Targeting the IL1β Pathway for Cancer Immunotherapy Remodels the Tumor Microenvironment and Enhances Antitumor Immune Responses

Chronic inflammation is often implicated in cancer development and progression (1–3). Dysregulated or unresolved inflammation plays a major role in multiple tumor-associated processes including invasion, progression, and metastasis (4–6). Inflammation-related mediators such as cytokines and growth factors are produced by the cells that form the tumor microenvironment (TME), including macrophages, neutrophils, dendritic cells (DC), and myeloid-derived suppressor cells (MDSC; ref. 7).

IL1β is a potent proinflammatory cytokine with a key role in amplifying inflammatory responses, given its ability to stimulate the production of proangiogenic and proinflammatory mediators (8). Although IL1β is an important mediator in acute inflammation, high levels of this cytokine can also result in chronic inflammation, which can promote tumor growth and metastasis (8). For example, transplanted melanoma, breast, and prostate cancer cells do not grow or metastasize in IL1β knockout (KO) mice, suggesting that this cytokine is necessary for tumor proliferation and spread (9). High levels of tumor-promoting inflammatory cytokines such as IL1β and IL6 correlate with advanced malignancy and are associated with increased invasion and poorer outcomes in patients with cancer (10, 11).

Although the precise mechanisms behind the effects of IL1β are not entirely understood, alterations in the TME have been suggested to play a major role. IL1β-driven inflammation can suppress antitumor immune responses by promoting the infiltration of immunosuppressive cells, including an influx of regulatory T (Treg) cells and MDSCs (12). These expanded and activated MDSCs can in turn enhance tumor cell proliferation and migration, angiogenesis, and immune escape (13). IL1β is also associated with stromal aberrations in the TME, including increased vascular permeability (14) and induction of inflammatory cancer-associated fibroblasts (CAF; ref. 15). Inhibition of IL1β could therefore be a promising therapeutic option in the treatment of cancer (16–19).

Canakinumab is a fully human IgG1/κ mAb with high affinity and specificity for IL1β (20, 21). Canakinumab showed promise in the Canakinumab Anti-inflammatory Thrombosis Outcomes Study (CANTOS, NCT01327846) clinical trial, which set out to evaluate the effects of IL1β inhibition in secondary prevention of major cardiovascular adverse events. The preplanned safety analysis showed that canakinumab was associated with a significant, dose-dependent reduction in the incidence of lung cancer and lung cancer-associated mortality compared with placebo (22). Because approximately half of the patients diagnosed with lung cancer during the trial harbored circulating tumor DNA at enrollment (23), it is hypothesized that some participants on the study had preexisting tumors, which were undetectable by conventional diagnostic methods at study entry.

Gevokizumab is a humanized IgG2 mAb targeting IL1β with high affinity and specificity (24, 25). Although gevokizumab has not yet been tested in cancer models, it has shown promising results in early clinical trials for inflammation-mediated diseases such as uveitis due to Behçet disease and autoimmune nonnecrotizing anterior scleritis (26–28). A clinical trial with gevokizumab in combination with standard-of-care anticancer therapies is currently ongoing in patients with metastatic colorectal cancer (mCRC), gastroesophageal cancer, and renal cell carcinoma (RCC; ref. 29).

Because tumor-associated inflammation can generate a proangiogenic and immunosuppressive environment that promotes tumor growth, targeting IL1β has the potential to alter the TME with beneficial effects. The effect of IL1β inhibition on the TME is likely to synergize with angiogenesis inhibition or checkpoint blockade to amplify their effects on growing tumors (30, 31). Many chemotherapeutic agents promote the secretion of IL1β (32), which suggests that inhibition of this cytokine could potentially enhance the effects of chemotherapy when used in combination.

Here, we report the effects of IL1β inhibition alone or in combination with docetaxel, anti–PD-1, anti-VEGFα, or anti-TGFβ in mouse models of cancer.

All cell lines used in this study were obtained from the ATCC, expanded to a low passage number and frozen down to create the working stocks; a SNP test was carried out to confirm the authenticity of human cell lines, while no cell authentication assays were carried out for mouse cell lines. All cell lines were grown at 37°C with 5% CO₂ and were tested and found to be free of Mycoplasma and viral contamination in the IMPACT 1 PCR assay panel (RADIL, MU Research Animal Diagnostic Laboratory). Murine lung LL2 cells (obtained in 2014) were cultured in DMEM (catalog no. 11965–092, Gibco) containing 10% heat-inactivated FBS (catalog no. 97068–085, VWR) without any antibiotics. Murine triple-negative breast cancer (TNBC) 4T1 (obtained in 2014) and human non–small cell lung cancer (NSCLC) H358 cells (obtained in 2019) were cultured in RPMI1640 (catalog no. 10–040-CM, Corning) containing 10% heat-inactivated FBS, 1% antibiotic-antimycotic (catalog no. 15240–062, Gibco) and 2 mmol/L glutamine (catalog no. 25030–081, Gibco). These cells were passaged 2 to 3 times post thaw before implanting into mice. Human TNBC MDA-MB-231 cells (obtained in 2019) were cultured in RPMI1640 supplemented with 10% heat-inactivated FBS, 2 mmol/L glutamine, and 1% antibiotic-antimycotic. Human colorectal SW480 cells (obtained in 2019) were cultured in McCoy 5A medium (catalog no. 30–2007, ATCC) supplemented with 10% heat-inactivated FBS, 2 mmol/L glutamine, and 1% antibiotic-antimycotic.

Animals were handled in accordance with Novartis and University of California, Los Angeles Institutional Animal Care and Use Committee (UCLA IACUC) regulations and guidelines; all animal studies were approved by our IACUC. For syngeneic tumor models, mouse lung carcinoma LL2 and breast cancer 4T1 cells were harvested in the exponential growth phase. For the LL2 model, 1 × 10⁶ cells in 100 μL sterile PBS were implanted subcutaneously in the right flank of approximately 8-week-old female C57BL/6 mice (Jackson Laboratories); for the 4T1 model, 1 × 10⁵ cells in 100 μL sterile PBS were implanted subcutaneously in the right flank of approximately 8-week-old female BALB/c mice (Jackson Laboratories). For xenograft studies, humanized Bone Marrow Liver Thymic (BLT) mice were used. These mice were generated according to standard protocols (33, 34) and transferred to the Slamon Laboratory at UCLA from the HIV Research Core at UCLA. Briefly, batches of 22 female 6- to 8-week-old NOD/SCID/IL-2RΥ^−/− (NSG) mice (Jackson Laboratories) were surgically implanted with human fetal liver and thymus fragments (Advanced Biosciences Resources) beneath the kidney capsule followed by engraftment of matched ex vivo expanded CD34⁺ stem cells via tail vein injection. Reconstitution of human immune lymphocytes by the presence of human T cells was confirmed by flow cytometry at 8 weeks postengraftment. Xenografts were established in the humanized BLT mice by subcutaneous injection of 1.0 × 10⁷ cells per mouse for each cell line injected subcutaneously (right flank) in a mixture of 50% Matrigel (catalog no. 354234, BD Biosciences) in culture media.

Tumor volume was determined by measurement with calipers and calculated using a modified ellipsoid formula, where tumor volume (TV; mm³) = [ ((l × w²) × 3.14159)/6], where l is the longest axis of the tumor and w is perpendicular to l. Tumor measurements were made 2 to 3 times per week depending on model growth.

Mice were randomized into treatment groups based on tumor volume at 60 to 100 mm³. Investigators were not blinded to group assignments. For humanized BLT models, when tumors reached a volume of 200 to 300 mm³, they were treated every 5 days with different agents. For syngeneic models, treatment was initiated when the tumors reached 100 mm³ (∼8–9 days postimplant). Isotype-matched antibodies or vehicle were used as controls, as appropriate. Five days after the single treatment dose or the second dose (in the case of 2-dose treatment), tumors were harvested and analyzed for changes to the infiltrating immune cell populations. Mice were monitored for body weight, clinical symptoms, and tumor growth over the course of 20 days.

All compounds were obtained from commercial sources except canakinumab, gevokizumab, anti-mouse IL1β antibody (clone 01BSUR), anti-mouse PD-1 (clone 1D2), and anti-human/mouse TGFβ (35), which were synthesized by Novartis, and anti-human PD-1 (pembrolizumab), which was synthesized by UCLA; isotype controls for these antibodies were also synthesized by Novartis and UCLA. Canakinumab and gevokizumab were used at 10 mg/kg every 5 days (Q5D, i.p.); anti-mouse IL1β (clone 01BSUR) was also used at 10 mg/kg, i.p., with a dosing regimen that depended on the specific experiment. Combination treatments included the chemotherapeutic agent docetaxel [#NDC 0409–0367–01, Hospira; 6.25 mg/kg every week (QW), i.v.], PD-1 pathway inhibitors (anti-human PD-1, pembrolizumab 10 mg/kg Q5D, i.p. or anti-mouse PD-1, clone 1D2 10 mg/kg QW, i.p.) and anti-mouse VEGFα blocking antibody (clone 4G3, 5 mg/kg Q5D, i.p.). For TGFβ blockade experiments, anti-mouse IL1β (01BSUR) and anti-human/mouse TGFβ were used at 10 mg/kg 3QW, i.p. Appropriate isotype controls were used. In all experiments, dosing was initiated after tumor implantation.

Changes in peripheral blood and tumor-infiltrating immune populations were analyzed by flow cytometry using T-cell and myeloid-cell markers. For humanized BLT mouse models, peripheral blood mononuclear cells (PBMC) were isolated from 50 μL of whole mouse blood collected by retro-orbital bleeding. Red blood cell (RBC) lysis buffer (catalog no. A1049201, Gibco) was added to the blood, mixed, and incubated at room temperature for 10 minutes. Lysis was stopped by addition of RPMI with 10% FBS and 1% penicillin/streptomycin (catalog no. 15140–122, Gibco), followed by centrifugation at 1,000 × g. PBMC samples were stained with antibody cocktail (Supplementary Tables S1 and S2) for 30 minutes on ice in the dark before washing with FACS buffer (2 mmol/L EDTA, 2% FBS in PBS) and fixing with 1% paraformaldehyde (PFA). The samples were run on a BD LSR Fortessa instrument (BD Biosciences).

For syngeneic models, tumor-infiltrating lymphocytes (TIL) were analyzed via flow cytometry analysis of tumor tissue five days after the first treatment dose and at the study endpoint. Samples were split into two separate flow cytometry panels, one for myeloid cells and one for T cells; for anti-TGFβ experiments, TIL samples were split into three separate flow cytometry panels, one for myeloid cells, one for T cells, and one for stromal cells. Tumors were harvested and then processed both mechanically and enzymatically into a single-cell suspension as previously described (36). The digestion process involved 4 to 5 consecutive digestion cycles with digestion buffer containing DNase I (catalog no. 10104159001, Roche/Sigma), Collagenase P (catalog no. 11249002001, Roche/Sigma), and Dispase (catalog no. 17105–041, Gibco Life Tech) in each cycle. At the end of this process, the cell suspension was seeded into 96-well plates and blocked with a 1:50 dilution of mouse Fc block (catalog no. 130–092–575, Miltenyi Biotec) for 20 minutes on ice. The samples were then stained with antibody cocktails (Supplementary Tables S3–S5) for 30 minutes on ice in the dark. Cells were fixed overnight using a fixation and permeabilization kit (catalog no. 00–5523–00, eBioscience) and then stained with antibodies for intracellular markers for 30 minutes on ice in the dark before running on a BD LSR Fortessa FACS machine (BD Biosciences). Gating strategies are shown in Supplementary Fig. S1. Data were analyzed using FlowJo analysis software (version 10.6, TreeStar, Inc.) and graphed using GraphPad Prism (version 9.4.1, GraphPad Software Inc.).

Xenograft and organ tissues from tumor-cell implanted mice were collected and either snap frozen in liquid nitrogen or fixed in 4% PFA for 24 hours followed by a switch to 70% ethanol for no longer than 96 hours before processing into tissue blocks. IHC analysis was performed on formalin-fixed, paraffin-embedded (FFPE) xenograft tissue using antibodies against PD-L1 22C3 pharmDx (Dako), CD3 (Clone 2GVG, Ventana), CD8 (Clone SP57, Ventana), CD163 (Clone MRQ-26, Ventana), and myeloperoxidase (catalog no. LS-B4741, LSBio) antibodies. IHC was performed on a Dako autostainer for PD-L1 staining (28–8 pharmDx assay, Dako), whereas a Ventana BenchMark autostainer (iView Detetction kit, catalog no. 05266157001, Ventana) was used for CD3, CD8, CD163, and MPO IHC following manufacturers’ specific recommendations and technical procedures. An Olympus BX41 microscope (20× and 40× objectives) and a slide scanner (Aperio AT2) were used for IHC imaging. IHC slides were microscopically assessed by a pathologist using manual scoring, except for CD8 quantification, which was carried out using HALO software version 2.3 (Indica Labs). PD-L1 staining was marked as positive if 1% tumor cells or more were showing positive reactivity with specific membrane staining. Other cell-specific markers (CD3, CD8, CD163, and MPO) were assessed using grades 0 to 3, where grade 0 represents negative staining; grade 1+ represents that <1% of marker-specific positive cells were observed in the entire tumor section; grade 2+ represents a range of positive cells between 1% and 5%, and grade 3+ represents >5% positive cells. Digital image analysis for CD8 IHC staining was performed using HALO software (Indica Lab) to define the number of CD8⁺ cells in whole tumor samples, as well as within tumor and stroma compartments.

Syngeneic tumors were harvested and digested as previously described for flow cytometry analysis. For stromal enrichment, single cells from the tumor digestion were resuspended in mouse Fc block diluted 1:10 (Miltenyi Biotec), followed by positive selection with biotinylated antibodies specific for CD31 (clone 390, BioLegend) and CD90 (clone 53–2.1, BioLegend) using the EasySep Magnetic Selection Kit (catalog no. 18709, STEMCELL Technologies). Cells were then washed into cold PBS twice, counted, and resuspended to approximately 10⁶/mL for sequencing using the 10× Genomics Chromium Single-Cell 3′ Reagent v2 kit (v2 Gel Bead Kit PN-120237, A Chip Kit PN-1000009, i7 Multiplex Kit PN-120262, 10× Genomics) under standard conditions and volumes for 3′ transcriptional profiling. Cell suspension volumes were calculated for a target cell recovery of 6,000 cells and loaded onto the Chromium kit as per manufacturer's instructions. Purified cDNAs were then quantified on the Agilent Tapestation using High Sensitivity D1000 ScreenTapes (catalog no. 5067–5584) and Reagents (catalog no. 5067–5585). The final single-cell 3′ libraries were quantified using Agilent High Sensitivity D5000 ScreenTapes (catalog no. 5067–5592) and Reagents (catalog no. 5067–5593). Libraries were diluted to 10 nanomolar in Qiagen Elution Buffer (catalog no. 1014609, Qiagen), denatured, and loaded on an Illumina MiSeq (Illumina, Inc.) at 12 picomolar with the MiSeq Reagent Kit v3 (catalog no. MS-102–3001) to assess sample quality and loading normalization for the HiSeq4000. Normalized libraries were loaded at a range of 2.5 to 4.0 picomolar on an Illumina cBOT using the HiSeq 4000 PE Cluster Kit (catalog no. PE-410–1001). Single-cell 3′ libraries were then sequenced on a HiSeq 4000 for 26 base pairs on the first read, followed by an 8 base pair index read, and a 98 base pair second read, using 2 HiSeq 4000 SBS kits (catalog no. FC-410–1001), at 50 cycles. All the sequence intensity files were generated on instrument using the Illumina Real-Time Analysis software. Resulting intensity files were then demultiplexed and aligned to the transcriptome using the 10× Genomics Cell Ranger software package.

The Seurat R package (version 3.2.3) was used for data preprocessing and analysis of mouse single-cell RNA sequencing (scRNA-seq) data. The raw count matrixes created by the 10× Cell Ranger (version 3.1) pipeline were converted into Seurat objects, where cells with <250 genes and genes expressed by <0.1% of all cells were filtered out. The percentage of mitochondrial content was calculated for each cell, and those that fell above the 99.5th percentile were accounted for as dead cells and filtered out. Events that could potentially be doublets were also filtered out based on the number of genes that fell above the 99.5th percentile. Raw gene expression for each cell was normalized by total expression before log-transformation. From the log-normalized data, the top 2,000 most highly variable genes were identified for principal component and clustering analysis. The data was then scaled by mean centering the expression for each gene before dividing by the SD. Principal component analysis (PCA) was done on the scaled data to reduce the dimensions, and number of principal components used was based on a cumulative proportion (accumulated amount of explained variance) of 95%. The shared nearest neighbor (SNN) algorithm was used to identify cell clusters. Cell type annotations were assigned to clusters based on the expression of canonical features in a minimum. The likelihood-ratio test (37) was used for the identification of differentially expressed genes (DEG) between cell types. The genes that were selected for could be detected in at least 10% of clusters, had an absolute fold-change >1.5, and a Bonferroni adjusted P < 0.05. Sequencing data have been deposited in the NCBI Sequence Read Archive: PRJNA901455.

Dissociated human microsatellite-stable colorectal carcinoma (MSS-CRC) samples were obtained from Discovery Life Sciences from five donors. Between 25,000 and 90,000 cells were cultured in 100 μL RPMI1640 supplemented with 2% FBS. Cells were treated with 10 pg/mL IL1β and 10 μg/mL canakinumab, gevokizumab, or isotype control and incubated at 37°C for 24 hours. Treated cells were then recovered and processed for scRNA-seq analysis using the 10× Genomics platform, with 10,000 cells per treatment condition, as described above.

The raw count matrixes were created using the 10× Cell Ranger (version 3.1, with GRCh38 as reference genome) pipeline, where droplets with less than 800 unique molecular identifiers (UMI) and percentage of mitochondrial content above 20% were filtered out. The Seurat R package (version 3.2.3) was used for data preprocessing and analysis of human scRNA-seq data. SCTransform was used to normalize and transform the gene expression count matrix. PCA was done on the scaled data to reduce the dimensions, followed by Uniform Manifold Approximation and Projection (UMAP) visualization. The SNN algorithm was used to identify cell clusters. Cell type annotations were assigned to clusters based on the expression of canonical features. Seurat function FindMarkers was used to identify DEGs between treatment conditions; logFC > 0.2 and P were used as cutoff (further details are included in the figure legends). Sequencing data have been deposited in the NCBI Sequence Read Archive: PRJNA900230.

In-life measurements and data were analyzed using Indigo software (Novartis); all other data were analyzed by GraphPad Prism 9.4.1 (GraphPad Software Inc.). Tumor growth results are presented as mean volumes for each group. Error bars represent the SEM. Statistical differences between mean tumor volumes at specific time points were performed using a one-way ANOVA with Dunnett posttest, while statistical differences between numbers or proportions of different cell types were assessed by two-tailed Student t test. Differences between groups were considered statistically significant at P < 0.05.

The data generated in this study are available within the article and its Supplementary Data files or upon request from the corresponding author. The sequencing data are available in NCBI Sequence Read Archive: PRJNA901455 and PRJNA900230.

The effects of canakinumab and gevokizumab were evaluated as single agents and in combination with drugs commonly used in the clinic (and used in the CANOPY trials) in models of lung, breast, and colon cancer in humanized BLT mice. Treatment with canakinumab alone induced a nonsignificant decrease in tumor volume compared with the isotype control in the H538 NSCLC cancer model (Fig. 1A, left). Pembrolizumab (anti–PD-1) alone had a more pronounced effect (albeit also nonsignificant) on reducing tumor growth, while the combination of canakinumab and pembrolizumab caused a 45.6% decrease in the mean tumor volume versus control at day 19 (P = 0.05). In the MDA-MB-231 TNBC model, canakinumab and pembrolizumab elicited similar, albeit nonsignificant reductions in tumor growth (22.1% and 25.1% vs. control, respectively; Fig. 1A, right), which were slightly improved when the treatments were used in combination (34.7% reduction vs. control, P = 0.0174).

The effects of gevokizumab were tested in a colon cancer model (SW480) in humanized BLT mice. Although gevokizumab alone induced a decrease in tumor volume (26.9% vs. control at day 13, P = 0.22) and combination with anti-VEGFα enhanced this effect (38.7% vs. control, P = 0.09), neither of these effects were significant (Fig. 1B). The effect of the combination was similar to the effect of anti-VEGFα single treatment (40.4% vs. control, P = 0.08), suggesting that the observed trend in reduction in tumor volume was driven primarily by VEGFα inhibition.

No significant body weight loss was recorded following treatment with canakinumab or pembrolizumab, alone or in combination (Supplementary Fig. S2A). No other overt signs of toxicity or graft versus host disease (GvHD) were observed in any of the mice during the study. Overall, these results show that canakinumab and gevokizumab can reduce growth in xenograft models of cancer of different tissue origins.

IL1β has been reported to induce changes in the TME, resulting in increased numbers of immunosuppressive cells (12). To assess the effect of IL1β blockade, immune cell infiltration was analyzed in FFPE tissues from lung and breast xenograft tumors by IHC.

Staining of tumors for CD8 showed that canakinumab alone and/or in combination with pembrolizumab induced robust infiltration of CD8⁺ cells in lung xenografts (Fig. 1C). Furthermore, canakinumab as a single agent induced a higher increase in the numbers of CD8⁺ TILs as compared with either pembrolizumab alone or in combination with canakinumab (Fig. 1D). Changes in circulating immune populations in peripheral blood were also assessed in mice bearing TNBC xenografts by flow cytometry. Pembrolizumab alone drove a decrease in the frequency of circulating monocytes, while treatment with either canakinumab or pembrolizumab reduced the frequency of DC-10 tolerogenic DCs (CD14⁺ CD16⁺ CD141⁺ CD163⁺) and monocytic MDSCs (mMDSC; Supplementary Fig. S2B).

Changes in immune populations in peripheral blood were also assessed in mice bearing SW480 colon xenografts by flow cytometry. Blockade of IL1β and VEGFα induced a nonsignificant increase in the frequency of CD45⁺ immune cells and a decrease in the frequency of monocyte-derived DCs (CD14⁺ CD11b⁺ CD68⁺), the latter only significant for anti-VEGFα (Supplementary Fig. S2C). On the other hand, gevokizumab in combination with anti-VEGFα induced a significant decrease in the frequency of mMDSC and DC-10 cells compared with isotype control, although the effect appeared to be driven mainly by anti-VEGFα alone. These results showed that the decrease in tumor growth following blockade of IL1β alone, albeit nonsignificant, was associated with TME remodeling, with increased tumor infiltration of immune effector cells.

Although humanized BLT mice can provide a wealth of information, the extent to which the reconstituted immune cells compare with a fully functioning immune system is still not completely clear. Furthermore, species-specific differences may lead to less effective interactions between the human immune cells and the mouse host cells, thus potentially altering immune cell trafficking and activation (38, 39). For these reasons, the mechanisms behind the slower tumor growth observed following IL1β blockade in humanized BLT mice were studied in syngeneic mouse models using a surrogate anti-mouse IL1β.

Treatment with anti-IL1β resulted in significant decreases in neutrophils infiltrating the NSCLC LL2 tumors (Fig. 1E). In addition, two-dose treatment resulted in reduced numbers of granulocytic MDSCs (gMDSC) and increased numbers of DCs. IL1β blockade also resulted in numerous changes in tumor-infiltrating immune cell populations in TNBC 4T1 tumors, with significant decreases in the numbers of neutrophils, gMDSCs, mMDSCs, and tumor-associated macrophages (TAM; Fig. 1F, top). Treatment with two doses of anti-IL1β further resulted in significant decreases in the numbers of gMDSCs (Fig. 1F, bottom). These results confirm the data from humanized BLT mice, showing that IL1β blockade results in changes within the tumor towards a less immunosuppressive phenotype.

Docetaxel has been shown to increase secretion of IL1β in tumor cells (32). Treatment with docetaxel alone did not substantially decrease NSCLC LL2 tumor growth; however, slower tumor growth was observed for docetaxel in combination with IL1β blockade (37.5% vs. 9.6% with docetaxel alone at day 14; Fig. 2A, top). Reduced tumor growth was accompanied by significant reductions in the numbers of tumor-infiltrating neutrophils, monocytes, TAMs, and mMDSCs following combination treatment (Fig. 2B, top). Similar results were obtained in TNBC 4T1 tumors, which could reflect additive effects of single-agent activity (Fig. 2A and B, bottom). Although both IL1β blockade and docetaxel contributed to the effect in the NSCLC model, the reduction in tumor growth as well as the decrease in neutrophils and gMDSCs observed in the TNBC model appeared to be driven mostly by docetaxel, although IL1β inhibition appeared to drive the observed reduction in TAMs.

The combined effect of IL1β and PD-1 blockade was also assessed. Although inhibition of IL1β alone decreased the numbers of neutrophils, monocytes, gMDSCs, and mMDSCs infiltrating TNBC tumors, combination with anti–PD-1 did not enhance this effect (Fig. 2C). Moreover, the decrease in suppressive cell populations was not observed following treatment with the anti–PD-1 alone.

Immunosuppressive cell migration into tumors mediated by IL1β can promote angiogenesis (13). Consistent with this, combination treatment of TNBC 4T1 tumors with anti-IL1β and anti-VEGFα resulted in significantly slower tumor growth (45.9% reduction vs. control at day 13, P < 0.001; Fig. 3A); similar to the results obtained in SW480 colon tumors in humanized BLT mice, this reduction appeared to be driven primarily by VEGFα blockade, with little additional effect from IL1β blockade. Treatment with anti-IL1β and anti-VEGFα as single agents significantly decreased the numbers of TAMs, but this effect was not observed with combined treatment (Fig. 3B). There was also a trend to increased numbers of effector CD4⁺ T cells, CD8⁺ T cells, and natural killer (NK) cells in TNBC tumors treated with the combination of anti-IL1β and anti-VEGFα, as well as significant increases in the numbers of monocytes and CD103⁺ DCs, which were not observed following treatment with single agents (Fig. 3B). Addition of anti-VEGFα to anti-IL1β reversed the decrease in numbers of CD4⁺ T cells and NK cells observed after treatment with anti-IL1β as a single agent.

Inhibition of IL1β significantly reduced the proportion of FoxP3⁺ Treg cells infiltrating the tumor; this effect was not observed following treatment with anti-VEGFα alone or in combination (Fig. 3C). In a similar way, the decreased proportion of CD4⁺ and CD8⁺ T cells expressing the immuno-suppressive transcription factor Helios was driven mainly by blockade of IL1β (Fig. 3C). On the other hand, treatment with anti-IL1β and anti-VEGFα, alone or in combination, resulted in significant increases in the proportion of CD62L^loCD45RB^loCD127^lo effector CD8⁺ cells infiltrating the tumor, although only treatment with anti-IL1β caused a similar increase in CD62L^loCD45RB^loCD127^lo effector CD4⁺ cells (Fig. 3D). It should be noted that these results were compiled from animals across multiple studies, suggesting repeating trends between different animal cohorts. Altogether, these results support the reduction of tumor-mediated immunosuppression following IL1β inhibition, providing a potential mechanism of action for combination therapies.

Results from our combination experiments suggested a role for IL1β in TME remodeling. To find the specific target/s of IL1β, the effects of IL1β blockade were evaluated in freshly dissociated samples of human MSS-CRC, which have shown resistance to immunotherapy in clinical trials. Cells from these tumors were treated in vitro with IL1β, canakinumab, gevokizumab, or isotype control for 24 hours and analyzed for gene expression by scRNA-seq (Fig. 4A). Data from all cells (n = 41,138) from five independent colorectal cancer samples after a 24-hour treatment were clustered and projected onto a UMAP plot, followed by cell type annotation on the basis of lineage-specific and canonical cell markers (Fig. 4B). Sample composition was considerably heterogeneous in terms of cell types, with distinct proportions of tumor, endothelial, stromal, and immune cells in each patient sample (Fig. 4C). Analysis of genes associated with the IL1β pathway (IL1B, IL6, NLRP3, and CXCL8) revealed strong expression in myeloid cells (Fig. 4D). On the other hand, IL1R1, which encodes the receptor for IL1β, was expressed at high levels in fibroblasts, in addition to its expression in DCs and Treg cells. Treg cells also expressed the gene encoding the IL1β decoy receptor IL1R2. This expression pattern was consistent with previous results in human NSCLC (40).

When comparing cell type and cluster proportions between culture conditions (isotype control vs. canakinumab vs. gevokizumab) after a 24-hour treatment, no substantial changes in cell populations were observed (Supplementary Fig. S3A and S3B), suggesting that cell populations were not greatly affected by canakinumab or gevokizumab within the 24-hour in vitro treatment window. Therefore, to better explore the effects of IL1β blockade, differential gene expression analysis between treatment conditions was performed for each cell type. Consistent with their prominent expression of the IL1β receptor, treatment with canakinumab and gevokizumab resulted in substantial gene expression changes in fibroblasts, as depicted by the large number of genes affected (Fig. 4E). There was also considerable overlap in the genes modulated by canakinumab and gevokizumab in fibroblasts, reflecting a common mechanism of action (Fig. 4F). Upregulated pathways in fibroblasts following IL1β blockade included stress response and protection of proper protein folding, while immune signaling related to IL10, IL4, IL13, and IFNγ was downregulated (Fig. 4G). Genes related to neutrophil and myeloid-cell recruitment were among the top genes that were downregulated by canakinumab and gevokizumab treatment (Fig. 4H), in agreement with the observed decrease in the numbers of these cells in all tumor models analyzed. Given the consistent modulation of fibroblasts by canakinumab and gevokizumab, fibroblasts were further analyzed to investigate functional subtypes. Five subtypes of fibroblast were identified, each of which showed distinct gene expression (cluster 0, FBN1⁺; cluster 1, ACTA2⁺; cluster 2, PI15⁺; cluster 3, PDGFRA⁺LOX⁺; and cluster 4, CXCL10⁺; Supplementary Fig. S4). Fibroblasts in clusters 0, 1, and 2 all showed high expression of the IL1β receptor (IL1R1) and downstream inflammatory genes (IL6, CXCL1, etc.); similar results have been reported in cross-tissue mouse and human fibroblast datasets (41, 42).

Previous studies have shown that TGFβ blockade results in extensive remodeling of the CAF populations within the TME, which can result in the development of chemo- or immunotherapy resistance (43, 44). To determine whether IL1β blockade synergizes with TGFβ inhibition, changes in CAF subpopulations following blockade of IL1β and TGFβ were analyzed by flow cytometry in TNBC 4T1 tumor models (Supplementary Fig. S5). Inhibition of TGFβ resulted in significant changes in CAF phenotypical markers, driving significant increases in expression of CD44 and CD73 as well as a significant decrease in expression of α-smooth muscle actin (α-SMA), consistent with the role of TGFβ in myofibroblast differentiation in tumors (refs. 41, 43; Fig. 5A). Although the blockade of IL1β only showed modest effects on these markers, IL1β pathway perturbation did drive a significant decrease in the expression of cell surface CD26 (also known as DPPIV), a marker associated with inflammatory CAFs (iCAF). Simultaneous blockade of IL1β and TGFβ did not enhance these effects, which appeared to be mainly driven by inhibition of IL1β or TGFβ alone.

Differential modulation of CAF subpopulations in the TME following IL1β and TGFβ blockade was confirmed using scRNA-seq, which identified different CAF subtypes based on the expression of key marker genes (refs. 41, 43, 45, 46; Fig. 5B and C; Supplementary Fig. S6). Analysis of stromal-enriched samples from 4T1 tumors showed that individual blockade of IL1β and TGFβ resulted in changes in the CAF profile within the tumor (Fig. 5D and E). Treatment with anti-TGFβ induced a decrease in the frequency of myofibroblastic CAFs (myCAF) accompanied by a simultaneous increase in the frequency of interferon-licensed CAFs (ilCAF), a recently described CAF population that shows increased expression of IFN-responsive genes (43). Inhibition of IL1β caused a decrease in the iCAF subpopulation with no major changes in ilCAFs, in agreement with flow cytometry results on CD26 and CD73 expression following IL1β inhibition. An amalgamation of these effects followed combination treatment, with an increase in the frequency of ilCAFs and decreases in the frequencies of iCAFs and myCAFs.

Fibroblasts, like other stromal cells, play a critical role in recruiting immune cells to the TME (47). The effects of IL1β and TGFβ blockade on tumor growth and immune populations in the TME were evaluated. As shown in Fig. 6A, inhibition of IL1β or TGFβ alone significantly reduced TNBC 4T1 tumor growth (19.1% and 18.7% vs. control at day 12, respectively, P < 0.01 for both), with minimal additive effects on terminal tumor weight (29.4% vs. control, P < 0.01) from simultaneous blockade of IL1β and TGFβ. Slower tumor growth was accompanied by extensive remodeling of the TME. Combined blockade of IL1β and TGFβ resulted in increased numbers of neutrophils and decreased numbers of TAMs infiltrating the tumor; both of these effects appeared to be driven mainly by TGFβ blockade (Fig. 6B). In contrast, the observed increase in CD11b⁺ DCs was mediated largely by IL1β inhibition. Interestingly, the increase in the number of tumor-infiltrating gMDSCs mediated by TGFβ inhibition was reversed by combination treatment, suggesting a strong effect of IL1β in modulating gMDSC levels in the tumor. The numbers of CD4⁺ T cells, CD8⁺ T cells, and NK cells infiltrating the tumor were also increased by combined inhibition of IL1β and TGFβ.

Combination treatment with anti-IL1β and anti-TGFβ also resulted in a significant increase in the proportion of PD-L1–expressing cells across multiple myeloid subsets including neutrophils, TAMs, mMDSCs, and CD103⁺ DCs in 4T1 syngeneic tumors (Fig. 6C; Supplementary Fig. S7). This effect was accompanied by a concomitant decrease in PD-1 expression in CD4⁺ and CD8⁺ T cells, suggesting functional alterations in the TME following simultaneous blockade of IL1β and TGFβ. These results show extensive remodeling of immune cell populations in the TME as a result of IL1β and TGFβ blockade.

The purpose of this work is to support ongoing clinical trials by assessing the effects of IL1β blockade in combination with other anticancer therapies and to better understand the mechanism of action of immune modulation upon treatment, which can only be assessed in a limited manner in clinical trials due to the lack of longitudinal tissue sample collections. The data presented here show that IL1β blockade has profound effects on the TME, driving phenotypical changes in its composition that result in a more conducive environment for immune activity, thus enhancing the effectiveness of various cancer therapies. These effects were observed across multiple models of cancer, which is suggestive of a broad utility of anti-IL1β therapy for cancer treatment. The results also increase our understanding of the role of IL1β in TME remodeling caused by chronic inflammation.

IL1β inhibition did not show notable efficacy as single-agent therapy; this could be due to the models used growing too rapidly to monitor differences in efficacy or timing of therapy. It is known that the rapid growth of tumors in these models does not enable the development of the chronic inflammatory environment observed in human tumors (48). Furthermore, both syngeneic models used here, LL2 and 4T1, are poorly immunogenic and resistant to immunotherapy. Alternatively, the lack of significant efficacy could suggest a major role being played by other components of the inflammasome pathway. Targeting the NLRP3 inflammasome complex, for example, might have a larger impact on tumor growth, as it would affect the transcription and activation of both IL1β and IL18 (49). Combined blockade of IL1β and IL18 may also achieve more significant effects on tumor growth inhibition, because IL18 drives immunosuppression in the TME by increasing MDSCs and decreasing NK cell–mediated surveillance (50, 51).

Notably, IL1β blockade enhanced the effectiveness of other treatments such as chemotherapy and immunotherapy. These findings, together with previous observations (9, 30, 32, 52, 53), provide a new therapeutic principle for the use of canakinumab/gevokizumab in clinical practice, placing IL1β inhibition as a complementary therapy to enhance the efficacy of standard-of-care treatments. In the CANOPY-N phase II trial (NCT03968419), canakinumab is being evaluated in the neoadjuvant setting alone and in combination with pembrolizumab (anti–PD-1), while in the CANOPY-A phase III trial (NCT03447769), the setting is assessing canakinumab as an adjuvant after cisplatin-based chemotherapy. For the phase III trials CANOPY-1 (NCT03631199) and CANOPY-2 (NCT03626545), canakinumab is being evaluated as a first-line treatment in combination with pembrolizumab and cisplatin, and as a second/third-line treatment in combination with docetaxel, respectively. Recent results from the CANOPY-A, CANOPY-1, and CANOPY-2 studies reported no significant increase in disease-free survival or overall survival (OS; refs. 54–56); this is in agreement with our results, which showed that combined treatment with anti-IL1β and docetaxel modulated the TME but had only a small effect on tumor growth.

Mechanistically, IL1β blockade was associated with significant remodeling of the TME, with decreased numbers of myeloid suppressive cells (neutrophils, gMDSCs, mMDSCs, and TAMs) and increased tumor infiltration by DCs (both DC1 and DC2) and effector T cells. Treatment with anti-IL1β also decreased the proportion of immune cells with regulatory function, which directly impacts prognosis: lower CD8⁺/Treg ratios are associated with poorer outcomes in multiple tumor types (57). These changes appear to result in an enhanced antitumor immune response. This unique mechanism of action can potentially synergize with the effects of several different types of anticancer therapies. Combination with docetaxel further enhanced the decrease in immunosuppressive cells, while combined treatment with anti-VEGFα resulted in increased infiltration of DCs, CD8⁺ T cells, and NK cells into the TME, and blockade of IL1β and TGFβ caused an increase in CD4⁺ and CD8⁺ T-cell infiltration. The latter combination also showed potential for enhancing the efficacy of anti–PD-L1, as it increased the proportion of PD-L1–expressing myeloid cells; this effect had been observed previously (43). As reported in the KEYNOTE-042 clinical trial, among others, higher expression of PD-L1 within the tumor resulted in better outcomes for patients treated with pembrolizumab (58). As resistance to immune checkpoint inhibitors is associated with an immunosuppressive TME, our data support a role for the IL1β pathway and more broadly chronic inflammation in immune dysregulation.

The data presented herein suggest that the TME changes elicited by IL1β blockade may stem, at least in part, from direct imprinting on stromal cell elements, and in particular, mesenchymal cells. Stromal cells do play an important role in immune suppression within the TME (59–62); one of the key mechanisms is the recruitment of myeloid cells through secretion of chemokines by different types of stroma cells (63), including fibroblasts. As new data continue to emerge on the phenotypic and functional heterogeneity of CAFs in tumors, it is becoming increasingly recognized that changes to the fibroblast landscape in tumors may lead to profound consequences in cancer progression and response to therapies. For example, blockade of TGFβ was previously shown to specially abate myofibroblasts in the TME while inducing the formation of ilCAFs, promoting immune cell infiltration and activity (43). More recently, iCAFs have been identified as key mediators of chemoresistance to neoadjuvant therapy in rectal cancer (44).

In our studies, human colon cancer samples treated ex vivo with canakinumab or gevokizumab underwent extensive changes in CAF populations. Indeed, fibroblasts were shown to be the cells most affected by treatment in terms of change in gene expression, reflecting their imprinting by IL1β signaling. Blockade of IL1β in vivo in the mouse models employed here also resulted in alterations in CAFs, with a stark decrease in the frequency of iCAFs. As this population of CAFs has been previously suggested to rely on signals such as IL1β/TNFα (15), our study provides an important demonstration of their dependency in vivo. The observed downregulation in CAFs of genes involved in myeloid-cell recruitment may likely contribute to the effects described in our studies, and in particular may provide a mechanism for the reduced numbers of neutrophils and MDSCs in mice treated with IL1β antibodies. Altogether, these results suggest that CAFs, and in particular iCAFs, may represent the primary targets of IL1β blockade, with downstream repercussions on a variety of other cellular elements of the TME. Importantly, IL1β blockade did not alter the formation of myCAFs, in line with the fact that these fibroblasts are strikingly dependent on TGFβ levels. This finding underlines the existence of nonredundant roles for TGFβ and IL1β in CAF biology, suggesting that blockade of TGFβ/IL1β may result in different stromal remodeling potential in the TME compared with the single agents alone.

Overall, the preclinical results presented here support the use of IL1β inhibition in cancer treatment; further exploration in ongoing clinical studies will help identify the best combination partners for different disease stages and lines of treatment.

N.A. O'Brien reports grants from Novartis during the conduct of the study; other support from 1200 Pharma and other support from TORL Biotherapeutics outside the submitted work. A.L. Grauel reports personal fees from Novartis Institutes for BioMedical Research during the conduct of the study, and personal fees from Novartis Institutes for BioMedical Research outside the submitted work. D.A. Ruddy reports work for Novartis Pharmaceuticals which produces the molecular entities discussed in the publication. A. Savchenko is a full-time employee of Novartis Pharmaceuticals. V. Rodrik-Outmezguine reports other support from Loxo Oncology at Eli Lilly outside the submitted work. C.C. Wong reports other support from Novartis outside the submitted work, in addition, C.C. Wong has a patent for EP3898674A1 pending and a patent for WO2020128637A1 issued. A. Martin reports other support from Novartis Pharmaceuticals and other support from GSK outside the submitted work, and at the time the work was conducted, A. Martin was a paid, full-time employee of Novartis Pharmaceuticals. Currently, A. Martin is a paid, full-time employee of GSK. G. Dranoff reports other support from Novartis during the conduct of the study, and other support from Novartis outside the submitted work. V. Cremasco reports other support from Novartis during the conduct of the study, other support from Novartis outside the submitted work. C. Sabatos-Peyton reports other support from Larkspur Biosciences and other support from CoStim/Novartis outside the submitted work. No disclosures were reported by the other authors.

R. Diwanji: Conceptualization, data curation, formal analysis, investigation, writing–review, and editing. N.A. O'Brien: Resources, investigation, writing–review, and editing. J.E. Choi: Data curation, formal analysis, writing–review, and editing. B. Nguyen: Data curation, formal analysis, investigation, writing–review, and editing. T. Laszewski: Data curation, formal analysis, investigation, writing–review, and editing. A.L. Grauel: Data curation, formal analysis, investigation, writing–review, and editing. Z. Yan: Data curation, formal analysis, writing–review, and editing. X. Xu: Data curation, formal analysis, writing–review, and editing. J. Wu: Data curation, formal analysis, investigation, writing–review, and editing. D.A. Ruddy: Data curation, formal analysis, writing–review, and editing. M. Piquet: Data curation, formal analysis, investigation, writing–review, and editing. M.R. Pelletier: Data curation, formal analysis, investigation, writing–review, and editing. A. Savchenko: Data curation, formal analysis, investigation, writing–review, and editing. L. Charette: Data curation, formal analysis, investigation, writing–review, and editing. V. Rodrik-Outmezguine: Conceptualization, data curation, formal analysis, writing–review, and editing. J. Baum: Conceptualization, data curation, formal analysis, writing–review, and editing. J.M. Millholland: Conceptualization, data curation, formal analysis, writing–review, and editing. C.C. Wong: Data curation, writing–review, and editing. A. Martin: Data curation, formal analysis, writing–review, and editing. G. Dranoff: Data curation, formal analysis, writing–review, and editing. I. Pruteanu-Malinici: Data curation, formal analysis, writing–review, and editing. V. Cremasco: Conceptualization, data curation, formal analysis, investigation, writing–review, and editing. C. Sabatos-Peyton: Data curation, formal analysis, writing–review, and editing. P. Jayaraman: Conceptualization, data curation, supervision, investigation, writing–review, and editing.

This study was funded by Novartis Pharmaceuticals. The authors would like to thank Lin Fan and Tyler Burk from ASI for scRNA sequencing support; Valerie Rezek and Tong Luo from the Geffen School of Medicine at UCLA for their work in generating the BLT mice; and Muchun Wang from NIBR for support with in vivo experiments. The authors would also like to thank Vanesa Martinez Lopez, PhD, of Novartis Ireland Limited, Dublin, Ireland for providing medical writing support.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.