The optimal management of non-metastatic muscle invasive urothelial cancer (MIUC) incorporates cystectomy, extended lymph node dissection, and peri-operative chemotherapy for those able to tolerate such intensive regimens. For those presenting or progressing to metastatic disease, palliative cisplatin-based chemotherapy is utilized but with limited response rates and generally poor overall improvement in survival up to 5–6% at 5 years [1]. The presence of high tumor mutational burden in urothelial cancers (UC) confers high immunogenicity, which makes these tumors good candidates for immunotherapy strategies, such as immune checkpoint blockade (ICB) [2]. Indeed, ICB therapy targeting the programmed death ligand-1 (PD-L1)/PD-1 immune checkpoint axis has shown some promising results in the treatment of UC; however, only a small subset of patients with metastatic disease had durable responses [3–5]. The major hurdles in achieving optimal survival benefits from ICB include tumor intrinsic heterogeneity, variations in pre-treatment tumor immune contexture, lack of predictive biomarkers, and optimal drug sequencing strategies. There is an urgent need to develop treatment combinations to enhance the proportion of chemotherapy and ICB responsive patients.

Several molecular classification schemes have been reported for MIUC subtyping, opening opportunities towards patient stratification for subtype tailored treatments [6]. There is a consensus that UC can be broadly divided into basal and luminal subtypes, which show distinct prognosis and response to chemotherapy [7]. Through in silico immune transcriptomic profiling, we recently reported that the four MIUC clusters earlier defined by The Cancer Genome Atlas (TCGA) network exhibit distinct immune gene expression patterns [8]. Robertson et al., 2017 reported the presence of five molecular subtypes in the TCGA MIUC cohort, a newer classification scheme, which incorporated the neuronal mRNA subtype [6].

The classification for treatment naïve immune contexture, which is critical in the context of current immunotherapy, divides solid tumors into “T cell-inflamed or hot” or “T cell non-inflamed or cold” categories [9]. An immunologically hot tumor is characterized by higher expression of IFN genes and corresponding higher density of activated CD8⁺ tumor infiltrating lymphocytes (TILs) [10]. Cold tumors usually show higher levels of immunosuppressive genes, lower density of activated CD8⁺ TILs and increased FoxP3⁺ or regulatory populations of immune cells. In the context of MIUC, luminal tumors are generally immunologically cold (with the exception of ‘luminal infiltrated’ subtype). It is intriguing that patients with luminal tumors show an increased overall survival, which is contradictory to their cold tumor state [11]. In contrast, patients with basal MIUC tumors exhibiting a hot tumor state have poorer outcomes, albeit their increased sensitivity to chemotherapy. The precise mechanisms underlying these counterintuitive associations remain to be fully elucidated although adaptive immune resistance mechanisms could be one of the potential underlying factors.

One interesting mechanism that regulates cellular Type I IFN responses, is the loss of function mutations in DNA damage repair (DDR) genes [12–18]. Approximately 40% of MIUC exhibit mutations in DDR genes [6]. It is widely established that pre-treatment breast (basal) and ovarian tumors with DDR gene (BRCA1/2) mutations are immunologically hot with high CD8⁺ TILs [19–21]. Moreover, these tumors exhibit increased response to platinum-based chemotherapy and have longer progression-free survival [22, 23]. Higher mutational burden leading to more neo-antigens in tumors with DDR deficiency has been suggested as one of the mechanisms that triggers spontaneous TIL infiltration in breast and ovarian cancers [21, 24]. This cancer agnostic association between DDR deficiency and treatment response is also seen in MIUC [21, 25–27]. Several reports in UCs have confirmed the strong association between mutations in DDR genes including, ERCC2, ATM, FANCD2, PALB2, BRCA1, BRCA2, RB1 and sensitivity to platinum-based neoadjuvant chemotherapy and chemoradiation therapy [28–33]. Most recently, Teo et al. confirmed that approximately 47% of advanced/metastatic UCs have at least one DDR gene mutation and exhibited higher sensitivity to platinum based chemotherapy [26].

A comprehensive view of the DDR mutation associated pre-treatment immune landscape is currently not available for MIUC. This knowledge is key to the future design of biomarker guided immunotherapy treatment combinations. To test our hypothesis that pre-treatment immune contexture and subsequent response of MIUC is potentially dictated by cancer cell intrinsic events such as DDR deficiency, we interrogated the subtype specific expression profiles of a panel of “immune regulatory” (immune-stimulatory and immune-inhibitory/checkpoint) genes using whole transcriptomic data from the TCGA (n = 408). We further correlated the expression profiles of immune regulatory genes with most prevalent DDR gene mutations in MIUC.

Recent success of immune checkpoint blockade therapies has re-directed the focus of molecular subtyping in MIUC to obtaining a deeper understanding of the cellular and secreted factors of the TME and its evolution by cancer cell intrinsic genomic alterations. Bladder cancer is amongst the few solid tumors where durable responses from ICB treatment have been observed in a subset of patients. A wide array of ICB trials targeting the PD-1/PD-L1 and CTLA-4 immune checkpoints are in progress with combinations using both adjuvant and neo-adjuvant chemotherapy in metastatic UC [37]. There is an urgent need to increase the proportion of patients responding to ICB. This can only be accomplished via the development of robust predictive biomarkers and superior combinatorial treatment approaches.

Amongst the cancers affecting the genitourinary system such as prostate and renal cell carcinoma, the bladder tumor immune contexture appears distinct, with higher mutational burden and features ranging from absence of TILs (immune desert) to a highly infiltrated tumor (immunologically hot) [37–39]. Relevant to our study findings, are the results of the recently completed ICB trials including IMvigor210 and Check Mate 275 trials that demonstrated the efficacy of PD-L1 and PD-1 inhibitors in platinum refractory MIUC patients [4]. Upon subtyping using the TCGA MIUC subtypes, these trials confirmed that tumors belonging to the luminal subtype and those lacking PD-L1 expression did not respond to atezolizumab (PD-L1 inhibitor). Another key finding from this trial was the cancer cell-specific expression of PD-L1 exclusively seen in tumors belonging to the TCGA basal subtype, which did not correlate with objective response rate [4]. These findings are suggestive of diverse impacts of immune cell versus cancer cell-specific expression of PD-L1 on ICB response.

Our analyses revealed the highest expression of immune checkpoint genes including CTLA-4, LAG3, TIGIT, PD-1, PD-L1, IDO1, TGFB, TGFBR1, TIM-3 and others in the basal squamous tumors. Based on these findings it can be speculated that an ICB treatment unresponsive tumor state is either due to lack of target expression or due to increased PD-L1 expression that imparts aggressive properties to cancer cells via activation of cellular proliferation pathways such as ERK and mTOR [40]. Our speculation is supported by previous findings, including our report, that showed the cancer cell intrinsic role of PD-L1 in mediating drug resistance, autophagy, and activation of aggressive pathways [40–42]. Within the basal MIUC tumors, it is also possible that immunogenic cell death induced by chemotherapy leads to enhanced tumor antigen cross-presentation to the pre-existing TILs leading to increased chemosensitivity. Activated TILs producing IFN-γ could further induce the expression of immune checkpoint genes with subsequent evolution of aggressive disease phenotype. These observations are also suggestive of counter-regulatory mechanisms in cancer cells (via possibly simultaneous expression of immune checkpoints PD-L1 and IDO1) putatively driven by genetic defects such as DDR mutations, that could eventually modulate the response to ICB [10]. Based on similar rationale, several ongoing trials in solid tumors are investigating the combinatorial effect of PARP inhibitors with immune checkpoint blockade [43]. Given the co-expression of immune checkpoint genes, predictive biomarker studies should also include evaluating simultaneous expression of proteins such as TIGIT, IDO1, LAG3, TIM-3 and others in addition to PD-L1.

Surprisingly, the results of KEYNOTE-045 Phase III trial using PD-1 targeting pembrolizumab, did not show a correlation between responses and tumor PD-L1 expression status [44]. Indeed, the high TGFβ expression in tumors leading to immune exclusion may not be discounted, which is also evident from our findings. This notion is supported by the recent report by Mariathasan et al., which showed tumors from aetozolizumab non-responsive patients had higher expression of TGFβ and its receptors and lacked CD8⁺ TILs in the tumor epithelial compartment [45]. Characterization of MIUC via subtyping prior to treatment and introduction of TGFβ inhibitors or immunostimulatory agents such as IFN inducing drugs, could thus potentially sensitize these tumors to ICB.

The outcome of CheckMate275 Phase II trial evaluating nivolumab (PD-1 inhibitor) in 270 UC patients, confirmed higher expression of a 25 gene IFN-γ signature in tumors from responders [3]. The most important and critical finding from this study, is the significantly higher expression of IFN activated STAT1, its downstream target TIL recruiting chemokine genes, CXCL9, CXCL10, CXCL11 in addition to the checkpoint molecules IDO1, LAG3, PD-1, markers of CD8⁺ T cell activation such as PRF1, GZMB and CD8A in a subset of basal tumors [3]. Interestingly, the treatment responsive group of patients was also enriched in a subgroup of basal subtype. Our findings in the basal squamous subtype and the presence of a group with CD8⁺ high TILs and high expression of PD-1, LAG-3, IDO1, CTLA-4 and PD-L1 showing better prognosis and decreased disease recurrence, are in concordance with this finding. Interestingly, basal breast cancer molecular subtypes that overlap in classification schemes applied in bladder, also exhibit features of higher chemosensitivity, presence of two basal subgroups and higher frequency of mutations in DDR genes and higher immune infiltrates [46, 47]. Notably, these features are also common in serous ovarian cancer where DDR deficiency associates with high immune activity in the TME [48]. A cancer type agnostic concurrent evolution of adaptive immune resistance due to increased immune checkpoint gene expression in these tumors, could potentially be a factor contributing to overall poor prognosis in basal UCs. Future chemo-immunotherapy trials should thus use combined subtyping by immunohistochemistry and immune checkpoint gene expression assay based biomarker signatures as tools for patient selection.

DDR mutation status of tumors is an important predictive biomarker of ICB response [45]. It is now well established that cancers with DDR mutations are immunologically hot and responsive to ICB. Three recent trials in UC including, NCT02553642, NCT01928394, and NCT02108652 (www.clinicaltrials.gov) have confirmed that mutations in DDR genes, are predictors of response to PD-1/PD-L1 ICB [27, 45]. Furthermore, it is now widely established that cancer cells with DDR deficiency exhibit constitutive activation of cellular IFN responses and secretion of TIL recruiting chemokines, CCL5 and CXCL10 [25, 49]. DDR mutation status of a tumor could also be combined with complementary biomarker assays for patient selection for ICB. Based on evolving concepts on the activation of cytosolic nucleic acid sensing IFN activating pathways such as cGAS-STING or RIG that ultimately lead to production of chemokines CCL5 and CXCL10 [50], it is possible that spontaneous TIL infiltration occurs simultaneous to malignant progression of tumors with DDR deficiency leading to an immunologically hot pre-treatment state. Although mechanistic studies are warranted to establish these associations further, our findings revealed increased expression patterns of the key IFN mediators STAT1 and/or STAT3 in tumors with DDR mutations such as those in ATM, ERCC1, RB1, BRCA2, POLE and TP53, reflective of IFN pathway activation.

Regardless of the mechanisms of evolution, in the context of current ICB treatments, information on DDR mutation status combined with assays to determine cancer and immune cell specific expression of immune checkpoint expression could further allow rational design of combinatorial immunotherapy trials. In summary (Fig. 6), the recent large and partially successful ICB trial outcomes re-emphasize the urgent need for the development of novel biomarker guided and combinatorial immunomodulatory treatments that can be used for MIUC patients with or without standard systemic chemotherapies. Given the development to novel immune stimulatory agents and the significance of DDR in activating cytosolic innate sensing pathways, these agents could be incorporated in treatment regimes, to sensitize MIUC tumors to ICB [51].Fig6Fig. 6

Proposed scheme for combination biomarkers to stratify MIUC patients for chemo-immunotherapy. Tumor mRNA subtype, DDR mutation status and immunoregulatory gene expression levels should be included in one combination biomarker assay to select patients for chemo-immunotherapy. Image design by http://www.designsthatcell.ca/

In conclusion, our results are suggest the potential co-activation of multiple compensatory immune checkpoint pathways in pre-treatment MIUC and thus provide rationale for use of combination ICB treatment. Findings from our comprehensive analyses will aid in the rational design of subtype specific combination immunomodulatory treatment approaches in UC.