Cell death-induced immunogenicity enhances chemoimmunotherapeutic response by converting immune-excluded into T-cell inflamed bladder tumors

Chemoimmunotherapy has recently failed to demonstrate significant clinical benefit in advanced bladder cancer patients; and the mechanism(s) underlying such suboptimal response remain elusive. To date, most studies have focused on tumor-intrinsic properties that render them “immune-excluded”. Here, we explore an alternative, drug-induced mechanism that impedes therapeutic response via disrupting the onset of immunogenic cell death. Using two immune-excluded syngeneic mouse models of muscle-invasive bladder cancer (MIBC), we show that platinum-based chemotherapy diminishes CD8+ T cell tumor infiltration and constraines their antitumoral activity, despite expression of activation markers IFNγ and granzyme B. Mechanistically, chemotherapy induces the release of prostaglandin E2 (PGE2) from dying cancer cells, which is an inhibitory damage-associated molecular pattern (iDAMP) that hinderes dendritic cell maturation. Upon pharmaceutical blockade of PGE2 release, CD8+ T cells become tumoricidal and display an intraepithelial-infiltrating (or inflamed) pattern. This “iDAMP blockade” approach synergizes with chemotherapy and sensitizes bladder tumors towards anti-PD1 immune checkpoint inhibitor therapy. These findings provide a compelling rationale to evaluate this drug combination in future clinical trials.

B ladder cancer is a devastating disease with severe morbidity at its advanced stages. Muscle-invasive bladder cancer (MIBC) represents an aggressive stage of bladder cancer and accounts for roughly 200,000 deaths worldwide annually 1 . The current standard of care for managing MIBC includes neoadjuvant chemotherapy followed by radical cystectomy. The most commonly used chemotherapy regimen is the gemcitabine and cisplatin (GC) combination. However, the absolute benefit of neoadjuvant chemotherapy is a modest~5% improvement in overall survival [2][3][4][5] . To date, most studies investigating chemotherapeutic response in MIBC have conventionally focused on the direct cytotoxic effect(s) of drugs on tumor cells. In contrast, emerging evidence from other tumor types has supported the notion that select chemotherapies can recruit immune cells to target tumor cells, via inducing immunogenic cell death [6][7][8][9] . This process is facilitated by the release of immunostimulatory damage-associated molecular patterns (DAMPs; e.g., CRT, HMGB1, HSP70) by dying cancer cells, which function as immunological adjuvants for recruiting and activating dendritic cells (DCs) to subsequently promote an adaptive (T-cell-mediated) antitumoral immune response [6][7][8][9][10] . In contrast, as we have previously demonstrated, inhibitory DAMPs (iDAMPs; e.g., PGE 2 ) functionally attenuate the immunogenicity of cell death 11,12 . Indeed, mechanistic studies have revealed that tumorinfiltrating T cells are not mere bystanders, but are important determinants that profoundly impact chemotherapeutic response 13 . However, the role(s) of immune cells (e.g., T lymphocytes and dendritic cells) in modulating chemotherapeutic efficacy in bladder cancer remain insufficiently investigated and poorly understood.
The recent FDA approval of immune-checkpoint inhibitors (ICIs) has revolutionized the treatment of advanced and metastatic MIBC; especially, for patients presenting with tumors that are pre-infiltrated with T cells (or "T-cell-inflamed" tumors) 14 . Yet, the vast majority (70-80%) of patients are nonresponsive towards ICIs; whereas the peripheral exclusion of T cells from tumor cell clusters is currently postulated as a mechanism for ICI resistance 15 . Although immune exclusion is a clinically significant phenomenon that is connected to immune-checkpoint response in bladder cancer, research progress has been significantly hampered by the lack of immune-excluded MIBC models 16 .
Aiming to improve the poor patient response towards chemotherapy or ICIs, a number of ongoing or recently completed clinical trials are combining the two as a new treatment modality (i.e., chemoimmunotherapy). Disappointingly, two highly anticipated Phase 3 clinical trials (e.g., IMvigor130 and KEY-NOTE-361) evaluating the efficacy of chemoimmunotherapy in metastatic and locally advanced bladder cancer failed to improve the overall survival 17,18 . In general, there was no significant clinical benefit when comparing the overall survival or objective response of patients receiving chemoimmunotherapy to those receiving standard chemotherapy 17,18 . In retrospect, inadequate preclinical studies were performed to evaluate the immunomodulatory effects of GC chemotherapy in bladder cancer, before its hasty clinical evaluation in combination with ICIs. These recent clinical challenges underscore the necessity for a thorough preclinical study to investigate how GC chemotherapy modulates immune effector cells in bladder cancer.
Here, we report two immune-excluded murine MIBC syngeneic models that exhibit intrinsic resistance towards platinumbased chemotherapy and ICI therapy. "Inhibitory DAMP" blockade synergizes with chemotherapy in these models by inducing immunogenic cell death; thereby converting immuneexcluded tumors into T-cell-inflamed. Mechanistically, iDAMP blockade skews DCs into immunogenic maturation and promotes the recruitment of antitumoral CD8+ T cells. Understanding the immunomodulatory roles of GC chemotherapy in bladder cancer is timely since these insights provide plausible explanations for the recent clinical failures of chemoimmunotherapy. Our proofof-concept preclinical study builds upon our recent report that drug-induced immunogenic cell death is governed by an intricate balance of DAMP-to-iDAMP ratio in the tumor microenvironment 11,12 , and provides a rationale for combining "iDAMP blockade" with chemoimmunotherapy in future clinical trials.

Results
Generating immune-excluded murine MIBC models that are unresponsive to chemotherapy and immune-checkpoint inhibitors. Most studies investigating therapeutic response towards ICIs in bladder cancer have focused on tumor-intrinsic mechanisms (e.g., TGF-β1, or Wnt/β-catenin) 15,19,20 . These pathways create an immunosuppressive microenvironment that precludes CD8+ T cells from the tumor parenchyma, resulting in immune-excluded (or "T-cell-non-inflamed") bladder tumors 15,19,20 . Currently, the field faces a paucity of available murine bladder cancer models with an immune-excluded phenotype 21 , thereby, limiting the investigation of bladder cancer immunomodulation in immunocompetent hosts. We therefore generated two transplantable murine MIBC cell lines-with molecular and physiological features that are akin to the human disease-that display an immune-excluded phenotype in vivo using the following method. Mouse bladder tumors were first induced using a conventional chemical carcinogenesis model 22,23 ; i.e., N-butyl-N-(4-hydroxybutyl)-nitrosamine (BBN) exposure via water intoxication as previously described 24 . Next, BBN-induced bladder tumor tissues were enzymatically processed to generate single-cell suspensions 25,26 and were grown in vitro to establish tissue culture cell lines that allow serial transplantation in immunocompetent syngeneic hosts (Fig. 1a).
Immunofluorescence evaluation of tumor tissues from the established syngeneic MIBC lines (designated as G69 and G7 hereinafter) revealed an overall non-T-cell-inflamed phenotype (Fig. 1b, G69; Supplementary Fig. 1A, G7). The majority of CD3+ T cells (green) were excluded from the tumor cores, and instead, were contained within the tumor periphery (Fig. 1b, Area 1). Epithelial tumor regions were predominantly T cell excluded (Fig. 1b, Area 2). Minimal T cells were observed but restricted to stromal regions (αSMA+, red) intercalating between tumor clusters (Fig. 1b, Area 3 and Area 4). Quantification of CD3+ Tcell localization confirmed a stroma-restrictive pattern (Fig. 1c). In accordance with the criteria set forth by Chen & Mellman, G69 and G7 bladder tumors qualify as T-cell-non-inflamed or immune-excluded 27 .
Next, to examine how these syngeneic bladder tumors respond to standard-of-care therapies, we evaluated the efficacy of gemcitabine-cisplatin (GC) chemotherapy and ICI therapy (i.e., anti-PD1) on G69 and G7 bladder tumors in vivo. Chemotherapy failed to inhibit tumor growth (Fig. 1d, e, Supplementary Fig. 1B) and as anticipated for non-inflamed tumors, ICI therapy was also unable to suppress tumor growth (Fig. 1f, Supplementary Fig. 1C, D). Herein, we introduce and characterize two unique immuneexcluded MIBC models 21 that are unresponsive to both chemotherapeutic and ICI interventions. These represent clinically relevant murine MIBC models that recapitulate the human malady for further mechanistic investigations 21 .
CD8+ T cells are not tumoricidal despite expressing granzyme B and IFNγ. A recent study using bioinformatics analysis revealed that a high T-cell signature (i.e., CD3Z, CD8a, and CXCL9) in a subset of human MIBCs receiving radical cystectomy and chemotherapy positively correlated with better survival 28 . However, the immunomodulatory roles of GC chemotherapy in MIBC patients or preclinical models of MIBC are poorly characterized. To address this research gap, we sought to directly evaluate how GC chemotherapy impacts the frequency and function of T lymphocytes in our immune-excluded murine MIBC models. In brief, mice bearing G69 and G7 tumors were treated with two cycles of GC chemotherapy with a rest period in between to recapitulate the clinical management of MIBC 25 (Fig.  2a). Forty-eight hours after the completion of the second GC cycle, tumor tissues were harvested and analyzed for the presence of tumor-infiltrating lymphocytes by flow cytometry (Fig. 2b). Treatment with GC chemotherapy significantly decreased overall tumor-infiltrating CD3+ T cells, as well as CD4+ and CD8+ T-cell subsets when compared to vehicle-treated tumors (Fig. 2c-e; *P < 0.05, **P < 0.001). While GC chemotherapy decreased the overall frequency of tumor-infiltrating T cells, immunofluorescence analysis revealed further spatial information which showed an increase in "immune desert" areas toward the core of the chemotherapy-treated tumors (Fig. 2g, Area 4). The majority of CD3+ T cells were constrained at the periphery (Fig. 2g, Area 1), while GC chemotherapy seemed to slightly increase T-cell infiltration into the epithelial region in certain areas (Fig. 2g, Area 2 and Area 3).
Given the importance of CD8+ T cells as effectors for eliminating cancer cells, we evaluated the co-expression of IFNγ and GZMb in CD8+ TILs-two commonly used markers to indicate their activation status and antitumoral activities 29 . b Immunofluorescence analysis of CD3+ T-cell infiltration and aSMA expression in G69 treatment-naïve tumor tissues (Scale bar: 100 μm; Images representative of n = 3 independent experiments). c Relative localization of CD3+ T cells in treatment naïve G69 tumors. CD3+ T cells in regions of interest were quantified and normalized to average CD3+ T-cell infiltration (n = 9 biologically independent samples). Data are presented as mean values ± SEM. (**-Two-tailed, unpaired t-test p = 0.0045) d, e Weight and growth curve of G69 tumors treated with vehicle (n = 8 biologically independent animals) or gemcitabine (120 mg/kg) and cisplatin (6 mg/kg) (n = 10 biologically independent animals) for two cycles. Data are presented as mean values ± SEM. (Two-tailed, unpaired t-test at endpoint values). f Growth curve of vehicle (n = 8 biologically independent animals) or anti-PD1-treated (n = 3 biologically independent animals) (200 mg/mouse) G69 tumors monitored for 21 days. Data are presented as mean values ± SEM. (Two-tailed, unpaired t-test at endpoint values). Source data are provided as a Source data file. ns: not significant.
CD8+ TILs from GC-treated tumors exhibited a significant increase in the expression of IFNγ and GZMb when compared to vehicle treatment alone (Fig. 2h, i and j, k, respectively; *P < 0.05, ***P < 0.005). Despite the expression of these active marks, functional evaluation of CD8+ T cells via antibody depletion in vivo (Fig. 2l) failed to exhibit changes in tumor growth when compared to IgG-treated controls (Fig. 2m), indicating they were nontumoricidal functionally. Taken together, these results demonstrate that GC chemotherapy fails to mount an effective antitumoral T-cell response in immune-excluded MIBC.
Gemcitabine-cisplatin chemotherapy attenuates the maturation of CD11c + MHCII + dendritic cells. To understand why these CD8+ T cells were nontumoricidal, we explored the impact of GC chemotherapy on modulating dendritic cell maturation within G69 bladder tumors. Since proper priming of CD8+ T cells by mature dendritic cells (DCs) is essential for the generation of antigen-specific cytotoxic T lymphocytes with antitumor activities 30,31 , we sought to explore this interaction. In brief, immunogenic DCs prime antitumoral CD8+ T cells through providing T-cell receptor activating signal (signal 1; MHCI), co-stimulation (signal 2; CD86, CD40) and paracrine stimuli (signal 3; cytokines); important for the proper activation, propagation, and polarization of naive CD8+ T cells into effector cells 32 . We, therefore, analyzed the infiltration and maturation status of DCs in vehicle-and GC-treated tumors using flow cytometry (Fig. 3a). Between the treatment groups, DCs were detected within all tumors and no difference in frequency was observed (Fig. 3b, c). However, further evaluation revealed that DCs in GC-treated tumors exhibited a significantly reduced expression of H2-k (i.e. MHCI; Signal 1) and CD86 (Signal 2) (Fig. 3d), indicative of diminished DC activation. In addition, the expression of a key Th1-polarizing cytokine, IL12, was significantly decreased in DCs from GC-treated tumors as well ( Fig. 3e, left panel). Conversely, the expression of a pro-tumoral immunosuppressive, class 2 cytokine, IL10 33 , was significantly increased (Fig. 3e, middle panel), while no difference was observed in the expression of the Th2-polarizing cytokine, IL4 (Fig. 3e, right panel). These results revealed that dendritic cells in GC-treated tumors expressed markers indicative of a diminished maturation and Th1-promoting cytokine 34 .
Gemcitabine-cisplatin chemotherapy promotes inhibitory DAMP release to attenuate dendritic cell maturation. To assess how GC chemotherapy diminished DC maturation, we first investigated whether GC chemotherapy impacts the release of DAMPs and inhibitory DAMP (iDAMP). The success of drug-induced immunogenecity is contingent on the cell-surface expression or extracellular release of immunological adjuvants (i.e., DAMPs) by dying cancer cells to facilitate the maturation of DCs, leading to the priming and recruitment of CD8+ T cells 8,35 . Most recently, we highlighted an intricate balance between immunostimulatory and inhibitory DAMP release that converges as "signal 0" to dictate the maturation of dendritic cells 11 . Considering GC chemotherapy resulted in a diminished infiltration of CD8+ T cells (  Fig. 2A); and accordingly, increased HMGB1 extracellular release into the cultured media ( Supplementary Fig. 2B). Furthermore, both G69 and G7 syngeneic tumors treated with GC chemotherapy revealed an increased serum expression of HMGB1 ( Supplementary Fig. 2C, ELISA), when compared with vehicle-treated controls. Taken together, these results confirmed GC chemotherapy induces the release of immunostimulatory DAMPs in vitro and in vivo.
Coinciding immunostimulatory DAMP release, GC chemotherapy-treated G69 cancer cells displayed increased COX-2 protein expression (i.e., upstream prostaglandin metabolizing enzyme, Fig. 4a), resulting in the elevated release of the inhibitory DAMP (i.e., PGE 2 ) into the cultured medium, when compared to vehicle-treated cells (Fig. 4b). Gemcitabine-cisplatin chemotherapy-treated tumors in vivo also exhibited increased protein expression of COX-2 ( Fig. 4c), and upregulated release of PGE 2 into serum when compared to vehicle-treated mice (Fig. 4d). Furthermore, COX-2 upregulation was validated by immunofluorescence analysis comparing vehicle-and GC-treated tumor tissues (Fig. 4e). Taken together, these results supported our previous finding that extracellular release of the inhibitory DAMP (or PGE 2 ) counteracted the adjuvanticity of immunostimulatory DAMPs in precluding DC maturation 11 .
Next, we assessed the immunomodulatory effects of DAMPs and iDAMP on DC maturation by two independent methods. First, we collected conditioned media from GC chemotherapy-treated G69 cells (Fig. 4f). We then generated bone-marrow-derived DCs (BMDCs) by culturing bone marrow cells following a standard protocol using GM-CSF, as previously described 36 (Fig. 4g). These BMDCs were subsequently cultured in conditioned media from G69-tumor cells pre-treated with GC chemotherapy in the presence or absence of iDAMP blockade (Fig. 4f, g). The inhibitory DAMP blockade was achieved by two methods: (i) treatment of G69 cells with GC chemotherapy in the presence of celecoxib-a COX-2 specific inhibitor, and (ii) a PGE 2 neutralizing antibody to sequester PGE 2 in the culture media. BMDCs were treated with these conditioned media for 24 h and harvested for flow cytometric analyses of DC maturation markers (Fig. 4g). Inhibitory DAMP blockade by both approaches consistently induced increased expression of maturation markers (i.e., CD40, CD86, H2-k) in BMDCs (Fig. 4g, red histograms), which was compared with LPSstimulated BMDCs (as positive control) and untreated tumor cells (as NT negative control) (Fig. 4g, light gray and dark gray histograms, respectively). As anticipated, BMDCs cultured in conditioned media without iDAMP blockade failed to upregulate DC maturation markers (Fig. 4g, dark gray histograms). The mean fluorescent intensities (MFIs) for each of these conditions are also presented in Fig. 4h as bar graphs. As a complementary approach, we adapted and performed a classical in vivo vaccination assay 11,37 . Gemcitabine-Cisplatin chemotherapy-treated G69 cancer cells (IC50 dose to induce cell death) in the presence or absence of iDAMP blockade were injected into the mouse footpad as "vaccines". The maturation status of dendritic cells from vaccine-treated draining lymph nodes was evaluated 3 days post-injection ( Fig. 4i, j). Flow cytometric analysis of popliteal lymph node revealed iDAMP blockade increased the expression of DC maturation markers (i.e., CD40, CD86, and H2-k), when compared to GC chemotherapytreated group control (Fig. 4j). Taken together, these results demonstrated that functional inhibition of the COX-2/PGE 2 axis promotes DC maturation in vitro and in vivo.
iDAMP blockade inhibits tumor growth by promoting proinflammatory DC maturation. To determine whether iDAMP blockade in preclinical models recapitulate its effects on modulating DC maturation, we treated G69 and G7 tumor-bearing mice with two cycles of GC chemotherapy in the presence or absence of celecoxib. In vivo celecoxib treatment was initiated 2 days prior to the first chemotherapy treatment and was administered daily for the duration of the experiment. The combination of GC chemotherapy with celecoxib (GC+ Celex) resulted in a significantly reduced tumor growth (and tumor weight at endpoint), when compared to GC chemotherapy treatment alone (Fig. 5a, b).
To evaluate the mechanisms underlying these significant differences in tumor volume, we examined whether celecoxib or iDAMP blockade promotes DC maturation in vivo. Co-treatment with celecoxib did not alter DC infiltration into tumors (Fig. 5c,  d), as assessed by flow cytometry. Remarkedly, DCs from GC + celecoxib-treated tumors exhibited significantly increased H2-k expression, when compared to DCs from GC-treated tumors (Fig. 5e, left panel). CD86 expression also exhibited a similar trend and was comparable to the vehicle control group; however, some variations existed between samples (Fig. 5e, middle panel). No difference was observed in CD40 expression (Fig. 5e, right panel). Importantly, PGE 2 blockade in vivo normalized IL12 expression to vehicle levels, while decreasing IL10 expression in tumor DCs (Fig. 5f left and middle panel, respectively). No difference was observed in the expression of IL4 (Fig. 5f, right panel). Evidence of DC interactions with CD8+ T cells was observed in tissues from both GC and GC+ Celex-treated tumors as expected (Fig. 5g), indicative of CD8+ T-cell priming at the tumor site.
iDAMP blockade-mediated tumor suppression is dependent on functional CD8+ T cells. To further evaluate how GC+ celecoxib mediated DC maturation (Fig. 5e, f) might impact T cells in vivo, we investigated whether iDAMP/PGE 2 blockade affects infiltration CD3+ T cells into tumors or polarization of CD8+ cytotoxic T cells into (Tc)-1 or 2 phenotypes. To evaluate this, we analyzed tumor tissues treated with GC+ celecoxib and other control groups by flow cytometry. Indeed, iDAMP blockade resulted in a significantly increased CD3+ T-cell infiltration as well as CD4+ and CD8+ T-cell subsets when compared with the GC-treated bladder tumors (Fig. 6a-c). Further immunofluorescence analysis revealed additional spatial redistribution of CD3+ TILs in mice with iDAMP blockade (Fig. 6d, e, GC + Celex), which are not reflected by flow cytometry analysis alone (Fig. 6a). Celecoxib enhanced the infiltration of T cells into the tumor core (Fig. 6e, Area 1-4) with a simultaneous decrease of CD3+ TILs in the tumor periphery when compared to GCtreated tumor tissues (Fig. 6d, e, quantified in 6F-G). Furthermore, we also evaluated the activation status of CD8+ T cells upon iDAMP blockade (i.e., GC+ Celex group). Flow cytometric analysis of CD8+ TILs revealed an increased expression of effector molecules such as IFNγ and GZMb upon iDAMP blockade (Fig. 6h, i and Supplementary Fig. 3A, B), as well as an increase in Tbet expression ( Fig. 6j and Supplementary Fig. 3C) compared to their GC-treated counterparts (i.e., GC group). Importantly, these reinvigorated CD8+ TILs were functionally tumoricidal, and indeed responsible for the improved efficacy of GC chemotherapy, since their functional depletion (via antibody) nullified the effects of iDAMP blockade (Fig. 6k). Collectively, the above results illustrate that iDAMP blockade reprograms the tumor microenvironment by facilitating a CD8+ Tc1-mediated antitumor immune response and tissue infiltration.
iDAMP blockade synergizes with chemotherapy to enhance immune-checkpoint response. While recent FDA approval of ICIs revolutionized the treatment of advanced and metastatic MIBCs, only a subset of MIBC patients is considered responders to ICIs (~15-30%) 38 . To address this unmet clinical need, several clinical trials are underway or have concluded in evaluating the combination of ICIs with GC chemotherapy. Disappointingly, two Phase 3 clinical trials combining ICIs with GC chemotherapy [IMvigor 310 and KEYNOTE-361] recently failed to meet their primary endpoint on improving the overall survival or objective response over monotherapies 17,18 , posing a new clinical challenge. Based on our preclinical findings that GC chemotherapy-induced a significant reduction of tumorinfiltrating T cells (Fig. 2a-g) and rendering them nontumoricidal (Fig. 2m), it might not be unreasonable that these Phase 3 clinical trials have resulted in disappointing efficacies.
Given the impact of iDAMP blockade in improving the infiltration of tumoricidal CD8+ T cells into the tumor core when combined with GC chemotherapy (Fig. 6e), we hypothesize that iDAMP blockade likely poses a superior option to combine with  (Fig. 7a). We treated G69 and G7 tumorbearing mice with different drug combinations (Fig. 7a), following the same treatment scheme in Fig. 2a with the addition of anti-PD1 drug on day 7, 9, and 11 of cycle 1 and days 21, 23, 25 of cycle 2 (Fig. 7a). The anti-PD1 treatment scheme was conceived after considering the time needed for the immune system to mount an initial response to tumor neo-antigens (~8 days) 39 . As expected, a combination of GC chemotherapy with anti-PD1 exhibited no effect on tumor growth when compared to celecoxib-or GC-treated tumors (Fig. 7b), recapitulating the disappointing results from recent Phase 3 clinical trials 17 .
Remarkably, iDAMP blockade via celecoxib demonstrated prolonged suppression of tumor growth when combined with GC chemotherapy and anti-PD1 (Fig. 7b, purple line; left panel), significantly superior to GC chemotherapy and celecoxib combination (Fig. 7b, blue line; left panel) in the G69 model. For the G7 syngeneic model, it is intriguing that the GC chemotherapy and celecoxib combination is as efficient as the GC chemotherapy plus anti-PD1 and celecoxib (Fig. 7b, right panel). It is also important to highlight that GC+ Celex (Fig. 7b, blue line) is significantly superior to GC plus anti-PD1 (Fig. 7b, yellow line), which essentially has no effects over GC chemotherapy alone (Fig. 7b, black line) for both G69 and G7 cancer models. Further immunofluorescence analyses of the treated tumors revealed that CD3+ T cells are primarily excluded along the tumor periphery (Fig. 7c, Area 1 and 2) or within stromal regions (Fig. 7c, Area 3 and 4) in the GC+ PD1 group. Remarkedly, iDAMP blockade induced T-cell infiltration into the epithelial tumor cell regions (Fig. 7d, Area 1-4; GC+ Celex+PD1 group), supportive of the remarkable therapeutic response in Fig. 7b.
High PTGS2 expression in pretreatment tissues correlates with poor ICI response in MIBCs. We have previously shown that PTGS2 (gene that encodes COX-2) expression is significantly upregulated in matching pre-versus post-chemotherapy-treated MIBC patients that are chemotherapy nonresponders 25,40 . To evaluate whether PTGS2 expression exhibits an association with ICI response in MIBC patients, we analyzed its expression in the IMvigor210 study where patients received ICI after they failed platinum-based chemotherapy 41 . Intriguingly, PTGS2 expression in pretreatment tissues significantly correlated with response to ICI therapy. Bladder cancer patients with stable or progressive disease (SD/PD, n = 230) exhibited a significantly higher PTGS2 gene expression when compared to patients demonstrating complete or partial response (CR/PR, n = 68) (p = 0.03, Fig. 7e).
Collectively, in conjunction with the preclinical studies in Fig. 7a-d, the clinical correlative studies in Fig. 7e confirmed PTGS2/COX-2 as a clinically relevant target, and pose "iDAMP blockade" as a rational combination with GC chemotherapy and ICI to be evaluated in future clinical trials (Fig. 7f).

Discussion
The advent of immune-checkpoint inhibitors (ICIs) has revolutionized the management of advanced and metastatic bladder cancer patients, which has remained stagnant for decades 38 . Despite the impressive survival benefit and durable response in a subset of patients (15-30%) characterized as responders to ICIs, they are usually limited to those with "inflamed tumors" (i.e., high T-cell infiltration in the tumor parenchyma) 38,42 . Recent efforts to increase the proportion of bladder cancer patients that can benefit from ICIs have primarily focused on combining ICIs with cisplatin-based chemotherapies. This combinatorial treatment modality (i.e., chemoimmunotherapy) has been tested in various cancer types--such as lung (IMpower130, Keynote 189), breast (IMpassion130), as well as head and neck cancers (Keynote 048) -where a significant improvement in overall patient survival was achieved [43][44][45][46] . However, at least one Phase 2 (NCT01524991) and two Phase 3 clinical trials in metastatic or locally advanced bladder cancer (IMvigor130 and KEYNOTE-361) have failed to demonstrate significant clinical benefit, when comparing the overall survival of bladder cancer patients who received chemoimmunotherapy to those who received standard chemotherapy 17,18,47 . These adverse outcomes suggest that additional, unexplored mechanisms such as the understudied bladder tumor microenvironment may influence the therapeutic efficacy of chemoimmunotherapy in this cancer type 48 .
In hindsight, insufficient preclinical studies were performed to evaluate the immunomodulatory roles of GC chemotherapy in bladder cancer, before its combination with ICIs in large-scale human clinical trials. One major technical challenge that has hampered such research in the field is the paucity of preclinical MIBC models that recapitulate the intratumoral immune profiles of human MIBC. For instance, in a high-profile study of bladder cancer, immune exclusion was postulated to be a clinically significant phenomenon that correlated with immune-checkpoint resistance 19 ; yet, the authors had to defer to a murine breast cancer model for mechanistic studies. Similarly, in another elegant study that evaluated the immunogenic effects of mitomycin C chemotherapy in the non-muscle-invasive bladder cancer setting, the authors deferred to a murine colon cancer model for mechanistic evaluation 49 . These are important studies Fig. 4 Gemcitabine-cisplatin chemotherapy promotes an overwhelmingly high release of inhibitory DAMP to attenuate DC maturation. a Western blot analysis of COX-2 expression in G69 cell lysates treated with either vehicle or GC for 48 h. GAPDH expression was used as loading control (n = 3 biologically independent experiments). b ELISA for PGE 2 release in the culture medium of G69 cells treated with either vehicle control or GC for 48 h (n = 4 independent biological samples). Data are presented as mean values ± SEM. (Two-tailed, unpaired t-test **p = 0.0016). c COX-2 expression in G69-tumor tissue lysates treated with either vehicle or GC for two chemotherapy cycles. GAPDH expression was used as a loading control (n = 3 biologically independent experiments). d ELISA for PGE2 release in the serum of G69-tumor-bearing mice treated with either vehicle or GC for 2 cycles (n = 5 and n = 4 biologically independent samples for vehicle and GC, respectively). Data are presented as mean values ± SEM. (Two-tailed, unpaired t-test ****p < 0.0001). e Immunofluorescence analysis of COX-2 expression in G69-tumor tissues treated with either vehicle or GC for two chemotherapy cycles (Scale bar: 100 μm; Images representative of n = 3 independent experiments). f Schematic representation of BMDC treatment with conditioned media in the presence or absence of iDAMP blockade. g Flow cytometric analysis of CD40, CD86, and MHCI expression in BMDCs treated with G69-conditioned media with or without iDAMP blockade. h Geometric mean fluorescence intensity (gMFI) of CD40, CD86, and MHCI molecule expression in BMDCs upon treatment with condition media as in g (n = 2 biologically independent samples). Data are presented as mean values. i Schematic representation of footpad vaccination assay. j Flow cytometric analysis of CD40, CD86 and MHCI expression in vaccination-draining lymph node CD11c + MHC2 + DCs upon injection of G69 cells treated with GC in the presence or absence of iDAMP blockade (n = 5 biologically independent samples). Non-vaccination-draining lymph node DCs were used as control (n = 3 biologically independent samples). Data are presented as mean values ± SEM. (One-way ANOVA-Tukey's multiple comparisons test. CD40: GC vs GC+ Celex, p = 0.045; GC vs GC+ aPGE2, p = 0.036; CD86: GC vs GC+ Celex, p = 0.046; GC vs GC+ aPGE2, p = 0.0007; H2-K: GC vs GC+ Celex, p = 0.0004; GC vs GC+ aPGE2, p = 0.005). Source data are provided as a Source data file. ns not significant.
investigating clinically relevant phenomena in human MIBCs. Therefore, this lack of relevant preclinical MIBC models underscores the significance of our current study in the development and characterization of autochthonous, immune-excluded MIBC mouse models (i.e., G69 and G7) for mechanistic investigations.
Using these immune-excluded MIBC models (i.e., G69 and G7), we found that standard GC chemotherapy significantly suppressed the frequency of intratumoral T-cell infiltration and increased the overall immune desert areas towards the center or core of the tumors. Although these tumor-infiltrating CD8+ T cells expressed classical markers indicative of their activation, functional evaluation via antibody depletion unexpectedly revealed they were nontumoricidal. These results suggest several plausible explanations. First, these CD8+ T cells were indeed activated but excluded to the tumor periphery; and therefore, unable to eliminate cancer cells due to their lack of tissue penetration into the tumor core. Secondly, since dendritic cells (DCs) are important partners for proper priming and activation of CD8+ T cells 50 , these tumor-infiltrating CD8+ T cells might not be properly primed by DCs. Indeed, the mechanistic investigation revealed that chemotherapy exerted negative effects to skew DC away from maturation and immunogenic polarization, which was also accompanied by an increased expression of the immunosuppressive cytokine, IL10, and a decreased expression of the Th1-polarizing cytokine, IL12. The upregulation of IL10 and downregulation of IL12 has been linked to antiinflammatory responses that support the development of a protumoral microenvironment 33 . Further mechanistic dissection also revealed an overwhelming increase of extracellular PGE 2 release from chemotherapy-treated cancer cells, a mechanism to hinder immunogenic DC maturation. These preclinical findings in vivo substantiated our recent findings posing extracellular PGE 2 as an inhibitory DAMP to counterpoise immunostimulatory DAMPs, negating their downstream signaling on dendritic cells to activate immunogenic cell death 11 . Using both in vitro and in vivo models, we demonstrated that pharmaceutical blockade of PGE 2 synthesis into the tumor microenvironment enhanced immunogenic DC maturation; thereby, reverting the tolerogenic status of DCs within the GC chemotherapy group. These effects of the COX-2/PGE 2 axis on dendritic cells corroborate with a previous study in a different context and cancer type 51 . In brief, tumorderived PGE 2 has been reported to inhibit CD103+ DC recruitment into the tumor microenvironment in the treatment naïve setting by promoting melanoma tumor growth through immune evasion 51 . In other MIBC xenograft and patient-derived xenograft models, our previous study and others reported chemotherapy induction of the COX-2/PGE 2 axis in promoting chemoresistance, via inducing compensatory proliferation of residual cancer stem cells 25 , and enhancing stemness via promoter methylation of let-7 and upregulation of the self-renewal gene SOX2 52 . Furthermore, in canine models of bladder cancer, COX-2 inhibition significantly enhanced chemotherapeutic response by~40% 53 .
Collectively, these prior studies and our current study convincingly show that pharmaceutical blockade of the COX-2/ PGE 2 axis could become a promising therapeutic approach in bladder cancer (in combination with chemotherapy and ICIs), since blockade of this axis has demonstrated consistent effects in targeting epithelial cancer stem cell repopulation, and reshaping the tumor microenvironment to allow CD8+ T-cell penetration into the tumor core. It was reported elsewhere that exogenous PGE 2 contributes to a loss of pericyte coverage along the vasculature by breaking down pericyte-endothelial cell interactions 54 . Pharmaceutical blockade of the COX-2/PGE 2 axis also restores pericyte-endothelial cell interactions to promote vasculature maturation (unpublished observations), which represents a plausible mechanism enabling more prominent T-cell penetration. Furthermore, our results here draw an important mechanistic connection between tumor cell-derived COX-2/PGE 2 , immunogenic DC activation, and T-cell-inflamed tumors 55,56 . The importance of matured DCs in priming intratumoral T-cell activation has been demonstrated mechanistically in other cancer types 57,58 . For instance, DCs secrete CXCL9 and CXCL10 as chemoattractors for recruiting CXCR3+ T cells into the tumor microenvironment 59 . Intriguingly, blockade of PGE 2 synthesis has been shown to increase the expression of CXCL9 and CXCL10 in murine models of cutaneous melanoma 51,60 , as well as our murine bladder cancer models (unpublished observations). However, the underlying mechanisms are not well studied. On the other hand, a lack of DC infiltration and maturation has been linked to defective CD8+ T-cell priming, leading to decreased IFNγ expression 61 . These studies have underscored the importance of DC presence in the tumor microenvironment as facilitators of antitumoral T-cell immunity. Nevertheless, the mechanisms of DC recruitment to the tumor bed and their activation are not fully understood. These recent studies underscore the importance of our current study and highlight the significance of tumor-derived PGE 2 in regulating DC activation in MIBCs-in the chemotherapy treatment context-a notion that has not been clinically evaluated in this cancer type. Independently, these findings also highlight the concept to exploit drug-induced immunogenic cell death in bladder cancer.  Although accumulating evidence has revealed a handful of chemotherapies exert beneficiary immunomodulatory effects by inducing immunogenic cell death in breast, colon, and non-small cell lung cancers [62][63][64] , the immunogenic effects of gemcitabinecisplatin combination remained controversial and has not been fully elucidated until the current study. Indeed, unpublished observations by others indicate another more toxic chemotherapy regimen-i.e., MVAC (methotrexate, vinblastine, adriamycin, and cisplatin) seems to exert more clinical benefits than GC chemotherapy when combined with ICI in bladder cancer patients. Our unpublished observations revealed adriamycin (or doxorubicin) is likely responsible for the immunogenic effects of the MVAC chemotherapy combination, since adriamycin induces a relatively high membrane exposure of immunostimulatory DAMPs with a low iDAMP/PGE 2 release into the tumor microenvironment.
Our findings pose "iDAMP blockade" as the next-generation drug combination to be evaluated with chemoimmunotherapy clinically, by converting immune-excluded tumors into T-cellinflamed tumors when coadministered. Indeed, a new class of antagonists that target PGE 2 receptors (EP) on downstream effector cells (e.g., dendritic cells or fibroblasts) might provide a next-generation approach for "iDAMP blockade". However, it remains unclear whether this therapeutic combination would be more efficacious if chemoimmunotherapy was given in sequence 65 . New evidence from recent Phase 2 and 3 clinical trials suggested that a "switch maintenance" approach might provide superior clinical benefit when compared to the concurrent treatment approach 66,67 . In these trials, advanced bladder cancer patients that responded to or had stable disease after treatment with platinum-based chemotherapy were switched to maintenance immunotherapy, who demonstrated a superior overall (OS) and progression-free survival (PFS) compared to patients assigned to best supportive care. Similar results were obtained in stage III lung cancer patients that did not progress after platinumbased chemotherapy and were switched to maintenance immunotherapy 68 . These recent findings implicate that treatment sequence is another important determinant to a successful therapeutic response, that should be carefully considered.
Collectively, the current findings provide evidence that therapeutic targeting of COX-2 can be explored to convert immuneexcluded into T-cell-inflamed bladder tumors; thereby, potentially expanding the response of MIBC patients towards chemoimmunotherapy. And thus, poses a promising strategy for future clinical evaluation.
Mice. All animal procedures were performed in ethical compliance and with approval by the Institutional Animal Care and Use Committees at Cedars-Sinai Medical Center. Wild-type male FVB/NJ mice at 6-to 8-week-old of age were utilized for all in vivo experiments. Mice have been housed in Cedars-Sinai Medical Center vivarium at ambient room temperature (22 ± 2°C) with humidity of 40-70% and a light/dark cycle of 12 h/12 h.
Tumor growth. G69 and G7 cells were harvested at their exponential growth phase (i.e., 70-80% confluency) using TrypLE express enzyme (Thermo, 12604013) and pelleted by centrifugation at 1200 × g for 5 min and washed once with DPBS (Sigma, D8537). 5 × 10 5 cells were suspended in 20 uL of DPBS and injected into the left flank of mice. Tumor growth was recorded twice a week with caliper. Tumor volume was calculated using the standard formula (width × length × length/ 2). For tumor growth studies, the endpoint was 2 days after the last treatment unless mice were in moribund status, defined as signs of prolonged distress, tumor interfering with ambulation (normal movement), or access to food and water, or when tumor displayed signs of ulceration, or tumor size of >15 mm in diameter. The maximal tumor size was not exceeded.
Treatment of G69 tumors with chemotherapy and anti-PD1 in vivo. Mice with established G69 or G7 tumors (i.e., tumor volume of~50 mm 3 ) were randomized into treatment groups (using RANDOM.com) and injected intraperitoneally (i.p.) with either vehicle (10% DMSO in 0.05% NaCl solution) or GC (gemcitabine at 120 mg/Kg; Fisher Scientific and cisplatin at 6 mg/Kg; Sigma-Aldrich) on day one and gemcitabine on day 8 to complete one cycle of chemotherapy. Mice were left to recover from chemotherapy for seven days upon which they received the second cycle of GC chemotherapy. Two days after the last gemcitabine treatment, tumors were harvested and processed for subsequent analyses. For the iDAMP blockade group, mice received celecoxib (10 mg/Kg; Sigma-Aldrich) daily until the completion of the experiment. For single therapy treatment, anti-PD1 antibody (200 ug/mouse, RMP1-14, Bio x Cell) was injected i.p. every other day for a total of three doses. For combination treatments, anti-PD1 was injected i.p. on days 9, 11, 13 of each chemotherapy cycle.
Footpad vaccination assay. G69 cells were seeded at 4.7 × 10 4 cells per mm 2 and were treated with IC50 concentrations of cisplatin and gemcitabine with or without celecoxib for 24 h in vitro. After 24 h, adherent cells were washed with DPBS (Sigma, D8537) thoroughly, dissociated with TrypLE express enzyme, and pelleted by centrifugation (1300 × g for 5 min at room temperature). Cell pellets were washed with DPBS once more to ensure clearance of residual enzyme and pharmacological agent(s). 5 × 10 5 cells were suspended in 15 μL of DPBS and were injected into the footpad of mice. Five days following vaccination mice were euthanized and the vaccine-draining lymph nodes (i.e., popliteal) were collected for subsequent flow cytometric analysis.
CD8 + cell depletion. Mice were injected subcutaneously with aCD8 mAb (BioXcell, BE0004-1-A005) at a concentration of 200 μg per 30 mg mouse. Mice were injected every other day with the aCD8 mAb for the duration of the chemotherapy treatment. Efficacy of CD8-depletion was checked using murine tailvain blood sampling coupled with flow cytometry analysis. Fig. 7 iDAMP blockade synergizes with chemotherapy to enhance immune-checkpoint response. a Schematic representation of ICBT incorporation into the GC treatment cycles. Anti-PD1 (clone: 29F-1A12) was injected ip on days 7, 9, and 11 of each GC cycle. b G69 (left panel) and G7 (right panel) tumor growth curves of vehicle (n = 8 biologically independent animals), Celecoxib (n = 7 biologically independent animals), anti-PD1 (n = 7 biologically independent animals), Celecoxib+anti-PD1 (n = 6 biologically independent animals), GC (n = 8 biologically independent animals) and GC+ anti-PD1 (n = 6 biologically independent animals), GC+ Celecoxib (n = 7 biologically independent animals), and GC+ Celecoxib + anti-PD1-treated mice (n = 8 biologically independent animals) for two cycles of chemotherapy. Data are presented as mean values ± SEM. (One-way ANOVA-Tukey's multiple comparisons test. Left: *p = 0.045, **p = 0.005; Right: *p = 0.013, **p = 0.007). c, d Immunofluorescence analysis of CD3+ T-cell infiltration in GC+ PD1, and GC+ Celecoxib+PD1-treated G69 tumors. Selected ROIs (1)(2)(3)(4) are representative areas that depict CD3+ T-cell infiltration in the tumor core and the tumor periphery (Scale bar: 100 μm; Images representative of n = 3 independent experiments). e PTGS2 (COX-2 gene) expression in complete/partial responders (CR/PR, n = 68) and stable/progressive disease (SD/PD, n = 230) MIBC patients from chemotherapy-ineligible or refractory MIBC patient cohort (Imvigor210) treated with atezolizumab (anti-PD-L1). (Wilcoxon paired two-sided rank-sum test p = 0.03). Boxplots are drawn as the inter-quartile range (IQR) with a line indicating the median, and outliers defined as points that fall outside of the range demarcated by 1.5×IQR. f Schematic representation of COX-2 blockade as a valid therapeutic target to improve chemoimmunotherapy response for MIBC patients with immune-privileged tumors. Source data are provided as a Source data file. Processing vaccine-draining lymph nodes for flow cytometry. Vaccine-draining lymph nodes were harvested 5 days post-vaccination of the murine footpad. Nondraining lymph nodes (i.e., opposite foot) were harvested as controls. Harvested lymph nodes were passed through a 70 μ mesh filter. Flow-through immune cells were then centrifuged. Lymph node cells were then collected, washed with DPBS, and processed for flow cytometric analysis.
Generation of bone-marrow-derived dendritic cells. Generation of bonemarrow-derived dendritic cells followed a stepwise protocol previously described by Helft et al. 36 . In short, bone marrow cells were collected from the murine femur and tibia. In all, 10 × 10 6 bone marrow cells were cultured in RPMI media (Sigma, R8758) supplemented with 50 μM beta-mercaptoethanol (Santa Cruz Biotechnology, sc-202966), 10% heat-inactivated fetal bovine serum (Thermo, A3840001), and 20 ng/mL of GM-CSF (Miltenyi,130-094-043) for 2 days. On day 2, half of the media was discarded and new fully supplemented media with GM-CSF (2×, 40 ng/ ml) was added to cultures. On day 3 the entire culture media was discarded and replaced with fresh fully supplemented media with GM-CSF (20 ng/mL) until day 6.
Treating BMDCs with cultured media. Non-adherent BMDCs in the culture supernatant and loosely adherent BMDCs from 6 days of culture were harvested using 5 μM EDTA solution. 1 × 10 5 cells were suspended in GC pre-treated G69 cultured media (20% of total volume and 80% base DC media as described above) in the absence or presence of iDAMP blockade. Cultured media PGE 2 sequestration was achieved by treating the cultured media with 100 nM of anti-PGE 2 mAb (Cayman, 10009814) for at least 30 min on ice. LPS (InvivoGen; tlrl-3pelps; 100 ng/ ml) treatment was used as a positive control. BMDCs were cultured and collected at 24-h post-activation with appropriate culture medium treatment.
Quantification of CD3+ T-cell density. Immunofluorescence images were acquired using Zeiss Axio Imager 2 at ×20 magnification and analyzed using Zen software (version 3.3, blue edition). Regions of interest (ROI's) were defined randomly from the whole tissue section and 3 to 5 ROI's were selected per tumor tissue section. CD3+ cells were defined based on CD3 and DAPI positive costaining. Immune cell infiltration was determined by performing linear regression, comparing the density of CD3+ cells within each ROI to the normalized mean distance of the ROI to the tumor periphery. The density of CD3+ cells was determined by calculating the number of CD3+ cells per area of ROI. The mean distance of the ROI to the tumor periphery was determined by calculating the mean nearest distance of the ROI border to the tumor periphery. Mean distances between samples were then normalized based on the maximum distance between the tumor core and tumor periphery 19 . . Concurrently, these DAMP-stained cells were labeled with a cell viability dye (DAPI at a 3 μM concentration) for dead cell exclusion. All antibody cocktails were diluted in ice-cold FACS buffer or Briliant Stain buffer (BD, 563794) when more than one brillian violet-conjugated antibody was used. Cells were incubated with the appropriate antibody cocktails on ice for 20 min and washed thoroughly with ice-cold FACS buffer. Flow cytometric analysis of immune cells was performed on BD LSRFortessa™ and Cytek™ Northern Lights cell analyzers. Flow cytometric analysis of cell-surface CRT and HSP70 was performed on BD LSRFortessa™. All flow cytometry data were processed using FlowJo software v.10.7.1.
ELISA. ELISA analysis of PGE2 (Cayman, 514010) and HMGB1 (Arigo, ARG81310) released by cancer cells and serum from G69 and G7 tumor-bearing mice was performed according to the manufacturer's protocol.
PTGS2 expression in ICB-treated bladder cancer patients. PTGS2 expression data were obtained from RNA-seq analysis of patient samples of a large phase II clinical trial (IMvigor210) that investigated the clinical activity of Atezolizumab (PD-L1 monoclonal antibody) in patients with locally advanced and metastatic urothelial carcinoma 41 . There were 68 samples from patients classified as complete or partial responders (CR/PR) and 230 samples from patients classified as having a stable disease or progressive disease (SD/PD). Two sample T-test was performed to compare PTGS2 gene expression between the CR/PR and SD/PD groups using the normalized RNA counts.
Statistics and reproducibility. All data were evaluated and graphed using Prism GraphPad ver. 9 software. Statistical comparison between control and experimental groups was performed utilizing two-tailed Student's t-test, one-way analysis of variance (ANOVA) with multiple comparisons test when appropriate. Quantified data shown were repeated at least three times in independent experiments (unless specified differently). Data shown are represented as mean ± SEM. P-value < 0.05 was considered statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The Mariathasan et al. 19 data used in this study are available in the European Genome Phenome Archive under accession number EGAS00001002556. In addition, source code and processed data used for all analyses presented in Mariathasan et al. 19 are available in IMvigor210CoreBiologies and can be downloaded from [http://research-pub.gene.com/ IMvigor210CoreBiologies]. All the source data and uncropped scans of blots supporting the findings of this study are provided with this paper as Source data file. The remaining data are available within the Article and Supplementary Information. Source data are provided with this paper.