Disruption of the Gut Microbiota Confers Cisplatin Resistance in Epithelial Ovarian Cancer

Epithelial ovarian, peritoneal, and fallopian tube carcinomas (EOC) are a leading cause of gynecologic cancer–related death (1). Following EOC diagnosis, patients are treated with a combination of cytoreductive surgery and platinum-taxane–based chemotherapy (2, 3). Despite many women achieving remission with first-line therapy, up to 80% of patients will recur and require additional treatment (2, 3). After disease recurrence, the interval from last treatment with platinum chemotherapy has important therapeutic and prognostic implications. Patients with platinum-resistant EOC have fewer treatment options, reduced response rate to chemotherapy, and poor prognosis compared with those with platinum-sensitive disease (4). There is a significant unmet need to further understand the mechanisms of platinum resistance in EOC and to advance current therapeutic options for these patients.

The gut microbiome has many roles in maintenance of human health, and has been increasingly linked with many disease states, including cancer (5–8). Recent evidence suggests that the gut microbiome may modulate responses to cancer treatments, including traditional chemotherapy and immunotherapy (5–9). The impact of oxaliplatin treatment was attenuated in mice treated with antibiotics showing decreased tumor regression and worse survival, compared with nonantibiotic control animals.

Antibiotic therapy is frequently used during cancer treatments for both prophylaxis and treatment of infections. Studies have demonstrated that receipt of antibiotics during both systemic chemotherapy and immunotherapy negatively impacts oncologic outcomes. In a study of patients with relapsed lymphoma and leukemia receiving either cisplatin or cyclophosphamide on two clinical trials, those that received antibiotics against gram-positive bacteria during platinum chemotherapy had a significant reduction in overall treatment response, time to recurrence, and overall survival. Among patients with EOC, studies have demonstrated that treatment for surgical-site infection following primary cytoreductive surgery is associated with worse survival (10). In a recent retrospective analysis, patients with EOC who received antibiotics, primarily against gram-positive bacteria, during primary platinum chemotherapy had reduced overall (45.6 vs. 62.4 months) and progression-free survival (17.4 vs. 23.1 months) compared with patients who received no antibiotic intervention (11). These studies suggest that an intact gut microbiome provides a protective microenvironment, and disruption leads to accelerated tumor growth and chemotherapy resistance, including platinum agents. Here, we investigated the impact of antibiotic-mediated disruption of the gut microbiome on EOC tumor growth and platinum chemotherapy response in preclinical models of EOC.

ID8 (RRID:CVCL_IU16), ID8-VEGF (C57Bl/6 syngeneic), and OV81 (human) EOC cell lines were cultured in DMEM media containing 5% heat-inactivated FBS (Atlas Biologicals, catalog no. F-0500-D, Lot F31E18D1) and grown under standard conditions. ID8 and ID8-VEGF are mouse ovarian surface epithelial cancer cells and were a gift from Dr. Vince Tuohy at the Cleveland Clinic Lerner Research Institute (Cleveland, OH) and were obtained from the University of Kansas. Cells were authenticated in the Tuohy lab. OV81 was a gift from Dr. Analisa DiFeo at the University of Michigan (Ann Arbor, MI). OV81 is a platinum-sensitive high-grade serous ovarian cancer derived from ascites and established as a patient-derived xenograft. OV81 has a p53 mutation (R121S) and does not harbor a BRCA1/2 mutation. Mycoplasma testing on cells was performed every 6 months with latest test in January 2022. HEK293T/17 (ATCC CRL-11268, RRID:CVCL_1926) cells were plated and cotransfected with Lipofectamine 3000 (L3000015 Invitrogen); 3rd generation packaging vectors pRSV-REV #12253 (RRID:Addgene_12253), pMDG.2 #12259, and pMDLg/pRRE #12251 (RRID:Addgene_12251); and lentiviral vector directing expression of luciferase reporter pHIV-Luciferase #21375 4.5 μg (Addgene). Viral particles were harvested, filtered through a 0.45-μm Durapore polyvinylidene difluoride Membrane (Millipore SE1M003M00) and added to cell line culture media. Viral infections were carried out over 72 hours and transduced cells were selected by their resistance to 2-μg/mL puromycin (MP Biomedicals 0219453910).

Female C57BL/6J (BL/6; RRID:IMSR_JAX:000664) mice were purchased from Jackson Laboratories (Bar Harbor, ME) at 6 weeks of age. Female NOD.Cg-Prkdc<scid>Il2rg<tm1Wjl>SzJ (NSG) mice were purchased from the Cleveland Clinic Biological Resources Unit at 6 weeks of age. Experimental animals were housed and handled in accordance with Cleveland Clinic Lerner Research Institute Institutional Animal Care and Use Committee (IACUC) guidelines under an approved protocol. Mice (BL/6 or NSG) were given either control or antibiotic (0.5 g/L vancomycin, 1 g/L neomycin sulfate, 1 g/L metronidazole, 1 g/L ampicillin; ABX; Fisher Scientific) containing water for 2 weeks prior to intraperitoneal injection of ID8-LUC (5×10⁶) ID8-VEGF (5×10⁶) or OV81 (5×10⁶) cells. The ABX containing water has been previously shown to be sufficient to deplete all detectable commensal bacteria (12, 13). Mice remained on either control or ABX containing water for the duration of the study. Two weeks after cancer cell injection, mice were treated intraperitoneally with either cisplatin (0.5 mg/kg, Spectrum Chemical) or vehicle (PBS) twice weekly until humane endpoint criteria of either total tumor burden in excess of 150 mm³ or debilitating ascites development was reached.

Following 2 weeks of ABX or control water treatment of mice, necropsies were performed and ceca were harvested and transported to a Coy Anaerobic Chamber for further processing. Ceca were dispersed with anaerobic PBS by vortexing for 10 minutes and left to sit for 10 minutes to allow debris to settle. The top portion was aliquoted into 1 mL crimp-top tubes and stored at −80°C until used. ABX cecal microbial transplant (CMT) mice received 1 week of ABX treatment as described above, followed by cessation of ABX and 2 doses of CMT material from ABX-treated mice through oral gavage, and then resumption of ABX water treatment. H₂O CMT mice were also treated for 1 week with ABX and similarly received 2 doses of CMT material from H₂O-treated mice through oral gavage. All mice received an intraperitoneal injection of 5×10⁶ ID8-LUC cells on day 27 of the study and monitored for tumor progression via In Vivo Imaging System (IVIS) imaging as described below as well as predetermined endpoint criteria of debilitating ascites development or a body composition score (BCS) < 2.

While gently restraining a mouse by holding the base of its tail, the observer (blinded to treatment group) used the thumb and index finger of the other hand to palpate the degree of muscle and fat over the sacroiliac region. A score from 1 to 5 was given to each mouse weekly following intraperitoneal tumor cell injection based on previous literature (14). On the basis of established IACUC protocols #2018–2003, a BCS score of 2 or lower was defined as meeting endpoint criteria.

For the ID8-LUC and ID8-VEGF cohorts, transabdominal ultrasound (TAUS) surveillance was initiated 7 days following intraperitoneal tumor injection as previously described (15). TAUS was performed weekly until study endpoint. Mice were anesthetized using isoflurane (DRE Veterinary) and placed in the supine position. Following the removal of abdominal hair using Nair (Church & Dwight Co. Inc.), sterile ultrasound gel was applied to the abdomen. TAUS was performed using Vevo2100 (VisualSonics) using the abdominal imaging package and MS550D probe (40 Hz). For each mouse, the abdomen was assessed for tumor in four quadrants. Tumors were noted to be absent or present at each assessment. Tumor length and width were recorded, and tumor volume was calculated using the formula: (Length×Width²)/2.

For the OV81-LUC and CMT study ID8-LUC cohort of mice, bioluminescence images were taken with IVIS Lumina (PerkinElmer) using D-luciferin as previously described (16). Mice were placed under inhaled isoflurane anesthetic and administered an intraperitoneal injection of D-luciferin (Goldbio LUCK-1G, 150 mg/kg in 150 μL). Images were analyzed (Living Image Software) and total flux reported in photons/second corrected for baseline measurement at implantation for each mouse abdomen. All images were obtained using the automatic exposure feature and mice were imaged individually.

Mouse tumor tissue was harvested at necropsy, fixed in formaldehyde, and embedded in the paraffin. Paraffin blocks were sectioned at 5-μm thickness, deparaffinized, and cleared in Histo-Clear. Cleared sections were rehydrated in graded alcohols (100%, 95%, 80%, 60%, and 50%) and water 3 times followed by antigen retrieval in Tris EDTA (pH 9) in a 60C water bath. Slides were cooled for 40 minutes and incubated in 0.1% triton X-100 for 10 minutes. Slides were washed in 1X TBS at pH 8.0 and incubated in 3% H₂O₂ solution to inhibit endogenous peroxidase activity. Sections were blocked in 5% goat serum at room temperature for 1 hour. Primary antibodies (CD31 and 53BP1) were added to the sections and incubated overnight at 4°C. Slides were washed 3 times in 1X TBS Tween (TBST). Fluorescent conjugated secondary antibody (Alexa Fluor 488) were added to the slides and incubated for 1 hour at room temperature then washed three times in 1X TBST. Slides were mounted with VECTASHIELD antifade mounting medium with 4',6-diamidino-2-phenylindole (DAPI). Images were captured in confocal microscope at 63× magnification.

Tissue slides were dewaxed and rehydrated as above. Slides were stained in Mayers Hematoxylin for 1 minute, washed in tap water 4 times, and were incubated in 1X TBS for 5 minutes. Slides were counter stained with eosin for 1 minute and washed 3 times in tap water. Slides were then dehydrated in graded alcohol 60%, 80%, 90%, and 100% ethanol baths for 5 minutes each and finally rinsed in Histo-Clear for 5 minutes. Slides were mounted in Cytoseal with cover glass. Bright field images were captured at ×20 magnification.

To detect apoptotic DNA fragmentation, terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay was performed. Tissue sections were prepared and dewaxed in Histo-Clear, dehydrated, and antigens processed as above. TUNEL enzyme solution and label solution were mixed according to the manufacturer's instruction (In Situ Cell Death Detection Kit, Sigma). Reaction mix was added to the slides, which were incubated for 60 minutes at 37°C and washed three times with 1X TBST. Slides were mounted with VECTASHIELD antifade mounting medium with DAPI. Images were captured in confocal microscope at 63x magnification.

Total mRNA was isolated using commercially available RNA isolation kit (NucleoSpin RNA plus, Takara). RNA samples were measured by NanoDrop to check quality and concentration. First strand cDNA was synthesized from RNA samples using a cDNA synthesis kit (Primescript Takara). Primers for selected target genes (PAX2, DLX5, MSX1, NANOG, Sox2, OCT4, EPCAM, and GAPDH) were designed with Primer-BLAST (Supplementary Table S1). Real-time PCR was carried out using these primers and SYBR Green qPCR Master Mix (Applied Biosystems). Relative gene expression was calculated as 2^−ΔΔCT and plotted in GraphPad (RRID:SCR_002798) Software.

At 5 predetermined time points (baseline, 2 weeks, cisplatin initiation, 5 weeks, and endpoint as in Fig. 1A schematic), stool was collected from each mouse and frozen in microcentrifuge tubes at −80°C. Upon completion of the study, 16S rDNA was isolated following the standard protocol and (QIAamp PowerFecalPro Kit; Qiagen). Isolated samples were sent to Miami University of Ohio on dry ice for 16S V4 rDNA processing. Data analysis was completed by the Microbiome and Composition Analytics Core facility. As previously described by Weiss and colleagues (17), individual FASTQ files without non-biological nucleotides were processed using the Divisive Amplicon Denoising Algorithm (DADA) pipeline (18). The output of the dada2 pipeline [feature table of amplicon sequence variants (an ASV table)] was processed for alpha and beta diversity analysis using the phyloseq and microbiomeSeq (http://www.github.com/umerijaz/microbiomeSeq) packages in R. Alpha diversity estimates were measured within group categories using the estimate_richness function in phyloseq. Multidimensional scaling [also known as principal coordinate analysis (PCoA)] was performed using Bray–Curtis dissimilarity matrix between groups and visualized using the ggplot2 (RRID: SCR_014601) package (19). We assessed the statistical significance (P < 0.05) throughout and, whenever necessary, we adjusted P values for multiple comparisons according to the Benjamini and Hochberg method to control FDR (20) while performing multiple testing on taxa abundance according to sample categories. We performed an ANOVA among sample categories while measuring the α-diversity using the plot_anova_diversity function in microbiomeSeq. Permutational multivariate ANOVA with 999 permutations was performed on all principal coordinates obtained during PCoA with the ordination function of the microbiomeSeq package. Linear regression (parametric test), and Wilcoxon (nonparametric) tests were performed on ASVs abundances against coprostanol levels using their base functions in R (21).

At necropsy, ID8 tumors were harvested from the omentums of NSG mice and snap-frozen in liquid nitrogen. Prior to RNA extraction (Takara Nucleospin), tumors were crushed using ice-cold mortar and pestle under sterile conditions. Following RNA isolation, cDNA libraries were prepared by the LRI Genomics Core facility and sent to Macrogen Inc. for RNA sequencing (RNA-seq). The RNA-seq results were analyzed by the Cleveland Clinic Quantitative Health Sciences’ Bioinformatics Consulting Service.

Raw sequence FASTQ files underwent read quality assessment with FASTQC (RRID:SCR_014583; version 0.11.8, https://www.bioinformatics.babraham.ac.uk/projects/fastqc/; ref. 22). Ribosomal RNA content was evaluated with SortMeRNA (version 2.1; ref. 23). Adapter removal and quality trimming was performed with Trimgalore (version 0.5.0, https://github.com/FelixKrueger/TrimGalore). The adapter was removed, trimmed reads were aligned to the Mus musculus genome assembly (GRCm38 version 99, ftp://ftp.ensembl.org/pub/release99/fasta/mus_musculus/dna/Mus_musculus.GRCm38.dna.primary_assembly.fa.gz) using STAR (RRID:SCR_004463; version 2.6.1d; ref. 24), and duplicates were marked with picard (RRID:SCR_006525; version 2.18.27, http://broadinstitute.github.io/picard/). Alignments were assessed for quality with qualimap (version 2.2.2) and rseqc (version 3.0.0; ref. 25), and for library complexity saturation with preseq (version 2.0.3). Aligned reads were counted at the exon level and summarized at the gene level with the featureCounts tool from the Subread package (version 1.6.4; ref. 26), using annotations for build GRCm38 (version99, ftp://ftp.ensembl.org/pub/release99/gtf/mus_musculus/Mus_musculus.GRCm38.99.gtf.gz). Normalization and differential expression analysis was performed with the R (version 3.6.3; ref. 27) package DESeq2 (RRID:SCR_000154; version 1.26.0; ref. 28). Size factor and dispersion estimations were performed using default settings. Comparison estimate P values for H₂O-Cisplatin versus H₂O-Placebo, Antibiotic-Cisplatin versus Antibiotic-Placebo, Antibiotic-Cisplatin versus H₂O-Cisplatin, and Antibiotic-Placebo versus H₂O-Placebo were extracted using Benjamini and Hochberg multiple testing adjustments (29) and an independent filtering significance cutoff of 0.05 (30). Log₂-fold change estimates for each comparison were shrunken with the function lfcShrink from the DESeq2 package, using default settings. The web tool DAVID bioinformatics database was used to assess the top 100 significantly expressed genes from each group comparison for enrichment of gene ontology terms (31).

Upon murine necropsy, ascites, splenocytes, and bone marrow were collected and processed into single-cell suspensions through filtration and stored in freezing media (10% DMSO in FBS; Atlas Biologicals) at −80°C. All single-cell suspensions were thawed on ice, washed once with 2% BSA in PBS and stained with live/dead UV stain (Invitrogen), and subsequently blocked in FACS buffer (PBS, 2% BSA) containing FcR blocking reagent at 1:50 (Miltenyi Biotec) for 15 minutes. After live/dead staining and blocking, antibody cocktails for myeloid or lymphoid panels (Supplementary Table S2) were incubated with cells at a 1:50 dilution for 20 minutes on ice before being washed and suspended in FACS buffer for analysis.

Cell populations were analyzed using an LSRFortessa (BD Biosciences) and were separated and quantified using FlowJo software (RRID:SCR_018933; Tree Star Inc.). Gating methods for myeloid and lymphoid populations were performed following standardized gating strategies previously described (32). For complete gating strategies, please see Supplementary Fig. S1.

Upon murine necropsy, total tumor collected per mouse was dissociated using standard methods with a Papain Dissociation Kit (Worthington Biologicals). Following filtration through a 40-μm filter, single cells were counted and cultured in serial dilutions in a non-adherent 96-well plate (Sarstedt) with 200-μL serum-free DMEM/F12 medium supplemented with 20-ng/mL basic FGF (Invitrogen), 10-ng/mL EGF (BioSource), and 2% B27 (vol/vol; Invitrogen). Tumorsphere formation was scored following 2 weeks incubation using a phase contrast microscope. The sphere initiation frequency was calculated using an extreme limiting dilution algorithm (RRID:SCR_018933; http://bioinf.wehi.edu.au/software/elda/).

ID8-LUC and ID8-VEGF cells were plated at 1,000 cells per well and treated with ampicillin, vancomycin, metrinodazole, and neomycin in the same proportions used in the murine studies. After 72 hours, proliferation was measured using CellTiter-Glo (Promega) and compared with vehicle-treated control. In addition, following incubation for 7 days with ampicillin, vancomycin, metrinodazole, and neomycin, the IC₅₀ of cisplatin (Spectrum Chemical) was assessed compared with vehicle-pretreated controls using CellTiter-Glo (Promega).

Twenty microliter of plasma was mixed with 80 μL of cold methanol containing an isotope-labeled internal standard mix composed of d4-indoxyl sulfate (Toronto Research Chemicals), d2-indole-3-propionic acid, and N-Benzoyl-d5-glycine (d5-hippuric acid; C/D/N Isotopes Inc.; each 2.5 μmol/L). The mixture was vortexed and centrifuged at 20,000 × g and 4C for 10 minutes. 0.5 μL of supernatant was injected into a CD-C18 column (50×3 mm, CD032, Imtake) at a flow rate of 0.35 mL/min using a Shimadzu high-performance liquid chromatography system interfaced with a mass spectrometer (LCMS-8050CL, Shimadzu). A liquid chromatography gradient generated from two solvents (A: 0.1% propionic acid in water; B: 0.1% acetic acid in methanol) to resolve analytes. Targeted metabolites and isotope-labeled internal standards were monitored using electrospray ionization in either negative-ion [indoxyl sulfate (Cayman Chemical), d4-indoxyl sulfate] or positive-ion [indole-3-propionic acid (Sigma), hippuric acid, d2-indole-3-propionic acid, d5-hippuric acid] multiple reaction monitoring mode with parent to daughter transitions of m/z 211.90 → 79.85 for indoxyl sulfate, 215.90 → 79.80 for d4-indoxyl sulfate, 190.10 → 130.15 for indole-3-propionic acid, 180.00 → 105.10 for hippuric acid, 191.80 → 130.05 for d2-indole-3-propionic acid, 185.00 → 110.15 for d5-hippuric acid. Parameters for ion monitoring were optimized for individual metabolites and internal standards and dwell time was set at 10 milliseconds. Argon was used as the collision-induced dissociation gas and nitrogen (99.95% purity) was used otherwise. Various concentrations of standards were mixed with fixed concentrations of an isotope-labeled internal standard mix to prepare calibration curves for quantitation of indoxyl sulfate, indole-3-propionic acid, and hippuric acid, respectively.

All data are presented as mean ± SEM. All statistical analysis was performed in GraphPad Prism v8 unless otherwise noted. Replicate numbers and P values are presented in figure legends.

All murine studies were completed in accordance with the Institutional Animal Care and Use Committee guidelines, approval # 2018–2003. All studies using lentiviral particle generation were completed in accordance with the Institutional Biosafety Committee guidelines, approval #IBC0920.

The data generated in the study are available within the article and Supplementary Data files. The sequencing data generated in this study are publicly available. RNA-seq data is available from the NCBI Gene Expression Omnibus (GSE200025). 16S rDNA sequencing is available at the NCBI with Bioproject Number PRJEB50898.

We tested the hypothesis that antibiotic-driven disruption of the microbiome impacts tumor development and chemotherapy sensitivity in EOC. To this end, the following study paradigm (Fig. 1A) was implemented wherein C57BL/6J female mice at 6 weeks of age were provided water or antibiotic-containing water (ABX; ampicillin, neomycin, metronidazole, and vancomycin; refs. 13, 33), which was sustained for the duration of the study. After 2 weeks, mice were injected intraperitoneally with murine EOC lines ID8 or ID8-VEGF, syngeneic with BL/6 mice. Tumor growth was monitored by TAUS weekly for the duration of the study (15). ID8 and ID8-VEGF cell lines were used as they are highly characterized and closely recapitulate human ovarian cancer progression. The ID8-VEGF cell line is slightly more severe, causing an increase in angiogenesis and ascites development (34). At 2 weeks post tumor cell injection, mice were injected intraperitoneally with cisplatin or vehicle weekly for the remainder of the study. Mice met study endpoint at tumor burden of > 150 mm³ or humane endpoint, including debilitating ascites development. ID8 and ID8-VEGF EOC tumor growth is significantly increased in ABX-treated mice compared with controls (Fig. 1B and C, respectively). Moreover, in the presence of ABX, cisplatin therapy exhibited attenuated efficacy compared with control-treated mice, indicating development of cisplatin resistance (Fig. 1B and C). Median survival was decreased in the ABX treatment groups compared with controls (Fig. 1D and E). The ID8 cohort exhibited a median survival of 66 and 64 days in the ABX placebo and cisplatin groups compared with 68.5 and 84 days in the H₂O placebo and cisplatin groups, respectively (Fig. 1D). This result was paralleled in the ID8-VEGF cohort of mice with a median survival of 39 and 34 days in the ABX placebo and cisplatin groups compared with 42 and 53 days in the H₂O placebo and cisplatin groups respectively (Fig. 1E). The ID8-VEGF mice reached the endpoint earlier than the ID8 cohort secondary to large volume ascites development. Upon necropsy, EOC tumor phenotype was confirmed through histologic assessment of hematoxylin and eosin–stained tissue sections as well as benign adjacent omentum (Fig. 1F). We performed 53BP1 immunofluorescence analysis on the ID8 tumors harvested from the ID8 H₂O + Cisplatin and ABX + Cisplatin cohort and found elevated 53BP1 in the H₂O cohort compared with ABX (Fig. 1F). Consistent with these findings we showed an increase in TUNEL staining in the H2O cisplatin group compared with the ABX cisplatin group (Fig. 1G). These results are consistent with an increase in DNA damage in the H₂O cohort treated with cisplatin that is attenuated in the ABX cisplatin-treated mice (Fig. 1G). To validate the histologic findings and further establish the mechanisms of increased tumor growth and chemoresistance in tumors from ABX mice, we performed qRT-PCR for 53BP1, BRCA1, and CD31 (angiogenesis). We found suppression of 53BP1 in tumors from ABX cisplatin-treated mice compared with tumors from H₂O cisplatin cohort (Fig. 1H). BRCA1 expression was increased in tumors from ABX cisplatin-treated mice compared with tumors from H₂O cisplatin cohort (Fig. 1H) indicative of increased DNA damage repair. Further, we found increased CD31 expression in tumors from ABX cisplatin-treated mice compared with tumors from H₂O cisplatin cohort (Fig. 1H). Collectively, our findings indicate ABX treatment promotes accelerated tumor growth via suppression of apoptosis, increase in DNA damage repair, and enhanced angiogenesis.

The majority of patients with EOC present with advanced disease (stage III or stage IV) that may include ascites. Previous studies have reported that immune cells such as natural killer cells and macrophages found in peritoneal ascites may play a key role in tumor cell invasiveness and growth. We analyzed the immune cell populations within the peritoneal ascites fluid through flow cytometry. Our analysis focused on both myeloid and lymphoid populations (immune panels and gating strategy shown in Supplementary Table S1 and Supplementary Fig. S1). No significant difference in myeloid (Supplementary Fig. S2A) or lymphoid (Supplementary Fig. S2B) cell populations was observed in the ascites of mice following ABX therapy in the presence or absence of cisplatin therapy in the C57Bl/6J ID8-VEGF cohort. ID8-VEGF cells were analyzed as they consistently developed ascites. In addition, no significant differences were observed between myeloid or lymphoid cell populations within collected splenocytes (Supplementary Fig. S3A and S3B) or bone marrow (Supplementary Fig. S3C and S3D) at endpoint.

To assess whether the observed augmentation of tumor growth and cisplatin resistance following ABX treatment is dependent on an intact immune system, the same study paradigm was used in NOD.Cg-Prkdc<scid>Il2rg<tm1Wjl>SzJ (NSG) immunodeficient mice. The NSG cohorts inoculated with either ID8 or ID8-VEGF and treated with ABX displayed accelerated tumor growth with more rapid onset when compared with the H₂O C57Bl/6J cohort, with no benefit of cisplatin (Fig. 2A and C). The time to tumor progression was accelerated and the effect on median survival was even more significant in the ID8 cohort, which exhibited a median survival of 41 and 45 days in the ABX placebo and cisplatin groups in comparison with a median survival of 55 and 69 days in the H₂O placebo and cisplatin groups respectively (Fig. 2B). A similarly reduced overall survival was observed in NSG mice inoculated with ID8-VEGF (Fig. 2D). Peritoneal ascites fluid harvested at endpoint was analyzed by flow cytometry. NSG mice have immature T cells, dendritic cells, and macrophages, but functional neutrophils. NSG mice in the ID8-VEGF cohort exhibited ascites with no significant alterations in myeloid (Supplementary Fig. S2C) populations across treatment groups. Collectively, the findings indicate that ABX treatment impacts the kinetics and overall survival of murine EOC tumor-bearing mice.

The ABX treatment regimen we applied was previously shown to be sufficient to deplete detectable commensal bacteria (13). However, to define the response of the gut microbiome to broad-spectrum antibiotic therapy, stool was collected at baseline, at 2 weeks, at cisplatin initiation, at 5 weeks, and at endpoint necropsy and processed for 16S rRNA gene sequencing. The microbiome was analyzed from the following cohorts: C57Bl/6J ID8, C57Bl/6J ID8-VEGF, NSG ID8, and NSG ID8-VEGF.

In total, 2,801,471 and 2,411,554 high-quality and usable reads were obtained from fecal samples of 5 mice per treatment group in the ID8 and ID8-VEGF BL/6 cohorts from sequencing the 16S rRNA gene, with an average length of 210 base pairs (bps). There were 3,184 ASVs in all ID8 and ID8-VEGF BL/6 samples. Pairwise comparisons within both the ID8 and ID8-VEGF cohorts revealed statistically significant differences in alpha diversity between temporal collections, as analyzed by the Shannon diversity index (ANOVA; Supplementary Fig. S4A and S4B). In addition, the Bray–Curtis dissimilarity-based beta diversity comparisons between collection time points in the ID8 and ID8-VEGF cohorts were significant at P = 0.003 and P = 0.001, respectively (Supplementary Fig. S4C).

Antibiotic-treated groups displayed significantly less diverse fecal microbial communities compared with control groups regardless of cisplatin therapy. Specifically, the relative abundance of some Proteobacteria including Enterobacteriaceae and Parasutterella were increased in the antibiotic-treated ID8 groups, while the abundance of Enterobacteriaceae was increased in the antibiotic-treated ID8-VEGF groups (Fig. 3A and B). Being gram-negative and mostly facultative anaerobes, the Proteobacteria are generally resistant to vancomycin, which affects mainly gram-positive bacteria, and are also less susceptible to metronidazole, which mostly affects anaerobes. Furthermore, Enterobacteriaceae have multiple antibiotic resistance genes allowing for their survival throughout ABX therapy (35). Increased abundance of Parasutterella has been associated with dysbiosis of the gut microbiome, but the mechanism behind this association has not been fully interrogated (36).

In total, 4,784,498 and 2,822,423 high-quality and usable reads were obtained from fecal samples of 5 mice per treatment group in the ID8 and ID8-VEGF NSG cohorts from sequencing the 16S rRNA gene, with an average length 210 bps. There were 6,484 ASVs in all ID8 and ID8-VEGF NSG samples. As with the BL/6 cohort, the pairwise comparisons revealed statistically significant (ANOVA) patterns for differences in alpha diversity between the collection time points in both the ID8 and ID8-VEGF cohorts (Supplementary Fig. S5A and S5B). In addition, the Bray–Curtis dissimilarity-based beta diversity between collection time points was significant at P = 0.001 for both the ID8 and ID8-VEGF cohorts (Supplementary Fig. S5C).

In addition, there was also a marked relative increase in Paenibacillus and Enterococcus (Fig. 4A and B).

To determine if tumor growth was directly affected by ABX interaction with the ID8 and ID8-VEGF cell lines, we performed tissue culture analyses. ID8 and ID8-VEGF cell lines were cocultured with various concentrations of ABX (metronidazole, vancomycin, ampicillin, and neomycin) in the same respective ratios used in the murine studies. Following 7 days in culture, there were no appreciable differences in ABX-treated cells when compared with cells in normal media (Supplementary Fig. S6A and s6B). In addition, there were no changes in the IC₅₀ of cisplatin for ID8 or ID8-VEGF cells following ABX treatment when compared with controls (Supplementary Fig. S6C and S6D). We performed coculture studies with conditioned media of bacterial cultures from the cecum of H₂O and ABX-treated mice with ID8 cells. No differences in ID8 cell growth or spheroid frequency could be detected.

To ensure our observed phenotype was not unique to syngeneic EOC cell lines (ID8 and ID8-VEGF). We repeated our study paradigm in NSG mice with a human-derived OV81 cell line. Following 2 weeks of ABX or control water, mice were intraperitoneally injected with 5×10⁶ OV81 cells. To monitor tumor growth is this cohort of mice, cells were transduced with a luciferase reporter prior to intraperitoneal injection. After 2 weeks post injection, mice were randomized into 4 groups (H₂O vehicle, H₂O cisplatin, ABX vehicle, and ABX cisplatin) and treated as previously outlined. Mice underwent imaging via the IVIS (Perkin Elmer), and total flux of photons/second was calculated to determine total tumor burden. Overall, ABX therapy resulted in a significant increase in total tumor burden when normalized to initial tumor burden in comparison with H₂O controls for both the vehicle (Supplementary Fig. S7A) and cisplatin-treated groups (Supplementary Fig. S7B).

To investigate the mechanisms of ABX-induced microbiome disruption on EOC tumor growth and sensitivity to cisplatin, we tested for presence of bacteria in the tumors and when available in the ascites. We cultured samples from tumors of H₂O and ABX-treated mice on agar plates prepared with Brain Heart Infusion (BHI). The inoculated plates were then incubated both under aerobic and anaerobic conditions for 4 days. As expected, we got bacterial growth for the H₂O-treatment samples but no bacterial growth was observed in tumors from ABX-treated mice. The findings are consistent with the findings in Figs. 3 and 4 indicating attenuation of bacteria in the gut and related reduction in the tumors. To elucidate the mechanisms underlying the impact of antibiotic therapy on tumor cells and their growth in the presence and absence of cisplatin, bulk RNA-seq was performed on tumors collected from H₂O (placebo, cisplatin) and ABX (placebo, cisplatin) treated NSG mice at endpoint (8 weeks post ID8 cell injection). The top 100 most significantly affected genes by P value following a pairwise comparison between the H₂O placebo and ABX placebo groups (Fig. 5A) were analyzed using DAVID software for the enrichment of GO terms. Interestingly, the most enriched GO terms included cell differentiation, proliferation, and locomotor activity in line increased cancer stem cells (CSC; Fig. 5B). We determined that genes commonly associated with CSCs were also increased in the ABX treatment group compared with the H₂O control group including the SOX2, WNT7a, HOXb5, HOXb6, DLX5, MSX1, EFNA4, SALL1, and PAX2 genes. SHOX2 and GATA5, genes that promote cell differentiation, were reciprocally regulated (Fig. 5C; refs. 37–39). We validated the RNA-seq data by performing qRT-PCR analysis of DLX5, MSX1, and PAX2 (Fig. 5D). Gene set enrichment analysis showed that epithelial–mesenchymal transition, NFkB signaling via TNFα, and hypoxia were all enriched in ABX-treated tumors compared with H₂O-treated controls (Fig. 5E; refs. 40, 41).

As we observed increased stem cell markers based on RNA-seq of tumors, we performed self-renewal assays to assess stem cell frequency in tumors from H₂O and ABX-treated mice with and without cisplatin therapy. Following endpoint necropsy, tumor tissue from each group was dissociated to single cells and plated in a limiting dilution assay. Following 14 days of incubation, sphere initiation frequency was determined, which indicated a significant increase in sphere initiation frequency of ID8 cells from ABX-treated BL/6 mice compared with the H₂O cohort that was highly significant in the cisplatin treatment group (Fig. 6A). We validated the change in stem cell frequency in a second cohort of mice by qRT-PCR analysis of treated mice (Fig. 6B–D). We determined that ABX treatment plus cisplatin increased expression of NANOG (Fig. 6B), OCT4 (Fig. 6C), and SOX2 (Fig. 6D) in a parallel manner to the spheroid analysis. The impact of ABX on CSCs was replicated in the NSG cohorts, demonstrating an increase in the stem cell population following ABX therapy in a manner augmented by cisplatin (Fig. 6E). Of note, the increase in CSCs in response to cisplatin was unmasked in the H₂O cohort of NSG mice, as cisplatin did not increase the frequency of stem cell formation in ID8 cells derived from the H₂O cohort of C57Bl/6J mice (compare Fig. 6A and E).

We tested the hypothesis that recolonization of the gut microbiome of ABX-treated mice with microbiome of H₂O-treated mice can suppress the accelerated EOC tumor growth and cisplatin resistance. Cecal content was harvested from either H₂O- or ABX-treated mice following the study paradigm outlined in Fig. 7A. C57Bl/6J mice treated with ABX were orally gavaged with cecal content of H₂O- or ABX-treated mice, followed by intraperitoneal injection with ID8 EOC cells and monitored for tumor progression with biweekly IVIS imaging. Endpoint criteria was predetermined to be: total tumor burden >150 mm³, debilitating ascites development, and/or a standard BCS < 2. At onset of endpoint for ABX CMT-treated mice, all groups were compared for total tumor progression from baseline (Fig. 7B). Mice receiving H₂O CMT had tumor burden comparable with mice that did not receive any ABX treatment (Supplementary Fig. S8A), whereas mice receiving ABX CMT had tumor burden nearly the same as mice receiving ABX throughout (Supplementary Fig. S8B). Survival analysis showed that ABX CMT-treated mice were not responsive to cisplatin therapy; the median survival of vehicle and cisplatin-treated mice was 61 and 57 days post cell injection respectively. The H₂O CMT-treated mice were sensitized to cisplatin therapy, showing median survival of 83 and 91 days in vehicle and cisplatin-treated groups respectively (Fig. 7C).

The suppression of tumor growth and sensitization to cisplatin by recolonization led us to screen plasma samples collected from C57Bl/6J mice treated with and without ABX as well as with and without cisplatin. Plasma samples were screened for 29 previously identified gut-derived metabolites via mass spectrometry. We identified indole-3-propionic acid, indoxyl sulfate, and hippuric acid in H₂O-treated mice that were suppressed by ABX (Fig. 8A–C). We next assessed the metabolites from the CMT study and found indole-3-propionic acid and indoxyl sulfate levels returned to baseline in ABX mice recolonized with H₂O CMT (Fig. 8A and B). Hippuric acid levels were not restored in H₂O CMT mice (Fig. 8C), indicating it may not participate in suppression of ovarian cancer growth in intact or CMT recolonized gut microbiome. The findings identify candidate gut-derived metabolites that may connect the gut microbiome to ovarian cancer growth in the peritoneum. Collectively, our studies point to a deleterious impact of antibiotics on the gut microbiome and microbial metabolites with subsequent impact on ovarian cancer growth and diminished response to chemotherapy (Fig. 8D). Tumors from ABX-treated mice exhibit increased angiogenesis and CSCs.

In recent years, there is growing evidence of a link between the gut microbiome, carcinogenesis, and response to cancer therapy (5–8, 42, 43). Studies have emerged supporting, in both preclinical models and patient cohorts, that antibiotic therapy-associated gut microbial disruption may negatively impact the efficacy of immune checkpoint inhibitors and systemic anticancer drugs, including platinum chemotherapy (6, 7, 42, 43). To date, the impact of the gut microbiome upon response to chemotherapy in women with gynecologic malignancies is yet to be explored. In women diagnosed with EOC, platinum chemotherapy remains the standard treatment in primary and newly recurrent disease. Disease prognosis and treatment efficacy depends upon platinum sensitivity, with platinum resistance portending a poor prognosis with limited active treatment options. Interventions to increase platinum sensitivity and prevent development of resistance are essential to improving the care of women with EOC.

The preclinical studies described here demonstrate that antibiotic therapy, and concomitant changes within the gut microbiome, results in accelerated tumor growth, attenuated cisplatin sensitivity, and decreased survival. In mice that received ABX, 16S rDNA analysis demonstrated that the gut microbiome was markedly suppressed in comparison with controls. ABX treatment led to significant reduction in diversity in the fecal microbial communities, regardless of cisplatin therapy. Similarly, cultured ascites, if available, and tumors from ABX-treated mice including cisplatin treatment on BHI agar plates incubated for 72 hours under either aerobic or anaerobic conditions showed no bacterial colonies. In contrast, tumors from H₂O-treated mice did identify bacterial colonies. The findings corroborate the 16S sequencing results, indicating significant reduction in bacteria upon ABX treatment. The coculture studies of ID8 cells with conditioned media of bacterial cultures from the cecum of H₂O and ABX-treated mice showed no differences in ID8 cell growth or spheroid frequency. This analysis excludes the possibility of bacteria leaking into the peritoneal cavity due to prolonged ABX treatment and supports our conclusion of a gut-derived suppressive activity in an intact microbiome.

ABX treatment led to chemoresistance and poor overall survival. Our analysis of the tumors derived from ABX plus cisplatin cohort revealed an increase in apoptosis based on TUNEL staining that should have resulted in anticancer activity. There are several counter mechanisms observed in the tumors from ABX and cisplatin-treated mice that could underlie the chemoresistance including reduced DNA damage based on 53BP1, increased angiogenesis based on CD31, and an increase in CSCs based on self-renewal in cells derived from the treated tumors. The mechanism of induction of chemoresistance in tumors from ABX-treated mice requires further investigation.

We found that restoration of the gut microbiome in ABX-treated mice via CMT from H₂O-treated mouse cecum is sufficient to improve overall survival. Moreover, we provide evidence that circulating gut-derived metabolites are found in plasma of H₂O-treated mice and are disrupted by ABX and restored by cecal microbiome recolonization. This supports a role for a gut-derived activity/metabolite underlies the suppression of tumor growth and maintenance of chemosensitivity in EOC.

Our findings are consistent with studies investigating the impact of the gut microbiome on response to platinum chemotherapy and immune checkpoint inhibitor treatment for non–gynecologic cancers. Indeed, Routy and colleagues showed resistance to immune checkpoint inhibitors was linked to abnormal gut microbiome composition (8). Following disruption of the gut microbiome through ABX treatment, response to CpG-oligonucleotide immunotherapy and cyclophosphamide was impaired secondarily to reduced cytokine production, lowered production of reactive oxygen species, and diminished cytotoxicity. In a separate study, Iida and colleagues demonstrated that disruption of the gut microbiome through ABX impaired platinum response, leading to decreased tumor regression and survival in animal models of colon cancer and lymphoma (7). Notably, the studies of Chen and colleagues (44) indicate that disruption of the gut microbiome with ABX is sufficient to delay induced spontaneous ovarian cancer development.

In their studies, Chen and colleagues (44) used a genetically engineered spontaneous mouse ovarian cancer model BPRN (Ovgp1-iCreERT2; Brca1fl/fl; Trp53fl/fl; Rbfl/fl; Nf1fl/fl) and monitored tumor induction after previous exposure to ABX. The data indicate ABX disruption leads to long-term impact on the gut and vaginal microbiome and is associated with subsequent reduction in ovarian tumors. This would appear to contrast with our findings. There are differences in the implementation of our studies that may account for the differences. Our study used an alternative experimental paradigm with C57BL/6J female mice who were similarly treated with antibiotic therapy and then underwent intraperitoneal injection of syngeneic mouse-derived ID-8 or ID-8 VEGF cell lines, and a human-derived platinum-sensitive OV81 high-grade serous ovarian cancer patient-derived xenograft model. OV81 has a p53 mutation (R121S) and does not harbor BRCA 1/2 mutations. The differences in models and the treatment paradigm may account for the observed outcomes in our respective studies. Collectively, our studies and those of Chen and colleagues highlight the complexity of elucidating the mechanism and impact of ABX and by extension the microbiome on ovarian cancer development and chemoresistance.

Similar findings to ours have been documented in cohorts of patients undergoing platinum chemotherapy for non–gynecologic cancers. Among 800 patients enrolled on clinical trials receiving cyclophosphamide or cisplatin for chronic lymphocytic leukemia or lymphoma, receipt of ABX targeting gram-positive species during chemotherapy was associated with significantly decreased treatment response, time to recurrence, and survival (45). These findings suggest that presence of specific bacterial populations may be essential for treatment response. In our studies, we identified significantly higher relative abundance of potentially pathogenic Enterobacteriaceae, and Parasutterella species. Notably, among animals treated with cisplatin, these bacteria were further increased in the absence of facultative gram-positive bacteria compared with those that received placebo. Although not yet studied in relation to cancer, an increase in the abundance of Parasutterella has been associated with dysbiosis of the gut microbiome and alterations in liver metabolism (36).

Our findings identify candidate gut-derived microbial metabolites that may facilitate suppression of ovarian cancer growth and sensitivity to cisplatin. Indole-3-propionic acid and indoxyl sulfate were identified in our metabolomics analysis and may provide a starting point for establishing the link between the gut microbiome and ovarian cancer. Future studies will focus on in vitro and preclinical studies to address the role of these candidate metabolites or other may act in suppression of ovarian cancer and chemotherapy sensitivity.

We determined that disruption of the immune system does not directly contribute to increased tumor growth or reduced platinum sensitivity in the setting of microbial disruption. No differences in myeloid and lymphoid populations were observed in the ascitic fluid of mice following antibiotic treatment when compared with controls. We acknowledge the significant contributions of the macrophage populations that were not included in our flow panel and acknowledge the limitation of their exclusion. Our results identify that, in the setting of a disrupted microbiome, response to platinum chemotherapy is reduced and populations of chemoresistant CSCs are increased on the basis of tumor sphere initiation frequency and transcriptional reprogramming to an increased CSC-like state. Our findings also unmask an interaction between cisplatin and ABX in gut dysbiosis.

Our findings unmask a role for the immune system in chemotherapeutic induction of CSCs. Our published studies and those of others indicate that chemotherapeutics can induce CSCs in culture. The studies are performed in culture, so lack an intact tumor microenvironment. Using spheroid frequency analysis from tumors in immune-competent mice, we found no difference in CSC frequency between H₂O placebo and H₂O cisplatin-treated mice. In contrast, in immunodeficient mice, we find a significant increase in CSC frequency in H₂O cisplatin-treated mice compared with H₂O placebo. ABX treatment unmasks the increase in cisplatin-induced stemness independent of immune status. The cisplatin-induced CSCs in tumors of immunodeficient mice, when compared with immunocompetent mice in the H₂O treatment cohort, implicates immune surveillance in suppression of CSCs induced by cisplatin.

The relationship between enhanced populations of CSCs and platinum resistance in EOC is well documented in the literature (40, 46–49). Specifically, we identified SOX2, WNT7a, HOXb5, HOXb6, DLX5, MSX1, EFNA4, SALL1, and PAX2 as being significantly upregulated in ID8 tumors of the ABX group over the control H₂O group. Many of these genes are markers of pluripotency and regulation as well as of long-term stemness (37–39, 47, 50). SALL1 interacts with NANOG, a well-established pluripotency transcription factor, to suppress differentiation (50). Conversely, SHOX2 and GATA5, genes that favor differentiation, were significantly decreased, providing further evidence that ID8 tumor cells from ABX mice are more stem-like and undifferentiated than ID8 tumor cells from H₂O-treated mice (51). Upon KEGG pathway analysis of enriched genes in the ABX group, multiple signaling pathways including P13K/AKT/mTOR and WNT, were identified and currently under investigation as targets for ovarian cancer therapies (41). Further investigation is needed to understand the mechanism through which the gut microbiome interacts with CSCs and how this specifically drives platinum resistance in EOC.

Preclinical studies have demonstrated that supplementation with specific bacterial species and fecal microbiota transplant can restore the efficacy of chemotherapy and immunotherapy in several cancer types. In a landmark, phase I clinical trial by Baruch and colleagues, 10 patients with programmed programmed cell death protein 1 refractory metastatic melanoma underwent microbiota depletion, followed by fecal microbiota transplant with stool from responding patients (52). Notably, clinical responses were restored in 30% of patients, with one complete response. Therefore, it is plausible that these approaches may yield similar benefits for women with gynecologic cancer. Our findings of changes in indole-3-propionic acid and indoxyl sulfate levels provide candidate molecules to test as potential therapeutics for overcoming ABX accelerated ovarian cancer growth and chemoresistance.

The goal of our studies was to substantially change the microbiota composition and provide proof of concept for impact of the gut microbiome on ovarian cancer growth. While the goal of this approach is to substantially change the microbiota composition and provide important proof-of-concept results, this also does have clinically relevant implications. For instance, when patients first arrive in septic shock or have neutropenic fever during chemotherapy, typically broad-spectrum antibiotics with gram-positive, gram-negative, and anaerobic coverage are empirically started until a definitive source of infection is identified. In these patients, it is typical that after culture information has resulted (which may have 3–4 days) that antibiotic therapy is de-escalated to single agent therapy. The impact of antibiotic prophylaxis upon oncologic outcomes has not been well explored. Prophylaxis in these settings generally constitutes 1 to 2 doses of antibiotic therapy, leaving it unclear whether or not this treatment causes appreciable changes in microbiota. Our group has published the results of an accompanying retrospective clinical study (summarized in the introduction) that demonstrates that antibiotic use is associated with decreased efficacy, worse progression-free survival, and worse overall survival in patients during platinum chemotherapy, with this effect most pronounced in patients who received antibiotics that target gram-positive species. While these clinical findings are hypothesis-generating and limited by lack of correlative stool specimens, they provide essential clinical context that alterations to the microbiota via antibiotic therapy are associated with significant alterations in platinum chemotherapy efficacy and poorer oncologic outcomes.

The clinical implications of the gut microbiome impacting platinum response in EOC are significant. Antibiotic therapy is often unavoidable in the care of patients with EOC following cytoreductive surgery or during chemotherapy. However, here the evidence supports the need for judicious selection of antibiotics and duration of dosing, as changes to the gut microbiome may negatively impact oncologic outcomes. Clinicians should carefully weigh the benefits and potential adverse effects of prescribing broad-spectrum antibiotics during platinum chemotherapy. Antibiotic therapy should be prescribed for the shortest duration necessary. While treatment with antibiotics is often unavoidable during chemotherapy, providers should be aware of institutional antibiograms and directed data from cultures to tailor therapy. It also supports the concept that measures to prevent infection in women with EOC, and therefore avoid antibiotic treatments, should be prioritized. Most importantly, understanding that disruption of gut microbiota impacts EOC growth and platinum chemotherapy introduces the potential for development of targeted therapeutics, designed to restore an intact gut microbiome, which may represent promising strategies to treat EOC and combat platinum resistance in the future.

We identify a role for the disruption of gut microbiota with concomitant reduction in gut microbial metabolites via antibiotic treatment results in enhanced tumor growth and reduced sensitivity to platinum chemotherapy in preclinical models of EOC (Fig. 8D). Further investigation is critically needed to understand how the tumor microenvironment, and CSCs specifically, communicate with the gut microbiome to drive platinum resistance in EOC. Answers to these questions will provide important insights as to whether microbe-directed interventions can be used to impact the response to platinum chemotherapy.

Z. Wang reports grants from NHLBI during the conduct of the study and personal fees from Procter & Gamble and Cleveland Heart Lab outside the submitted work. P.P. Ahern reports personal fees from Novome Biotechnologies outside the submitted work. J. Claesen reports personal fees from Seed Health Inc. outside the submitted work. No disclosures were reported by the other authors.

L.M. Chambers: Conceptualization, data curation, investigation, methodology, writing–original draft, writing–review and editing. E.L.E. Rhoades: Investigation, methodology, writing–original draft, writing–review and editing. R. Bharti: Formal analysis, validation, investigation, methodology. C. Braley: Validation, investigation, writing–review and editing. S. Tewari: Validation, investigation. L. Trestan: Investigation, methodology. Z. Alali: Formal analysis, validation, investigation, methodology. D. Bayik: Data curation, formal analysis, investigation. J.D. Lathia: Formal analysis, writing–review and editing. N. Sangwan: Data curation, software, investigation, methodology. P. Bazeley: Data curation, formal analysis, investigation, methodology. A.S. Joehlin-Price: Methodology, writing–review and editing. Z. Wang: Formal analysis, validation, investigation. S. Dutta: Resources, formal analysis, validation, methodology. M. Dwidar: Formal analysis, investigation, methodology. A. Hajjar: Resources, formal analysis, investigation, methodology, writing–review and editing. P.P. Ahern: Validation, methodology, writing–review and editing. J. Claesen: Validation, methodology, writing–review and editing. P. Rose: Validation, writing–review and editing. R. Vargas: Validation, writing–review and editing. J.M. Brown: Validation, investigation, methodology, writing–review and editing. C.M. Michener: Validation, writing–review and editing. O. Reizes: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing.

The authors would like to acknowledge members of the Reizes laboratory for their collaborative effort and insight toward the completion of the studies within this manuscript. They thank Alex Myers in the Reizes lab for careful editing of the manuscript. The studies were facilitated by the LRI Shared Lab Resources that played a role in data collection, analysis, and/or interpretation of the findings presented within the manuscript, including: Microbiome Analytics and Composition Core Facility, Image Core, Histology Core, and Microbial Culture and Engineering Core. D. Bayik received support from NIH F32 CA243314. Dr. Reizes is the Laura J. Fogarty Endowed Chair for Uterine Cancer Research. The Reizes laboratory is funded by VeloSano Bike to Cure, Case Comprehensive Cancer Center, Center of Research Excellence in Gynecologic Cancer, and the Department of Defense.

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