Ample evidence indicates that the gut microbiome is a tumor-extrinsic factor associated with antitumor response to anti-programmed cell death protein-1 (PD-1) therapy, but inconsistencies exist between published microbial signatures associated with clinical outcomes. To resolve this, we evaluated a new melanoma cohort, along with four published datasets. Time-to-event analysis showed that baseline microbiota composition was optimally associated with clinical outcome at approximately 1 year after initiation of treatment. Meta-analysis and other bioinformatic analyses of the combined data show that bacteria associated with favorable response are confined within the Actinobacteria phylum and the Lachnospiraceae/Ruminococcaceae families of Firmicutes. Conversely, Gram-negative bacteria were associated with an inflammatory host intestinal gene signature, increased blood neutrophil-to-lymphocyte ratio, and unfavorable outcome. Two microbial signatures, enriched for Lachnospiraceae spp. and Streptococcaceae spp., were associated with favorable and unfavorable clinical response, respectively, and with distinct immune-related adverse effects. Despite between-cohort heterogeneity, optimized all-minus-one supervised learning algorithms trained on batch-corrected microbiome data consistently predicted outcomes to programmed cell death protein-1 therapy in all cohorts. Gut microbial communities (microbiotypes) with nonuniform geographical distribution were associated with favorable and unfavorable outcomes, contributing to discrepancies between cohorts. Our findings shed new light on the complex interaction between the gut microbiome and response to cancer immunotherapy, providing a roadmap for future studies.
This is a preview of subscription content, access via your institution
Open Access articles citing this article.
Potential non-invasive biomarkers in tumor immune checkpoint inhibitor therapy: response and prognosis prediction
Biomarker Research Open Access 02 June 2023
Molecular Cancer Open Access 26 April 2023
Nature Communications Open Access 18 March 2023
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 12 print issues and online access
$209.00 per year
only $17.42 per issue
Rent or buy this article
Prices vary by article type
Prices may be subject to local taxes which are calculated during checkout
The corresponding authors will comply with all requests for raw and analyzed data and materials after verification whether the request is subject to any patients’ confidentiality obligation. Patient-related data not included in the paper were generated as part of clinical trials and might be subject to patient confidentiality. All sequencing (human and microbiome) data and de-identified metadata that support the findings have been deposited in NCBI databases and are all accessible via BioProject accession no. PRJNA762360. AGP data are available at the ENA database (https://www.ebi.ac.uk/) under accession no. PRJEB11419. Access to publicly available sequencing data of the other cohorts analyzed in this study was obtained through BioProject accession nos. PRJNA399742 (Chicago), PRJNA541981 (New York), PRJNA397906 (Dallas) and PRJEB22893 (Houston). Source data are provided with this paper. The Sequence Read Archive accession numbers for each sample from each of these cohorts is the stated name of the sample on the spreadsheet in the source data. All other data are provided within the paper and its Supplementary Information files.
All codes used for shotgun sequencing analysis can be found within the in-house JAMS_BW package, version 1.5.7, publicly available on GitHub (https://github.com/johnmcculloch/JAMS_BW/). GSEA analysis was done in R using fgsea package 1.19.4. Codes for transkingdom network analysis are available at https://github.com/richrr/TransNetDemo/. Additional codes used are part of the packages mentioned in the text or can be found on GitHub at https://github.com/trinchierilab/microbiotapd1melanoma2021/.
Sivan, A. et al. Commensal Bifidobacterium promotes antitumor immunity and facilitates anti-PD-L1 efficacy. Science 350, 1084–1089 (2015).
Vetizou, M. et al. Anticancer immunotherapy by CTLA-4 blockade relies on the gut microbiota. Science 350, 1079–1084 (2015).
Dzutsev, A., Goldszmid, R. S., Viaud, S., Zitvogel, L. & Trinchieri, G. The role of the microbiota in inflammation, carcinogenesis, and cancer therapy. Eur. J. Immunol. 45, 17–31 (2015).
Gopalakrishnan, V. et al. Gut microbiome modulates response to anti-PD-1 immunotherapy in melanoma patients. Science 359, 97–103 (2018).
Matson, V. et al. The commensal microbiome is associated with anti-PD-1 efficacy in metastatic melanoma patients. Science 359, 104–108 (2018).
Routy, B. et al. Gut microbiome influences efficacy of PD-1-based immunotherapy against epithelial tumors. Science 359, 91–97 (2018).
Chaput, N. et al. Baseline gut microbiota predicts clinical response and colitis in metastatic melanoma patients treated with ipilimumab. Ann. Oncol. 28, 1368–1379 (2017).
Dubin, K. et al. Intestinal microbiome analyses identify melanoma patients at risk for checkpoint-blockade-induced colitis. Nat. Commun. 7, 10391 (2016).
Andrews, M. C. et al. Gut microbiota signatures are associated with toxicity to combined CTLA-4 and PD-1 blockade. Nat. Med. 27, 1432–1441 (2021).
Baruch, E. N. et al. Fecal microbiota transplant promotes response in immunotherapy-refractory melanoma patients. Science 371, 602–609 (2021).
Davar, D. et al. Fecal microbiota transplant overcomes resistance to anti-PD-1 therapy in melanoma patients. Science 371, 595–602 (2021).
Frankel, A. E. et al. Metagenomic shotgun sequencing and unbiased metabolomic profiling identify specific human gut microbiota and metabolites associated with immune checkpoint therapy efficacy in melanoma patients. Neoplasia 19, 848–855 (2017).
Peters, B. A. et al. Relating the gut metagenome and metatranscriptome to immunotherapy responses in melanoma patients. Genome Med. 11, 61 (2019).
Gharaibeh, R. Z. & Jobin, C. Microbiota and cancer immunotherapy: in search of microbial signals. Gut 68, 385–388 (2019).
Limeta, A., Ji, B., Levin, M., Gatto, F. & Nielsen, J. Meta-analysis of the gut microbiota in predicting response to cancer immunotherapy in metastatic melanoma. JCI Insight 5, e140940 (2020).
Shaikh, F. Y. et al. A uniform computational approach improved on existing pipelines to reveal microbiome biomarkers of non-response to immune checkpoint inhibitors. Clin. Cancer Res. 27, 2571–2583 (2021).
Byrd, A. L. et al. Gut microbiome stability and dynamics in healthy donors and patients with non-gastrointestinal cancers. J. Exp. Med. 218, e20200606 (2021).
Derosa, L. et al. Gut bacteria composition drives primary resistance to cancer immunotherapy in renal cell carcinoma patients. Eur. Urol. 78, 195–206 (2020).
Viaud, S. et al. The intestinal microbiota modulates the anticancer immune effects of cyclophosphamide. Science 342, 971–976 (2013).
Ogluszka, M., Orzechowska, M., Jedroszka, D., Witas, P. & Bednarek, A. K. Evaluate Cutpoints: adaptable continuous data distribution system for determining survival in Kaplan–Meier estimator. Comput. Methods Prog. Biomed. 177, 133–139 (2019).
Capone, M. et al. Baseline neutrophil-to-lymphocyte ratio (NLR) and derived NLR could predict overall survival in patients with advanced melanoma treated with nivolumab. J. Immunother. Cancer 6, 74 (2018).
Valero, C. et al. Pretreatment neutrophil-to-lymphocyte ratio and mutational burden as biomarkers of tumor response to immune checkpoint inhibitors. Nat. Commun. 12, 729 (2021).
Ascierto, P. A. et al. Proteomic test for anti-PD-1 checkpoint blockade treatment of metastatic melanoma with and without BRAF mutations. J. Immunother. Cancer 7, 91 (2019).
Yang, A. P., Liu, J. P., Tao, W. Q. & Li, H. M. The diagnostic and predictive role of NLR, d-NLR and PLR in COVID-19 patients. Int. Immunopharmacol. 84, 106504 (2020).
Knight, J. M. et al. Noninvasive analysis of intestinal development in preterm and term infants using RNA-sequencing. Sci. Rep. 4, 5453 (2014).
Whitfield-Cargile, C. M. et al. The noninvasive exfoliated transcriptome (exfoliome) reflects the tissue-level transcriptome in a mouse model of NSAID enteropathy. Sci. Rep. 7, 14687 (2017).
Rodrigues, R. R., Shulzhenko, N. & Morgun, A. Transkingdom networks: a systems biology approach to identify causal members of host–microbiota interactions. Methods Mol. Biol. 1849, 227–242 (2018).
Yambartsev, A. et al. Unexpected links reflect the noise in networks. Biol. Direct 11, 52 (2016).
Das, S. et al. Immune-related adverse events and immune checkpoint inhibitor efficacy in patients with gastrointestinal cancer with food and drug administration-approved indications for immunotherapy. Oncologist 25, 669–679 (2020).
Matsuoka, H. et al. Correlation between immune-related adverse events and prognosis in patients with various cancers treated with anti PD-1 antibody. BMC Cancer 20, 656 (2020).
Suo, A. et al. Anti-PD-1-induced immune-related adverse events and survival outcomes in advanced melanoma. Oncologist 25, 438–446 (2020).
Horvath, A. et al. Biomarkers for oralization during long-term proton pump inhibitor therapy predict survival in cirrhosis. Sci. Rep. 9, 12000 (2019).
Cortellini, A. et al. Integrated analysis of concomitant medications and oncological outcomes from PD-1/PD-L1 checkpoint inhibitors in clinical practice. J. Immunother. Cancer 8, e001361 (2020).
Stacy, A. et al. Infection trains the host for microbiota-enhanced resistance to pathogens. Cell 184, 615–627 (2021).
Kelsey, C. M. et al. Gut microbiota composition is associated with newborn functional brain connectivity and behavioral temperament. Brain Behav. Immun. 91, 472–486 (2021).
Benito, M. et al. Adjustment of systematic microarray data biases. Bioinformatics 20, 105–114 (2004).
Lazar, C. et al. Batch effect removal methods for microarray gene expression data integration: a survey. Brief. Bioinform. 14, 469–490 (2013).
Segata, N. et al. Metagenomic biomarker discovery and explanation. Genome Biol. 12, R60 (2011).
Subramanian, A. et al. Gene-set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).
Kang, D. D., Froula, J., Egan, R. & Wang, Z. MetaBAT, an efficient tool for accurately reconstructing single genomes from complex microbial communities. PeerJ 3, e1165 (2015).
Kang, D. D. et al. MetaBAT 2: an adaptive binning algorithm for robust and efficient genome reconstruction from metagenome assemblies. PeerJ 7, e7359 (2019).
Balomenou, S. et al. Distinct functions of polysaccharide deacetylases in cell shape, neutral polysaccharide synthesis and virulence of Bacillus anthracis. Mol. Microbiol. 87, 867–883 (2013).
Bessman, N. J. et al. Dendritic cell-derived hepcidin sequesters iron from the microbiota to promote mucosal healing. Science 368, 186–189 (2020).
Kjer-Nielsen, L. et al. MR1 presents microbial vitamin B metabolites to MAIT cells. Nature 491, 717–723 (2012).
Fan, S. et al. Cloning, characterization and production of three alpha-l-fucosidases from Clostridium perfringens ATCC 13124. J. Basic Microbiol. 56, 347–357 (2016).
Wright, D. P., Rosendale, D. I. & Robertson, A. M. Prevotella enzymes involved in mucin oligosaccharide degradation and evidence for a small operon of genes expressed during growth on mucin. FEMS Microbiol. Lett. 190, 73–79 (2000).
Arumugam, M. et al. Enterotypes of the human gut microbiome. Nature 473, 174–180 (2011).
Brooks, A. W., Priya, S., Blekhman, R. & Bordenstein, S. R. Gut microbiota diversity across ethnicities in the United States. PLoS Biol. 16, e2006842 (2018).
Dwiyanto, J. et al. Ethnicity influences the gut microbiota of individuals sharing a geographical location: a cross-sectional study from a middle-income country. Sci. Rep. 11, 2618 (2021).
Gorvitovskaia, A., Holmes, S. P. & Huse, S. M. Interpreting Prevotella and Bacteroides as biomarkers of diet and lifestyle. Microbiome 4, 15 (2016).
Knights, D. et al. Rethinking ‘enterotypes’. Cell Host Microbe 16, 433–437 (2014).
Zhang, R., Walker, A. R. & Datta, S. Unraveling city-specific signature and identifying sample origin locations for the data from CAMDA MetaSUB challenge. Biol. Direct 16, 1 (2021).
McDonald, D. et al. American Gut: an open platform for citizen science microbiome research. mSystems 3, e00031–18 (2018).
Levine, J. H. et al. Data-driven phenotypic dissection of aml reveals progenitor-like cells that correlate with prognosis. Cell 162, 184–197 (2015).
Costea, P. I. et al. Enterotypes in the landscape of gut microbial community composition. Nat. Microbiol. 3, 8–16 (2018).
He, Y. et al. Regional variation limits applications of healthy gut microbiome reference ranges and disease models. Nat. Med. 24, 1532–1535 (2018).
Luke, N. R. et al. Identification and characterization of a glycosyltransferase involved in Acinetobacter baumannii lipopolysaccharide core biosynthesis. Infect. Immun. 78, 2017–2023 (2010).
Freeman-Keller, M. et al. Nivolumab in resected and unresectable metastatic melanoma: characteristics of immune-related adverse events and association with outcomes. Clin. Cancer Res. 22, 886–894 (2016).
Weber, J. S. et al. Safety profile of nivolumab monotherapy: a pooled analysis of patients with advanced melanoma. J. Clin. Oncol. 35, 785–792 (2017).
Weber, J. S., Kahler, K. C. & Hauschild, A. Management of immune-related adverse events and kinetics of response with ipilimumab. J. Clin. Oncol. 30, 2691–2697 (2012).
Asnicar, F. et al. Microbiome connections with host metabolism and habitual diet from 1,098 deeply phenotyped individuals. Nat. Med. 27, 321–332 (2021).
Gevers, D. et al. The Human Microbiome Project: a community resource for the healthy human microbiome. PLoS Biol. 10, e1001377 (2012).
Human Microbiome Project Consortium. A framework for human microbiome research. Nature 486, 215–221 (2012).
Human Microbiome Project Consortium. Structure, function and diversity of the healthy human microbiome. Nature 486, 207–214 (2012).
Amir, A. et al. Correcting for microbial blooms in fecal samples during room-temperature shipping. mSystems 2, e00199–16 (2017).
Davar, D. et al. Phase Ib/II study of pembrolizumab and pegylated-interferon alfa-2b in advanced melanoma. J. Clin. Oncol. 36, JCO1800632 (2018).
Eisenhauer, E. A. et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur. J. Cancer 45, 228–247 (2009).
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).
Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359 (2012).
Li, D., Liu, C. M., Luo, R., Sadakane, K. & Lam, T. W. MEGAHIT: an ultra-fast single-node solution for large and complex metagenomics assembly via succinct de Bruijn graph. Bioinformatics 31, 1674–1676 (2015).
Li, D. et al. MEGAHIT v1.0: a fast and scalable metagenome assembler driven by advanced methodologies and community practices. Methods 102, 3–11 (2016).
Wood, D. E., Lu, J. & Langmead, B. Improved metagenomic analysis with Kraken 2. Genome Biol. 20, 257 (2019).
Seemann, T. Prokka: rapid prokaryotic genome annotation. Bioinformatics 30, 2068–2069 (2014).
Gu, Z., Eils, R. & Schlesner, M. Complex heat maps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32, 2847–2849 (2016).
Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 2498–2504 (2003).
Haratani, K. et al. Association of immune-related adverse events with nivolumab efficacy in non-small-cell lung cancer. JAMA Oncol. 4, 374–378 (2018).
Robert, C. et al. Long-term safety of pembrolizumab monotherapy and relationship with clinical outcome: a landmark analysis in patients with advanced melanoma. Eur. J. Cancer 144, 182–191 (2020).
Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8, 118–127 (2007).
Ho, T. K. Random decision forests. in Proceedings of the Third International Conference on Document Analysis and Recognition, vol. 1, 278–282 (IEEE Computer Society, 1995).
Boser, B. E., Guyon, I. M. & Vapnik, V. N. A training algorithm for optimal margin classifiers. in Proceedings of the Fifth Annual Workshop on Computational Learning Theory, 144–152 (Association for Computing Machinery, 1992).
McCullagh P. & Nelder, J. A. & Generalized Linear Models (Chapman and Hall, 1989).
We acknowledge all patients and families affected by metastatic melanoma. We acknowledge J. Wargo for providing unpublished metadata of the Houston melanoma patient cohort. This work was supported in part by the Intramural Research Program of the NIH, NCI, Center for Cancer Research. D.D. is supported by the Melanoma Breakthrough Foundation Breakthrough Consortium. H.M.Z. is supported by the NIH/NCI (R01 CA228181 AND R01 CA222203) and the James W. and Frances G. McGlothlin Chair in Melanoma Immunotherapy Research. M.V. was supported by an Irvington postdoctoral fellowship from the Cancer Research Institute. Work of the University of Pittsburgh Medical Center HCC Microbiome Shared Facility and Cytometry Facility is supported by the NIH NCI Comprehensive Cancer Center Support Core grant (P30 CA047904). This research was supported in part by the University of Pittsburgh Center for Research Computing and Unified Flow Cytometry Core of the University of Pittsburgh’s Department of Immunology through the resources provided.
D.D. reports the following disclosures: Arcus, Bristol-Myers Squibb, Checkmate Pharmaceuticals, CellSight Technologies, Merck, GlaxoSmithKline/Tesaro (research support); Array Biopharma, Checkmate Pharmaceuticals, Finch, Incyte, Immunocore, Merck; Shionogi (consulting); and Vedanta Biosciences (scientific advisory board). H.M.Z. reports the following disclosures: Bristol-Myers Squibb, Checkmate Pharmaceuticals, GlaxoSmithKline (research support); Bristol-Myers Squibb, Checkmate Pharmaceuticals, GlaxoSmithKline, Vedanta (consulting). D.D., H.M.Z., J.A.M., R.R.R., G.T. and A.K.D. are inventors on a patent application (US patent no. 63/208,719) submitted by the University of Pittsburgh that covers methods to enhance checkpoint blockade therapy by the microbiome. The other authors declare no competing interests.
Peer review information
Nature Medicine thanks R. Dummer, A. Bhatt and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Editor Recognition Statement: Javier Carmona and Saheli Sadanand are the primary editors on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended Data Fig. 1 Kaplan-Meier plots of progression-free survival and overall survival in the Pittsburgh early sample cohort and progression-free survival after dichotomization for abundance of select bacterial species.
a and b. Kaplan-Meier plots of probability of progression-free survival (PFS) (a) and overall survival (OS) (b) of PD-1-treated Pittsburgh early cohort melanoma patients. Vertical ticks show censored data. Central line is median OS or PFS probability, shaded area shows 95% confidence interval. c. Optimal cutpoints of bacterial abundance determined using Evaluate Cutpoints. Different plots show the effect on PFS of abundance (high vs. low) of the top four most significantly increased (left) and decreased (right) individual bacterial species in non-progressors at 10 months, determined using Mann-Whitney U test (Fig. 1c). Number of people at risk in in either group (high vs. low abundance) is shown below each panel. Vertical ticks show censored data. Hazard Ratio (HR) and score (logrank) test two-tailed p-value from Cox proportional hazards regression analysis.
Extended Data Fig. 2 Microbiota composition of non-progressing patients in the Pittsburgh cohort whose stool samples were collected 4–41 months after initiation of therapy is not predictive of late therapy failure but is enriched for similar bacterial taxa as observed in the initial microbiome of patients who did not progress at 10 months.
a. Plot of time of stool sample acquisition from 31 patients whose samples were collected after >4 months from therapy initiation. b. Progressor (P) and non-progressor (NP) groups identified at serial timepoints after late stool collection (top panel) were used to calculate the significance (two-tailed p-value) of compositional differences of the late-collected fecal microbiome using PERMANOVA (bottom panel). Fecal microbiota composition was determined using metagenomic sequencing. Progression during continued therapy was evaluated using RECIST v1.1 every 3 months or by clinical observation during follow-up visits. Number of patients on follow-up at each timepoint in relation to response status is shown in top panel. c. t-distributed uniform manifold approximation and projection (t-UMAP) plot depicts fecal microbiota compositional differences between early-collected patients who progressed (red) or did not progress (blue) in the first 10 months after initiation of therapy and late-collected long-term responders (green). Distance between centroids calculated as described in Fig. 1a, and significance (two-tailed p-value) of the differences by PERMANOVA are shown in lower table. d. Heatmap shows differentially abundant taxa (p < 0.05 and FC > 2) between the late Pittsburgh cohort compared with Ps (top) and NPs (bottom) at 10 months from the early Pittsburgh cohort. Columns denote patients grouped by each cohort before clustering; rows denote bacterial taxa enriched (black) or depleted (red) in early-sampled P versus late-sampled long-term NP clustered based on microbiota composition. Two-tailed p-values were calculated using two-tailed Mann-Whitney U test. e. ROC curve for manual model trained on the organisms associated with increased and decreased PFS in the Pittsburgh cohort from Supplementary Tables 4 and 5. Note that the model predicts late Pittsburgh samples well even though they were not included in the data used in training.
Extended Data Fig. 3 Entire time-to-event progression data analysis by Cox regression method of baseline fecal microbiome composition identifies additional favorable and unfavorable taxa linked with response to anti-PD-1 immunotherapy.
a. Volcano plot depicting bacteria identified by effect on progression-free survival (PFS) in the Pittsburgh early sample cohort using Cox regression analysis in Evaluate Cutpoints software. Taxa with q < 0.05 are shown as red dots. b. Cladogram visualization (favorable taxa – blue; unfavorable taxa – red) of bacterial taxa at different phylogenetic levels identified using approach described in (a).
Extended Data Fig. 4 Differential abundance analysis reveals relationship of baseline gut microbial taxa with high vs. low neutrophil-lymphocyte ratio in Pittsburgh early sample cohort.
a. t-distributed uniform manifold approximation and projection (t-UMAP) plots depicting fecal microbiota compositional differences between patient groups with high (≥3.82; orange) and low (<3.82; green) pre-treatment neutrophil-lymphocyte ratio (NLR). Optimal cutoff for NLR (3.82) was determined by time serial PERMANOVA as shown in Fig. 1a. Two-tailed p-value was calculated using PERMANOVA. b. Heatmap of differentially abundant taxa (p < 0.05 and FC > 2) in high-pre-treatment NLR (orange) and low-NLR (green) patients, using optimal cutoff (3.82). Columns denote patients grouped by NLR status and clustered within each group; rows denote bacterial taxa enriched (red) in patients with high NLR clustered based on microbiota composition; no bacterial taxa significantly enriched in the low-NLR patients were identified. Statistical significance was calculated using two tailed Mann‐Whitney U test. Bar plot to left of heatmap indicates extent of association between corresponding taxa and PFS probability [scaled hazard ratio (HR)] with Storey’s q-values <0.1 displayed within cells. Proportion of Gram-negative bacteria among those associated with high NLR was 58%, significantly higher than the average proportion of Gram-negative in patients’ fecal microbiome (28%, Chi-squared p = 0.0004).
Extended Data Fig. 5 Gut microbial gene differences discriminate between non-progressors and progressors during anti-PD-1 therapy in the Pittsburgh early sample cohort.
a. t-distributed uniform manifold approximation and projection (t-UMAP) plot depicting genetic differences of gut microbiomes between non-progressors (NPs; blue) and progressors (Ps; red) at time of maximal difference from start of therapy (10 months). Filled circles represent centroids, with connecting lines corresponding to samples from each group. Two-tailed p-value was calculated using PERMANOVA. b. Metagenomic shotgun sequencing of fecal microbiota samples identifies differentially abundant genes in Ps vs. NPs at 10 months from start of therapy. Heatmap shows differentially abundant genes identified by metagenomic shotgun sequencing (FDR < 0.2 and FC > 1.5). Columns denote patients grouped by progression status and clustered within P/NP groups; rows denote bacterial genes significantly upregulated (red) or downregulated (blue) in Ps versus NPs. c and d. Select genes involved in representative microbial processes of lipopolysaccharide (LPS) processing (c) and iron metabolism (d).
Extended Data Fig. 6 Metagenomic sequencing identifies distinct taxa associated with various immune-related adverse events in PD-1-treated melanoma patients in Pittsburgh early cohort.
Heatmap depicts metagenomic compositional differences between patients with a given immune-related adverse event (irAE) as compared to patients with other irAEs using scaled fold differences (high – red; low – blue) in abundances of specific bacteria. Values in individual cells represent unadjusted p-values calculated using two-tailed Mann-Whitney U test, with p-values <0.1 displayed within cells. Bar plot to left of heatmap indicates extent of association between corresponding taxa and progression-free survival probability [scaled hazard ratio (HR)], with Storey q-values <0.1 displayed within cells (from Supplementary Tables 4 and 5).
Extended Data Fig. 7 Reanalysis of four previously published individual cohorts using the same bioinformatic pipeline.
a. Analysis of α-diversity from five PD-1-treated melanoma patient cohorts (n = 185), including the Pittsburgh early sample cohort (n = 63), using either shotgun metagenomic (5 cohorts, red) or 16S rRNA gene amplicon (4 cohorts, black) sequencing. Details of each individual cohort are summarized in Supplementary Table 3. Forest plots depict α-diversity-based association tests including inverse Simpson, Shannon, and observed operational taxonomic units. Within each fixed-effect plot, names of each cohort are shown on a separate line, while log odds ratio of α-diversity (squares, size proportional to sample size used in meta-analysis) and associated 95% confidence intervals (bars) are shown, along with the dotted vertical line of no effect. The p-values reported for each cohort are two-tailed p values computed from the z statistic. To control for unobserved heterogeneity, we separately evaluated α-diversity using a random effects model on both pooled shotgun and 16S sequencing data from the 5 cohorts and performed I2 test for heterogeneity as shown. The p-value reported for heterogeneity is a one-tailed Cochran’s Q-test. b. t-distributed uniform manifold approximation and projection (t-UMAP) plot before (left) and after (right) correction for study-related batch effect using ComBat R package for all cohorts together including Pittsburgh cohort. P-values were calculated using PERMANOVA. c. t-UMAP plot of batch-corrected pooled metagenomic sequencing data from five separate cohorts of melanoma patients treated with anti-PD-1 therapy depicting fecal microbiota compositional differences with two-tailed p-value calculated using PERMANOVA between responders (Rs) and non-responders (NRs). d. Heatmap of differentially (p-values were calculated using non-parametric two-tailed Mann-Whitney U test) abundant gut microbiome taxa (p < 0.05, FC > 2) evaluated with shotgun sequencing in five melanoma patient cohorts, including Pittsburgh early sample cohort. Study-related batch effect was removed using ComBat R package. Response to therapy in published cohorts was determined as described in each study (Supplementary Table 3). Response to therapy in the Pittsburgh early sample cohort was defined as non-progression at 10 months after initiation of treatment. Columns represent patients grouped by clinical response and clustered within R/NR groups; rows depict bacterial taxa enriched (black) or depleted (red) in Rs versus NRs clustered based on gut microbiota composition.
Extended Data Fig. 8 Meta-analysis of all cohorts using random effects model identifies organisms differentially enriched in melanoma patients treated with anti-PD-1 therapy in separate cohorts by response status.
a. Random effect model meta-analysis of differentially abundant bacteria between responders (Rs) and non-responders (NRs) from five cohorts (n = 185) including Pittsburgh early sample cohort (n = 63) using shotgun metagenomic sequencing. All significant bacterial taxa enriched in Rs and NRs are shown. b. Forest plots depict association of representative bacterial species with response to anti-PD-1 therapy. Within each plot, names of various cohorts are shown on separate lines, while Hedge’s g (squares, standardized mean differences, size proportional to sample size) and associated 95% confidence intervals (bars) are shown, along with the dotted vertical line of no effect. To control for unobserved heterogeneity, we separately evaluated Hedge’s g using random effect model on metagenomic data and performed I2 test for heterogeneity as shown. P-values were calculated using random effect model.
Extended Data Fig. 9 Expression levels of selected taxa in different American Gut Project enteric microbiotypes.
t-distributed stochastic neighbor embedding (t-SNE) plots depicting American Gut Project (AGP) dataset with visualization of abundances of select taxa (blue – low; red – high).
a. t-distributed stochastic neighbor embedding (t-SNE) plots depicting mapping of individual melanoma patient cohorts to American Gut Project (AGP) dataset revealed distinct compositional differences between them. Each cohort is represented by a different color that is maintained in the overlay. Colors in the overlay are semi-translucent and were stratified starting from the Pittsburgh to New York cohort. Gut microbiota compositions of the different cohorts were significantly different (PERMANOVA, two-tailed p < 0.001). b. and c. Heatmaps represent scaled abundances of each enteric microbiotypes across 28 states from which AGP data were available on the left, with the four states from which anti-PD-1-treated melanoma cohorts originated separately on the right using three individuals per county per cluster as a cutoff. Data were scaled by number of individuals per state and per cluster and are depicted in relation to the 28 states that met the cutoff and in relation to the four states from which the four studied cohorts originated. b. Heatmaps are (b) scaled only by number of samples from state (to reflect local abundance of microbiotypes) or (c) scaled by both number of samples per state and by number of samples per microbiotype (to reflect distribution of different microbiotypes across the US). d. Geographic representation in the US of four representative enteric microbiotypes, with most uneven distribution between four states (right panel in c).
About this article
Cite this article
McCulloch, J.A., Davar, D., Rodrigues, R.R. et al. Intestinal microbiota signatures of clinical response and immune-related adverse events in melanoma patients treated with anti-PD-1. Nat Med 28, 545–556 (2022). https://doi.org/10.1038/s41591-022-01698-2
This article is cited by
Potential non-invasive biomarkers in tumor immune checkpoint inhibitor therapy: response and prognosis prediction
Biomarker Research (2023)
Molecular Cancer (2023)
Nature Medicine (2023)
Nature Communications (2023)
Nature Reviews Clinical Oncology (2023)