembrolizumab (all patients, 24%; PD-L1 CPS ≥1, 38%; PD-L1 CPS ≥10, 38%; ref. 13). Conversely, in IMvigor210, which used the PD-L1 SP142 IHC assay, first-line atezolizumab did not demonstrate a substantial increase in ORR with increasing PD-L1 enrichment (IC0, 21%; IC1, 21%; IC1/2/3, 24%; IC2/3, 28%; ref. 16). In the salvage setting, ORRs were comparable in the overall population and in the CPS ≥10 subgroup receiving second-line pembrolizumab (21.1% vs. 21.6%, respectively) in KEYNOTE-045 (3). Conversely, patients receiving salvage atezolizumab or nivolumab demonstrated ORRs ranging from 13% to 20% in all patients and 23% to 28% in patients with PD-L1–positive tumors (IC2/3 or tumor PD-L1 ≥1% or ≥5%; refs. 17–19). In our analysis, PD-L1 CPS was associated with improved outcomes in patients given first-line pembrolizumab in KEYNOTE-052. In KEYNOTE-045, this finding for testing PD-L1 CPS as a continuous variable was not replicated and treatment benefit was observed regardless of PD-L1 status. A trend toward lower OS HR was observed in patients with CPS ≥10 tumors.

TMB is independently associated with response to PD-1/PD-L1 inhibitor across several tumor types and patients with TMB-H tumors are currently approved for treatment with pembrolizumab regardless of tumor histology in the advanced setting where other treatment options do not exist (6, 7). Several analyses have been performed in advanced UC trials to explore the relevance of TMB. In IMvigor210, mutation load was independently associated with response to atezolizumab in a post hoc analysis; median mutation load was increased in responders (12.4 vs. 6.4 mutation load/Mb; P < 0.0001; refs. 19, 20). Similarly, tumor mutation load was significantly higher in responders and was associated with OS in cohort 2 of IMvigor210 (19). In IMvigor 211, median TMB was comparable between atezolizumab and chemotherapy groups; numerically longer OS was observed in patients with TMB-H with atezolizumab versus chemotherapy (17). Furthermore, a positive trend in OS was observed in IMvigor130 for TMB-H tumors (>10 mut/Mb); interim OS HRs were 0.71 (95% CI, 0.49–1.03) for atezolizumab monotherapy versus chemotherapy and 0.82 (95% CI, 0.58–1.17) for atezolizumab plus chemotherapy versus chemotherapy; an association with combination therapy has not been established (21). Finally, higher TMB was associated with improved clinical outcomes with nivolumab in CheckMate-275 (22). Consistent with these findings, the current analysis demonstrated that TMB, as a continuous variable, was associated with improved outcomes of pembrolizumab. Pembrolizumab improved OS versus chemotherapy regardless of TMB cutoff in KEYNOTE-045, although the observed trend was toward lower HR in the TMB ≥175 mut/exome subgroup.

Exploration of gene expression panels using RNA-sequencing can help define an inflamed TME and determine tumor inflammatory status (23). TcellinfGEP has been validated using tumor samples from >22 tumor types in PD-1 inhibitor–treated patients (8, 9). In CheckMate-275, a 25-gene IFNγ signature was associated with higher response to nivolumab; patients with high IFNγ signature scores were more likely to respond than patients with low IFNγ signature scores (18). Consistent with previous findings, the current exploratory analysis demonstrated that TcellinfGEP was positively associated with outcomes to pembrolizumab in both KEYNOTE-052 and with improved clinical outcomes for pembrolizumab versus chemotherapy in KEYNOTE-045. In the subset of patients with available GEP data, pembrolizumab showed a greater OS benefit versus chemotherapy in the TcellinfGEP-nonlow versus TcellinfGEP-low subgroup in KEYNOTE-045. Associations between clinical outcome and T-cell inflammation and IFNγ-driven gene expression in the TME, as captured by TcellinfGEP, appear to predict response in both the first-line and salvage settings represented by KEYNOTE-052 and KEYNOTE-045, supporting the anticipated mechanism of action of pembrolizumab.

A GEP signature representing convergent biology related to stroma/EMT/TGFβ pathways was developed to test for association with pembrolizumab efficacy (12). Stromal signature was negatively associated with outcomes of pembrolizumab in KEYNOTE-052, consistent with published results (10, 11). This was not replicated in KEYNOTE-045. Putative changes in the TME driven by the therapeutic intervention before pembrolizumab may potentially explain the lack of association with outcomes. However, Figs. 2D and 3D do not suggest that there are notable differences in the distribution of the stromal signature score between the two studies. Further, the correlation between the four biomarkers was highly similar between the two studies (Supplementary Fig. S1). Along with other general limitations cited below, the data at hand, resulting from a mixture of archival and freshly obtained formalin-fixed paraffin-embedded samples in both studies, are not ideally suited to explain TME differences between the two study cohorts proximal to the time of treatment. Similar stromal-type signature analyses have demonstrated a negative association with PD-1/PD-L1 response in advanced UC. Lack of response to atezolizumab was associated with a TGFβ signaling signature in fibroblasts (10). Lower response rates and shorter PFS and OS with nivolumab were observed with higher EMT/stromal-related gene expression (11).

Strengths of this analysis include the application of non-IHC molecular platforms (WES and transcriptome-wide RNA) and the prespecified retrospective approach to hypothesis testing of the same biomarkers in two large clinical trials that led to regulatory approval of pembrolizumab in advanced UC in two treatment settings. Limitations include lack of ascertainment of exploratory biomarkers on the full randomized population, small sample sizes of subgroups when evaluating biomarkers by cutoff, and relatively low number of events in specific subset analyses. Intrinsic biological differences between first-line and salvage populations and KEYNOTE-045 and KEYNOTE-052 study designs may partially account for the different biomarker outcomes reported; timing from tumor tissue acquisition, the therapy administered, and potential interval therapeutic intervention pressure in the TME may also contribute. In addition, it is possible that other nonstromal pathways impact immune evasion, such as T-cell exclusion, in these patient populations, among them the WNT (24), CDK4 (25), and PAK4 (26) pathways. A previous study has shown that cytotoxic CD4 T cells (27) and tertiary lymphoid structures are important in establishing an effective antitumor immune response. Furthermore, the immune contexture may change after chemotherapy treatment; thus, it would be of interest for a future analysis to explore TcellinfGEP of patients enrolled in KEYNOTE-045 based on tumor sample (archival, prechemotherapy vs. fresh biopsy, postchemotherapy). Future analyses may include the assessment of cell-free circulating tumor DNA and generation of a multi-omics biomarker model in larger randomized trials in which sufficient sample size per subgroup can lead to meaningful conclusions.

The exploratory analysis presented here in advanced UC indicates that multiple characteristics of TME may play a role in response/resistance to anti–PD-1/PD-L1 monotherapy and that better methods to describe the TME at the time of therapy are needed. Our results suggest that PD-L1, TMB, TcellinfGEP, and potentially other axes of gene expression may play independent or complementary roles in predicting clinical outcomes. Further studying these dimensions in the monotherapy setting may provide insights that support the potential to combine PD-1/PD-L1 inhibitor therapy with other novel agents to overcome the effects of immunosuppressive microenvironmental features that can be tested in clinical trials. Moreover, innovative clinical trial designs may help to evaluate the clinical utility of integrated biomarkers.

J. Bellmunt reports personal fees from Pfizer-Merck (Germany), Merck, Genentech, AstraZeneca, and BMS; grants from Pfizer-Merck and Takeda outside the submitted work; and author and section editor of UpToDate. R. de Wit reports personal fees from Merck during the conduct of the study; grants and personal fees from Sanofi and Bayer; personal fees from Astellas and Orion outside the submitted work. Y. Fradet reports grants and other support from Merck; personal fees from Astellas, Sanofi, Janssen, AstraZeneca, and Ferring; grants and personal fees from TerSera; and grants from IMV Inc. outside the submitted work. M.A. Climent reports personal fees from Pfizer, MSD, Roche, BMS, EUSA, Ipsen, Astellas, Janssen, Astra, and Merck outside the submitted work. D.P. Petrylak reports other support from Ada Cap (Advanced Accelerator Applications) Amgen, Astellas, AstraZeneca, Bayer, Bicycle Therapeutics, Boehringer Ingelheim, BMS, Clovis Oncology, Eli Lilly, Exelixis, Gilead Sciences, Incyte, Ipsen, Janssen, Mirati, Monopteros, Pfizer, Pharmacyclics, Regeneron Pharmaceuticals, Roche, Seattle Genetics, and Urogen, grants from Ada Cap (Advanced Accelerator Applications), Agensys Inc, Astellas, AstraZeneca, Bayer, BioXcel Therapeutics, BMS, Clovis Oncology, Eisai, Eli Lilly, Endocyte, Genentech, Gilead Sciences, Innocrin, MedImmune, Medivation, Merck, Mirati, Novartis, Pfizer, Progenics, Replimune, Roche, Sanofi Aventis, and Seattle Genetics, and other support from Bellicum (Sold 7/2020), Tyme (sold 10/2019) during the conduct of the study. L. Fong reports grants and personal fees from Merck during the conduct of the study; grants from Abbvie, Bavarian Nordic, and Dendreon and grants and personal fees from BMS, Janssen, and Roche/Genentech outside the submitted work. A. Necchi reports grants from Merck and Gilead, personal fees from Roche and Bayer, Pfizer, BMS, and Ipsen, and grants and personal fees from AstraZeneca, Janssen, and Incyte during the conduct of the study; grants and personal fees from Merck, Janssen, AstraZeneca, and Incyte, and personal fees from Roche, Bayer, Pfizer, and BMS outside the submitted work. C.N. Sternberg reports personal fees from Pfizer, BMS, Sanofi-Genzyme, MSD, Merck, Roche-Genentech, Incyte, Foundation Medicine, Immunomedics, Medscape, UroToday, Janssen, CCO Clinical, NCI, Bayer, AstraZeneca, and IMPACT Therapeutics outside the submitted work. P.H. O'Donnell reports grants from Merck during the conduct of the study; grants, personal fees, and other support from Genetech/Roche, Astellas Pharma, Seattle Genetics, and Janssen; other support from Allergan, Nektar, NIH, Dragonfly Therapeutics, and G1 Therapeutics; personal fees from Atheneum Partners, Health Advances, Dedham Group, CLD, Axiom Healthcare Strategies, EMD Serono, IntrinsiQ Specialty Solutions, ISMIE, NAMCP, and Merck; grants and personal fees from Pfizer; grants from Boehringer Ingeleim, AstraZeneca/MedImmune, Acerta Pharma and BMS outside the submitted work. T. Powles reports personal fees from AstraZeneca, BMS, Exelixis, Incyte, Ipsen, Merck, MSD, Novartis, Pfizer, Seattle Genetics, Merck Serono, Astellas, Johnson & Johnson, Eisai, and Roche; grants from AstraZeneca, Roche, BMS, Exelixis, Ipsen, Merck, MSD, Novartis, Pfizer, Seattle Genetics, Merck Serono, Astellas, Johnson & Johnson, and Eisai; other support from Roche, Pfizer, MSD, AstraZeneca, and Ipsen outside the submitted work. E.R. Plimack reports grants and personal fees from Merck during the conduct of the study; personal fees from AstraZeneca, Infinity Pharma, Pfizer, Astellas, AstraZeneca, Aveo, BMS, Calithera, Exelexis, Flatiron, Genentech, Janssen, MEI, Merck Natera, Pfizer, Regeneron Pharmaceuticals, and Seattle Genetics, and grants from Astellas, BMS, Genentech, and Merck outside the submitted work; also E.R. Plimack has a patent for Methods for screening muscle invasive bladder cancer patients for neoadjuvant chemotherapy responsiveness (US Patent 14/588,503) pending. D.F. Bajorin reports grants, personal fees, and other support from Merck, Inc. during the conduct of the study; grants and personal fees from BMS and AstraZeneca; personal fees from Dragonfly and Fidia Farmaceutici outside the submitted work. A.V. Balar reports grants from Merck during the conduct of the study; grants and personal fees from Merck, Genentech, BMS, Seagen, Immunomedics/Gilead, and Pfizer; personal fees from Incyte and Janssen outside the submitted work. D. Castellano reports personal fees from Pfizer, Roche, BMS, AstraZeneca, Novartis, MSD, Ipsen, Eisai, Astellas, Janssen, Bayer, and Sanofi outside the submitted work. T.K. Choueiri reports grants, personal fees, nonfinancial support, and other support from Merck, BMS, Pfizer, Roche, and EMD Serono during the conduct of the study; grants, personal fees, non-financial support, and other support from NCCN panel outside the submitted work; and has a patent for Biomarkers of activity/toxicity from IO pending and issued and a patent for ctDNA pending. T.K. Choueiri also reports institutional and personal, paid and unpaid support for research, advisory boards, consultancy, and honoraria from AstraZeneca, Aravive, Aveo, Bayer, BMS, Calithera, Circle Pharma, Eisai, EMD Serono, Exelixis, GlaxoSmithKline, IQVA, Infinity, Ipsen, Jansen, Kanaph, Lilly, Merck, Nikang, Nuscan, Novartis, Pfizer, Roche, Sanofi/Aventis, Surface Oncology, Takeda, Tempest, Up-To-Date, CME events (Peerview, OncLive, MJH and others), outside the submitted work; Institutional patents filed on molecular mutations and immunotherapy response, and ctDNA; Equity in Tempest, Pionyr, Osel, NuscanDx; Committees NCCN, GU Steering Committee, ASCO/ESMO; Medical writing and editorial assistance support may have been funded by Communications companies in part; No speaker's bureau; Mentored several non-US citizens on research projects with potential funding (in part) from non-US sources/foreign components; and the institution (Dana-Farber Cancer Institute) may have received additional independent funding of drug companies or/and royalties potentially involved in research around the subject matter. T.K. Choueiri is supported in part by the Dana-Farber/Harvard Cancer Center Kidney SPORE (2P50CA101942-16) and Program 5P30CA006516-56, the Kohlberg Chair at Harvard Medical School and the Trust Family, Michael Brigham, and Loker Pinard Funds for Kidney Cancer Research at DFCI; CV provided upon request for scope of clinical practice and research. S. Culine reports personal fees from Bayer, Merck, and Astellas outside the submitted work. W. Gerritsen reports other support from MSD during the conduct of the study; grants from Bayer and Astellas and other support from IqVia outside the submitted work. H. Gurney reports personal fees from BMS, MSD, Pfizer, Merck Serono, Roche, and Ipsen during the conduct of the study; personal fees from Astellas and Janssen outside the submitted work. D.I. Quinn reports personal fees from Astellas, Bayer, Pfizer, AstraZeneca, and Merck outside the submitted work; and employment with Abbvie Research and Development. J. Vuky reports grants from Merck during the conduct of the study; grants from Agendia, Arvinas, Astellas Pharma, Blueprint Medicines, BMS, Clovis Oncology, Deciphera, Eisai, Exelixis, Fortis, Genentech, Ignyta, Incyte, Innocrin Pharma, Lilly, Loxo, Novartis, Pfizer, Polyphor, and Rgenix outside the submitted work. N.J. Vogelzang reports personal fees from Merck (consulting regarding FDA submission; advisory board; and advisory board for Lenvima in collaboration with AstraZeneca), Aveo, Exelixis, Pfizer, BMS, and PER during the conduct of the study. R. Cristescu reports other support from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, during the conduct of the study, and owns stock in Merck & Co., Inc., Kenilworth, NJ, USA. J. Lunceford reports other support from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, outside the submitted work; also has a patent for Angiogenesis and mMDSC gene expression based biomarker of tumor response to PD-1 antagonists pending and a patent for patent WO 2020/167619 pending. A. Saadatpour reports personal fees and other support from Merck & Co., Inc., Kenilworth, NJ, USA, during the conduct of the study. A. Loboda reports other support from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, during the conduct of the study. J. Ma reports other support from Merck & Co, Inc., Kenilworth, NJ, USA, outside the submitted work. M. Rajasagi reports grants from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and stock in Merck & Co., Inc., Kenilworth, NJ, USA, outside the submitted work. J.L. Godwin reports other support from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and Agenus outside the submitted work. B. Homet Moreno reports personal fees from Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, outside the submitted work. P. Grivas reports grants from Merck during the conduct of the study; grants and personal fees from Merck; grants and personal fees from BMS and Clovis Oncology, GlaxoSmithKline, Immunomedics/Gilead, Mirati Therapeutics, Pfizer, and QED Therapeutics; personal fees from Dyania Health, EMD Serono, Exelixis, Foundation Medicine, Genentech/Roche, Genzyme, Guardant Health, Infinity Pharmaceuticals, and Janssen, Regeneron Pharmaceuticals, Seattle Genetics, 4D Pharma PLC, UroGen, AstraZeneca, Astellas Pharma, and Bayer; grants from Bavarian Nordic, Debiopharm, and G1 Therapeutics outside the submitted work. No disclosures were reported by the other authors.

J. Bellmunt: Conceptualization, data curation, formal analysis, validation, writing–original draft, writing–review and editing. R. de Wit: Conceptualization, data curation, formal analysis, validation, writing–original draft, writing–review and editing. Y. Fradet: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. M.A. Climent: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. D.P. Petrylak: Validation, writing–original draft, writing–review and editing. J.-Y. Lee: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. L. Fong: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. A. Necchi: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. C.N. Sternberg: Validation, writing–original draft, writing–review and editing. P.H. O'Donnell: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. T. Powles: Formal analysis, validation, writing–original draft, writing–review and editing. E.R. Plimack: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. D.F. Bajorin: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. A.V. Balar: Formal analysis, validation, writing–original draft, writing–review and editing. D. Castellano: Formal analysis, validation, writing–original draft, writing–review and editing. T.K. Choueiri: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. S. Culine: Data curation, validation, writing–original draft, writing–review and editing. W. Gerritsen: Formal analysis, validation, writing–original draft, writing–review and editing. H. Gurney: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. D.I. Quinn: Conceptualization, data curation, formal analysis, validation, writing–original draft, writing–review and editing. J. Vuky: Formal analysis, validation, writing–original draft, writing–review and editing. N.J. Vogelzang: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. R. Cristescu: Data curation, formal analysis, validation, writing–original draft, writing–review and editing. J. Lunceford: Conceptualization, validation, writing–original draft, writing–review and editing. A. Saadatpour: Formal analysis, validation, writing–original draft, writing–review and editing. A. Loboda: Formal analysis, validation, writing–original draft, writing–review and editing. J. Ma: Formal analysis, validation, writing–original draft, writing–review and editing. M. Rajasagi: Formal analysis, validation, writing–original draft, writing–review and editing. J.L. Godwin: Formal analysis, validation, writing–original draft, writing–review and editing. B. Homet Moreno: Formal analysis, validation, writing–original draft, writing–review and editing. P. Grivas: Conceptualization, formal analysis, validation, writing–original draft, writing–review and editing.

The authors thank the patients and their families and caregivers as well as all primary investigators and site personnel for participating in the study. Medical writing and/or editorial assistance was provided by Holly C. Cappelli, PhD, CMPP, and Dana Francis, PhD, of ApotheCom. This assistance was funded by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bray
F
,
Ferlay
J
,
Soerjomataram
I
,
Siegel
RL
,
Torre
LA
,
Jemal
A
.
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2018
;
68
:
394
424
.
2.
National Comprehensive Cancer Network
.
NCCN clinical practice guidelines in oncology. bladder cancer. version 4.2021
. https://www.nccn.org/professionals/physician_gls/pdf/bladder.pdf.
3.
Bellmunt
J
,
de Wit
R
,
Vaughn
DJ
,
Fradet
Y
,
Lee
JL
,
Fong
L
, et al
.
Pembrolizumab as second-line therapy for advanced urothelial carcinoma
.
N Engl J Med
2017
;
376
:
1015
26
.
4.
Boorjian
SA
,
Sheinin
Y
,
Crispen
PL
,
Farmer
SA
,
Lohse
CM
,
Kuntz
SM
, et al
.
T-cell coregulatory molecule expression in urothelial cell carcinoma: clinicopathologic correlations and association with survival
.
Clin Cancer Res
2008
;
14
:
4800
8
.
5.
Faraj
SF
,
Munari
E
,
Guner
G
,
Taube
J
,
Anders
R
,
Hicks
J
, et al
.
Assessment of tumoral PD-L1 expression and intratumoral CD8+ T cells in urothelial carcinoma
.
Urology
2015
;
85
:
703
.
e1–6
.
6.
KEYTRUDA® (pembrolizumab) injection, for intravenous use
.
12/2021
.
Whitehouse Station, NJ, USA
:
Merck Sharp & Dohme Corp.
;
2021
.
7.
Marabelle
A
,
Fakih
M
,
Lopez
J
,
Shah
M
,
Shapira-Frommer
R
,
Nakagawa
K
, et al
.
Association of tumour mutational burden with outcomes in patients with advanced solid tumours treated with pembrolizumab: prospective biomarker analysis of the multicohort, open-label, phase 2 KEYNOTE-158 study
.
Lancet Oncol
2020
;
21
:
1353
65
.
8.
Ayers
M
,
Lunceford
J
,
Nebozhyn
M
,
Murphy
E
,
Loboda
A
,
Kaufman
DR
, et al
.
IFN-gamma-related mRNA profile predicts clinical response to PD-1 blockade
.
J Clin Invest
2017
;
127
:
2930
40
.
9.
Cristescu
R
,
Mogg
R
,
Ayers
M
,
Albright
A
,
Murphy
E
,
Yearley
J
, et al
.
Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy
.
Science
2018
;
362
:
eaar3593
.
10.
Mariathasan
S
,
Turley
SJ
,
Nickles
D
,
Castiglioni
A
,
Yuen
K
,
Wang
Y
, et al
.
TGFbeta attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells
.
Nature
2018
;
554
:
544
8
.
11.
Wang
L
,
Saci
A
,
Szabo
PM
,
Chasalow
SD
,
Castillo-Martin
M
,
Domingo-Domenech
J
, et al
.
EMT- and stroma-related gene expression and resistance to PD-1 blockade in urothelial cancer
.
Nat Commun
2018
;
9
:
3503
.
12.
Cristescu
R
,
Nebozhyn
M
,
Zhang
C
,
Albright
A
,
Kobie
J
,
Huang
L
, et al
.
Pan-tumor analysis of the association of cancer and immune biology-related gene expression signatures with response to pembrolizumab monotherapy
.
J ImmunoTher Cancer
2019
;
7
.
Abstract P324
.
13.
Balar
AV
,
Castellano
D
,
O'Donnell
PH
,
Grivas
P
,
Vuky
J
,
Powles
T
, et al
.
First-line pembrolizumab in cisplatin-ineligible patients with locally advanced and unresectable or metastatic urothelial cancer (KEYNOTE-052): a multicentre, single-arm, phase 2 study
.
Lancet Oncol
2017
;
18
:
1483
92
.
14.
Aurora-Garg
D
,
Albright
A
,
Qiu
P
,
Li
Y
,
Liu
X
,
Fabrizio
D
, et al
.
Large-scale evaluation of concordance of genomic scores in whole exome sequencing and Foundation Medicine comprehensive genomic platform across cancer types
.
J ImmunoTher Cancer
2019
;
7
:
Abstract P318
.
15.
Udall
M
,
Rizzo
M
,
Kenny
J
,
Doherty
J
,
Dahm
S
,
Robbins
P
, et al
.
PD-L1 diagnostic tests: a systematic literature review of scoring algorithms and test-validation metrics
.
Diagn Pathol
2018
;
13
:
12
.
16.
Balar
AV
,
Dreicer
R
,
Loriot
Y
,
Perez-Gracia
JL
,
Hoffman-Censits
JH
,
Petrylak
DP
, et al
.
Atezolizumab (atezo) in first-line cisplatin-ineligible or platinum-treated locally advanced or metastatic urothelial cancer (mUC): long-term efficacy from phase 2 study IMvigor210
.
J Clin Oncol
2018
;
36
(suppl): Abstr 4523
.
17.
Powles
T
,
Duran
I
,
van der Heijden
MS
,
Loriot
Y
,
Vogelzang
NJ
,
De Giorgi
U
, et al
.
Atezolizumab versus chemotherapy in patients with platinum-treated locally advanced or metastatic urothelial carcinoma (IMvigor211): a multicentre, open-label, phase 3 randomised controlled trial
.
Lancet
2018
;
391
:
748
57
.
18.
Sharma
P
,
Retz
M
,
Siefker-Radtke
A
,
Baron
A
,
Necchi
A
,
Bedke
J
, et al
.
Nivolumab in metastatic urothelial carcinoma after platinum therapy (CheckMate 275): a multicentre, single-arm, phase 2 trial
.
Lancet Oncol
2017
;
18
:
312
22
.
19.
Rosenberg
JE
,
Hoffman-Censits
J
,
Powles
T
,
van der Heijden
MS
,
Balar
AV
,
Necchi
A
, et al
.
Atezolizumab in patients with locally advanced and metastatic urothelial carcinoma who have progressed following treatment with platinum-based chemotherapy: a single-arm, multicentre, phase 2 trial
.
Lancet
2016
;
387
:
1909
20
.
20.
Balar
AV
,
Galsky
MD
,
Rosenberg
JE
,
Powles
T
,
Petrylak
DP
,
Bellmunt
J
, et al
.
Atezolizumab as first-line treatment in cisplatin-ineligible patients with locally advanced and metastatic urothelial carcinoma: a single-arm, multicentre, phase 2 trial
.
Lancet
2017
;
389
:
67
76
.
21.
Galsky
MD
,
Banchereau
R
,
Rahman Hamidi
H
,
Leng
N
,
Harris
W
,
O'Donnell
PH
, et al
.
Tumor, immune, and stromal characteristics associated with clinical outcomes with atezolizumab (atezo) + platinum-based chemotherapy (PBC) or atezo monotherapy (mono) versus PBC in metastatic urothelial cancer (mUC) from the phase III IMvigor130 study
.
J Clin Oncol
2020
;
38
(15 suppl). Abstract 5011
.
22.
Galsky
MD
,
Saci
A
,
Szabo
PM
,
Han
GC
,
Grossfeld
G
,
Collette
S
, et al
.
Nivolumab in patients with advanced platinum-resistant urothelial carcinoma: efficacy, safety, and biomarker analyses with extended follow-up from CheckMate 275
.
Clin Cancer Res
2020
;
26
:
5120
28
.
23.
Aggen
DH
,
Drake
CG
.
Biomarkers for immunotherapy in bladder cancer: a moving target
.
J Immunother Cancer
2017
;
5
:
94
.
24.
Spranger
S
,
Bao
R
,
Gajewski
TF
.
Melanoma-intrinsic beta-catenin signalling prevents anti-tumour immunity
.
Nature
2015
;
523
:
231
5
.
25.
Jerby-Arnon
L
,
Shah
P
,
Cuoco
MS
,
Rodman
C
,
Su
MJ
,
Melms
JC
, et al
.
A cancer cell program promotes T cell exclusion and resistance to checkpoint blockade
.
Cell
2018
;
175
:
984
97
.
26.
Abril-Rodriguez
G
,
Torrejon
DY
,
Liu
W
,
Zaretsky
JM
,
Nowicki
TS
,
Tsoi
J
, et al
.
PAK4 inhibition improves PD-1 blockade immunotherapy
.
Nat Cancer
2020
;
1
:
46
58
.
27.
Oh
DY
,
Kwek
SS
,
Raju
SS
,
Li
T
,
McCarthy
E
,
Chow
E
, et al
.
Intratumoral CD4(+) T cells mediate anti-tumor cytotoxicity in human bladder cancer
.
Cell
2020
;
181
:
1612
25
.

Supplementary data