Intratumoral bacteria are immunosuppressive and promote immunotherapy resistance in head and neck squamous cell carcinoma

Abstract

Despite the promise of immune checkpoint blockade (ICB) in head and neck squamous cell carcinoma (HNSCC), mediators of response are poorly understood. To address this, here we analyzed oropharyngeal HNSCCs treated with neoadjuvant durvalumab (anti-PDL1) alone or in combination with tremelimumab (anti-CTLA4) from the CIAO clinical trial ( NCT03144778 ). We found that only the total abundance of intratumoral bacteria predicted ICB response, which was validated in multiple independent cohorts. High intratumoral bacteria abundance was associated with an immunosuppressive tumor microenvironment, characterized by an accumulation of neutrophils coupled with depletion of T cells and other adaptive immune cells. Experimental elevation or reduction in intratumoral bacteria abundance in orthotopic models of HNSCC in female mice recapitulated immunological associations observed in participant tumors. Increasing intratumoral bacteria abundance was sufficient to induce resistance to anti-PDL1 ICB, irrespective of bacterial species tested. Together, these findings demonstrate that high intratumoral bacteria abundance is a key suppressor of antitumor immunity and promotes immunotherapy resistance.

Fig. 1. High intratumoral bacteria is associated with ICB resistance.

a, Overview of study cohort and association of clinicopathological variables with response to neoadjuvant immunotherapy. P values were determined using a two-tailed Fisher’s exact test, except for stage, which was analyzed by a chi-squared test (n = 28). b, Comparison of PDL1 combined positivity score between responders (PR) and nonresponders (SD and PD). Data are shown as the mean and s.d. Statistical analysis was conducted using a two-tailed Welch’s t-test (responder, n = 15; nonresponder, n = 12). c, Comparison of TMB, defined as mutations per megabase of sequenced DNA, between responders (PR) and nonresponders (SD and PD). Data are shown as the median and interquartile range. Statistical analysis was conducted using a two-tailed rank-sum test (responder, n = 16; nonresponder, n = 12). d, Comparison of immune cell populations from RNA sequencing deconvolution using ssGSEA with signatures from Bindea et al.^, between responders (PR) and nonresponders (SD and PD). Statistical analysis was conducted using a two-tailed Welch’s t-test with false discovery rate (FDR) determined by Benjamini–Hochberg procedure (responder, n = 16; nonresponder, n = 12). e, Comparison of HPV viral burden, defined as the number of HPV reads per million human reads, between responders (PR) and nonresponders (SD and PD). Data are shown as the mean and s.d. Statistical analysis was conducted using a two-tailed Welch’s t-test (responder, n = 16; nonresponder, n = 12). f, Comparison of TBB between responders (PR) and nonresponders (SD and PD). Data are shown as the mean and s.d. Statistical analysis was conducted using a two-tailed Welch’s t-test (responder, n = 16; nonresponder, n = 12). g, Comparison of TBB between pathological responders and nonresponders. Data are shown as the mean and s.d. Statistical analysis was conducted using a two-tailed Welch’s t-test (responder, n = 14; nonresponder, n = 11). h, Comparison of intratumoral microbiome diversity, quantified by Shannon diversity index, between responders (PR) and nonresponders (SD and PD). Data are shown as the mean and s.d. Statistical analysis was conducted using a two-tailed Welch’s t-test (responder, n = 16; nonresponder, n = 12). i, Correlation of TBB with relative Fusobacterium. Statistical analysis was conducted using a two-tailed Spearman correlation coefficient (n = 28). j, Univariable regression assessing response to ICB as a function of TBB and relative Fusobacterium. Error bars indicate the 95% confidence interval (responder, n = 16; nonresponder, n = 12). k, Multivariable regression assessing response to ICB as a function of TBB and relative Fusobacterium. Error bars indicate the 95% confidence interval (responder, n = 16; nonresponder, n = 12). T_EM cells, effector memory T cells. Source data

Fig. 2. Validation of TBB as a biomarker for ICB response.

a, Progression-free survival in an internal cohort of participants with HNSCC treated with immunotherapy stratified by median TBB determined from WES. Statistical analysis was conducted using a two-tailed log-rank test (n = 19). HR, hazard ratio. b, Progression-free survival in an internal cohort of participants with HNSCC treated with immunotherapy stratified median Fusobacteria abundance from using WES. Statistical analysis was conducted using a two-tailed log-rank test (n = 19). c, Regression coefficient for TBB, relative Fusobacterium or total Escherichia to predict response (CR or PR) to ICB in Hartwig NSCLC samples. Values were log-transformed and z-normalized before regression. Cancer type and biopsy site were taken as fixed effects and the sequencing platform was taken as a random effect. Error bars indicate the 95% confidence interval (n = 72). d, Receiver operating characteristic curve for ability of TBB, relative Fusobacterium or total Escherichia to predict response (CR or PR) to ICB in Hartwig NSCLC samples (responder, n = 12; nonresponder, n = 60). e, Percentage of responders (CR or PR) to ICB in Hartwig NSCLC samples stratified by positivity for Escherichia. Inset numbers indicate the number of responders among total participants. Statistical analysis was conducted using a two-tailed Fisher’s exact test. f, Multivariable regression for ability of TBB, relative Fusobacterium and total Escherichia to predict response to ICB in Hartwig NSCLC, HNSCC, urothelial carcinoma, mesothelioma and small intestine or colorectal cancer samples. Values were log-transformed and z-normalized before regression. Cancer type and biopsy site were taken as fixed effects and the sequencing platform was taken as a random effect. Error bars indicate the 95% confidence interval (responder, n = 23; nonresponder, n = 122). g, Multivariable regression for ability of TBB, relative Fusobacterium and total Escherichia to predict response (CR or PR) to chemotherapy or targeted therapies in Hartwig NSCLC, HNSCC, urothelial carcinoma, mesothelioma and small intestine or colorectal cancer samples. Values were log-transformed and z-normalized before regression. Cancer type and biopsy site were taken as fixed effects and the sequencing platform was taken as a random effect. Error bars indicate the 95% confidence interval (responder, n = 69; nonresponder, n = 273). Source data

Fig. 3. Microbial, clinical and molecular correlates of intratumoral bacteria burden.

a, Correlation of WGS TBB with intratumoral microbiome diversity. Statistical analysis was conducted using a two-tailed Spearman correlation coefficient (n = 130). b, Correlation of WGS TBB with relative fractions of indicated genera, quantified as fraction of reads for each genera among all mapped bacterial reads. Statistical analysis was conducted using a two-tailed Spearman correlation coefficient. Highlighted genera indicate an FDR < 0.05 (n = 130). c, Correlation of WGS TBB with absolute abundances of indicated genera, quantified as the number of reads per million human reads. Statistical analysis was conducted using a two-tailed Spearman correlation coefficient. Highlighted genera indicate an FDR < 0.05 (n = 130). d, Correlation of WGS TBB with relative fraction of F. nucleatum across cancers (HNSCC, n = 157; stomach adenocarcinoma (STAD), n = 128; esophageal carcinoma (ESCA), n = 62; colorectal adenocarcinoma and rectal adenocarcinoma (COADREAD), n = 170). Statistical analysis was conducted using a two-tailed Spearman correlation coefficient. e–g, WGS TBB in tumors from participants with HNSCC analyzed on the basis of pathological T stage (e; T1, n = 15; T2, n = 41; T3, n = 31; T4, n = 49), tumor location (f; oral cavity, n = 106; larynx, n = 24; oropharynx, n = 27) or tumor HPV status (g; HPV-negative, n = 110; HPV-positive, n = 38). The center white point is the median and the box is the interquartile range. Statistical analysis was conducted using a Kruskal–Wallis (e,f) or two-tailed rank-sum test (g). h, Difference in WES TBB in HNSCC tumors based on mutations in specific genes relevant to HNSCC. Statistical analysis was conducted using a two-tailed Welch’s t-test. Highlighted values indicate P < 0.05 (n = 507). i, Correlation between WES TBB and protein expression from reverse-phase protein array targeted proteomics in HNSCC tumors. Highlighted values indicate P values < 0.05 (n = 349). Statistical analysis was conducted using a two-tailed Spearman correlation coefficient. j, GSEA of genes correlated with WES TBB using Hallmark gene sets. Highlighted values indicate an FDR < 0.05 (n = 503). Source data

Fig. 4. Changes in the tumor immune microenvironment associated with intratumoral bacteria burden.

a, Correlation coefficients of TBB determined from WGS with immune cell populations in HNSCC tumors. Statistical analysis was conducted using a two-tailed Spearman correlation. Highlighted values indicate an FDR < 0.05 (n = 157). T_FH cells, T follicular helper cells; T_CM cells, central memory T cells. b, Correlation coefficients of TBB determined from WES in independent samples not analyzed by WGS (excluding samples from a) with immune cell populations in HNSCC tumors. Statistical analysis was conducted using a two-tailed Spearman correlation. Highlighted values indicate an FDR < 0.05 (n = 348). c, Correlation coefficients of TBB with immune cell populations in tumors from CIAO trial. Statistical analysis was conducted using a two-tailed Spearman correlation. Highlighted values indicate P < 0.05 (n = 28). d, Correlation of TBB with log-transformed tNLR in samples from CIAO trial. Statistical analysis was conducted using a two-tailed Spearman correlation (n = 28). e, Comparison of log-transformed tNLR between responders (PR) and nonresponders (SD and PD) from CIAO trial. Data are shown as the mean and 95% confidence interval. Statistical analysis was conducted using a two-tailed Welch’s t-test (nonresponder, n = 16; responder, n = 12). f, Spearman correlation coefficients of TBB with tNLR across cancer types for HNSCC by WGS (n = 157) and WES (n = 503), STAD by WGS (n = 114) and WES (n = 413), ESCA by WGS (n = 62) and WES (n = 183) and COADREAD by WGS (n = 155) and WES (n = 582). Statistical analysis was conducted using a two-tailed Spearman correlation. g–i, Correlation of imaging-based quantification of bacteria by 16S rRNA in situ hybridization with T cells detected by CD3 (g), neutrophils detected by CD66b (h) and difference between neutrophils and T cells (i). Inset values indicate the two-tailed Spearman correlation coefficient (n = 17). j, Changes induced by in vitro exposure of HNSCC cell lines to either Fusobacterium or Prevotella, with each dot representing an individual gene. Four separate cell lines were exposed to bacteria and the average fold change is shown. The inset value indicates the two-tailed Pearson correlation coefficient. Values represent the average of n = 4 cell lines, each analyzed in duplicate. k, Comparison correlation coefficients between individual genes and TBB in participant tumors (n = 503) with average change induced by in vitro exposure of HNSCC cell lines to Fusobacterium and Prevotella (n = 4 cell lines). In a–c, immune populations were determined by ssGSEA using signatures from Bindea et al.^,. Source data

Fig. 5. Depletion of intratumoral bacteria remodels the tumor microenvironment in preclinical models.

a, Experimental schematic. Syngeneic oral cancer cell lines were implanted either subcutaneously or in the orthotopic (tongue) location. For antibiotic studies, antibiotic treatment (ABX) was initiated 1 week before tumor implantation. b, Representative immunofluorescence staining detecting bacteria in tumors for the treatment groups (green, bacterial 16S rRNA); cell nuclei were stained with DAPI (blue). Scale bar, 100 μm. c, Quantification of relative bacteria in the ABX versus control treatment groups in the flank versus orthotopic location (n = 5). Data are shown as the mean and s.d. d–g, Comparison of tumor volume growth curves with and without ABX bacterial depletion for control (n = 10) and ABX (n = 8) MOC2 orthotopic tumors (d), control (n = 10) and ABX (n = 10) MOC2 subcutaneous tumors (e), control (n = 9) and ABX (n = 10) MOC1 orthotopic tumors (f) and control (n = 6) and ABX (n = 6) MOC1 subcutaneous tumors (g). All tumor volumes are presented as the mean and s.e.m. Statistical analysis was conducted using an unpaired two-tailed t-test. NS, not significant. h, Immune deconvolution of mRNA expression profiling using ssGSEA with signatures from Bindea et al.^, from orthotopic tumors in mice with or without ABX demonstrating increased neutrophils and decreased CD8 T cells (n = 5). Statistical analysis was conducted using a two-tailed t-test. The box represents the median and interquartile range and whiskers represent the range. i, Comparison of changes in intratumoral immune cell populations quantified by immune deconvolution using ssGSEA with signatures from Bindea et al.^, induced by ABX in mice bearing either orthotopic tumors (x axis) or subcutaneous tumors (y axis). Blue-highlighted populations indicate populations significantly changed in orthotopic tumors. Inset values indicate the two-tailed Spearman correlation coefficient (n = 5). j, Changes in cytokine expression induced by ABX in orthotopic tumors or subcutaneous tumors, as well as the interaction effect between ABX and tumor site. The dot color indicates the direction and magnitude of change; nominally significant relationships (P < 0.05) are encircled in black. Significance was determined using a two-way ANOVA (interaction effect, ABX × site), followed by Šidák post hoc analysis to compare antibiotic-induced changes in the orthotopic and subcutaneous sites (n = 5). k,l, Comparison of tumor volumes after depletion of CD4-positive and CD8-positive cells (k; n = 6) or Ly6G-positive cells (l; n = 9) in orthotopic MOC1 tumors. All tumor volumes are presented as the mean and s.e.m. Statistical analysis was conducted using an unpaired two-tailed Welch’s t-test. Source data

Fig. 6. Increasing intratumoral bacteria burden induces resistance to ICB.

a, Schematic of experimental bacterial inoculation and immunotherapy treatment. b, Representative images of control tumors or tumors inoculated with indicated microbes stained for CD3 (T cells, cyan), Ly6G (neutrophils, red), 16S rRNA (bacteria, green) and nuclear counterstain (DAPI, white). Scale bar, 100 μm. c–f, Quantifications of images in b for 16S rRNA staining (c), neutrophils (d), T cells (e) and log₂ ratio of neutrophil to T cell density (f) (control, n = 8; otherwise, n = 7). Statistical analysis was conducted using a one-way ANOVA with Benjamini–Hochberg correction for multiple comparisons. The box center indicates the median, box edges indicate the interquartile range and whiskers indicate the minimum and maximum. g–j, Tumor volumes in orthotopic MOC1 tumors with and without anti-PDL1 treatment or IgG control for control mice (g) and mice orally inoculated with F. nucleatum (h), P. scopos (i) or C. rectus (j) (n = 8). Tumor volumes are presented as the mean and s.e.m. Statistical analysis was conducted using an unpaired two-tailed Welch’s t-test. k,l, Tumor volumes in subcutaneous MOC1 tumors with anti-PDL1 treatment or IgG control for control mice (k) and mice orally inoculated with F. nucleatum (l) (n = 8). Tumor volumes are presented as the mean and s.e.m. Statistical analysis was conducted using a unpaired two-tailed Welch’s t-test. Panel a created with BioRender.com. Source data

Extended Data Fig. 1. Quantification of tumor bacteria burden from bulk DNA sequencing.

(A) Comparison of average relative abundance of bacterial genera in oral cavity tumors quantified from TCGA WGS data in this study (n = 87) to results from 16S rRNA sequencing from Michikawa et al. (n = 33). Two-tailed Pearson correlation coefficient. (B) Comparison of average relative abundance of bacterial genera in oral cavity tumors quantified from TCGA WGS data from the retracted work by Poore et al. (n = 157) to results from 16S rRNA sequencing from Michikawa et al. (n = 33). Two-tailed Pearson correlation coefficient. (C) Comparison of relative bacteria abundance by 16S rRNA sequencing in oral cavity samples (n = 66) compared to sham controls (n = 22). Red highlighted taxa indicate a subset of contaminants identified. See Table S1. (D) Comparison of average relative bacterial abundances at the phylum level in a cohort of samples analyzed by both whole exome sequencing and 16S rRNA sequencing. Two-tailed Pearson correlation coefficient. n = 22 (E) Sample-wise two-tailed Pearson correlation coefficients of average relative bacterial abundance at the phylum level in samples profiled by both whole exome sequencing and 16S rRNA sequencing. Shaded area indicates 95% confidence interval of correlation coefficients for unmatched samples (for example random pairs). n = 22. (F) Comparison of average relative bacterial abundances at the genus level in a cohort of samples analyzed by both whole exome sequencing and 16S rRNA sequencing. Two-tailed Pearson correlation coefficient. n = 22. (G) Sample-wise two-tailed Pearson correlation coefficients of average relative bacterial abundance at the genus level in samples profiled by both whole exome sequencing and 16S rRNA sequencing. Shaded area indicates 95% confidence interval of correlation coefficients for unmatched samples (for example random pairs). n = 22. (H) Comparison of tumor bacteria burden determine by whole exome sequencing with 16S rRNA qPCR quantification of intratumoral bacteria. Two-tailed Pearson correlation coefficient. n = 22. (I) Pearson correlation coefficients (two-tailed) of TBB determined from imaging 16S rRNA in situ hybridization with TBB determined by qPCR for 16S rRNA. Shaded region indicates detection limit. n = 17. (J) Pearson correlation coefficients (two-tailed) of TBB determined from imaging 16S rRNA in situ hybridization on two slides of the same tumor. Shaded region indicates detection limit. n = 17. Source data

Extended Data Fig. 2. Evaluation of tumor bacteria burden in TCGA samples.

(A) Comparison of TBB determined based on whole genome sequencing (WGS) and whole exome sequencing (WES) in samples with both approaches available. Two-tailed Spearman correlation coefficient. n = 155. (B) Correlation of average relative microbial abundances between WGS and WES in matched samples. Analysis was restricted to WES samples with ≥50 total microbial reads for calculation of relative abundance values. Two-tailed Pearson correlation coefficient. n = 67. (C) Sample-wise comparison of relative microbial abundances determined by WGS and WES in matched samples. Each dot represents the two-tailed Pearson correlation coefficient. correlation coefficient from an individual tumor sample. Analysis was restricted to WES samples with ≥50 total microbial reads for calculation of relative abundance values. Median with interquartile range. The shaded region represents the mean with 95% confidence interval for correlation of unmatched samples. n = 67. (D) Comparison of tumor bacteria burden determined by whole genome sequencing (WGS) based on tissue collection site in oral cavity HNSC samples. Linear regression shown +/- 95% confidence interval. Raw P-value and P-value adjusted for multiple comparisons by Benjamini-Hochberg procedure shown. (E) Comparison of tumor bacteria burden determined by whole genome sequencing based on DNA sequencing site in oral cavity HNSC samples. Each dot represents a single tumor; mean +/- stdev. One-way ANOVA. HMS n = 62; MDACC n = 20; BCM n = 16; BI n = 8. (F) Comparison of tumor bacteria burden determined by whole exome sequencing (WES) based on tissue collection site in oral cavity HNSC samples. Linear regression shown +/- 95% confidence interval. Raw P-value and P-value adjusted for multiple comparisons by Benjamini-Hochberg procedure shown. Source data

Extended Data Fig. 3. Correlation of bacterial species with tumor bacteria burden.

(A) Correlation of relative fractions of Fusobacterium, quantified as fraction of reads out of all mapped bacterial reads with TBB, quantified as bacterial reads per million human reads, in all TCGA HNSCC samples. Two-tailed Spearman correlation. n = 130. (B) Correlation of relative fractions of Fusobacterium, quantified as fraction of reads out of all mapped bacterial reads, with tumor bacteria burden in only oropharynx HNSCC samples from TCGA. Two-tailed Spearman correlation. n = 50. (C) Breakdown of relative proportion of bacteria in TCGA WGS HNSCC samples into anaerobic, aerobic, or not classified (unknown or genera with mixture of both aerobic and anaerobic). Inset numbers indicate mean +/- std. n = 130. (D) Correlation of tumor bacteria burden, defined as bacteria reads per million human reads, with total anaerobic bacteria, defined as anaerobic bacteria reads per million human reads, in TCGA HNSCC samples profiled by WGS. Two-tailed Spearman correlation coefficient. n = 130. (E) Correlation of tumor bacteria burden, defined as bacteria reads per million human reads, with total aerobic bacteria, defined as aerobic bacteria reads per million human reads, in TCGA HNSCC samples profiled by WGS. Two-tailed Spearman correlation coefficient. n = 130. (F) Correlation of tumor bacteria burden, defined as bacteria reads per million human reads, with relative anaerobic bacteria, defined as anaerobic bacteria reads per total bacteria reads, in TCGA HNSCC samples profiled by WGS. Two-tailed Spearman correlation coefficient. n = 130. (G) Correlation of tumor bacteria burden, defined as bacteria reads per million human reads, with relative aerobic bacteria, defined as aerobic bacteria reads per total bacteria reads, in TCGA HNSCC samples profiled by WGS. Two-tailed Spearman correlation coefficient. n = 130. Source data

Extended Data Fig. 4. Clinical correlates with tumor bacteria burden in HNSCC.

Comparison of TBB based on (A) Hypoxia score determined from RNAseq (n = 157), (B) pathological stage (Stage I n = 10; Stage II n = 25; Stage III n = 23; Stage IV n = 74), (C) clinical stage (Stage I-II n = 38; Stage III n = 31; Stage IV n = 87), (D) tumor grade (Grade 1-2 n = 108; Grade 3 n = 44), (E) tumor grade in only oral cavity tumors (Grade 1-2 n = 79; Grade 3 n = 26), (F) sex (Male n = 114, Female n = 43), (G) ethnicity (White n = 141, Black/Afr. American n = 12, Asian n = 2), (H) patient age (n = 157), (I) subsite within the oral cavity (oral tongue n = 55, not otherwise specified n = 25, floor of mouth n = 12, other = 8, alveolar ridge n = 6), (J) HPV status for only oropharynx tumors (HPV+ n = 23, HPV- n = 3), (K) alcohol history (high n = 50, no/low n = 21), or (L) smoking status (n = 155). (M) Multivariable regression of HPV status and smoking status (current versus never), error bars represent 95% confidence interval. Unless otherwise noted, comparisons between two groups were made with a rank-sum test, comparisons between more than two groups were made with a Kruskal-Wallis test, and comparisons between two continuous variables were made using a Spearman correlation coefficient. Center white point is median, box is interquartile range. Source data

Extended Data Fig. 5. Association of tumor bacterial burden with mutations.

(A) Difference in WES tumor bacteria burden (TBB) in HPV-negative HNSCC tumors based on mutations in specific genes relevant to HNSCC. Two-tailed Welch’s t-test. Highlighted values indicate P-values<0.05. n = 415. (B) Association of WES TBB with mutations in specific genes relevant to HNSCC and HPV status assessed by multivariable regression of log-transformed TBB with mutation status + HPV status. Highlighted values indicate P-values<0.05. n = 467. Source data

Extended Data Fig. 6. Gene expression changes in cell lines infected with tumor-relevant bacteria in vitro.

(A-D) Comparison of gene expression changes following infection with either Fusobacterium nucleatum or Prevotella scopos in human HNSCC cell lines OQ01 (A), SCC25 (B), FaDu (C), and PCI-15B (D). Pearson correlation coefficients. n = 2 technical replicates per cell line from one representative experiment. (E-J) Expression of indicated genes in OQ01 cells infected with either live or dead (heat-killed) Fusobacterium nucleatum compared to mock-infected control. n = 4. Mean with s.e.m. (K) Comparison of gene expression changes following infection with either Fusobacterium nucleatum or Prevotella scopos in mouse cell line MOC1. Pearson correlation coefficients. n = 3 technical replicates from one representative experiment. (L) Comparison of average gene expression alterations following exposure to Prevotella scopos and Fusobacterium nucleatum in mouse (n = 1 cell line in technical triplicate) and human (n = 4 cell lines in technical duplicate from one representative experiment per cell line) HNSCC cell lines. Inset value indicates Spearman correlation coefficient. (M) Comparison of average gene expression alterations following exposure to Prevotella scopos and Fusobacterium nucleatum in MOC1 mouse cancer cells in vitro (n = 1 cell line in technical triplicate from one representative experiment) and orthotopic MOC1 tumors treated with antibiotics in vivo (n = 5). Dotted line indicates 0 for change post-antibiotics. Source data

Extended Data Fig. 7. Murine intratumoral microbiome.

(A) Relative abundance of indicated bacteria in MOC1 orthotopic tumors, as determined by 16S rRNA sequencing. Showing taxa detected in at least two tumors. n = 5. (B) Average relative abundance for all detected taxa in MOC1 orthotopic tumors, as determined by 16S rRNA sequencing. n = 5. Source data

Extended Data Fig. 8. Effect of antibiotics on immune checkpoint blockade response in HNSCC.

Clinical benefit rate for patients with HNSCC receiving anti-PD(L)1 immune checkpoint blockade in the presence or absence of concurrent antibiotics. Inset fraction indicates number with clinical benefit over total sample size. n = 111 total. Two-tailed Chi-square test. Source data

Extended Data Fig. 9. In vivo immune cell depletion.

C57BL/6 mice were treated with depletion antibodies and whole blood was collected week 3 for flow cytometry analysis. (A) Blood was stained with anti-CD45, anti-CD4 and anti-CD8, followed by flow cytometry sorting of CD45+CD4+ and CD45+CD8+ population. Flow Cytometry gating strategy to identify CD45+CD4+ and CD45+CD8+ cells shown. (B and C) Quantification of percentage of subpopulations of sorted CD4+ and CD8+ cells in the CD45+ population. Data are presented as mean ± SEM (n=3). P values by two-tailed unpaired t test. (D) Blood was stained with anti-CD45, anti-CD11b, anti-Ly6G followed by flow cytometry sorting of CD45+CD11b+Ly6G+ population with the gating strategy. (E) Quantification of percentage of subpopulations of sorted Ly6G+ cells in the CD45+CD11b+ population. Data are presented as mean ± SEM (n=3). P values by two-tailed unpaired t test. Source data

Extended Data Fig. 10. Depletion of oral microbiome with antibiotics.

(A) Culture of bacterial swabs in either aerobic or anerobic conditions from control mice and mice treated with antibiotics (ABX) for four days prior to tumor injection. Three representative technical replicates from two biological replicates shown. (B) Quantitative PCR on bacterial swabs from control mice and mice treated with antibiotics (ABX) for four days prior to tumor injection. Each dot represents an individual mouse. Median +/- interquartile range. Rank-sum test. n = 5. Source data