To determine whether gasdermin D (GSDMD) gene expression within the tumor was associated with differences in the anticancer immune response, the intratumoral expression of GSDMD in patients was correlated with clinical outcome and with immune cell infiltration using the TCGA repository. The expression of GSDMD mRNA in different types of cancer was first compared with its expression in the corresponding healthy tissues. GSDMD mRNA expression was upregulated in several types of cancer, including breast, kidney, liver, and prostate cancer ( Figure 1A ). The potential impact of intratumoral GSDMD expression on survival outcomes and immune cell-related genes was examined in, primary breast invasive carcinoma (BRCA) and liver hepatocellular carcinoma (LIHC). Higher GSDMD expression correlated with increased survival in BRCA, but not in LIHC patients ( Figures 1B, D ). A comparative analysis of the expression of genes associated with immune cell types was conducted in patients in the top and bottom 25% of GSDMD expression levels (18) ( Figures 1C, E ). For patients with BRCA, differences in GSDMD expression were mainly linked to differential expression patterns of macrophage-associated genes. Elevated GSDMD expression significantly correlated with decreased expression of markers such as FN1, EMP1, and PCOLCE2 ( Figure 1C , Supplementary Table S1 ), which are typically associated with M2-type macrophages and are indicative of a poor prognosis (19–21). This analysis did not reveal substantial differences in markers for other immune cell types in the context of BRCA. For LIHC patients, less expression of immune-related genes was detected and no clear myeloid signature was seen in LIHC patients with high GSDMD expression, in contrast to what was observed in BRCA patients. However, GSDMD-high patients exhibited an elevated expression of CD8 T cell-related genes, including CDKN2AIP, DNAJB1, and SLC16A7. Meanwhile, genes associated with NK cells, such as SMEK1 and YNF747, were predominantly downregulated in this context ( Figure 1E ). In summary, high GSDMD expression was associated with improved survival in BRCA but not in LIHC patients. In addition, BRCA patients with high GSDMD expression showed a distinct downregulation of the M2-type macrophage signature, which was not the case for LIHC patients.

To determine which cell types expressed GSDMD, we used publicly available human single-cell RNA data sets from human breast cancer (n=84) (22) and liver cancer (n=21) (23). In breast cancer, we observed that GSDMD was expressed both in cancer cells and in the microenvironment, including endothelial cells, fibroblasts and most immune cell types, such as myeloid cells and B and T lymphocytes ( Figure 2A ). Expression of EpCAM and PTPRC (CD45) are shown as positive controls for cancer cells and immune cells, respectively. Average gene expression for individual samples is shown in Supplementary Figure S1 . Similarly, in hepatocellular carcinoma, we observed expression of GSDMD in hepatocytes, endothelial cells, fibroblasts and different immune cell subsets ( Supplementary Figure S2 ).

In order to confirm the expression of GSDMD at the protein level in breast cancer, we stained breast cancer samples for GSDMD by immunohistochemistry. In five patients with invasive ductal carcinoma, we found expression of GSDMD in tumor cells as well as in immune cells and endothelial cells ( Supplementary Figure S3 , Table 1 ). In all samples examined, GSDMD expression colocalized with the macrophage marker CD68 and the B-cell lineage marker CD79A ( Figure 2B ). A lymph node was used for control staining ( Figure 2C ).

To specifically examine the role of GSDMD expression by host cells in the antitumor immune response, we used two different tumor models, one for breast cancer and one for hepatoma in GSDMD-deficient mice. The murine breast cancer cell line EO771 was selected as a breast cancer model, since the growth of EO771 tumors is dependent on IL-1β derived from the intra-tumoral myeloid compartment (24). First, to verify that myeloid cells from Gsdmd ^-/- mice did not release IL-1β upon classical stimulation of the inflammasome, bone marrow-derived macrophages (BMDM) from WT and Gsdmd ^-/- mice were stimulated using lipopolysaccharide (LPS) and adenosine triphosphate (ATP) (25). As expected, cells from WT, but not from Gsdmd ^-/- mice, secreted IL-1β following stimulation ( Supplementary Figure S4 ). EO771 cells were then implanted in the mammary fat pad of WT and Gsdmd ^-/- mice, and tumor growth and immune cell infiltration were compared. No discernible differences in either tumor growth or tumor weight were observed between WT and Gsdmd ^-/- mice ( Figures 3A, B ). To gain further insights, the immune composition of the tumor was analyzed by flow cytometry. No significant variations in the overall immune cell infiltration within the tumor was observed between Gsdmd ^-/- and WT mice ( Figure 3C ). Furthermore, intratumoral T cell, DC, macrophage and MDSC percentages did not differ between the two experimental groups ( Figure 3D ). In addition, the cellular composition of the spleen and lymph nodes did not show any changes between tumor-bearing WT and Gsdmd ^-/- mice, nor did the proportion of CD4⁺ and CD8⁺ naïve, effector and memory T cells in tumor-draining lymph nodes ( Supplementary Figures S5 – S7 ).

A similar experiment was performed in a murine hepatoma model. Here, subcutaneous (s.c.) Hepa1-6 tumors were established in Gsdmd ^-/- and WT mice. Neither tumor growth nor tumor weight were influenced by the deficiency in GSDMD ( Figures 3E, F ). Moreover, intratumoral immune cell infiltration ( Figure 3G ), T cell, DC, macrophage and MDSC percentages showed no significant differences between WT and Gsdmd ^-/- mice ( Figure 3H ). The lack of GSDMD also had no significant impact on immune populations in the spleen of tumor-bearing mice or on the phenotype of CD4⁺ and CD8⁺ T cells in the tumor-draining lymph nodes ( Supplementary Figure S8 ).

Subsequently, intratumoral gene expression levels for different cytokines and chemokines were assessed in EO771 and Hepa1-6 tumors. Il6 mRNA expression levels were not significantly different between Gsdmd ^-/- and WT mice in either tumor type ( Figure 3I ). Unsurprisingly, as IL-1β production is regulated post-transcriptionally, Il1b mRNA expression levels were also unaffected ( Figure 3J ). Tnfα expression levels also did not show major differences ( Figure 3K ). In contrast, Cxcl10 levels were significantly upregulated in mice lacking GSDMD in Hepa1-6 tumors, but not in EO771 tumors ( Figure 3L ). No significant difference was seen in the intratumoral expression levels of Gzmb, Prf1, Il1r1, Tnfrsf1a, Ifng and Cxcl9 mRNA between Gsdmd ^-/- and WT mice in either tumor type ( Supplementary Figure S9 ). In summary, the deficiency in GSDMD did not exert an impact on tumor growth, immune cell infiltration, or the immune landscape in either the EO771 breast cancer or the Hepa1-6 hepatoma model, although intratumoral Cxcl10 levels were elevated in Hepa1-6 tumors grown in Gsdmd^-/- mice.

The observed upregulation of Cxcl10 mRNA expression in Hepa1-6 tumors in Gsdmd ^-/- mice suggested that GSDMD deficiency might influence the activation of immune cells. To examine this, splenocytes and bone marrow cells from Gsdmd ^-/- and WT mice were stimulated with activators for different innate immune pathways. Stimulants for Toll-like receptor (TLR) 7 (R848), TLR4 (LPS), STING (3’3’cGAMP), TLR9 (CpG) and TLR3 (polyI:C) were used. As indicators of immune cell activation, the levels of CXCL10 and IL-6 were measured in the cell culture medium. While CXCL10 was not induced in bone marrow cells, R848 and CpG induced IL-6 release, with no difference between Gsdmd ^-/- and WT cells ( Figures 4A, B ). Several activation markers were evaluated by flow cytometry on different immune cell subsets. For bone marrow cells, CD83 was not clearly upregulated on either macrophages or dendritic cells (DC) ( Figure 4C , Supplementary Figure S10A ). MHC-II was more highly expressed after STING activation on Gsdmd ^-/- macrophages than on WT macrophages, but this difference was not seen in DCs ( Figure 4D , Supplementary Figure S10B ). For splenocytes, CXCL10 release was induced by all stimulants and IL-6 release was induced by R848, LPS and CpG ( Figures 4E, F ). However, no difference between the two experimental groups was found. Within splenocytes, the activation markers CD62L and CD44 on CD4⁺ and CD8⁺ T cells and CD62L and CD69 on B cells were examined. CD62L showed an expected upregulation following stimulation with TLR ligands, but showed no difference between the two experimental groups on both CD4⁺ and CD8⁺ T cells ( Figure 4G, H ). The remaining activation markers on B cells, CD4⁺ and CD8⁺ T cells did also not show any significant changes between the two experimental groups ( Supplementary Figure S10C-F ). No IL-1β or free IL-18 were detected in the supernatants of stimulated bone marrow cells and splenocytes in these conditions (data not shown). Next, GSDMD cleavage was assessed by Western blotting upon LPS or R848 stimulation in combination with ATP in BMDM. Only the combination of LPS and ATP led to secretion of IL-1β and to GSDMD cleavage ( Supplementary Figure S11 ). This differential sensitivity of inflammasome activation to LPS and R848 in BMDM is consistent with previous observations (26).

In conclusion, bone marrow and spleen cells isolated from Gsdmd^-/- and WT mice showed similar responses in terms of CXCL10 and IL-6 release, as well as activation marker levels on different immune cell subsets, following in vitro stimulation with various innate immune pathway activators. Notably, despite the observed upregulation of Cxcl10 in Hepa1-6 tumors in Gsdmd^-/- mice, in vitro stimulation did not unveil substantial differences in immune cell activation between the two experimental groups, suggesting that the relationship between GSDMD deficiency and immune response activation might be better studied in vivo.

TLR7 stimulation in mice leads to the activation of myeloid cells that produce proinflammatory cytokines and chemokines, such as IL-6, CXCL10 and IFNγ. This in turn leads to the activation of T cells (27, 28). B cells express TLR7 and are also directly activated by TLR7 agonists (29, 30). To examine whether Gsdmd ^-/- mice responded differently to TLR7 activation than WT mice, mice were injected with the TLR7 ligand R848 or with PBS and serum cytokine levels were assessed after 3h and 24h. Serum IL-6 levels were increased as expected 3h after R848 injection, but no difference was seen between WT and Gsdmd ^-/- mice ( Figure 5A ). Cxcl10 gene expression levels were increased in the lymph nodes and spleen 3h after R848 stimulation, but Gsdmd ^-/- mice exhibited no difference to WT mice ( Figure 5B ). A similar pattern was observed for Ifng and Tnfa gene expression in both the lymph nodes and spleen ( Figures 5C, D ). Immune cell activation was measured by flow cytometry in the lymph nodes, spleen and bone marrow of Gsdmd ^-/- and WT mice. The activation marker CD69 was increased on both CD4⁺ and CD8⁺ T cells in the lymph nodes 3h after R848 stimulation, but no difference was observed between experimental groups ( Figures 5E, F ). CD44, MHC-I and CD62L showed no increase at this time point upon stimulation in either group ( Figures 5E, F ). A similar pattern was observed on lymph node B cells and on CD4⁺ and CD8⁺ T lymphocytes and B cells in the spleen ( Supplementary Figures S14A-D ). In splenic macrophages, MHC-II levels increased following R848 stimulation, but no difference was seen between Gsdmd ^-/- and WT mice ( Figure 5G ). CD83, CD86 and CD80 only increased slightly following stimulation ( Figure 5G ). Findings were similar for splenic DCs ( Supplementary Figure S14E ). Considering these results, we did not test effects of TLR activation in tumor-bearing Gsdmd ^-/- and WT mice. In summary, Gsdmd^-/- and WT mice responded similarly to TLR7 activation, showing no significant differences in serum cytokine levels or immune cell activation markers. The expected upregulation of Il6, Cxcl10, Ifng, and Tnfa gene expression after R848 stimulation was comparable between both groups, indicating that GSDMD deficiency did not markedly affect the response to TLR7 activation in the assessed tissues.

TIMER2.0 (timer.cistrome.org) was used to analyse bulk tumor transcriptome data generated by the TCGA Research Network: https://www.cancer.gov/tcga. The statistical significance of GSDMD expression differences between healthy tissue and tumor tissue was computed by the Wilcoxon test using TIMER2.0. An adjusted partial Spearman’s test was used to calculate a rho value as the degree of gene correlation. Patients with the highest (top 25%) and lowest (bottom 25%) GSDMD expression were segregated into two cohorts, in which overall survival was compared with a Cox proportional-hazards model. A Wald test was used to determine the P value. MRNA expression data for BRCA and LIHC cancer types was downloaded from the TCGA portal. The signature genes previously described (18) were extracted from the expression data for the top 25% and bottom 25% GSDMD expressing patient cohorts and differential expression was visualized using Tibco Spotfire 12.5. For comparison the two-tailed t test was used.

To show the expression of GSDMD according to cell types, we used publicly available scRNA datasets from breast cancer (n = 84) (22) and hepatocellular carcinoma (n = 21) (23). The original cell type annotation from the respective authors was used. Pseudo-bulk summaries per patient were constructed by averaging the gene expression over all cells in a given cell type and plotted using the method of Bill et al. (2023) (45). Epithelial cell adhesion molecule (EpCAM) and protein tyrosine phosphatase receptor type C (PTPRC), also known as CD45, were used as controls to show specificity of the patterns of expression in epithelial and immune cells, respectively, under this analysis approach.

Immunohistochemical multiplex staining was performed on 4-μm sections from FFPE blocks using the fully automated Discovery Ultra system (ROCHE Diagnostics). The process included baking, deparaffinization, cell conditioning, staining, IHC, counterstaining, and titration. Antigen retrieval was conducted using CC1 for 64 minutes, followed by incubation with a 2% normal goat serum solution at 37°C for 32 minutes. A peroxidase blocking solution was then applied.

For single stainings, rabbit polyclonal anti-GSDMD antibody (20770-1-AP, Proteintech) was used at a 1:200 dilution and incubated for 60 minutes at 37°C. The primary antibody was detected with anti-rabbit HRP complex and revealed by diaminobenzidine using automated routine procedures. Brown coloration corresponds to a positive staining.

For double stainings, the primary antibody was followed by an anti-rabbit-AP system for 16 minutes and a Yellow detection kit. Antibody denaturation was performed using CC2 at 100°C for 24 minutes before the second antibody incubation. For the second marker, mouse monoclonal anti-CD79a antibody (M7050, DAKO) or mouse monoclonal anti-CD68 PG antibody (M0876, DAKO) were used at a 1:50 dilution and incubated for 60 minutes at 37°C. An amplification system was then applied, followed by OmniMap Mouse HRP for 24 minutes, and Teal detection reagents for 32 minutes. Hematoxylin and Bluing Reagent were used as counterstains.

Bone marrow cells were flushed out of freshly isolated tibias and femurs of 8-12 week-old male or female C57BL/6N-Gsdmd^em4Fcw/J (Gsdmd^-/- ; The Jackson Laboratory, 032410) or WT C57BL/6N mice, using PBS. The cell suspension was flushed through a 40 µM filter (Greiner, 7542041). Red blood cells were lysed using BD Pharm Lyse™ Lysing Buffer (BD, 555899). Cells were then cultured in differentiation medium composed of RPMI (GIBCO, 11875093), (Pan Biotech, p30-3300), 1% Penicillin-Streptomycin (Gibco, 15140122), 1% L-glutamine (GIBCO, 25030081), 0.5% sodium pyruvate (GIBCO, 11360070), 0.1% β-mercaptoethanol (GIBCO, 21985023), and M-CSF (10ng/ml, Miltenyi Biotec, 130-101-705). BMDM were cultured at a density of 80.000 cells/cm² in a 37°C, 5% CO₂ humidified environment. Medium was refreshed 2 days and 4 days after isolation. On day 6, BMDM were collected by scraping in PBS, seeded at 6.25x10⁵ cells/cm² and left to adhere overnight in a 37°C, 5% CO₂ humidified environment. Cells were stimulated with lipopolysaccharide (LPS; 1 ug/ml, Invivogen, tlrl-eblps) for 3 hours, before adenosine triphosphate (ATP; 10 mM, Sigma, A2383) in Opti-MEM™ (Gibco, 31985062) was added for 1 hour. Culture supernatants were collected for measurements of IL-1β levels. IL-1β levels were measured using ELISA MAX™ following manufacturer’s instructions.

EO771 (ATCC CRL-3461) and Hepa1-6 (ATCC CRL-1830) murine cell lines were cultured using respectively DMEM (Gibco, 41965-039), 10% FBS (Pan Biotech, p30-3300), 1% Penicillin-Streptomycin (Gibco, 15140122) or DMEM (GIBCO, 41965-039), 10% FBS (Pan Biotech, p30-3300) medium in a 37°C, 5% CO₂ humidified environment. Cells were detached using trypsin-EDTA (0.5%) (Gibco, 15400054). Cell suspensions were washed twice using PBS (Gibco, 10010015). Hepa1-6 cells were then injected at a concentration of 5x10⁶ cells/mL in a volume of 200 µL s.c. in the right upper flank of 8-12 week-old female Gsdmd^-/- or WT C57BL/6N mice. For EO771 cell injection, mice were anesthetized using isofluorane (Piranal, B03A16C). EO771 cells were injected into the 4^th mammary pad of 8-12 week-old female Gsdmd^-/- or WT C57BL/6N mice at a concentration of 0.5x10⁶ cells/mL in a volume of 20 µL. Tumor surface was calculated by multiplying the width and length of the tumor. Mice were sacrificed by CO₂ euthanasia (Tem Sega automate) when the tumor surface reached 200 mm². For mice bearing Hepa1-6 tumors, the tumor, spleen and the two axillary lymph nodes were collected. For mice bearing EO771 tumors, the tumor, spleen and the two subiliac lymph nodes were collected. Tissue was stored in TRIzol™ (ThermoFisher, 15596026) for RNA isolation. For flow cytometry, tumors were digested using a Tumor Dissociation Kit (130-096-730, Miltenyi) and the gentleMACS™ (130-096-427, Miltenyi) 37C_m_TDK_2 program. The digested tumors, spleen and lymph nodes were passed through a 40 µM filter (Greiner, 7542041) and red blood cells were lysed using BD Pharm Lyse™ Lysing Buffer (BD, 555899). All animal studies were approved by the Geneva cantonal authority for animal experimentation. Homozygous Gsdmd^-/- and WT C57BL/6N mice for all mentioned in vivo experiments were bred in house and maintained in the SPF facility of the Geneva University School of Medicine (Geneva, Switzerland).

Bone marrow cells (2x10⁶ cells/mL) and splenocytes (2x10⁶ cells/mL) were seeded in a 200 µL volume in a 96-well plate (Greiner, 655083) and left overnight in a 37°C, 5% CO₂ humidified environment. All cells were cultured in differentiation medium without M-CSF. The cells were stimulated using R848 (0.1 µg/mL, InvivoGen, tlrl-r848), LPS (0.1 µg/mL, Invivogen, tlrl-3pelps), 3’3’cGAMP (10 µg/mL, InvivoGen, tlrl-nacga), CpG (30 µg/mL, InvivoGen, ODN 1585) or poly(I:C) (200 µg/mL, InvivoGen, tlrl-picw). After an 18-hour incubation, the plate was centrifuged at 400 g for 5 min and conditioned medium was collected. IL-6 levels were measured using ELISA MAX™. CXCL10 was measured using a DuoSet ELISA (R&D systems, DY466). Activation marker expression was assessed by flow cytometry.

The samples were separated using SDS-PAGE and then transferred onto a membrane (Macherey-Nagel). The membrane was blocked in 0.05% Tris-buffered saline, Tween 20, containing 5% milk for 1 hour at room temperature and immunoblotting was performed with anti-GSDMD antibody (Abcam, EPR19828). Immunoreactive bands were visualized by Odyssey (LICOR) using appropriate secondary reagents.

For the in vivo R848 stimulation, R848 (invivogen, TLR-R848) was brought to a concentration of 2 µg/mL in PBS and 200 µl were injected s.c. into 8-12 week-old female Gsdmd^-/- or WT C57BL/6N mice. Mice were sacrificed after 3 or 24 hours and lymph nodes, blood, spleen, bone marrow and liver were collected. Tissues were prepared for RNA isolation or for flow cytometry as described above. IL-6 levels were measured in serum using ELISA MAX™ (Biolegend, 431316), as this cytokine is highly sensitive to measure activity of TLR7 ligands (46, 47).

Total RNA was extracted using Trizol (Thermo Fischer). cDNA was generated from 1 µg total RNA using a High Capacity cDNA Reverse Transcription kit (Thermo Fischer, 4368814). Non-reverse-transcribed RNA samples were used as negative controls. PCR reactions were executed on a QuantStudio 5 system, using PowerUp SYBR Green Master mix (Thermo Fisher, A25742) and the primers listed in Table 2 .

Cell suspensions were washed with PBS and treated with Fc Receptor block (1:200, TruStain FCX anti-mouse CD16/32, Biolegend, S17011E) in FACS buffer (PBS, 0.5% BSA (PAN Biotech, P06-1391500) and 2mM EDTA (Promega, V4231)). After 15 min of blocking, antibodies in FACS buffer were added to stain the cells. Antibodies and dilutions used for the staining are listed in Table 3 . Marker expression was assessed using a Novocyte Flow Cytometer 3000 (Agilent). Flow cytometry data was analysed using FlowJo 10. Gating strategies used are outlined in Supplementary Figures S4 , S5 , S8 and S9 .

Data represent the average and SD of individual experiments or mice. A 2-way ANOVA, using Graphpad Prism 10, was used to determine P values. (GraphPad Software, La Jolla, CA, USA; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001).