Gut Microbiota Promotes Tumor Growth in Mice by Modulating Immune Response

We studied the effects of gut microbiome depletion by oral antibiotics on tumor growth in subcutaneous and liver metastases models of pancreatic cancer, colon cancer, and melanoma. Gut microbiome depletion significantly reduced tumor burden in all the models tested. However, depletion of gut microbiome did not reduce tumor growth in Rag1-knockout mice, which lack mature T and B cells. Flow cytometry analyses demonstrated that gut microbiome depletion led to significant increase in interferon gamma–producing T cells with corresponding decrease in interleukin 17A and interleukin 10–producing T cells. Our results suggest that gut microbiome modulation could emerge as a novel immunotherapeutic strategy. We studied the effects of gut microbiome depletion by oral antibiotics on tumor growth in subcutaneous and liver metastases models of pancreatic cancer, colon cancer, and melanoma. Gut microbiome depletion significantly reduced tumor burden in all the models tested. However, depletion of gut microbiome did not reduce tumor growth in Rag1-knockout mice, which lack mature T and B cells. Flow cytometry analyses demonstrated that gut microbiome depletion led to significant increase in interferon gamma–producing T cells with corresponding decrease in interleukin 17A and interleukin 10–producing T cells. Our results suggest that gut microbiome modulation could emerge as a novel immunotherapeutic strategy. What You Need to KnowBackground and ContextGut microbiome has been implicated in the etiopathology of various disease-states like colitis, metabolic syndrome, ischemic stroke etc. but its role in cancer modulation is obscure.New FindingsDepletion of gut microbiome in mice using oral antibiotics attenuated cancer and metastases burden in multiple models, and activated an antineoplastic immune phenotype in the tumor microenvironment.LimitationsIt is unclear if dysbiosis in general or some specific gut microbe is responsible for the effects observed.ImpactThis study suggests manipulation of the gut microbiome may be an anti-cancer therapeutic strategy. Gut microbiome has been implicated in the etiopathology of various disease-states like colitis, metabolic syndrome, ischemic stroke etc. but its role in cancer modulation is obscure. Depletion of gut microbiome in mice using oral antibiotics attenuated cancer and metastases burden in multiple models, and activated an antineoplastic immune phenotype in the tumor microenvironment. It is unclear if dysbiosis in general or some specific gut microbe is responsible for the effects observed. This study suggests manipulation of the gut microbiome may be an anti-cancer therapeutic strategy. There are more resident microbes in the human body than there are “human” cells, and most of these microbes occupy an ambiguous niche in the gut. The gut microbiota, forming a unique metagenome, is dynamic and changes with a person’s nutrition state, geography, and even age. A growing body of evidence hints toward a co-evolved relationship between gut microbes and our immune system.1Hooper L.V. et al.Science. 2012; 336: 1268-1273Crossref PubMed Scopus (2687) Google Scholar In fact, some inflammatory diseases, like colitis, are characterized by a transition in the gut microbiome, which changes from a “eubiotic” to a “dysbiotic” state, with interesting therapeutic implications.2Sartor R.B. Gastroenterology. 2004; 126: 1620-1633Abstract Full Text Full Text PDF PubMed Scopus (872) Google Scholar Although several epidemiological studies associate dysbiosis with cancer, the exact role of gut bacteria in the pathogenesis of cancer is still unclear. We evaluated the impact of gut microbiome depletion on tumor growth in multiple mouse models. Gut microbiome was depleted in age and sex-matched C57BL/6J mice with a broad-spectrum cocktail of oral antibiotics (vancomycin, neomycin, metronidazole, ampicillin and amphotericin B) using a well-established protocol3Reikvam D.H. Erofeev A. et al.PLoS One. 2011; 6: e17996Crossref PubMed Scopus (115) Google Scholar (Figure 1A). Mice, with or without gut microbiome depletion, were used to establish cancer models by subcutaneous injection of KPC pancreatic cancer cells derived from tumors forming in KrasG12D/+; Trp53R172H/+; Pdx-1cre mice4Hingorani S.R. et al.Cancer Cell. 2005; 7: 469-483Abstract Full Text Full Text PDF PubMed Scopus (1689) Google Scholar; or melanoma cells derived from tumors forming in Tyr-CreER; BrafV600E/+; Ptenfl/fl mice,5Dankort D. et al.Nat Genet. 2009; 41: 544-552Crossref PubMed Scopus (857) Google Scholar and by splenic injection of KPC cells; B16-F10 melanoma cells; or MC38 colon cancer cells to induce liver metastases. Our results show that gut microbiome depletion led to a significant decrease in subcutaneous tumor burden in pancreatic cancer and melanoma models (Figure 1B and C). There was also a significant decrease in liver metastases burden in pancreatic cancer, colon cancer, and melanoma models (Figures 1D and E, and Supplementary Figure 1A). Interestingly, the tumor-suppressing effect of gut microbiome depletion was abolished when the subcutaneous experiments were carried out in Rag1 knockout mice lacking mature T (and B) lymphocytes (Figures 2A; Supplementary Figure 2A). This suggests that the tumor-decreasing effect of antibiotics was not an off-target cytotoxic action on cancer cells, but required active participation of adaptive immunity. We next evaluated the impact of gut microbiome depletion on the balance between pro- and anti-tumor T cells in tumor microenvironment (TME). It is known that naïve helper T cells (Th0), typically, mature into Th1, Th2, regulatory T cell, or Th17 lineage. The classical Th1 cytokine interferon (IFN) gamma plays an anti-tumorigenic role in TME, whereas the Th2/regulatory T cell cytokines interleukin (IL) 4, IL5, and IL10 mediate a pro-tumorigenic role. As would be expected, a high Th1/Th2 ratio in TME correlates with improved survival in pancreatic cancer.6De Monte L. et al.The Journal of Experimental Medicine. 2011; 208: 469-478Crossref PubMed Scopus (486) Google Scholar Moreover, Th17 cells are known to be pro-tumorigenic in pancreatic cancer,7McAllister F. et al.Cancer Cell. 2014; 25: 621-637Abstract Full Text Full Text PDF PubMed Scopus (265) Google Scholar melanoma,8Wang L. et al.J Exp Med. 2009; 206: 1457-1464Crossref PubMed Scopus (628) Google Scholar and colorectal cancer.9Grivennikov S.I. et al.Nature. 2012; 491: 254-258Crossref PubMed Scopus (916) Google Scholar IL17a is also intricately linked to the gut microbiome10Ivanov I.I. et al.Cell Host Microbe. 2008; 4: 337-349Abstract Full Text Full Text PDF PubMed Scopus (1274) Google Scholar and plays a key role in defending against fungal and bacterial pathogens. Our results indicate that gut microbiome depletion caused a significant increase in Th1(IFN gamma+CD4+CD3+) and Tc1(IFN gamma+CD8+CD3+) cells in TME (Figure 2B). Furthermore, gut microbiome depletion caused a significant increase in numbers of anti-tumor IFN-gamma–secreting T cells (IFN gamma+CD3+), with a corresponding decrease in numbers of pro-tumor IL17a(IL17a+CD3+) and IL10(IL10+CD4+CD3+) secreting immune populations (Figure 2C–F). Tumor-attenuating effect of antibiotics was abrogated in mice that were treated in vivo with IL17a neutralizing monoclonal antibody (Figure 2G), thereby confirming the essential role of IL17a in mediating this phenomenon. Analysis of stool samples from subcutaneous KPC-bearing mice, given antibiotics, revealed an expected ablation of 16S ribosomal DNA and decrease in relative abundance of the 2 phyla majorly found in mouse (and human) gut: Bacteroidetes and Firmicutes (Supplementary Figure 3A and E). Antibiotics also caused a significant decrease in α-diversity, a significant change in β-diversity, a reversed Bacteroidales to Clostridiales abundance ratio and colonization of gut by otherwise scarce (and likely antibiotic-resistant) Proteobacteria (mainly Alcaligenaceae and Enterobacteriaceae) and Tenericutes (mainly Mycoplasmataceae) (Supplementary Figure 3B–F). We also observed presence of 16S ribosomal DNA belonging to diverse microbial taxa in metastatic livers (Supplementary Figure 4A–E). The mechanism by which gut microbiome interacts with immune system and affects cancer progression is unclear, but some inferences can be drawn from literature. Bacterial products recognized by Toll-like receptors have been previously known to activate the IL23/IL17 axis and promote colon cancer development.9Grivennikov S.I. et al.Nature. 2012; 491: 254-258Crossref PubMed Scopus (916) Google Scholar Thus, it is possible that gut microbes interact with immune system via pattern recognition receptors in pancreatic and other cancers too. The exact cell type that participates in this interaction and the potential site of this interaction (gut vs intra-tumoral) will be deciphered in future studies. While the goal of using antibiotic cocktail was to deplete gut microbiome, Metronidazole has appreciable oral bioavailability and thus some systemic effects cannot be ruled out. However, the same antibiotic cocktail failed to reduce tumor size in Rag1 knockout mice or mice treated with IL17a-neutralizing antibody, thereby, suggesting that a direct cytotoxic effect of antibiotics is not responsible for mediating the anti-tumor phenomenon. Additionally, our studies suggest that depleting the gut microbiome leads to infiltration of pancreatic tumors with effector T cells. Conventional immunotherapeutic drugs like the modern checkpoint inhibitors have failed to show significant efficacy against pancreatic cancer, in part due to minimal effector T cell infiltration in this cancer. Inducing T cell immunity has been previously shown to overcome pancreatic cancer’s resistance to immune checkpoint inhibitors11Winograd R. et al.Cancer Immunol Res. 2015; 3: 399-411Crossref PubMed Scopus (304) Google Scholar and, therefore, future studies should evaluate whether a gut microbial modulation strategy can potentiate the efficacy of checkpoint inhibitors or cytotoxic drugs in pancreatic and other cancers with minimal adverse effects. In summary, our studies suggest that the gut microbiome modulates tumor progression and that manipulation of gut bacteria could emerge as a novel immunotherapeutic strategy, either alone or in combination with conventional immunotherapy. The authors thank the University of Miami’s Sylvester Comprehensive Cancer Center’s IVIS Facility and Flow Cytometry Core Facility as well as the University of Minnesota’s Genomic Core for valuable assistance in data acquiring. We also thank Ms Patricia L. Guevara, BS, MA, and Ms Sydney Paige Kopen, BSB, for important assistance during flow cytometry sample preparation and data acquiring. We also thank Yuguang James Ban, PhD and Xi Steven Chen, PhD from the Department of Public Health Sciences at Sylvester Comprehensive Cancer Center, University of Miami for assistance in statistical and microbiome analysis. Author contributions: Conception and design: Vrishketan Sethi, Ashok Saluja, and Vikas Dudeja. Acquisition and analyses of data: Vrishketan Sethi, Saba Kurtom, Mohammad Tarique, Shweta Lavania, Zoe Malchiodi, Leonor Hellmund, Li Zhang, Umakant Sharma, Bhuwan Giri, Bharti Garg, Anthony Ferantella, Sabita Roy, Sundaram Ramakrishnan, Selwyn Vickers, Ashok Saluja, and Vikas Dudeja. Microbiome data analyses: Vrishketan Sethi, Umakant Sharma, and Li Zhang. Writing, review and/or revision of the manuscript: Vrishketan Sethi, Saba Kurtom, Bhuwan Giri, Shweta Lavania, Rajinder Dawra, Selwyn Vickers, Sundaram Ramakrishnan, Sabita Roy, Ashok Saluja, and Vikas Dudeja. KPC cells were derived from spontaneous pancreatic tumors originating in KrasG12D/+; Trp53R172H/+; Pdx-1cre mice after plating tumor fragments in Dulbecco’s modified Eagle medium/Ham’s F12 (50:50) medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. Outgrowing cancer cells were then expanded and subsequently depleted of fibroblasts by magnetic selection and differential trypsinization. Cell lines from passages 5–15 were used for all experiments. Braf-Pten melanoma cells were derived from topical tamoxifen-induced tumors arising in Tyr-CreER; BrafV600E/+; Ptenfl/fl mice bred in C57BL/6 background. Briefly, under sterile conditions, tumors were collected and minced in cold phosphate-buffered saline. Tumor chunks were xenografted into Rag1 KO mice and allowed to grow. The tumor that formed was harvested, digested mechanically and enzymatically, and its single-cell suspension was plated in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. Fibroblasts were depleted magnetically following differential trypsinization over several passages. B16-F10 melanoma cells were a kind gift from Prof Sundaram Ramakrishnan (University of Miami, Miami, FL). RFP tagged MC38 colon cancer cells were kindly gifted by Dr Jaime R. Merchan (University of Miami, Miami, FL) All animal experiments were performed in compliance with protocols approved by Institutional Animal Care and Use Committee at the University of Miami, Miami, FL. Only 6- to 8-week old mice of either sex were used in the experiments. Wild-type C57BL/6J and Rag1 knockout mice (B6.129S7-Rag1tm1Mom/J) were purchased from The Jackson Laboratory (Bar Harbor, ME). KPC (KrasG12D/+; Trp53R172H/+; Pdx-1cre mice) were bred in our animal facilities by crossing KrasG12D/+, Trp53R172H/+, and Pdx-1cre mice. Tyr-CreER, BrafV600E/+, Ptenfl/fl mice were a kind gift of Prof Eli Gilboa (University of Miami, Miami, FL). These were housed in the same facility under specific pathogen-free conditions in a 12-hour day/night cycle and fed Teklad irradiated global 18% protein rodent diet (Envigo, Madison, WI). Gut microbiome was depleted as described previously.1Reikvam D.H. Erofeev A. et al.PLoS One. 2011; 6: e17996Crossref PubMed Scopus (325) Google Scholar For subcutaneous experiments, 1 × 105 Braf-Pten melanoma cells or 2 × 105 KPC cells were suspended in 100% growth factor reduced Matrigel (Corning, Corning, NY) and injected subcutaneously on the right flank. For intrasplenic injections, 1 × 106 KPC cells, 2.5 × 105 B16-F10 melanoma cells, and 5 × 105 MC38-RFP cells were injected in the spleen as described previously.2Soares K.C. et al.J Vis Exp. 2014; : e51677PubMed Google Scholar MC38-RFP injected animals were imaged in an in vivo imaging system facility. Weight of the liver metastases was calculated as the difference between experimental metastatic liver weight and mean liver weight of same-aged cancer-naïve mice given saline or antibiotics. Subcutaneous tumor volumes were measured serially and tumor volume was calculated by modified ellipsoid formula (0.5 × length × width2). For IL17a-depletion experiment, 500 μg antibody clone 17F3 or its isotype (both sourced from BioXcell, West Lebanon, NH) were injected intraperitoneally twice a week beginning 1 day before tumor inoculation. Tumors harvested from euthanized mice at end point were enzymatically digested to a single- cell suspension, washed in 1% bovine serum albumin and stained for various surface markers (1 × 106 cells with manufacturer-suggested concentration of fluorochrome-conjugated antibody). Subsequently, they were fixed and permeabilized with BD Cytofix/Cytoperm (BD Biosciences, San Jose, CA) and stained with intracellular antibodies. The following clones of antibodies were used (all from Biolegend, San Diego, CA): CD3 (17A2), CD4(RM4-5), CD8(53-6.7), IFN gamma (XMG1.2), IL17a (TC11-18H10.1), and IL10 (JES5-16E3). Flow cytometry data were acquired on BD LSRFortessa cytometer (BD Biosciences) and analyzed with FlowJo software (FlowJo, Ashland, OR). Metastatic livers or stool samples from distal rectum were aseptically collected immediately after euthanizing mice. All samples were snap frozen in liquid nitrogen vapor and then equal amounts were used for further DNA collection using PowerSoil DNA Isolation Kit (Mo Bio Labs, Carlsbad, CA) according to the manufacturer’s instructions. For polymerase chain reaction (PCR) and gel electrophoresis detection of 16S ribosomal DNA in stool samples, following primers were used for PCR amplification of 100 ng stool DNA run for 40 cycles: forward (5′→3′) AGAGTTTGATCCTGGCTCAG; reverse: (5′→3′) GGTTACCTTGTTACGACTT. For reverse transcriptase PCR of liver samples, V6 region of bacterial 16S ribosomal DNA was amplified and for mouse genomic DNA, pIgR mouse genomic region (internal control) was amplified as has been described previously.1Reikvam D.H. Erofeev A. et al.PLoS One. 2011; 6: e17996Crossref PubMed Scopus (325) Google Scholar, 3Andersson A.F. et al.PLoS One. 2008; 3: e2836Crossref PubMed Scopus (761) Google Scholar For amplification of V6 region of 16S ribosomal DNA, primers used were forward (5′→3′) AGGATTAGATACCCTGGTA and reverse (5′→3′) CRRCACGAGCTGACGAC. For amplification of mouse genomic DNA (internal control), mouse pIgR genomic region was amplified by primers used before1Reikvam D.H. Erofeev A. et al.PLoS One. 2011; 6: e17996Crossref PubMed Scopus (325) Google Scholar: forward (5′→3′) TTTGCTCCTGGGCCTCCAAGTT and reverse (5′→3′) AGCCCGTGACTGCCACAAATCA. PCR reactions were carried out in triplicates with 100 ng of liver DNA, SYBR Green I Real-Time PCR Master Mix (ThermoFisher Scientific, Waltham, MA), a 900-second activation step (95°C); 40 cycles of 30-second denaturation step (95°C) + 30-second annealing step (55°C) + 30-second extension step (72°C) on a Roche LightCycler 480 II machine (Roche Diagnostics Corporation, Indianapolis, IN). 16S ribosomal RNA gene sequencing from stool samples and livers was done at the University of Minnesota Genomic Core as described previously.4Gohl D.M. et al.Nat Biotechnol. 2016; 34: 942-949Crossref PubMed Scopus (382) Google Scholar Processing summary for V5V6 region amplification was provided by the Core and can be accessed at: https://drive.google.com/file/d/11sxdADoO_wWWm-iZ5-L1qwnKsdjhuFrS/view?usp=sharing. For bioinformatics analyses, sequencing adapter sequences were trimmed from the 3' ends of reads using Trimmomatic. PandaSeq was used to removed primer sequences from the beginning of reads and to stitch the overlapping paired reads together. Reads without primer sequences and reads that could not be stitched together were discarded. Stitched reads whose length were outside the expected length of the targeted variable region were discarded. Chimeric sequences were identified and removed using the identify_chimeric_seqs Qiime function with the usearch61 method. An operational taxonomic unit table was constructed using the Qiime open-reference operational taxonomic unit picking workflow using the greengenes reference database. Operational taxonomic unit table summary plots, α-diversity, β-diversity metrics, and Weighted UniFrac distance comparisons were generated using Qiime workflows. Unpaired Student t test with Welch’s correction, Mann-Whitney test, and 1-way analysis of variance with Tukey’s multiple comparison test were used for comparisons as described in corresponding figure legends. All statistical analyses were computed through Prism 7 for Mac OS X, version 7.0c (Graphpad Software, La Jolla, CA). Author names in bold designate shared co-first authorship.Supplementary Figure 2C57BL/6J mice carrying a Rag1tm1Mom mutation were either given oral antibiotics or saline and were subcutaneously implanted with Braf-Pten melanoma cells and tumor progression was serially followed (n = 10 per group). X-axis label in tumor kinetics represents days after tumor injection (unpaired Student t test with Welch’s correction was used for comparisons. Data are shown as mean ± SEM; ns, non-significant.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Figure 3Oral antibiotics lead to widespread gut microbial changes. (A–F) Saline and antibiotics-gavaged mice were implanted with KPC cancer cells subcutaneously and stool samples were collected from the distal rectum after euthanasia. These samples were then probed for bacterial 16S ribosomal DNA (rDNA) through polymerase chain reaction and gel electrophoresis (representative samples shown in [A] and also passed through a standardized 16S ribosomal RNA gene sequencing pipeline. For simplicity, only families with >1% relative abundance are depicted in [E]). (For 16S rRNA gene sequencing analysis, n = 8 for saline and n = 6 for antibiotics). A 1-way ANOVA with Tukey’s multiple comparison test was used for UniFrac distance comparisons in (D). Unpaired Student t test with Welch’s correction was used for all other comparisons. Data are shown as mean ± SEM. ∗P < .05; ∗∗∗∗∗P < .0005; ∗∗∗∗∗∗P < .0001. OTU, operational taxonomic unit.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Figure 4Oral antibiotics significantly change metastatic liver microbiome (A–E) saline and antibiotics-gavaged mice were given KPC hepatic metastases and metastatic livers were collected at end point. Livers were probed for relative abundance of 16S ribosomal DNA (rDNA)/mouse genomic DNA through reverse transcriptase PCR and also passed through a 16S ribosomal RNA gene sequencing pipeline. For simplicity, only families with >1% relative abundance are depicted in (E). (n = 8 for saline and n = 7 for antibiotics). A 1-way analysis of variance with Tukey’s multiple comparison test was used for UniFrac distance comparisons in (D). Unpaired Student t test with Welch’s correction was used for all other comparisons. Data are shown as mean ± SEM. ∗∗P < .01; ∗∗∗P < .005; ∗∗∗∗∗∗P < .0001. OTU, operational taxonomic unit.View Large Image Figure ViewerDownload Hi-res image Download (PPT) Innate Immune Cells Regulate Oncoimmunity and Cancer DevelopmentGastroenterologyVol. 155Issue 6PreviewWe read with great interest the recently published article titled “Gut microbiota promotes tumor growth in mice by modulating immune response.”1 Sethi et al1 reported that gut microbiome depletion by a combination of multiple antibiotics suppressed tumor growth and metastasis in wild-type mice, but not in RAG1-/- mice, the T-cell and B-cell deficient strain. They further observed significant increase of interferon (IFN)-γ–producing T cells with a decrease in IL-17–producing T cells in wild-type mice after gut microbiome depletion. Full-Text PDF