Graphical abstract

   graphic file with name fx1.jpg
   [95]Open in a new tab

Highlights

     * •
       High IL-6 associates with poor atezolizumab efficacy in patients
       with advanced cancer
     * •
       IL6R- and PD-L1-blocking antibodies combine synergistically to
       treat murine tumors
     * •
       IL-6-STAT3 signaling blocks cytotoxic effector differentiation of
       CD8^+ T cells
     * •
       Cell-intrinsic IL-6 signaling in CD8^+ T cells limits pre-clinical
       anti-PD-L1 activity
     __________________________________________________________________

   Identifying clinically actionable drivers of therapeutic resistance is
   a major objective for cancer immunotherapy. Here, Huseni et al.
   identify IL-6 as a correlate of poor clinical response to atezolizumab
   (anti-PD-L1) therapy and demonstrate that IL-6 impairs anti-PD-L1
   efficacy by restricting the anti-tumor functions of cytotoxic T cells.

Introduction

   Immune checkpoint inhibitors (ICIs) are approved therapies for multiple
   forms of cancer, but they fail to elicit durable clinical responses in
   the majority of patients.[96]^1 Identifying therapeutically actionable
   mechanisms of ICI resistance is therefore critical to maximize the
   benefit of cancer immunotherapy.

   Despite the well-described roles of many cytokines in tumor
   development, they are not widely exploited as therapeutic targets in
   oncology. The pleiotropic cytokine interleukin 6 (IL-6) is associated
   with tumor progression and is thought to influence anti-tumor immunity
   through various mechanisms.[97]^2^,[98]^3^,[99]^4 Interestingly, plasma
   IL-6 was associated prognostically with reduced survival of ICI-treated
   melanoma patients.[100]^5^,[101]^6 Recent pre-clinical studies have
   also implicated IL-6 as a potential driver of ICI
   resistance,[102]^5^,[103]^7^,[104]^8^,[105]^9^,[106]^10 but the
   underlying mechanisms remain unclear.

   In this study, we identified high IL-6 levels as a feature of
   atezolizumab (anti-PD-L1)-resistant disease in patients with advanced
   cancers. IL-6 restrained effector differentiation of CD8^+ T cells
   (also known as cytotoxic T lymphocytes, or CTLs), and high plasma IL-6
   correlated with reduced effector gene expression in CTLs from cancer
   patients. In pre-clinical tumor models, IL6R blockade or genetic
   ablation of CTL-intrinsic IL-6 signaling synergized with anti-PD-L1
   therapy to enhance anti-tumor CTL responses, leading to improved tumor
   control.

Results

IL-6 is associated with poor clinical activity of atezolizumab (anti-PD-L1)

   To identify potential drivers of anti-PD-L1 resistance, we evaluated
   gene expression by RNA-seq in pre-treatment tumor biopsies from
   IMmotion150, a randomized phase II trial of previously untreated
   metastatic renal cell carcinoma (RCC) comparing atezolizumab with or
   without bevacizumab versus the tyrosine kinase inhibitor
   sunitinib.[107]^11 Clinical cohorts evaluated in this study are
   described in [108]Figure S1A and [109]Tables S1–S3. Among patients
   treated with atezolizumab monotherapy, differential expression analysis
   comparing those who developed progressive disease (PD) versus those
   with evidence of disease control revealed various inflammatory factors
   associated with PD, including the cytokine IL6 ([110]Figure 1A). Among
   other cytokines and chemokines associated with PD, only CXCL5 and CXCL8
   were more highly ranked than IL6 ([111]Figure 1B). Ingenuity Pathway
   Analysis identified IL-6 as a putative upstream regulator of
   PD-associated genes (p = 3.16e−9), in addition to the IL-6 receptor
   (IL6R; p = 1.04e−11) and the IL-6-induced transcription factor STAT3
   (signal transducer and activator of transcription 3; p = 1.45e−9). IL6
   correlated strongly with STAT3 and the STAT3-response gene SOCS3,
   consistent with active IL-6 signaling ([112]Figure 1C). Tumor IL6 mRNA
   (evaluated in n = 59 evaluable cases of RCC using in situ hybridization
   [ISH]) was observed mainly in epithelial and stromal cells
   ([113]Figure 1D).

Figure 1.

   [114]Figure 1
   [115]Open in a new tab

   IL6 expression associates with poor clinical outcome in
   atezolizumab-treated patients with metastatic RCC

   (A–C) RNA-seq analysis of pre-treatment tumor samples from the
   atezolizumab monotherapy arm of IMmotion150. (A) Differential gene
   expression analysis (Limma-voom), comparing PD (progressive disease)
   versus SD (stable disease), PR (partial response), or CR (complete
   response). Nominal p values are shown. (B) Differentially expressed
   cytokine and chemokine genes. (C) Pearson correlation of IL6 with SOCS3
   and STAT3. ∗∗∗∗p < 0.0001.

   (D) ISH analysis of IL6 mRNA in RCC tumors from IMmotion150 (n = 59).
   Black arrows, epithelial cell expression; arrowheads, stromal cell
   expression. Scale bar, 50 μm. Pie chart: proportions of tumors with IL6
   expression (staining in ≥1% of cells) in epithelial cells only
   (yellow), stromal cells only (blue), or both epithelial and stromal
   cells (red).

   (E) Association of tumor IL6 mRNA with overall survival (OS) in
   IMmotion150.

   (F) Association of tumor IL6 mRNA with OS in patients with high
   intratumoral T cell signature expression (>median). In (E) and (F), HR
   (+/− 95% CI) and p values were adjusted after multivariate analysis
   including the following co-variates: MSKCC (Memorial Sloan Kettering
   Cancer Center) prognostic risk score, previous nephrectomy, and liver
   metastasis.

   To confirm the association of IL-6 with poor atezolizumab efficacy, we
   next evaluated long-term patient survival. After multivariate
   correction, high tumor IL6 (>median) was associated significantly with
   poor overall survival (OS) only in patients who received atezolizumab
   monotherapy (hazard ratio [HR]: 2.72, 95% confidence interval [CI]:
   1.30, 5.72, p = 0.008; [116]Figure 1E), including those with high
   expression of a CD8^+ T cell signature (HR: 4.30, 95% CI: 1.35, 13.71,
   p = 0.014; [117]Figure 1F).[118]^11^,[119]^12

   IL-6 protein was significantly higher in pre-treatment plasma samples
   from RCC (IMmotion150) patients compared with healthy controls
   ([120]Figure S1B) and correlated significantly with tumor IL6 mRNA (r =
   0.32, p < 0.0001; [121]Figure S1C). High plasma IL-6 (>10 pg/mL;
   [122]Figures S1D and S1E) was associated with reduced OS in all
   treatment arms ([123]Figure S1F). Similarly, in atezolizumab-treated
   patients with metastatic triple-negative breast cancer (TNBC) from the
   PCD4989g clinical trial,[124]^13 or with metastatic urothelial bladder
   carcinoma (UC) from the IMvigor210 and IMvigor211
   trials,[125]^14^,[126]^15^,[127]^16 plasma IL-6 was elevated compared
   with healthy controls ([128]Figure S1B) and associated with poor OS in
   multivariate survival analysis ([129]Figures S1G–S1I). The IL-6-driven
   acute phase response protein CRP (C-reactive protein) was similarly
   associated with plasma IL-6 and poor OS in all three cancer types
   ([130]Figures S1J–S1M).

Dual blockade of IL6R and PD-L1 improves tumor control and CTL function

   To determine if IL-6 affects anti-PD-L1 activity, we first examined the
   syngeneic EMT6 mouse model of TNBC ([131]Figure S2A). EMT6
   proliferation in vitro was unaffected by IL-6 ([132]Figure S2B).
   Compared with either anti-IL6R or anti-PD-L1 monotherapy, combined
   blockade of IL6R and PD-L1 significantly improved tumor control and
   progression-free survival (HR = 0.11, p < 0.0001; [133]Figures 2A and
   2B). Analysis of fluorescence-activated cell sorting (FACS)-purified
   tumor populations (EMT6 cells, T cells, myeloid cells, and fibroblasts)
   identified inflammatory Ly6C^+ cancer-associated fibroblasts as the
   main IL-6 producers ([134]Figures S2C–S2D), consistent with prior
   studies.[135]^17^,[136]^18^,[137]^19

Figure 2.

   [138]Figure 2
   [139]Open in a new tab

   IL-6 inhibits anti-PD-L1 efficacy and anti-tumor CTL response

   (A–H) Treatment of EMT6 tumor-bearing mice with antibodies targeting
   PD-L1 and/or IL6R. (A) Individual tumor growth curves (n = 10 per
   group) from one of three independent studies. PD, progressive disease;
   PR, partial response; CR, complete response. (B) Progression-free
   survival (time to 5x increase in tumor volume) of mice pooled from
   three independent studies, analyzed using log rank test. (C) Cellular
   composition of CD45^+ tumor-infiltrating leukocytes, from one of three
   experiments. (D) Balance of CD4^+ and CD8^+ T cells (CTLs) among total
   TCRβ^+ tumor-infiltrating T cells. Data are concatenated from n = 5
   mice per group, from one of three studies. (E) Relative abundance
   (normalized to tumor weight) of CTLs, conventional CD4^+ T cells, and
   regulatory (CD4^+ Foxp3^+) T cells in EMT6 tumors, pooled from three
   experiments. Groups compared using one-way ANOVA with Holm-Sidak’s
   multiple comparisons test. (F) Effector phenotype of tumor-infiltrating
   CTLs after re-stimulation with PMA/ionomycin. (G) Frequencies of
   polyfunctional cells among tumor-infiltrating CTLs (left panel), and
   their absolute abundance relative to isotype control (right panel).
   Data pooled from three experiments. Groups compared using one-way ANOVA
   with Holm-Sidak’s multiple comparisons test. (H) Relative abundance
   (mean ± SEM) of polyfunctional CTLs and IFN-γ^+TNF^+ CD4^+ T cells
   (multifunctional cells) vs. dysfunctional cells (IFN-γ^− TNF^− GzmB^−
   CD8^+ T cells, or IFN-γ^− TNF^− CD4^+ T cells), from one of three
   experiments with n = 3–5 per group. Groups compared using one-way ANOVA
   with Holm-Sidak’s multiple comparisons test. ∗p = 0.0143, ∗∗∗p =
   0.0001, ∗∗∗∗p < 0.0001.

   (I–K) Single-cell RNA-seq analysis of pre-treatment peripheral blood
   CD8^+ T cells from patients with RCC (from IMmotion150) or UC (from
   IMvigor210). Differential expression analysis was performed on CD8^+
   T cells from patients with high (>10 pg/mL) versus low plasma IL-6
   (separate analyses for each cancer type). Genes identified as
   concordant in both cohorts (“consensus genes”) were evaluated for
   Reactome pathway enrichment (I). Volcano plots of differential gene
   expression are shown for UC and RCC samples in (J) and (K),
   respectively.

   We next characterized tumor-infiltrating leukocytes at a time point
   prior to tumor regression (day 11). Combination therapy increased CTL
   frequency and abundance compared with anti-PD-L1 or anti-IL6R
   monotherapy, but it had little effect on regulatory T cells (Tregs),
   conventional CD4^+ T cells, or myeloid cells ([140]Figures 2C–2E,
   [141]S2E), resulting in a significantly increased CD8^+ to CD4^+ T cell
   ratio ([142]Figure 2D). Combination therapy also increased the
   frequency and abundance of polyfunctional (IFN-γ^+TNF^+GzmB^+) CTLs
   ([143]Figures 2F and 2G), with a comparatively modest effect on
   IFN-γ^+TNF^+ (Th1) CD4^+ T cells ([144]Figure 2H). Dysfunctional
   (IFN-γ^−TNF^−GzmB^−) CTL abundance was unaffected ([145]Figure 2H).
   While IL-6 can promote development of IL-17-producing CD4^+ (Th17) and
   CD8^+ (Tc17) T cells,[146]^2 these were rare in tumors and draining
   lymph nodes, and their abundance was unaffected by any treatment
   ([147]Figures S2F–S2H). We next evaluated two additional tumor models:
   CT26 (anti-PD-L1-resistant) and MC38 (partially anti-PD-L1-responsive).
   Consistent with EMT6, combination therapy in both models improved tumor
   control and enhanced anti-tumor CTL responses compared with anti-PD-L1
   alone ([148]Figures S2I–S2O).

High IL-6 is associated with impaired CTL function in cancer patients

   To explore the relationship between IL-6 and CTL in cancer patients, we
   selected patients from the IMmotion150 (RCC) and IMvigor210 (UC) trials
   with high (>10 pg/mL) or low plasma IL-6 and performed 10x single-cell
   RNA sequencing of their pre-treatment peripheral blood mononuclear
   cells. CTLs were identified from n = 8 RCC patients (n = 5 IL-6-high
   and n = 3 IL-6-low) and n = 10 UC patients (n = 5 IL-6-high and n = 5
   IL-6-low; [149]Figures S3A and S3B). Differential expression analysis
   comparing CTLs from IL-6-high versus IL-6-low patients identified
   numerous genes that were shared between the RCC and UC datasets
   ([150]Figures 2I–2K). Concordant genes associated with high IL-6 were
   related to glucose metabolism and protein synthesis pathways, the
   latter driven largely by ribosomal genes, suggesting a
   transcriptionally quiescent state ([151]Figure 2I). In contrast, CTLs
   from IL-6-low patients showed high expression of genes associated with
   activation and effector function (including CD69, IL32, IFNG, and
   GZMK), as well as enrichment of IFN signaling, the MHC-I pathway, and
   oxidative phosphorylation (OxPhos) ([152]Figure 2I, [153]Table S4).

   Higher resolution clustering revealed several CTL subpopulations that
   associated strongly with plasma IL-6 status ([154]Figures S3C–S3H).
   Clusters of polyfunctional effector cells expressing IFNG, GZMB, and
   various chemokines were identified in both the RCC and UC datasets (C5
   and C2, respectively), and they occurred almost exclusively in IL-6-low
   patients. An additional terminal effector-like population was
   associated with low IL-6 in UC patients (C0), while IL-6-low RCC
   patients harbored cells with high expression of GZMA, GZMK, and TCF7,
   suggesting a state of early activation (C4). In contrast, the majority
   of cells from IL-6-high patients appeared to be inactive or
   functionally impaired, with low expression of IFNG and GZMB (e.g., UC
   cluster C4; [155]Figures S3C–S3H).

IL-6 inhibits CTL effector differentiation

   We next performed a series of in vitro studies to explore the
   mechanistic basis of reduced CTL function under IL-6-high conditions.
   IL-6 can influence CTL metabolism and promote unconventional
   differentiation states (Tc17, Tc21, Tc22) under specific
   conditions,[156]^20^,[157]^21^,[158]^22^,[159]^23^,[160]^24 but its
   impact on classical effector differentiation is not well understood.
   Consistent with our findings in vivo, IL-6 inhibited expression of
   IFN-γ, TNF, and GzmB by CTL from OT-I TCR-transgenic mice after
   activation with SIINFEKL peptide ([161]Figure 3A). IL-6 also reduced
   the rate of cell division, but this effect was offset by improved cell
   viability ([162]Figure S4A). We confirmed these findings by activating
   splenic CTLs from wild-type (WT) or IL6R.ko mice with anti-CD3/CD28
   antibodies in the presence or absence of IL-6 (classical signaling) or
   hyper-IL-6 (a complex of IL-6 and soluble IL6R that activates trans
   signaling via direct engagement of gp130).[163]^3 As expected, IL6R.ko
   T cells were unaffected by IL-6 but remained susceptible to functional
   impairment by hyper-IL-6 ([164]Figures S4B and S4C). IL-6 inhibited
   secretion of IFN-γ, GM-CSF, and IFN-responsive chemokines (CXCL9 and
   CXCL10), while it induced production of IL-10 and factors related to
   Th2/Th17 responses (IL-13 and IL-17A; [165]Figure 3B). IL-6 repressed
   effector functions of both naive and memory CTLs ([166]Figure S4D), and
   it had a similar effect on naive human CTLs ([167]Figure S4E).
   Importantly, IL-6-conditioned OT-I CTLs were attenuated in their
   ability to kill SIINFEKL-loaded MC38 cells ([168]Figure 3C).

Figure 3.

   [169]Figure 3
   [170]Open in a new tab

   IL-6 blocks CTL effector differentiation

   (A) Splenocytes from OT-I mice were stimulated with SIINFEKL
   peptide +/− IL-6 and analyzed by flow cytometry on day 7. Boolean
   analysis of IFN-γ, TNF, and GzmB expression in CTLs is shown. Groups
   (mean +/− SEM of n = 4 technical replicates) were compared by t test
   and represent one of three independent experiments.

   (B) Cytokine secretion (measured by Luminex multiplex assay) by
   FACS-purified CTLs activated with anti-CD3/CD28 +/− IL-6 for 3 days
   ∗p < 0.05 (t test; n = 3 technical replicates per condition).

   (C) OT-I CTLs were activated as described in (A) and co-cultured with
   SIINFEKL-pulsed MC38.GFP cells at a 5:1 effector/target ratio. MC38
   destruction was quantified using Incucyte live-cell imaging. Groups
   compared using two-way ANOVA (n = 4 technical replicates per
   condition); data indicate mean ± SEM, and are representative of three
   independent experiments.

   (D–I) OT-I splenocytes activated with SIINFEKL peptide in the presence
   of IL-6, control IgG, or anti-IL6R antibody. CTLs were FACS-purified
   for RNA-seq analysis after 2 and 7 days (n = 3 technical replicates per
   condition/time point). Principal components analysis (PCA) is shown in
   (D). (E) Differentially expressed genes (FDR <0.05 and absolute fold
   change >1.5). (F) Heatmap of representative differentially expressed
   genes (day 7). (G) Boolean flow cytometry analysis of OT-I CTLs at day
   7. Groups (mean +/− SEM of n = 4 technical replicates) compared using t
   tests, from one of three independent experiments. (H) Reactome pathway
   analysis of differentially expressed genes between IL-6- and
   anti-IL6R-treated cells at day 7. (I) Distribution of differentially
   expressed genes (day 7) among differentiation modules from Best
   et al.[171]^25 ∗p = 0.0239, ∗∗∗∗p < 0.0001 (Fisher’s exact test). Fold
   differences refer to IL-6 versus anti-IL6R treatment.

   (J) Tumor RNA-seq analysis from IMmotion150. Gene modules (average Z
   scores) associated with effector or naive-like CTLs were used to
   calculate an effector/naive-like (Eff/N) ratio. IL6 expression across
   Eff/N quartiles (mean +/− SEM, n = 65–66 per group) was compared using
   one-way ANOVA.

   While IL-6 inhibited CTL activation in vitro ([172]Figure 3), IL6R
   blockade in vivo showed clear effects only in combination with
   anti-PD-L1 ([173]Figure 2). Optimized T cell activation in vitro is
   likely more efficient than activation in vivo, which could influence
   the impact of IL-6. Indeed, PD-1 blockade in vivo was recently shown to
   enhance TCR signal strength in CD4^+ T cells.[174]^26 To explore this
   concept further, we identified a sub-optimal SIINFEKL concentration
   (1 ng/mL) at which PD-L1 blockade enhanced OT-I CTL activation
   ([175]Figure S4F). IL-6 had little effect on T cells activated with
   1 ng/mL peptide alone but prevented expansion of IFN-γ^+ CTLs driven by
   anti-PD-L1 ([176]Figures S4G and S4H). Thus, the inhibitory effect of
   IL-6 is most apparent when T cells are activated by strong
   TCR/co-stimulatory signals.

IL-6 promotes naive/memory-like features in CTLs

   To further characterize the effects of IL-6, we activated CTLs in the
   presence of IL-6 or anti-IL6R antibody (to block endogenous signaling)
   and analyzed them by RNA-seq after 2 and 7 days ([177]Figures 3D–3F).
   IL-6 suppressed expression of effector genes (e.g., Ifng, Gzmb, Prf1,
   Klrg1), chemokines, and transcription factors necessary for effector
   differentiation including Tbx21 and Eomes ([178]Figures 3E and
   3F).[179]^27^,[180]^28^,[181]^29 Conversely, IL-6 promoted expression
   of transcription factors that oppose effector differentiation,
   including Foxo1 and Bach2,[182]^30^,[183]^31 and genes associated with
   naive or central memory (CM) cells such as Sell, Ccr7, and Il7r
   ([184]Figures 3E and 3F). We confirmed by flow cytometry that IL-6
   suppressed expression of Tbet, Eomes, and KLRG1, but it promoted
   retention of CD62L ([185]Figure 3G). Reactome pathway analysis
   suggested that IL-6 suppressed processes related to immune signaling,
   cell division, metabolism, and RNA processing ([186]Figure 3H,
   [187]Table S5). Mapping IL-6-regulated genes onto transcriptional
   modules associated with different stages of CTL differentiation (Best
   et al.)[188]^25 confirmed that IL-6 suppressed genes related to CTL
   activation and cell cycle engagement (modules 2/3), while it promoted
   genes related to naive, memory precursor, and late memory states
   (modules 4/7/8; [189]Figure 3I). We next used tumor RNA-seq data from
   IMmotion150 to evaluate gene expression associated with effector (Eff)
   versus naive-like (N) CTLs. Intriguingly, IL6 was associated inversely
   with the Eff/N expression ratio (p < 0.0001; [190]Figure 3J).

IL-6 controls expression of diverse T cell regulators

   To address whether IL-6 effects are reversible, we activated CTLs and
   stimulated them with IL-6 or hyper-IL-6 for the first 2 days only, or
   during days 3–7. IL-6/hyper-IL-6 signaling on days 0–2 attenuated
   effector function at day 7, whereas exposure after day 2 had no effect
   ([191]Figure S4I). CTL downregulated Il6r expression (and to a lesser
   extent Il6st/gp130) after TCR stimulation, suggesting that T cells may
   lose sensitivity to IL-6 after priming ([192]Figure S4J). However,
   primed cells remained capable of activating STAT3 in response to IL-6
   or hyper-IL-6 ([193]Figure S4K). Similarly, effector memory CTLs
   displayed lower IL6R and gp130 levels than naive or CM cells
   ([194]Figure S4L) but remained capable of activating STAT3 in response
   to IL-6 ([195]Figure S4M).

   To determine if IL-6 pre-conditioning affects CTL differentiation upon
   subsequent TCR stimulation, we treated CTLs with IL-6 for 1 day; then
   we withdrew IL-6 and stimulated with anti-CD3/CD28. IL-6 pre-treatment
   suppressed effector function at day 3 ([196]Figure S4N), suggesting
   that IL-6 alters expression of downstream regulators that provide
   durable control of T cell function, regardless of ongoing IL-6
   signaling. To identify candidate regulators, we performed RNA-seq
   analysis of naive CTLs stimulated with IL-6 alone, or with IL-6 and
   anti-CD3/CD28 for 4 h ([197]Figure 4A). Intriguingly, Ifng was induced
   by IL-6 at 4 h, confirming that proximal IL-6 signaling does not block
   effector function directly ([198]Figure S4O). IL-6 suppressed
   expression of numerous genes with known importance for CTL activation,
   including co-stimulatory receptors (Cd28, Cd226) and mediators of
   dendritic cell cross-talk (Cd40lg, Xcl1), while it induced expression
   of co-inhibitory receptors (Ctla4, Tigit), repressors of inflammatory
   signaling (Il10 and SOCS-family members), and diverse transcription
   factors (Batf, Foxo1, Hif1a, Tox2; [199]Figures 4B and 4C).

Figure 4.

   [200]Figure 4
   [201]Open in a new tab

   IL-6 regulates CTLs via STAT3-dependent BATF induction

   (A–C) RNA-seq analysis of WT naive CTLs stimulated with IL-6 +/−
   anti-CD3/CD28 antibodies for 4 h. (A) Differentially expressed genes
   (FDR <0.05 and absolute fold change >2).

   (B) IL-6-regulated genes with potential roles in CTL differentiation,
   and their functional categorization (C).

   (D) IFN-γ expression in WT or STAT3.ko CTLs (from CD4-Cre x
   Il6r^loxP^/loxP mice) activated with anti-CD3/CD28 +/− IL-6 or
   hyper-IL-6 for 3 days. Groups (mean ± SEM of n = 4 technical
   replicates) compared by t test; data representative of three
   independent experiments.

   (E) Batf mRNA expression (qRT-PCR) in WT or STAT3.ko CTLs activated +/−
   IL-6. ∗∗p < 0.01, ∗∗∗p < 0.001 (WT vs. STAT3.ko; t tests). Data points
   indicate mean +/− SEM of n = 4 technical replicates, from one of two
   independent experiments.

   (F) Western blot of BATF and p-STAT3 (Y705) in activated WT or STAT3.ko
   CTLs. Data represent one of two independent experiments.

   (G) BATF CRISPR-ko or control CTLs were activated +/− IL-6 and analyzed
   on day 3 (groups compared by t test; mean ± SEM of n = 4 technical
   replicates). Data representative of two independent experiments.

   To identify mediators of IL-6-driven CTL dysfunction, we first
   confirmed a requirement for STAT3, the primary transcription factor
   activated by IL-6.[202]^3 As expected, IFN-γ production by
   STAT3-deficient CTLs was unaffected by IL-6 ([203]Figure 4D). Among
   additional IL-6-regulated factors ([204]Figures 4A–4C), we were
   intrigued by BATF (basic leucine zipper transcription factor,
   ATF-like), which restricts expression of CTL effector genes (including
   IFN-γ and perforin)[205]^32 and can be induced by
   IL-6.[206]^23^,[207]^33 Indeed, IL-6 triggered rapid STAT3-dependent
   BATF expression in CTLs ([208]Figures 4E and 4F), and knockout of Batf
   using CRISPR-Cas9 abrogated the effect of IL-6 on CTL effector
   differentiation ([209]Figures 4G and [210]S4P). Thus, IL-6 may impair
   CTL function through STAT3-dependent induction of BATF.

CTL-intrinsic IL-6 signaling inhibits anti-PD-L1 efficacy

   To evaluate the impact of CTL-intrinsic IL-6 signaling in vivo, we
   adoptively transferred naive CTLs from CD90.1^+ OT-I mice that were
   Il6r^wt/wt or Il6r^−/− to WT CD90.2^+ recipients bearing
   ovalbumin-expressing EG.7 tumors. Combined with ovalbumin immunization
   ([211]Figures S5A and S5B), PD-L1 blockade increased the polyfunctional
   anti-tumor CTL response in mice that received Il6r^−/− OT-I cells, but
   not Il6r^wt/wt cells ([212]Figures S5C and S5D).

   We next crossed E8i.Cd8a.Cre and Il6r.loxP mice to generate animals
   with CTL-restricted IL6R deficiency (CD8^ΔIL6R; [213]Figures 5A and
   [214]S5E). CTLs from CD8^ΔIL6R mice were unresponsive to classical IL-6
   signaling, whereas CD4^+ T cells showed normal IL-6 responsiveness
   ([215]Figures S5F and S5G). MC38 tumors grew slowly in CD8^ΔIL6R mice
   ([216]Figure S5H); thus, only CD8^ΔIL6R mice with relatively large
   tumors (comparable in size to those of WT mice) were selected for
   therapeutic studies. Compared with WT littermates, CD8^ΔIL6R mice
   showed greater tumor control ([217]Figure 5B) and stronger induction of
   anti-tumor CTL responses during PD-L1 blockade ([218]Figures 5C–5F and
   [219]S5I). This phenotype was not due to obvious baseline differences
   in T cell development or function ([220]Figure S5J).

Figure 5.

   [221]Figure 5
   [222]Open in a new tab

   CTL-intrinsic IL-6 signaling impairs anti-PD-L1 efficacy

   (A) IL6R expression on lymph node T cells from CD8^ΔIL6R mice or WT
   littermate controls (n = 5 mice per group).

   (B) Tumor growth (mean +/− SEM) in CD8^ΔIL6R mice and WT littermates
   treated with anti-PD-L1 or control antibodies, pooled from two
   independent studies and compared using two-way ANOVA.

   (C–F) Cytokine expression in tumor-infiltrating CTLs after 1 week of
   treatment. Representative staining and Boolean analysis are shown in
   (C) and (D), respectively. Frequencies of IFN-γ^+ TNF^+ cells among
   tumor-infiltrating CTLs (E), and their absolute number (F), compared
   using one-way ANOVA with Holm-Sidak’s multiple comparisons test (n =
   9–12 mice per group). Data pooled from two independent studies.

   (G–I) RNA-seq analysis of FACS-purified tumor-infiltrating CTLs from
   CD8^ΔIL6R mice or WT littermates. Mice with established MC38 tumors
   (∼150 mm^3) were treated with anti-PD-L1 or control antibodies for
   7 days (n = 4–5 mice per group). (G) Differentially expressed genes
   between WT and CD8^ΔIL6R CTLs. Separate comparisons were made based on
   treatment. (H) Reactome pathway analysis of anti-PD-L1-driven
   protein-coding genes (p < 0.05) in WT and CD8^ΔIL6R CTLs. (I) Reactome
   pathway analysis of the top 500 protein-coding genes (ranked by p
   value) that were significantly associated with CD8^ΔIL6R CTLs during
   anti-PD-L1 treatment.

   RNA-seq analysis of intratumoral CD8^ΔIL6R versus WT CTLs revealed
   substantial differences in gene expression ([223]Figure 5G). Anti-PD-L1
   treatment induced genes related to IL-2, TGFβ, and VEGF signaling in WT
   CTLs, whereas it drove genes associated with IFN response and OxPhos in
   IL6R-deficient CTLs ([224]Figure 5H, [225]Table S5). We next evaluated
   genotype-dependent expression in CTLs from mice treated with
   anti-PD-L1. No differences were observed in co-inhibitory molecule
   expression ([226]Figure S5K). IL6R-deficient CTLs expressed high
   amounts of genes related to IFN response, OxPhos, mitochondrial
   biogenesis, cell cycle, DNA repair, and mRNA processing
   ([227]Figure 5I) and low amounts of naive-associated factors
   ([228]Figure S5L). Genes associated with CD8^ΔIL6R CTL were found
   predominantly in modules 2 and 3 (early activation and cell division)
   from Best et al.,[229]^25 whereas WT-associated genes were related to
   modules 7 and 8 (naive and memory precursor; [230]Figure S5M). Overall,
   these findings suggest that during PD-L1 blockade, IL-6 acts on CTLs to
   restrain their effector differentiation and impair therapeutic
   efficacy.

Discussion

   Our clinical observations corroborate the recently described prognostic
   association between plasma IL-6/CRP and poor survival of ICI-treated
   melanoma patients,[231]^6^,[232]^34 and they identify high tumor IL6
   expression as a predictor of poor atezolizumab monotherapy outcome in
   RCC. This relationship may be causal based on our experimental data,
   which demonstrate that IL-6 can impair anti-PD-L1 efficacy via direct
   functional inhibition of CTLs. Consistent with the association between
   plasma IL-6 and CTL dysfunction in patients with
   COVID-19,[233]^35^,[234]^36^,[235]^37^,[236]^38 we identified
   polyfunctional effector CTLs almost exclusively in cancer patients with
   low plasma IL-6.

   IL-6 was reported recently to also inhibit anti-CTLA4 efficacy in mouse
   tumor models.[237]^39 Notably, while IL-6 blockade promoted anti-tumor
   immunity in combination with anti-CTLA4, it ameliorated pathology
   driven by autoreactive Th17 responses,[238]^39 suggesting that
   IL-6/IL6R blockade may have the added benefit of attenuating
   immune-related adverse events.

   IL-6 can potentially influence anti-PD-L1 efficacy through several
   mechanisms, including support of cancer cell fitness,[239]^2^,[240]^4
   inhibition of Th1 responses,[241]^5^,[242]^8^,[243]^10 promotion of
   immunosuppressive myeloid
   cells,[244]^7^,[245]^9^,[246]^40^,[247]^41^,[248]^42 or disruption of
   conventional dendritic cells (cDCs).[249]^43 In our studies, IL6R
   blockade had little impact on myeloid populations, and anti-IL6R/PD-L1
   combination therapy did not affect Th1 responses reproducibly. In
   contrast, combination therapy enhanced CTL responses in several tumor
   models, consistent with the inhibitory effect of IL-6 on CTLs in vitro.
   While this could be related to impaired cDC1s,[250]^43 we observed only
   modest changes in cDCs during combination therapy, and CTL-restricted
   IL6R deficiency enhanced anti-PD-L1 efficacy. Thus, although IL-6 can
   potentially influence several cell types to impair ICI activity, direct
   effects on CTLs may also be important for this process.

   IL-6-mediated CTL inhibition required STAT3-dependent expression of
   BATF, consistent with prior observations implicating STAT3 and BATF as
   regulators of CTL
   function.[251]^22^,[252]^32^,[253]^44^,[254]^45^,[255]^46 However, the
   complete mechanism in vivo is likely complex and context dependent,
   since IL-6 controls expression of numerous T cell regulators, including
   the immunotherapy targets CTLA4 and TIGIT. IL-6 also promoted
   expression of factors necessary for memory formation, including Foxo1,
   IL-10, and
   Bach2,[256]^30^,[257]^31^,[258]^44^,[259]^47^,[260]^48^,[261]^49
   consistent with its role in promoting CD4^+ T cell memory.[262]^50
   Although CTLs with progenitor-like properties are important for
   sustained anti-tumor immunity,[263]^51 successful immunotherapy
   requires their differentiation into cytotoxic effector
   cells.[264]^52^,[265]^53^,[266]^54^,[267]^55 Thus, by promoting
   progenitor/memory-like properties, IL-6 may support the longevity of
   anti-tumor CTLs but nevertheless limit ICI efficacy by impairing their
   development into effector cells.

   Direct inhibition of CTLs by IL-6 is mechanistically distinct from
   other cytokine-dependent processes that can potentially inhibit ICI
   efficacy, including recruitment of inhibitory myeloid cells by
   VEGF,[268]^56 IL-1β,[269]^57 IL-8,[270]^58^,[271]^59 or LIF,[272]^60
   suppression of tumor CTL infiltration by TGFβ,[273]^61 and support of
   Treg function by IL-23.[274]^62^,[275]^63 Importantly, therapies
   targeting IL-6 and IL6R are approved for several conditions including
   rheumatoid arthritis, cytokine release syndrome, giant cell arteritis,
   and multicentric Castleman disease.[276]^64 Given the extensive
   clinical experience with IL-6 inhibitors, and the pressing need to
   improve upon existing immunotherapies, combination of ICIs with
   approved anti-IL-6/IL6R agents warrants investigation in cancer
   patients.

Limitations of the study

   Interpretation of our clinical data is limited by their retrospective
   nature, and prospective studies are necessary to fully evaluate the
   association of IL-6 with immunotherapy efficacy. IL-6 can potentially
   influence anti-PD-L1 response through many cell types, and our study
   does not provide a comprehensive assessment of its cell-intrinsic
   effects in non-CTLs. Further experiments with lineage-specific deletion
   of IL6R, combined with analyses such as single-cell RNA-seq, could
   provide additional mechanistic information. Finally, clinical studies
   of IL-6 pathway inhibitors in combination with ICIs will be necessary
   to fully evaluate the therapeutic translatability of our findings.

STAR★Methods

Key resources table

   REAGENT or RESOURCE SOURCE IDENTIFIER
   Antibodies
     __________________________________________________________________

   Anti-PD-L1 (mouse IgG2A, clone 6E11) Genentech N/A
   Anti-IL6R (mouse IgG2A, clone MR16-1) Genentech N/A
   Mouse IgG2A isotype control (anti-gp120) Genentech N/A
   Anti-mouse-DEC205-ovalbumin Genentech N/A
   Anti-mouse-CD40 (clone FGK4.5) Genentech N/A
   Rabbit anti-phospho-STAT3 (Y705; clone D3A7) Cell Signaling Technology
   Cat#9145; RRID:[277]AB_2491009
   Rabbit anti-STAT3 (clone D1B2J) Cell Signaling Technology Cat#30835;
   RRID:[278]AB_2798995
   Mouse anti-β-actin-HRP (clone AC-15) Sigma-Aldrich Cat#A3854;
   RRID:[279]AB_262011
   Rabbit anti-BATF (clone D7C5) Cell Signaling Technology Cat#8638;
   RRID:[280]AB_11141425
   Hamster anti-mouse CD3e (clone 145-2C11) BD Biosciences Cat#550275;
   RRID:[281]AB_393572
   Hamster anti-mouse CD28 (clone 37.51) BD Biosciences Cat#553295;
   RRID:[282]AB_394764
   Rat anti-mouse CD16/32 (clone 2.4G2) BD Biosciences Cat#553142;
   RRID:[283]AB_394657
   Anti-mouse CD45-BV510 (clone 30-F11) BD Biosciences Cat#563891;
   RRID:[284]AB_2734134
   Anti-mouse Thy1.2-efluor450 (clone 53-2.1) eBioscience Cat#48-0902-82;
   RRID:[285]AB_1272200
   Anti-mouse Thy1.2-alexafluor700 (clone 53-2.1) Biolegend Cat#140324;
   RRID:[286]AB_2566740
   Anti-mouse Thy1.1-alexafluor488 (clone OX-7) Biolegend Cat#202506;
   RRID:[287]AB_492882
   Anti-mouse CD3ε-PE/Cy5.5 (clone 145-2C11) Biolegend Cat#100310;
   RRID:[288]AB_312675
   Anti-mouse CD4-BUV395 (clone GK1.5) BD Biosciences Cat#563790;
   RRID:[289]AB_2738426
   Anti-mouse CD4-BV785 (clone GK1.5) Biolegend Cat#100453;
   RRID:[290]AB_2565843
   Anti-mouse CD8a-BB515 (clone 53-6.7) BD Biosciences Cat#564422;
   RRID:[291]AB_2738801
   Anti-mouse CD8a-PE (clone 53-6.7) Biolegend Cat#100708;
   RRID:[292]AB_312747
   Anti-mouse CD8a-BUV737 (clone 53-6.7) BD Biosciences Cat#612759;
   RRID:[293]AB_2870090
   Anti-mouse CD11b-alexafluor700 (clone M1/70) Biolegend Cat#101222;
   RRID:[294]AB_493705
   Anti-mouse Gr1-PE-Cy5.5 (clone RB6-8C5) eBioscience Cat#35-5931-82;
   RRID:[295]AB_469740
   Anti-mouse CD11c-PE/Dazzle594 (clone N418) Biolegend Cat#117348;
   RRID:[296]AB_2563655
   Anti-mouse MHCII (I-A/I-E)-FITC (clone M5/114.15.2) eBioscience
   Cat#11-5321-85; RRID:[297]AB_465233
   Anti-mouse CD64-PE/Cy7 (clone X54-5/7.1) Biolegend Cat#139314;
   RRID:[298]AB_2563904
   Anti-mouse CD169-PE/Cy7 (clone 3D6.112) Biolegend Cat#142412;
   RRID:[299]AB_2563911
   Anti-mouse B220-BUV737 (clone RA3-6B2) BD Biosciences Cat#612838;
   RRID:[300]AB_2870160
   Anti-mouse IL6R-PE (clone D7715A7) Biolegend Cat#115806;
   RRID:[301]AB_313677
   Anti-mouse gp130-PE (clone 4H1B35) Biolegend Cat#149403;
   RRID:[302]AB_2566644
   Anti-mouse Foxp3-efluor450 (clone FJK-16s) eBioscience Cat#48-5773-82;
   RRID:[303]AB_1518812
   Anti-mouse/human GzmB-Pacific blue (clone GB11) Biolegend Cat#515408;
   RRID:[304]AB_2562196
   Anti-mouse TNF-PE (clone MP6-XT22) Biolegend Cat#506306;
   RRID:[305]AB_315427
   Anti-mouse IFNγ-PE/Dazzle594 (clone XMG1.2) Biolegend Cat#505846;
   RRID:[306]AB_2563980
   Anti-mouse IL-17A-BV785 (clone C11-18H10) Biolegend Cat#506928;
   RRID:[307]AB_2629787
   Anti-human CD8a-BUV395 (clone RPA-T8) BD Biosciences Cat#563796
   Anti-human TNF-alexafluor488 (clone MAb11) Biolegend Cat#502917;
   RRID:[308]AB_493122
   Anti-human IFNγ-APC (clone B27) Biolegend Cat#506510;
   RRID:[309]AB_315443
   Anti-human CD3 (clone OKT3) eBioscience Cat#16-0037-85;
   RRID:[310]AB_468855
   Anti-human CD28 (clone CD28.2) BD Biosciences Cat#555725;
   RRID:[311]AB_396068
     __________________________________________________________________

   Biological samples
     __________________________________________________________________

   Pre-treatment PBMC from Atezolizumab trials (IMmotion150 and
   IMvigor210) Genentech N/A
   Healthy donor PBMC Genentech N/A
   Renal cell cancer tissue blocks from IMmotion150 Genentech N/A
     __________________________________________________________________

   Chemicals, peptides, and recombinant proteins
     __________________________________________________________________

   TrueCut Cas9 Protein v2 Thermo Fisher Scientific Cat#A36499
   Recombinant mouse IL-6 R&D Systems Cat#406-ML
   Recombinant mouse hyper-IL-6 R&D Systems Cat#
   9038-SR-025/CF
   Recombinant human IL-6 R&D Systems Cat#206-IL
   Recombinant human IL-2 R&D Systems Cat#202-IL
   SIINFEKL (OVA 257-264) peptide AnaSpec Cat#AS-60193-1
   Fixable Viability Dye eFluor™ 780 Invitrogen Cat#65-0865-14
   Cell Stimulation Cocktail plus protein transport inhibitors (500X)
   Invitrogen Cat#00-4975-03
   Brefeldin A (1000X) eBioscience Cat#00-4506-51
     __________________________________________________________________

   Critical commercial assays
     __________________________________________________________________

   Foxp3/Transcription Factor Staining Buffer Set eBioscience
   Cat#00-5523-00
   EasySep Mouse CD8^+ T cell Isolation Kit StemCell Technologies
   Cat#19853
   EasySep Mouse Naive CD8^+ T cell Isolation Kit StemCell Technologies
   Cat#19858
   EasySep Human Naive CD8^+ T cell Isolation Kit StemCell Technologies
   Cat#19258
   RNEasy Mini Kit Qiagen Cat#74104
   iScript cDNA synthesis kit Biorad Cat#1708891
     __________________________________________________________________

   Deposited data
     __________________________________________________________________

   Raw and analyzed bulk RNAseq data (mouse) This paper GEO:
   [312]GSE199048
   Tumor expression data (IMmotion150) EGA EGA: EGAS00001004387
   Human PBMC single-cell RNAseq count data EGA EGA: EGAS00001004451,
   EGAS00001004458
     __________________________________________________________________

   Experimental models: Cell lines
     __________________________________________________________________

   EMT6 Genentech N/A
   MC38 Genentech N/A
   MC38-GFP Genentech N/A
   CT26 Genentech N/A
   EG7.OVA Genentech N/A
     __________________________________________________________________

   Experimental models: Organisms/strains
     __________________________________________________________________

   Mouse: C57BL/6J The Jackson Laboratory JAX: 000664
   Mouse: BALB/c Charles River Model# 028
   Mouse: B6; SJL-Il6ra^tm1.1Drew/J The Jackson Laboratory JAX: 012944
   Mouse: Rosa26-Cre.ki (C57BL/6NTac-Gt(ROSA)26Sor^tm16(cre)Arte) Taconic
   Model# 12524
   Mouse: IL6R^−/− (B6; SJL-Il6ra^tm1.1Drew/J x
   C57BL/6NTac-Gt(ROSA)26Sor^tm16(cre)Arte This paper N/A
   Mouse OT-I: C57BL/6^Tg(TcraTcrb)1100Mjb/J The Jackson Laboratory JAX:
   003831
   Mouse: IL6R^−/− OT-1 (C57BL/6^Tg(TcraTcrb)1100Mjb/J x C57BL/6^Il6r−/−)
   This paper N/A
   Mouse: Cd4-Cre (B6.Cg-Tg(CD4-cre)1Cwi N9) Taconic Model# 4196
   Mouse: Stat3^flox (B6.129S1-Stat3^tm1Xyfu/J) The Jackson Laboratory
   JAX: 016923
   Mouse: CD4^ΔSTAT3 (B6.Cg-Tg(CD4-cre)1Cwi N9 x B6.129S1-Stat3^tm1Xyfu/J)
   This paper N/A
   Mouse: Cd8a-Cre (C57BL/6-Tg(Cd8a-cre)1Itan/J) The Jackson Laboratory
   JAX: 008766
   Mouse: CD8^ΔIL6R (C57BL/6-Tg(Cd8a-cre)1Itan/J x B6;
   SJL-Il6ra^tm1.1Drew/J) This paper N/A
     __________________________________________________________________

   Oligonucleotides
     __________________________________________________________________

   Rpl19 (Mm02601633_g1) Life Technologies Cat#4331182
   Il6ra (Mm01211445_m1) Life Technologies Cat#4331182
   Il6st (Mm00439665_m1) Life Technologies Cat#4331182
   Batf (Mm00479410_m1) Life Technologies Cat#4331182
   Guide targeting exon 1 of Batf 5′ GGGGGTACCTGTTTGCCAG-3′ IDT N/A
   Alt-R CRISPR-Cas9 tracrRNA IDT Cat#1072534
   Alt-R CRISPR-Cas9 negative control crRNA IDT Cat#1072544
   IL6 in situ hybridization probe Advanced Cell Diagnostics Cat#2-1082
     __________________________________________________________________

   Software and algorithms
     __________________________________________________________________

   FlowJo FlowJo [313]https://www.flowjo.com/
   Prism 9 GraphPad [314]https://www.graphpad.com/
   R The R Project for Statistical Computing
   [315]https://www.r-project.org/
   Microsoft Excel for Mac Microsoft
   [316]https://products.office.com/en-us/excel
   [317]Open in a new tab

Resource availability

Lead contact

   Further information and requests for resources and reagents should be
   directed to and will be fulfilled by the lead contact, Nathaniel West
   ([318]west.nathaniel@gene.com).

Materials availability

   This study did not generate new unique reagents.

Experimental model and subject details

In vivo animal studies

   C57BL/6J (JAX stock #000664), BALB/c (Charles River), C57BL/6.OT-I (JAX
   stock #003831),[319]^65 C57BL/6.Stat3^flox (JAX stock #016923),[320]^66
   C57BL/6.CD4-cre (Taconic stock #4196),[321]^67 C57BL/6J.Il6ra^flox (JAX
   stock #012944),[322]^68 and C57BL/6.E8i.Cd8a-cre (JAX stock
   #008766)[323]^69 mice were bred and housed at Genentech under specific
   pathogen free (SPF) conditions. Whole-body IL6R.ko mice were generated
   by crossing C57BL/6J.Il6ra^flox and Rosa26-Cre.ki mice. Female mice
   were used for all studies and were 8–10 weeks old at the start of
   experiments. Experimental animals were housed at Genentech in standard
   rodent microisolator cages. Specific animal genotypes are indicated in
   figures or figure legends. For studies of transgenic mice, littermate
   controls were used where appropriate. All animal studies were approved
   by the Genentech Institutional Animal Care and Use Committee.

Human clinical studies

   Patient demographics for the clinical trials analyzed in this study are
   provided in [324]Tables S1, [325]S2, and [326]S3. All patients gave
   informed consent and studies were approved by their respective ethical
   review committees. For specific details of ethical review and study
   designs, see original publications for IMmotion150,[327]^11
   PCD4989g,[328]^13 IMvigor210,[329]^14^,[330]^15 and IMvigor211.[331]^16

Cell lines and primary cultures

   The EMT6 murine mammary carcinoma (BALB/c), CT26 murine colon carcinoma
   (BALB/c), MC38 murine colon carcinoma (C57BL/6), and E.G7-OVA murine
   lymphoma (C57BL/6) cell lines were cultured in Roswell Park Memorial
   Institute (RPMI) 1640 medium plus 2 mM L-glutamine with 10% fetal
   bovine serum (FBS; Hyclone, Waltham, MA). Cell lines used in this study
   were negative for mycoplasma (based on Lonza Mycoalert and Stratagene
   Mycosensor assays) and authenticated by RNA-seq analysis. Sources of
   primary T cells used in ex vivo activation assays are specified in
   figure legends.

Method details

Clinical sample collection

   Plasma samples from TNBC patients were collected from PCD4989g
   ([332]NCT01375842), a single-arm Phase I study that evaluated the
   clinical activity of atezolizumab in patients with locally advanced or
   metastatic malignancies, including TNBC. Bladder cancer plasma samples
   were collected in IMvigor210, a single-arm Phase 2 study investigating
   atezolizumab in UC patients ([333]NCT02951767, [334]NCT02108652) and in
   the Phase 3 UC trial IMvigor211 ([335]NCT02302807) in which patients
   were treated with either chemotherapy (taxane or vinflunine) or
   atezolizumab as a second-line or higher treatment. RCC plasma and tumor
   samples were collected from IMmotion150 ([336]NCT01984242), a phase II
   multicenter, randomized, open-label study investigating activity of
   atezolizumab and atezolizumab + bevacizumab versus sunitinib in
   metastatic clear cell renal carcinoma. Specimens from RCC patients were
   acquired <12 months before study treatment.

RNAseq profiling of tumor tissue

   Whole-transcriptome profiles were generated using TruSeq RNA Access
   technology (Illumina). RNA-seq reads were first aligned to ribosomal
   RNA sequences to remove ribosomal reads. The remaining reads were
   aligned to the human reference genome (NCBI Build 38) using GSNAP
   version 2013-10-10 (Wu et al., 2016), allowing a maximum of two
   mismatches per 75 base sequence (parameters: ‘-M 2 -n 10 -B 2 -i 1 -N 1
   -w 200000 -E 1-pairmax-rna = 200000 –clip-overlap). To quantify gene
   expression levels, the number of reads mapped to the exons of each
   RefSeq gene was calculated using the functionality provided by the
   R/Bioconductor package GenomicAlignments. The CD8^+ T cell gene
   expression signature (GES) was defined in a previous publication for
   RCC.[337]^11

Plasma IL-6 assay

   EDTA plasma samples were collected from patients and stored at −80°C.
   Plasma IL-6 was evaluated by previously qualified immunoassays on the
   multi-analyte platform Simple Plex Ella.[338]^70 The samples were
   diluted twofold in sample diluent and loaded onto the cartridge for
   data acquisition.

PBMC isolation and scRNAseq analysis

   PBMC from UC and RCC patients were isolated using 50 mL Leucosep™ tubes
   (Greiner Bio-One International, Germany) and Ficoll-Paque™ PLUS (GE
   Healthcare, Sweden). Whole blood drawn into sodium heparin blood
   collection tubes was diluted 3x with phosphate-buffered saline (PBS)
   without calcium or magnesium (Lonza, Walkersville, Maryland). Diluted
   cell suspensions were layered on Leucosep tubes and centrifuged for
   15 min at 800 x g at room temperature (RT). Interphases containing
   PBMCs were harvested, washed with PBS, and centrifuged for 10 min at
   250 g at RT before further processing. PBMC were processed for scRNAseq
   analysis using the 10x Genomics platform (10x Genomics, Pleasanton,
   CA). Sample processing for single-cell RNA-seq was done using Chromium
   Single Cell 3’ Library and Gel bead kit v2 (PN-120237) following
   manufacturer’s user guide (CG00052, 10x Genomics, Pleasanton, CA). cDNA
   and libraries were prepared following manufacturer’s user guide (10x
   Genomics). cDNA amplification and indexed libraries were prepared using
   12 and 14 cycles of PCR, respectively. Libraries were profiled,
   quantified, and sequenced as described above (5′ single-cell gene
   expression libraries).

   Seurat (version 3.0) was used to perform basic quality control on the
   raw 50 GEX matrices output from Cell Ranger 2.2.0. The Cell Ranger
   Single Cell Software Suite v.2.2.1 was used to perform sample
   de-multiplexing, alignment, filtering, and UMI (i.e., universal
   molecular identifier) counting
   (https://support.10xgenomics.com/single-cell-gene
   expression/software/pipelines/latest/what-is-cell-ranger). The data for
   each respective subpopulation were aggregated for direct comparison of
   single cell transcriptomes. Then, gene dispersion analysis implemented
   in Seurat was used to select highly variable genes, preserving genes
   with logarithmic mean expression between 0.0 and 3.0 and with
   logarithmic dispersion less than 0.5. Seurat (version 3.0) was used to
   analyze the PBMC GEX data. Genes with detected expression in at least
   five cells, and cells with at least ten genes detected were used. The
   first 20 principal components were used for clustering (resolution =
   0.6) and for UMAP visualization. Clusters were identified based on
   genes that are enriched in a specific cluster. Immunophenotyping of
   PBMCs was inferred from the annotation of cluster-specific genes; Total
   T cells (CD3D, CD3E), CD8^+ T cells (CD3E, CD8A, CD8B), B cells
   (CD79A), CD14 monocytes (CD14), and NK cells (NKG7-positive and
   CD3E-negative).

   Differential gene expression analysis of CD8^+ T cells from IL-6-high
   (plasma IL-6 >10 pg/mL) versus IL-6-low patients used raw counts of the
   samples and was performed by edgeR in R (version 2.13.0) using the
   generalized linear model workflow described in the edgeR manual. First,
   the sequencing reads for duplicate sequencing libraries were combined
   to produce a single set of sequencing reads for each sample, and the
   raw read counts for each gene were used to produce a DGEList object in
   edgeR. Genes were only included if they were represented by at least
   one read in all of the samples. The calcNormFactors() function was used
   to account for differences in the library size for each sample, and an
   experimental design model consisting of the batch and HS status was
   established. The functions estimateCommonDisp() and
   estimateTagwiseDisp() were used to estimate dispersion. Following this,
   differential expression was tested using the exact test based on qCML
   methods. The Benjamini-Hochberg correction was used with a false
   discovery cut-off of 0.05. Reactome pathway enrichment analysis was
   used to identify processes associated with differentially expressed
   genes.[339]^71

Software versions related to clinical data analysis

   Computational analysis was performed using Cell Ranger software (10x
   Genomics, Pleasanton, CA) version 2.2.0, Perl version 5.18.4, R version
   3.6.0, and the following packages and versions in R for analysis:
   Seurat, 3.0.0; edgeR, 3.26.0; cluster, 2.0.8; dynamicTreeCut, 1.63-1;
   UMAP, WGCNA, 1.66; and survival, 2.42-6. Figures and tables were
   generated using the following packages and versions in R: RColorBrewer,
   1.1-2; ggplot2, 3.1.1; gridExtra, 2.3; ComplexHeatmap, 2.0.0;
   superheat, 1.0.0; colorspace, 1.3-2; dplyr, 0.7.8; and data for
   external datasets were obtained using GenomicDataCommons, 1.4.3;
   GEOquery, 2.48.0. The above R packages depended secondarily on the
   following support packages: Matrix, 1.2-17; Biobase, 2.40.0;
   BiocGenerics, 0.26.0; cowplot, 0.9.3; DDRTree, 0.1.5; edgeR, 2.13.0;
   irlba, 2.3.2; limma, 3.38.2; magrittr, 1.5; Matrix, 1.2-15; ranger,
   0.10.1; and VGAM, 1.0-6.

IL6 in situ hybridization

   For the detection of IL6 mRNA expression in RCC tumors, in situ
   hybridization was performed on 4μm thick formalin-fixed,
   paraffin-embedded tissue sections mounted on glass slides. The process
   was automated on the Leica BOND Rx platform (Buffalo Grove, IL). A 20
   base-pair probe to the target region of IL6 (2-1082) was used (Advanced
   Cell Diagnostics, Inc., Newark, CA). Tissue sections were pre-treated
   with heat and protease before hybridization with oligonucleotide
   probes. Detection and amplification were performed with the RNAscope
   2.5 LSx Reagent Kit in Red (Advanced Cell Diagnostics, Inc., Newark,
   CA). Tumor sections were analyzed by a qualified pathologist and
   considered IL6 positive if at least 1% of either tumor cell area or
   stromal area showed IL6 stain. Epithelial and stromal cells were
   distinguished based on a combination of features and also evaluation of
   an adjacent, hematoxylin/eosin-stained section. Tumor cells in
   [340]Figure 1D are characterized by larger nuclei with variable shape
   and outline and a more "open" chromatin pattern. Stromal cells show
   smaller and condensed nuclei. Tumor cells also have more abundant
   cytoplasm compared to stromal cells and show a more cohesive growth
   pattern of trabecular or acinar shapes.

In vivo tumor studies

   Tumor cell lines in log-phase growth were centrifuged, washed with
   Hank’s balanced salt solution (HBSS), counted, and resuspended in 50%
   HBSS and 50% Matrigel (BD Biosciences; San Jose, CA). EMT6 cells were
   inoculated in the left #5 mammary fat pad (1 × 10^5 cells in 100 μL of
   HBSS/Matrigel mixture). Mice were inoculated subcutaneously in the
   right flank with 1 × 10^5 CT26 cells, 1 × 10^6 MC38 cells, or 2 × 10^6
   E.G7-OVA cells in 100 μL of HBSS/Matrigel mixture. When tumors reached
   a volume of 130–230 mm^3 (approximately 8 days after inoculation),
   animals were distributed into treatment groups such that variance in
   tumor sizes between treatment groups was minimized. Mice were treated
   with isotype control antibodies, anti-PD-L1 (mouse IgG1 clone 6E11,
   10 mg/kg first dose followed by 5 mg/kg thereafter), anti-IL6R (mouse
   IgG2a clone MR16-1, 15 mg/kg), or a combination of anti-PD-L1 and
   anti-IL6R. Anti-PD-L1, anti-IL6R, and isotype control antibodies were
   produced in-house and were free of endotoxin contamination. For EMT6,
   CT26, and MC38 studies, mice were euthanized after 10–11 days (after 3
   doses of treatment) and tumors collected for flow cytometry analysis.
   For E.G7-OVA studies, tumors were evaluated one week after adoptive
   transfer of 1 × 10^6 OT-I CD8^+ T cells. For therapeutic efficacy
   studies, antibodies were administered 2 times per week for 3 weeks
   (intravenously for the first dose and intraperitoneally thereafter).
   Tumors were measured 2 times per week using digital calipers, and tumor
   volumes calculated using the modified ellipsoid formula, 1/2 x (length
   x width^2). When tumor volumes fell below 32 mm^3 (lower limit of
   detection) they were considered a complete response (CR; 100% tumor
   growth inhibition). Tumors that regressed to less than 50% of starting
   volume but eventually recurred were considered partial responders (PR),
   and tumors that never regressed were considered to be progressive
   disease (PD). Disease progression was defined as a 5x increase in tumor
   volume. Mice were euthanized if tumors ulcerated or volumes exceeded
   2000 mm^3. Early euthanasia of some mice due to tumor ulceration
   contributed to some variability in final sample sizes of immune
   pharmacodynamic studies. No mice met criteria for euthanasia due to
   body weight loss or adverse clinical signs.

DEC-OVA immunization

   Naive OT-I T cells were isolated from spleens and lymph nodes of OT-I
   or OT-I.IL6R.ko mice (both CD90.1+) by first mashing through 70 μm pore
   filters using the sterile blunt end of a plunger from a 1 mL syringe.
   Naive CD8^+ T cells were then isolated using the EasySep Mouse Naive
   CD8^+ T cell Isolation Kit (STEMCELL Technologies, Cambridge, MA).
   Cells were resuspended at 1 × 10^7 cells/mL in sterile HBSS and
   1 × 10^6 cells (0.1 mL) were injected intravenously via the lateral
   tail vein into wild type C57BL/6J (CD90.2) recipient mice bearing
   EG.7-OVA tumors. The next day mice were injected intravenously with a
   mixture of 50 μg/kg DEC-OVA (ovalbumin fused to anti-DEC205 antibody;
   produced in-house) and 2.5 mg/kg anti-CD40 antibody (produced in-house;
   clone FGK4.5).

EMT6 cell proliferation in vitro

   EMT6 cell responsiveness to IL-6 (10 ng/mL) and hyper-IL-6 (20 ng/mL)
   was quantified using a p-STAT3 (Y705) electrochemiluminescence assay
   after 15 min of stimulation (Meso Scale Diagnostics, Rockville, MD).
   Cell proliferation in the presence or absence of hyper-IL-6 (20 ng/mL)
   or anti-IL6R antibody (40 μg/mL) was determined by seeding cells into
   flat-bottom 96-well plates and tracking confluence over time using
   IncuCyte live-cell imaging (Essen BioScience, Ann Arbor, MI).

Tissue preparation and flow cytometry

   Tumors were weighed, minced, and enzymatically digested using a
   cocktail of dispase (Life Technologies, Carlsbad, CA), collagenase P,
   and DNaseI (Roche, Penzberg, Germany) for 45 min at 37°C to obtain a
   single-cell suspension. Cells were counted using a Vi-CELL XR (Beckman
   Coulter, Brea, CA). Cell suspensions were passed through 100 μm pore
   filters to remove clumps and debris. For analysis of cytokine
   expression, cells were re-stimulated ex vivo for 3 h at 37°C in T cell
   stimulation media composed as follows: RPMI 1640 medium with 10% FBS
   (Hyclone, Waltham, MA), 100 U/mL penicillin/100 μg/mL streptomycin
   (Gibco, Thermo Fisher Scientific, Waltham, MA), 55 μM β-mercaptoethanol
   (Gibco, Thermo Fisher Scientific, Waltham, MA), 2 mM L-glutamine
   (Gibco, Thermo Fisher Scientific, Waltham, MA), 1 mM sodium pyruvate
   (Gibco, Thermo Fisher Scientific, Waltham, MA), 0.1 mM non-essential
   amino acids (Gibco, Thermo Fisher Scientific, Waltham, MA), 10 mM HEPES
   (Gibco, Thermo Fisher Scientific, Waltham, MA), and 1x Cell Stimulation
   Cocktail with protein transport inhibitors (containing phorbol
   12-myristate 13-acetate (PMA), ionomycin, brefeldin A, and monensin;
   eBioscience, Thermo Fisher Scientific, Waltham, MA). For cell staining,
   cells were first incubated with anti-CD16/CD32 Fc block (5 μg/mL; BD
   Biosciences, San Jose, CA; clone 2.4G2) and LIVE/DEAD Fixable dead cell
   stain (efluor780; Invitrogen, Carlsbad, CA) in PBS for 20 min at 4–8°C.
   Cells were then washed and stained with combinations of the following
   antibodies: CD45-BV510 (2 μg/mL; BD Biosciences, San Jose, CA; clone
   30-F11), Thy1.2-efluor450 (2 μg/mL; eBioscience, Thermo Fisher
   Scientific, Waltham, MA; clone 53-2.1), Thy1.2-alexafluor700 (5 μg/mL;
   BioLegend, San Diego, CA; clone 53-2.1), Thy1.1-alexafluor488
   (2.5 μg/mL; BioLegend, San Diego, CA; clone OX-7), CD3ε-PE/Cy5.5
   (2 μg/mL; eBioscience, Thermo Fisher Scientific, Waltham, MA; clone
   145-2C11), CD4-BUV395 (2 μg/mL; BD Biosciences, San Jose, CA, clone
   GK1.5), CD4-BV785 (2 μg/mL; BioLegend, San Diego, CA; clone GK1.5),
   CD8a-BB515 (2 μg/mL; BD Biosciences, San Jose, CA, clone 53-6.7),
   CD8a-PE (2 μg/mL; BioLegend, San Diego, CA; clone 53-6.7), CD8a-BUV737
   (2 μg/mL; BD Biosciences, San Jose, CA; clone 53-6.7),
   CD11b-alexafluor700 (5 μg/mL; BioLegend, San Diego, CA; clone M1/70),
   Gr1-PE-Cy5.5 (1 μg/mL; eBioscience, Thermo Fisher Scientific, Waltham,
   MA; clone RB6-8C5), CD11c-PE/Dazzle594 (2 μg/mL; BioLegend, San Diego,
   CA; clone N418), MHCII (I-A/I-E)-FITC (2.5 μg/mL; eBioscience, Thermo
   Fisher Scientific, Waltham, MA; clone M5/114.15.2), CD64-PE/Cy7
   (2 μg/mL; BioLegend, San Diego, CA; clone X54-5/7.1), CD169-PE/Cy7
   (2 μg/mL; BioLegend, San Diego, CA; clone 3D6.112), B220-BUV737
   (2 μg/mL; BD Biosciences, San Jose, CA; clone RA3-6B2), IL6R-PE
   (2 μg/mL; BioLegend, San Diego, CA; clone D7715A7), and gp130-PE
   (2 μg/mL; BioLegend, San Diego, CA). Cells were stained for 20 min at
   4–8°C.

   Staining patterns used to define cell subsets were as follows: CD8^+
   T cells, Thy1^+ CD3ε^+ CD8a^+ CD4^− CD11b^−; CD4^+ Tconv cells, Thy1^+
   CD3ε^+ CD8a^− CD4^+ Foxp3^− CD11b^−; Treg, Thy1^+ CD3ε^+ CD8a^− CD4^+
   Foxp3^+ CD11b^−; macrophages, CD11b^+ CD64^+ Gr1^−; monocytes/Mo-MDSC,
   CD11b^+ Gr1^int CD64^int; neutrophils/Gr-MDSC, CD11b^+ Gr1^high CD64^−;
   B cells, Thy1^− CD11b^− B220^+ MHCII^+ CD4^−; plasmacytoid DC, Thy1^−
   CD11b^− B220^+ MHCII^int CD4^+ CD11c^+; classical DC1, Thy1^− B220^−
   CD64/CD169^− CD11c^+ MHCII^+ CD11b^− CD8a^+; classical DC2, Thy1^−
   B220^− CD64/CD169^− CD11c^+ MHCII^+ CD11b^+ CD8a^−.

   For analysis of intracellular proteins, surface-stained cells were
   fixed and permeabilized with the eBioscience Foxp3/Transcription Factor
   staining buffer set (Thermo Fisher Scientific, Waltham, MA). Cells were
   then stained for 30–60 min at 4–8°C with combinations of the following
   antibodies in 1x permeabilization buffer: Foxp3-efluor450 (2 μg/mL;
   eBioscience, Thermo Fisher Scientific, Waltham, MA; clone FJK-16s),
   GzmB-Pacific Blue (1 μg/mL; BioLegend, San Diego, CA; clone GB11),
   TNF-PE (1 μg/mL; BioLegend, San Diego, CA; clone MP6-XT22),
   IFNγ-PE/Dazzle594 (0.67 μg/mL; BioLegend, San Diego, CA; clone XMG1.2),
   and IL-17A-BV785 (1 μg/mL; BioLegend, San Diego, CA; clone TC11-18H10).

   Flow cytometry data were collected with a BD LSRFortessa or BD
   FACSymphony analyzer (BD Biosciences, San Jose, CA) and analyzed using
   FlowJo software (Version 10.5, FlowJo LLC, Ashland, OR). Cell sorting
   was performed on a BD Fusion/S6 (BD Biosciences, San Jose, CA).

T cell activation ex vivo

   Mouse spleens and/or lymph nodes were isolated and mashed through 70 μm
   pore filters. For standard peptide activation of OT-I CD8^+ T cells,
   0.2 million splenocytes were seeded in Falcon flat bottom 96 well
   plates (Corning Life Sciences, Corning, NY) and pulsed with 100 ng/mL
   SIINFEKL peptide (AnaSpec, Fremont, CA), allowing splenic
   antigen-presenting cells to present both peptide and PD-L1 to T cells.
   After 2 days, cells were analyzed or transitioned to T cell media
   (without SIINFEKL) containing 10 ng/mL recombinant human IL-2 and
   incubated for 3–4 days before use in cytotoxicity assays or
   re-stimulation with anti-CD3/CD28 antibodies. For flow cytometry
   analysis of cytokine expression, Brefeldin A (eBioscience, Thermo
   Fisher Scientific, Waltham, MA) was added 4 h before staining.

   For polyclonal T cell activation, bulk splenocytes or CD8^+ T cells
   isolated using the EasySep CD8^+ T cell Isolation Kit (STEMCELL
   Technologies, Cambridge, MA) were plated in T cell media containing
   10 ng/mL human IL-2 at 0.2 million cells per well in Falcon flat bottom
   96 well plates (Corning Life Sciences, Corning, NY) coated overnight
   with 5 μg/mL anti-CD3 antibody (BD Biosciences, San Jose, CA, clone
   145-2C11) and 2.5 μg/mL anti-CD28 antibody (BD Biosciences, San Jose,
   CA, clone 37.51). Cells were labeled in some experiments with Cell
   Trace Violet-421 (Molecular Probes, Thermo Fisher Scientific, Waltham,
   MA). In some assays, T cells were activated in the presence of
   recombinant mouse IL-6 (10 ng/mL; R&D Systems, Minneapolis, MN), mouse
   hyper-IL-6 (20 ng/mL; R&D Systems, Minneapolis, MN), isotype control
   mouse IgG2a antibody (5 μg/mL), mouse IgG2a anti-IL6R antibody
   (5 μg/mL; clone MR16-1), or mouse IgG1 anti-PD-L1 (5 μg/mL; clone
   6E11). CRISPR/Cas9-mediated deletion of Batf in primary T cells was
   performed according to a previously published protocol[341]^72 using
   the following gRNA targeting exon 1 (GGGGGTACCTGTTTGCCAG; IDT, San
   Diego, CA). BATF was detected in western blots using anti-BATF clone
   D7C5 (Cell Signaling Technology, Danvers, MA).

   For human T cell activation, naive human CD8^+ T cells were isolated
   from buffy coats using the EasySep Human Naive CD8^+ T cell Isolation
   Kit (STEMCELL Technologies, Cambridge, MA), and plated on 96-well
   flat-bottom plates pre-coated with 10 μg/mL anti-CD3 (clone OKT3,
   Thermo Fisher Scientific, Waltham, MA) in T cell medium containing
   1 μg/mL soluble anti-CD28 (clone CD28.2, BD Biosciences, San Jose, CA)
   and 10 ng/mL recombinant human IL-2 (R&D Systems, Minneapolis, MN).
   Recombinant human IL-6 (R&D Systems, Minneapolis, MN) was added to some
   wells at 10 ng/mL. Cells were re-stimulated with PMA/ionomycin before
   analysis.

T cell cytotoxicity assays

   OT-I CD8^+ T cells were activated with SIINFEKL peptide (AnaSpec,
   Fremont, CA) as described above in the presence or absence of
   recombinant mouse IL-6 (10 ng/mL) or hyper-IL-6 (20 ng/mL) for
   5–6 days. MC38-GFP cells (plated at 5,000 cells per well in 96-well
   flat-bottom plates) were pulsed with 10 ng/mL SIINFEKL peptide for 1 h
   at 37°C and washed with PBS, at which time T cells were added in
   complete T cell medium. MC38 cell death was quantified over time using
   IncuCyte Live Cell Analysis (Essen Bioscience, Ann Arbor, MI).

Bulk RNA-seq analysis of mouse T cells

   OT-I CD8^+ T cells were activated with SIINFEKL peptide as described
   above. Viable CD8^+ T cells were sorted to >99% purity on day 7 using a
   BD FACS Aria Fusion cell sorter (BD Biosciences, San Jose, CA). RNA was
   extracted using the RNEasy mini kit (Qiagen, Germantown, MD). 0.1 μg of
   total RNA was used for library preparation using TruSeq Stranded Total
   RNA Library Prep Kit (Illumina, San Diego, CA). Libraries were
   multiplexed and sequenced on Illumina HiSeq4000 (Illumina, San Diego,
   CA) to generate 30 million single-end 50 base pair reads. Data were
   analyzed using HTSeqGenie in BioConductor as follows: first, reads with
   low nucleotide qualities (70% of bases with quality <23) or rRNA and
   adapter contamination were removed. The reads that passed were then
   aligned to the reference genome GRCh38.p10 using GSNAP (Wu et al.,
   2016). Alignments of the reads that were reported by GSNAP as “uniquely
   mapping” were used for subsequent analysis. Gene expression levels were
   quantified as Reads Per Kilobase of exon model per Million mapped reads
   normalized by size factor (nRPKM), defined as number of reads aligning
   to a gene in a sample/(total number of uniquely mapped reads for that
   sample x gene length x size factor). Principal components analysis
   (PCA) and generation of differential expression heatmaps were performed
   using Partek Flow version 8.0.19.0710. Differential gene expression
   analysis was performed using voom from the limma R package.[342]^73

RNA extraction and qRT-PCR

   Cells were lysed in RLT buffer and homogenized by pipetting. RNA was
   isolated using RNEasy Mini or Micro kits (Qiagen, Germantown, MD)
   followed by reverse transcription using the iScript cDNA synthesis kit
   (Biorad, Hercules, CA). Quantitative RT−qPCR was performed using Taqman
   assays (Life Technologies, Carlsbad, CA) according to manufacturer
   instructions and a ViiA7 384-well real-time PCR detection system
   (Applied Biosystems, Waltham, MA). All expression levels were
   normalized to an internal housekeeping gene (Rpl19) and calculated as
   2^∧−(CTHK−CTgene).

Western blotting

   Cell pellets were lysed in RIPA buffer with protease and phosphatase
   inhibitors (Roche, Penzberg, Germany) at 4°C for 20 min. Supernatants
   were obtained after high-speed centrifugation and protein concentration
   measured using the BCA assay (Thermo Fisher Scientific, Waltham, MA).
   Lysates were denatured with reducing sample buffer and dithiothreitol
   (Invitrogen, Waltham, MA) at 95°C for 10 min. Proteins were separated
   by sodium dodecyl sulphate-polyacrylamide gel electrophoresis in a
   NuPAGE 4–12% gradient Bis-Tris gel and analyzed by western blotting
   with antibodies against phospho-STAT3 (clone D3A7, Cell Signaling
   Technology, Danvers, MA), total STAT3 (clone D1B2J, Cell Signaling
   Technology), BATF (clone D7C5, Cell Signaling Technology), and
   β-actin-HRP (Sigma-Aldrich, St Louis, MO).

Quantification and statistical analysis

   The Kaplan-Meier method was used to estimate the probability of OS. For
   OS analysis, surviving patients were censored at the time of the last
   contact. The hazard ratios and 95% confidence intervals for OS were
   estimated by a Cox regression model. Cox proportional hazards and
   linear regression model was performed to conduct multivariate analysis
   (co-variates are specified in figure legends).

   Statistical calculations were performed using GraphPad software or R.
   Statistical details for specific experiments including the tests used
   and value/definition of n are provided in figure legends.
   Box-and-whisker plots represent median, quartiles, and maximum/minimum
   values. Bar charts and error bars indicate means and SEM. In vitro
   assays were performed with 3–4 technical replicates per condition.
   Sample sizes for in vivo mouse studies were based on the number of mice
   routinely needed to establish statistical significance based on
   variability within study arms. Each data point represents either a
   technical replicate (in the case of in vitro studies) or a biological
   replicate (individual mouse from in vivo studies). p-values and FDR
   values < 0.05 were considered in all analyses to be statistically
   significant.

Additional resources

   Clinical trial information for IMmotion150 ([343]NCT01984242):
   [344]https://clinicaltrials.gov/ct2/show/NCT01984242.

   Clinical trial information for PCD4989g ([345]NCT01375842):
   [346]https://www.clinicaltrials.gov/ct2/show/NCT01375842.

   Clinical trial information for IMvigor210 ([347]NCT02951767 and
   [348]NCT02108652): [349]https://clinicaltrials.gov/ct2/show/NCT02951767
   and [350]https://clinicaltrials.gov/ct2/show/NCT02108652.

   Clinical trial information for IMvigor211 ([351]NCT02302807):
   [352]https://clinicaltrials.gov/ct2/show/NCT02302807.

Acknowledgments