Female and male BALB/c mice (7–8 weeks of age) were purchased from the Charles River Frederick colony. Mice were maintained on a low-fat chow diet (LFD) for 1 week to acclimate, then randomized to either LFD or a high-fat diet (HFD; Research Diets; catalog #12492) for 20 weeks and were allowed ad libitum access to food and water throughout. After 20 weeks, lean, diet-induced obese (DIO) and OB-resistant (OB-Res) mice were identified as described,8 20 then randomized to treatment groups. Mice were housed five to a cage in standard caging under specific-pathogen-free conditions in 12:12 light:dark cycles at 22°C (72 °F average).

Renca, a BALB/c syngeneic renal adenocarcinoma cell line, was purchased from American Type Culture Collection (ATCC), engineered by lentiviral transduction to stably express firefly Luciferase, and cultured as described.21 Cells were used at the same passage number throughout the study to minimize experimental variation. CMS5, a BALB/c syngeneic fibrosarcoma cell line, was a generous gift of Paul Allen (Washington University School of Medicine) and was maintained as described.22 Cell lines were not validated after purchase/receipt. Renca and CMS5 cells were tested and identified as mycoplasma-negative.

On day 6 post-tumor challenge, BLI was performed to confirm that tumor burdens were statistically equivalent across experimental groups prior to randomization to treatment. BLI was accomplished using an IVIS Lumina III (PerkinElmer) and analyzed as described.21 Tumor luminescence was calculated in radiance (photons/sec/cm²/sr) using the Living Image Software, V.4.5.2.

Murine tumors were excised and stored in RNAlater stabilization solution (ThermoFisher) at −80°C until bulk RNA isolation using the Qiagen RNeasy Mini Kit. RNA integrity was determined using an Agilent 2100 Bioanalyzer. Samples with RIN values >8 were used in iRepertoire analyses.

As a screening tool, immune-related gene expression patterns were interrogated on whole-tumor RNA using the nanoString murine PanCancer Immune Profiling Panel according to manufacturer’s instructions by the UAB nanoString facility.

Murine renal tumors were harvested, weighed, and disrupted using a Miltenyi GentleMacs Dissociator or pestle (USA Scientific) in Dulbecco’s Phosphate Buffered Saline (DPBS) at a concentration of 0.2 g of tissue per 1 mL of DPBS. Tissue supernatant was collected and stored at −80°C until processing. Murine cytokines/chemokines in tumor homogenates and human plasma leptin were quantified using the Bio-Plex Multiplex System (Bio-Rad) and analyzed on a Bio-Plex 200 (Bio-Rad) or MAGPIX (ThermoFisher) instrument.

Statistical analyses were performed utilizing Prism, V.7.00 (GraphPad), unless specifically noted elsewhere. Gaussian distribution was assessed using Shapiro-Wilk normality testing. Throughout, statistical differences between two groups were analyzed using parametric two-tailed unpaired Student’s t-tests or non-parametric two-tailed Mann-Whitney U tests, as appropriate. Statistical significance is denoted as asterisks (*) for parametric tests or as pound signs (#) for non-parametric tests. Statistical differences among multiple groups with one independent variable were analyzed using parametric or Kruskal-Wallis nonparametric one-way analysis of variance (ANOVA) with Dunn’s post hoc multiple comparison testing. In studies with multiple groups and two independent variables (ie, obesity status and therapy groups), two-way ANOVAs were performed with Bonferroni’s post hoc multiple comparison testing. Statistical differences in studies examining survival outcomes in mice were calculated using Log-rank (Mantel-Cox) testing. To examine linear correlations, linear regression analyses were performed where indicated. To assess changes in tumor weights over time, mixed-effects modeling with a Bonferroni correction was used. Note that in all figures, statistical comparisons between all groups were performed; however, only comparisons that were significantly different or trending (p<0.15) were shown on figures to improve data clarity.

Gene expression data from this study are available from the corresponding author on reasonable request.

Written informed consent was received from each subject prior to inclusion and participation in our study. All animal procedures were approved by the Institutional Animal Care and Use Committee at UAB, an AAALAC-accredited institution.

We began by examining survival outcomes of RCC patients with and without obesity following anti-PD-1 treatment. Recent retrospective studies investigating the effects of obesity on immunotherapy outcomes in RCC have provided intriguing but conflicting results: that increased adiposity (defined as BMI >25 kg/m²) trended toward being detrimental,18 that obesity (BMI >30 kg/m²) had no impact on outcomes in multivariate analysis,16 or that obesity improved PFS but not OS in patients who showed primary clinical benefit.19

We examined this issue by evaluating outcome data for all RCC patients treated with standard of care anti-PD-1 who had ≥6 months of follow-up from treatment initiation at the UAB Hospital (n=54) and the UI Hospitals and Clinics (n=18). All patients had an ECOG performance status of 0–2 and none had received prior immunotherapies. Demographic information for these patients can be found in online supplemental table S1. The WHO-defined BMI cut point of 30 kg/m² was used to identify patients with obesity (OB) or without (NOB). The mean BMI of the OB cohort was 38.74 kg/m² vs 25.07 kg/m² for the NOB. Obesity at treatment initiation was associated with a significant reduction in PFS (p=0.0448), as well as OS (p=0.0288) in Kaplan-Meier analysis (figure 1A,B). Specifically, obesity was associated with a 6.5 month-reduction in median PFS and a 12-month reduction in OS. Notably, after controlling for patients’ age, sex, number of prior treatments received and IMDC risk score, the Cox proportional hazard regression on PFS indicated that the hazard of disease progression in patients with BMI <30 kg/m² was 54% (95% CI (0.31% to 0.95%; p=0.0322)] of that seen in patients with BMI ≥30 kg/m²; for OS the risk was 48% (95% CI 0.24% to 0.96%; p=0.0371). When BMI ≥25 kg/m² was used as the cut-point instead of 30 kg/m², to permit evaluation of the combined effects of overweight and obesity, the Kaplan-Meier curves for PFS and OS were no longer statistically different (p=0.5520 and p=0.8279, respectively) (online supplemental figure S1). Therefore, host obesity was associated with unfavorable survival outcomes in RCC patients treated with anti-PD-1 immunotherapy.

We next explored the possibility that underlying differences in CD8⁺ tumor-infiltrating lymphocyte (TIL) phenotypes existed in the presence of obesity that might contribute to the divergent response to anti-PD-1. PD-1 expression on CD8⁺ T cells regulates T cell function and cancer immunotherapy outcomes.26 27 Although frequently regarded only as an indicator of T cell functional exhaustion, PD-1 is upregulated following antigen recognition during the earliest stages of T cell activation.28 Thus, in the tumor microenvironment, PD-1 expression is used as an indicator of T cell encounter(s) with antigens,27 29 as repeated TCR engagement increases PD-1 levels.26 30 31

We therefore analyzed PD-1 expression on CD8⁺ TILs from a subset of treatment-naive RCC subjects with (BMI ≥30 kg/m²; n=11) or without (BMI <30 kg/m²; n=8) obesity. Demographic information for these RCC subjects can be found in online supplemental table S2. This analysis revealed that obesity did not alter the activation status (CD45RA^-HLA-DR⁺) of CD8⁺ TILs; however, we observed significantly reduced percentages of activated PD-1⁺CD8⁺ TILs and PD-1^highCD8⁺ TILs with obesity (figure 2A, gating as per online supplemental figure S2A). No obesity-associated alterations in the frequency of activated PD-1⁺CD8⁺ T cells in the peripheral blood of RCC subjects (n=82) or tumor-free donors (n=19) were found (figure 2B). Demographic information for tumor-free donors is in online supplemental table S3. RCC subjects with obesity displayed the expected elevations in plasma leptin (figure 2C and online supplemental figure S2B), but we found negative correlations between the frequencies of circulating activated PD-1⁺CD8⁺ T cells and plasma leptin concentrations (figure 2D) or subject BMI (online supplemental figure S2C), contrasting with prior findings in melanoma.11

To address whether similar trends existed regarding PD-1 expression on CD8⁺ TILs in mice with renal tumors, we used our well-characterized model of DIO.8 20 In this model, BALB/c mice are randomized to either a HFD or a LFD for 20 weeks prior to orthotopic renal tumor challenge. Individual HFD-fed mice are identified as DIO only if their body weight is ≥3 SD above the mean body weight of the age-matched LFD-fed lean cohort (figure 2E). DIO mice identified in this manner display classic hallmarks of obesity (increased adiposity, serum leptin and insulin and adipose tissue crown-like structures).8 20 Following orthotopic, syngeneic renal tumor challenge, an equivalent percentage of CD8⁺ TILs cells were activated (CD44⁺) in DIO and lean animals, with ~75% of these CD44⁺ T cells expressing PD-1, regardless of obesity status at a late day 28 time point (figure 2F, gating as per figure 3). Similar to what was observed in RCC subjects, DIO mice had decreased frequencies of activated PD-1^highCD8⁺ TILs. Notably, PD-1 expression on activated CD8⁺ TILs from lean mice increased from day 15 to day 28 post-tumor challenge (figure 2G), reflecting prior reports that sustained antigen presence and repeated TCR engagement drive heightened PD-1 expression.30–32 In contrast, CD8⁺ TILs from DIO mice did not display increased PD-1 expression over time, resulting in significantly reduced PD-1 expression on CD8⁺ TILs from DIO mice versus lean mice at day 28 (figure 2G). Thus, we found that obesity was associated with reduced frequencies of PD-1^highCD8⁺ TILs in both human subjects with RCC and an orthotopic mouse model of renal cancer, implying that CD8⁺ TIL responses are indeed altered by obesity and illustrating the clinical relevance of our murine tumor model for further study.

We next assessed the functional capacity of PD-1^highCD8⁺ TILs from treatment-naive mice as prior studies indicated that, depending on the tumor type,33 PD-1⁺CD8⁺ TILs can exist either as functional effectors or as exhausted cells that are resistant to re-invigoration with anti-PD-1. We found a higher frequency of PD-1^highCD8⁺ TILs secreting perforin and tumor necrosis factor (TNF)α relative to PD-1^int or PD-1^neg subsets (figure 3A,B). This pattern was conserved within both lean and DIO cohorts. Similar patterns were observed for interferon (IFN)γ secretion in lean, but not DIO, mice (figure 3C). Thus, obesity is associated with reduced frequencies of functionally-competent, PD-1^highCD8⁺ TILs in the treatment-naive setting, suggesting the presence of a diminished endogenous T cell response.

Given the obesity-associated decrease in functional PD-1^highCD8⁺ TILs in our murine model, we hypothesized that host obesity would impair anti-PD-1 immunotherapy outcomes. However, single agent anti-PD-1 was unable to reduce day 28 excised renal tumor burdens in either lean or DIO mice (figure 4A,B, and online supplemental figure S3A, respectively). Notably, lean mice treated with anti-PD-1 did experience significant, but modest, improvements in survival relative to NT controls (figure 4C), an outcome reflecting survival benefits detected in RCC patients who receive anti-PD-1 monotherapy. Consequently, we developed a combinatorial therapeutic approach to evaluate the effect of host obesity on immunotherapy outcomes, wherein anti-PD-1 was administered downstream of our previously described in situ T cell priming therapy.8 34 This priming approach consists of adenovirus encoding murine TRAIL (AdTR) coadministered with the toll-like receptor 9 agonist CpG (AdTR/CpG) at day 7 post-tumor challenge (figure 4A), after renal tumors are established and when tumor burdens are statistically equivalent based on BLI (figure 4H).35 36 Both AdTR37 and CpG38 have been used in phase 1 clinical trials with low toxicity, illustrating potential translational relevance. Here, combinatorial AdTR/CpG/PD-1 reduced renal tumors by 75% in young, lean mice and was more efficacious than either AdTR/CpG or anti-PD-1 alone (figure 4B), leading to prolonged survival through day 100 (figure 4C). A similar analysis in older, DIO mice revealed that neither single agent anti-PD-1, AdTR/CpG, nor the combination of AdTR/CpG/PD-1 led to a significant reduction in excised renal tumor weights (online supplemental figure S3A). However, the combination of AdTR/CpG/PD-1 did reduce tumor weights by 42% in DIO mice, an improvement over the 3% and 4% reductions detected in mice treated with either anti-PD-1 or AdTR/CpG, respectively.

We therefore used this AdTR/CpG/PD-1 approach to evaluate the impact of obesity on therapeutic efficacy in age-matched mice. Lean and DIO mice (figure 4D) were established, challenged with orthotopic renal tumors, then treated with AdTR/CpG/PD-1 (figure 4E). Because AdTR/CpG/PD-1 reduced excised tumor weights by an average of 75% in young, lean mice relative to untreated controls (figure 4B), we classified individual AdTR/CpG/PD-1-treated mice as responders if they exhibited a ≥75% reduction in excised tumor weight, relative to the mean tumor weight of their respective untreated controls (figure 4F). These objectively determined responder thresholds are indicated by dotted lines in figure 4F. Lean mice maintained on LFD for 20 weeks displayed significant reductions in tumor weights after receiving AdTR/CpG/PD-1 immunotherapy (figure 4F), translating to a response rate of 73% (figure 4G). In contrast, immunotherapy did not significantly alter tumor weights in DIO mice (figure 4F) and a response rate of only 44% was achieved (figure 4G), demonstrating impaired immunotherapeutic outcomes in DIO mice. Next, we evaluated changes in excised renal tumor weights over time, and found that at day 15 post-tumor challenge, tumor weights were not significantly different among lean and DIO mice, in the absence or presence of AdTR/CpG/PD-1 therapy (online supplemental figure S4). However, by day 28, lean mice receiving AdTR/CpG/PD-1 exhibited a significant reduction in renal tumor burdens relative to lean and DIO NT controls and DIO mice receiving AdTR/CpG/PD-1(online supplemental figure S4). In DIO mice, no significant reductions in tumor burdens were observed following therapy administration (online supplemental figure S4). These outcomes reflect our clinical observations in RCC patients treated with anti-PD-1 (figure 1), further highlighting the clinical relevance of our preclinical renal tumor model.

As lean and DIO mice are fed different diets, we investigated the possibility that HFD administration, rather than obesity, was impairing therapeutic efficacy in DIO mice. To do this, we used our previously characterized OB-Res BALB/c mice.20 OB-Res mice are DIO cage-mates maintained on the same HFD for 20 weeks, but they demonstrate no weight gain relative to LFD-fed controls (figure 4D). More importantly, we have reported that OB-Res mice are physiologically comparable (ie, adiposity, serum leptin and insulin) to lean LFD-fed mice.20 AdTR/CpG/PD-1 efficacy was maintained in OB-Res mice (figure 4F), resulting in a response rate of 73% (figure 4G), equivalent to that observed in lean mice. Therefore, the loss of therapeutic efficacy in DIO mice was linked to obesity, rather than HFD use.

We then examined potential underlying factors that could contribute to decreased tumor clearance and response rates in DIO mice. These included: tumor burdens at treatment initiation (figure 4H), body weights of individual DIO mice at tumor challenge, the frequency of PD-L1⁺ CD45⁺ or CD45^- cells within renal tumors, and day 28 excised renal tumor weights in untreated mice (online supplemental figure S3B–E). No significant changes in any of the above parameters were found. We also interrogated intratumoral T cell receptor (TCR) diversity via deep sequencing at day 28 in untreated lean and DIO mice, as it was previously reported that obesity compromises TCR repertoire diversity.39 However, whole tumor αβ T cell repertoire sequencing revealed no significant differences in magnitude or diversity between DIO and lean mice (online supplemental figure S3F). Finally, as these studies were conducted using female mice, we asked if our observations extended to male mice. We determined that the reduced response rate to immunotherapy also occurred in male DIO mice, as only 17% of DIO males but 78% of lean males responded to AdTR/CpG/PD-1 (online supplemental figure S3G), demonstrating that loss of immunotherapeutic efficacy in DIO mice was consistent in both males and females. Therefore, none of the above factors appeared to contribute to differential AdTR/CpG/PD-1 response rates in DIO versus lean mice.

To identify potential mediators of obesity-associated immune dysfunction within tumors, we evaluated gene expression profiles (GEPs) from whole renal tumors in untreated lean and DIO mice. We performed nanoString Immune profiling of 750 immune-related genes in tumor-free or tumor-bearing kidneys from treatment-naive DIO versus lean mice. GEPs were remarkably conserved between lean and DIO treatment-naive tumors and tumor-free contralateral kidneys (online supplemental figure S5A,B), illustrating that underlying immune responses were minimally impacted by obesity.

Next, we asked if obesity altered immune-related GEPs following AdTR/CpG/PD-1 therapy. We therefore used nanoString immune profiling as a screening tool to identify broad immunogenetic changes in whole renal tumors from individual lean and DIO untreated mice, as well as treatment responders and NR. Here, unbiased two-way hierarchical analysis revealed that all lean and DIO treatment responders clustered together (figure 5A) and were characterized by a conserved group of highly expressed genes (denoted by the yellow dendrogram). Strikingly, DIO NR clustered with lean and DIO no therapy (NT) mice, demonstrating their lack of robust immune remodeling in response to therapy administration. Lean NRs were characterized by a conserved group of highly expressed genes (denoted by the green dendrogram).

To more fully understand GEPs that characterized a productive response to therapy, we used volcano plots to identify genes that were differentially expressed (DE; p<0.05 significance threshold indicated by dotted line) in lean or DIO responders versus their respective NR group (figure 5B). We identified several gene expression changes conserved in both lean and DIO responders. Of the shared responder genes (online supplemental table S4), many of those upregulated were indicative of CD8⁺ T cell immunity (ie, CD3e, CD3g, Cd8a, Cd8b, Ccl5, Gzma, Prf1). Conserved downregulated genes included those related to myeloid-lineage population dynamics (ie, Csf2, Ccl2, S100a8, Cxcr2, Il1r1, Il1b and Arg1). These patterns were reinforced through our analysis of nanoString cell type scores, which showed that tumors from lean and DIO responders had increased CD8⁺ T cell (‘CD8 T’) and cytotoxic cell scores (‘Cyto’, comprizing both T cells and NK cells) relative to NR. Responders also had simultaneous reductions in dendritic cell (DC) and macrophage (“MΦ”) scores (figure 5C). As the majority of DC in Renca tumors from DIO mice are myeloid lineage cells with potent suppressor functions,8 our GEP analyses suggested that productive immune responses to AdTR/CpG/PD-1 were characterized by strong CD8⁺ T cell responses with concomitant reductions in myeloid cell responses.

We next sought to validate and expand on the identified immunogenetic changes (figure 5) via flow cytometric cellular profiling of day 28 tumors from lean and DIO therapy-naive and therapy-treated mice. The percentages of activated (CD44⁺) CD8⁺ TILs were robustly and significantly increased in both lean and DIO AdTR/CpG/PD-1 responders relative to respective untreated controls (figure 6A), confirming our nanoString data. In contrast, lean and DIO NR exhibited a weak intratumoral CD44⁺CD8⁺ T cell presence. Furthermore, in lean and DIO responders to AdTR/CpG/PD-1 therapy, activated CD8⁺ TILs had low PD-1 expression (figure 6B), likely due to effective tumor clearance limiting repeated TCR engagement.26 30 31 We also observed significant increases in the frequencies of IFNγ⁺CD8⁺ TILs in lean and DIO responders, relative to their respective untreated controls and NR (figure 6C). We determined that reduced PD-1 expression in responders was driven by a relative shift toward PD-1^intCD8⁺ TILs in both lean and DIO mice (figure 6D). In both lean and DIO responders, IFNγ, TNFα and perforin were produced primarily by PD-1^intCD8⁺ TILs (figure 6E). Next, as CD4⁺ TILs are known to modulate the quality and magnitude of CD8⁺ TIL responses,40 we examined the CD4⁺ T cell response. Activated CD4⁺ TILs (online supplemental figure S6A) were elevated in both lean and DIO therapy responders (online supplemental figure S6B). However, CD4⁺Foxp3⁺ Treg percentages were unchanged by obesity or response to therapy (online supplemental figure S6B). These results indicate that immunotherapy administration in lean and DIO responders initiated a robust expansion of intratumoral PD-1^intCD8⁺ T cells that produced IFNγ, TNFα and perforin.

We previously reported that in this orthotopic Renca model, obesity is associated with heightened intratumoral immune suppression via myeloid-lineage DC and myeloid-derived suppressor cell (MDSC) which actively suppress CD8⁺ T cell proliferation when sort-purified and functionally evaluated ex vivo.6 8 Here, DIO therapy responders exhibited significant reductions in tumor-infiltrating CD11b⁺CD11c^highMHC II⁺ DC (online supplemental figure S7A), relative to NR (online supplemental figure S7C). DC subtype analysis (online supplemental figure S7B) showed no alterations toward or away from conventional DC1 (XCR1⁺Sirpα) or DC2 (XCR1^-Sirpα⁺) subsets with obesity or response to treatment (online supplemental figure S5D,E). Total intratumoral MDSCs (CD45⁺/CD11b⁺/CD11c^-/MHC II^-/Ly6G⁺ or Ly6C⁺) (gating strategy as per online supplemental figure S7A) were also significantly reduced in DIO and lean responders to AdTR/CpG/PD-1 therapy, relative to respective NR and untreated controls (figure 6F). MDSC subtype analysis showed a predominance of polymorphonuclear ‘PMN-MDSCs’ over monocytic ‘M-MDSCs’ (online supplemental figure S7F–G), although both subtypes are suppressive in this model.6 Notably, the ratio of CD44⁺CD8⁺ TILs to MDSCs was significantly and favorably increased in DIO and lean responders (figure 6G), confirming our nanoString data (figure 5). Thus, our cellular analyses illustrated that AdTR/CpG/PD-1 success in both lean and OB mice was defined by immune responses that culminate in similar, favorable ratios of activated CD8⁺ T cells to MDSCs within tumors. Therefore, obesity reduced the percentage of mice that respond favorably to this immunotherapy, without altering the quality of immune responses in DIO versus lean treatment responders.

Next, we sought to identify factors responsible for the differential response rates to therapy in DIO versus lean mice. We examined intratumoral immune responses at an early day 15 time point in treatment-naive mice, with the goal of identifying obesity-associated drivers of the divergent therapy outcomes we had identified (figure 4F,G). At this time, DIO mice displayed a trending but non-significant reduction in the frequency of activated CD44⁺CD8⁺ TILs versus lean mice (figure 7A), suggesting that baseline T cell priming and trafficking were predominantly intact in DIO hosts.

To more directly evaluate the effects of obesity on CD8⁺ T cell priming in tumor-bearing mice, we used a complementary CMS5 fibrosarcoma tumor model that permits tracking of tumor antigen-specific T cell responses to an endogenous tumor antigen.22 23 Equal numbers of naive, tumor antigen-specific, TCR transgenic DUC18 CD8⁺ TILs from lean mice were CFSE-labeled and transferred into lean or DIO recipients with early established tumors. Comparable percentages of activated (CD25⁺) proliferative DUC18 T cells were present in tumor-draining inguinal lymph nodes from lean and DIO mice at day 4 post-T cell transfer (online supplemental figure S8A), illustrating that T cell priming was not impaired by host obesity in this model. However, analysis of endpoint CMS5 tumor areas showed tumor regression (ie, no palpable tumor detected) in 75% of lean recipients, as expected,22 23 but only 33% of DIO recipients (online supplemental figure S8B). Therefore, despite intact T cell priming, transferred DUC18 T cells were less able to control CMS5 tumors outgrowth in obese animals, reflecting our findings in mice with renal tumors.

Having found no obvious defects in T cell priming in DIO mice, we examined MDSCs within renal tumors. At day 15 post-tumor challenge, DIO mice had significantly increased frequencies of tumor-infiltrating MDSCs in the absence of therapy (figure 7B). We therefore examined known soluble mediators of MDSC responses. Treatment-naive DIO mice had increased intratumoral concentrations of the MDSC-related proteins GM-CSF, IL-1β, CXCL1 and CCL26 41–43 at day 15, relative to lean mice (figure 7C). IL-1β is a known driver of MDSC generation, accumulation and suppressive capacity,44 45 and was decreased at the gene expression level in both lean and DIO responder mice (figure 5B). Investigating the IL-1β pathway is clinically relevant, as high intratumoral IL-1β gene expression is an unfavorable prognostic in renal cancer patients (p<0.001, thehumanproteinatlas.org) and elevated IL-1β in human renal tumors is associated with increased MDSC accumulation.43

In our renal cancer model, intratumoral IL-1β concentrations correlated positively with both tumor size and tumor-infiltrating MDSC abundance across all categories of lean and DIO treatment-naive and therapy-treated animals (figure 7D). We therefore asked if neutralization of IL-1β in AdTR/CpG/PD-1 treated DIO mice would improve treatment response rates. Indeed, administration of anti-IL-1β in DIO mice receiving AdTR/CpG/PD-1 led to a significant reduction in tumor weights vs NT controls, and improved the mean response rate to 58% vs 33% in DIO AdTR/CpG/PD-1-treated mice with intact IL-1ß (figure 7E). Administration of anti-IL-1β in DIO mice receiving AdTR/CpG/PD-1 reduced intratumoral concentrations of IL-1ß (figure 7F) and also lowered concentrations of other myeloid-related proteins within tumors, including GM-CSF, CXCL1 and CCL2 (online supplemental figure S9A–C). Use of anti-IL-1β did not alter intratumoral concentrations of IL-2, IL-4, or CCL20 (online supplemental figure S9D–F), illustrating the specificity of this approach. Importantly, neutralizing IL-1β did reduce the frequencies of total MDSCs within renal tumors (figure 7G). Thus, reduced AdTR/CPG/PD-1 response rates in DIO mice were mediated in part by early, heightened elevations in intratumoral IL-1β concentration and MDSCs, changes that negated the largely intact CD8⁺ T cell response present in OB hosts.

Finally, we asked if the mean response rate in lean mice could be impaired by neutralizing a chemoattractant mediator of effector T cell trafficking into tumors. In concordance with our finding that host obesity did not significantly alter day 15 CD44⁺CD8⁺ TIL accumulation, we observed no differences in the intratumoral concentrations of T cell-related chemokines (CCL4, CCL5, CXCL10 and CXCL11) in DIO versus lean animals (online supplemental figure S10A). However, CCL5 was recently identified as a key regulator of CD8⁺ TIL trafficking in human renal tumors,41 and we found that elevated CCL5 was associated with reduced tumor weights and increased CD44⁺CD8⁺ T cell infiltration into renal tumors of mice (online supplemental figure S10B). Neutralizing CCL5 in lean mice receiving AdTR/CpG/PD-1 therapy reduced the mean response rate from 86% (in mice with intact CCL5) to 46% (online supplemental figure S10C) and administration of isotype had no effect (online supplemental figure S10D). Collectively, our results implicate IL-1β and CCL5 as important regulators of AdTR/CPG/PD-1 outcomes in lean and DIO mice with renal tumors.