Antibiotic-induced disruption of the microbiome exacerbates chemotherapy-induced diarrhea and can be mitigated with autologous fecal microbiota transplantation

Chemotherapy is well documented to disrupt the gut microbiome leading to poor treatment outcomes and a heightened risk of adverse toxicity. While strong associations exists between its composition and gastrointestinal toxicity, its causal contribution remains unclear. Our inability to move beyond association has limited the development and implementation of microbial-based therapeutics in chemotherapy adjuncts with no clear rationale of how and when to deliver them. Here, we investigate the impact of augmenting the gut microbiome on gastrointestinal toxicity caused by the chemotherapeutic agent, methotrexate (MTX). Fecal microbiome transplantation (FMT) delivered after MTX had no appreciable impact on gastrointestinal toxicity. In contrast, disruption of the microbiome with antibiotics administered before chemotherapy exacerbated gastrointestinal toxicity, impairing mucosal recovery (P<0.0001) whilst increasing diarrhea severity (P=0.0007) and treatment-related mortality (P=0.0045). Importantly, these detrimental effects were reverse when the microbiome was restored using autologous FMT (P=0.03); a phenomenon dictated by the uptake and subsequent expansion of Muribacteraeae. These are the first data to show that clinically impactful symptoms of gastrointestinal toxicity are dictated by the microbiome, and provide a clear rationale for how and when to target the microbiome to mitigate the acute and chronic complications caused by disruption of the gastrointestinal microenvironment. Translation of this new knowledge should focus on stabilizing and strengthening the gut microbiome before chemotherapy, and developing new microbial approaches to accelerate recovery of the mucosa. By controlling the depth and duration of mucosal injury, secondary consequences of gastrointestinal toxicity may be avoided.


Introduction
Cancer therapy is well recognized for its detrimental effects on the host microbiota, causing profound and often chronic changes in the microbial community of the gastrointestinal tract 1 . While the exact microbial shifts differ between unique treatment regimens, chemotherapy is typically associated with a loss of key commensals and expansion of enteric pathogens; both of which are exacerbated by prophylactic and empirical antibiotic use 2 . Disruption of the host microbiota is now considered a critical event in the initiation of various treatment complications, many of which are catalyzed by breakdown of the mucosal barrier 3,4 . While mucosal barrier injury (MBI) is initiated by acute cytotoxic injury to the intestinal stem cell niche, it is perpetuated by innate immune responses that are thought to be dictated by host-microbe interactions at the mucosal interface 5 . However, these assumptions are largely based on associations between the gut microbiome composition and gastrointestinal symptoms, with robust causal data lacking 5 . As such, it remains unclear if microbial changes that occur after chemotherapy truly contribute to symptoms, or are simply a form of collateral damage resulting from breakdown of the mucosal barrier.
MBI has long been a major clinical obstacle without effective intervention. Acutely, it decreases the absorptive capacity of the intestine and manifests as diarrhea causing a debilitating loss in quality of life and thus, compliance and dose maintenance 6,7 . In the setting of hematopoietic stem cell transplantation (HSCT), the complications of MBI are particularly profound due to underlying immunosuppression driving immunological complications including graft-versus host disease (GvHD) by permitting translocation of endogenous danger signals that activate donor derived T cells 8 . This same process underpins the development of bloodstream infection (BSI), with the expansion and subsequent translocation of enteric pathobionts now reported to drive more than 50% of all BSI cases in stem cell transplant recipients 9,10 . Of particular interest, a growing body of evidence suggests that microbial disruption (i.e. caused by antibiotic use) at the time of chemotherapy increases the risk of adverse events linked to MBI, suggesting a causal role for the host microbiome 11 .
Fecal microbiota transplantation (FMT) is a powerful tool that can be used fundamentally to explore causal relationships between the host and their resident microbes. It is also recognized for its ability to deliver a diverse and metabolically active microbial community to the host with greater durability compared to commercial probiotics making it of great interest in supportive cancer care.
Its efficacy is well documented in the case of recurrent Clostridium difficile infection (CDI) 12 with experimental efficacy also shown for Inflammatory Bowel Disease (IBD) 13 , highlighting its luminal and mucosal benefits. Within oncology, its use has been met with caution due to the perceived risks of translocation 14 , however recent reports detail profound benefits in steroid refractory GvHD 15,16 . While it has been demonstrated that FMT is able to restore the microbial community after 5-fluoruoracil and antibiotics 17 , it remains unclear if FMT-induced microbial restoration is able to 72 Chapter 4 prevent or minimize MBI and its associated clinical manifestations. The current study therefore aimed to determine the impact of microbial manipulation, using antibiotics and FMT, on MBI and its clinical manifestations using a validated model of methotrexate-induced gastrointestinal toxicity 18,19 . By delivering autologous gut microbes after chemotherapy, we aimed to provide new data to aid in defining the causal contribution of the microbiome to MBI and associated symptoms, and identify translational methods aimed at strengthening the host microbiome before treatment.

Materials and Methods
This study is reported in accordance with the ARRIVE guidelines for the robust and reproducible reporting of preclinical research.

Ethical statement, husbandry specifications and experimental considerations
All experiments were conducted in accordance with ethical guidelines approved by the national Diet Services, The Netherlands) and sterile water. Sawdust bedding was provided in all cases as well as a toilet roll for enrichment. All cages were randomly arranged across racks to prevent potential bias. All experiments were performed in N=9 rats per group which is sufficiently powered to detect a 25% effect size at power 80% and alpha=0.05.

Preclinical model of MTX-induced mucosal barrier injury
Mucosal barrier injury (MBI) was induced with the chemotherapeutic drug, methotrexate (MTX) obtained from the Pharmachie Holding B.V. (The Netherlands) administered at 45-60 mg/kg as previously described 19 . Briefly, male Wistar rats (150-180 g, Charles River, Germany) were treated with a single intravenous dose of MTX on day 0 administered via the penile vein under anesthetic (3% isoflurane). A volume equivalent dose of 0.9% NaCl was used as vehicle control in all experiments as previously described 20 . All experiments were performed in a blinded manner, with rats given a unique identification number.
The primary endpoint of the study was plasma citrulline, a validated biomarker of small intestinal enterocyte mass and clinically utilized marker of MBI caused by chemotherapy 21,22 . Plasma citrulline was assessed in repeated blood samples (75 µl) collected from the tail vein into EDTA-treated hematocrit capillary tubes. Plasma was isolated from whole blood by centrifugation (4000 g for 10 4 min) and citrulline quantified using automated ion exchange column chromatography as previous described 23 . Data are presented as raw citrulline values assessed longitudinally and area under the curve (AUC). AUC was defined as the area below Y = Mean baseline citrulline value for all groups and above the individual animal citrulline curve. Accordingly, the more severe the mucositis, the higher the AUC value.

Clinical toxicity assessment
Clinical manifestations of MBI were assessed daily and included body weight, diarrhea, food intake, water intake and welfare assessments. Diarrhea was assessed using a validated grading system where 0 = no diarrhea, 1 = mild diarrhea indicated by soft but formed pellets, 2 = moderate diarrhea indicated by peri-anal staining and 3 = severe diarrhea with extensive staining and leakage 24 . Food and water intake were measured daily by manually weighing the contents of the chow hopper and water bottle. During the experimental period, HRW was responsible for maintaining animal water and chow eliminating the possibility of animal caretakers to change water bottles or refresh chow.

Histopathology
Routine haematoxylin and eosin (H&E) staining was performed to assess intestinal tissue morphology. Briefly, drop-fixed jejunal tissue was processed and embedded into paraffin wax. 4 µm sections were cut on a rotary microtome and mounted onto glass Superfrost® slides. H&E staining was performed as per routine protocols and slides were scanned using the Hamamatsu Photonics Digital Slide Scanner (NanoZoomer S60). Images were evaluated using the NDP.view2 software.
Villus height and crypt depth were measured using annotation tools in NDP.view2. 10 well oriented crypts/villi were measured per slide and an average calculated per animal.

Antibiotic-induced microbial disruption
Antibiotics (750 mg/l vancomycin, 350 mg/l imipenem and 1 g/l metronidazole) were administered in drinking water for 1 week to disrupt the intestinal microbiota prior to MTX treatment. Antibiotics were refreshed twice weekly.

Fecal microbiota transplantation
In all experiments, fecal microbiota transplantation (FMT) was performed, prepared using fecal pellets collected during acclimatization. To eliminate individual variation in the FMT product, fecal pellets were pooled from all rats. Fecal pellets were collected from rats by placing them into individual cages (without bedding) cleaned with 70% ethanol. Immediately following defecation, fecal pellets were collected with sterile forceps and placed directly into de-oxygenated amies transport solution with 10% glycerol at a concentration of 1 pellet/ml. Samples were then homogenized by vortexing at maximum speed for 5 min until no visible pellets were present. The fecal homogenate was then aliquoted under anaerobic conditions into individual (daily) portions (7 ml) and stored at -80oC 74 Chapter 4 until the day of administration. An additional FMT was also prepared for quality control (culture analysis of CFUs). On the day of administration, an aliquot of fecal homogenate was thawed at room temperature under anaerobic conditions before being centrifuged at 13,000 g for 10 min.
The supernatant was removed and the pellet resuspended in 7 ml of sterile, de-oxygenated 1X PBS. The sample was then homogenized by vortexing for 1 min and allowed to settle by gravity for 15 min. The supernatant was then administered to rats between 8-10am via oral gavage (500 µl).
Control rats received a volume equivalent dose of 1X PBS. FMT was evaluated as: 1) an adjunct to MTX, administered on day 1-7 after MTX, and 2) a restorative approach after antibiotic prophylaxis, administered on days -3, -2 and -1 prior to MTX (administered on day 0). Specific schedules for each experiment are shown in figure panels (Figures 3C and 5F).

Microbiome analysis
The fecal microbiota composition was assessed using 16S rRNA gene sequencing as previously described 19. Briefly, DNA was extracted using the double bead-beater procedure and the QIAamp DNA Stool Minikit guidelines. Isolated DNA was quantified using the NanoDrop UV Visible Light Spectrophotometer and the V3-V4 region amplified using polymerase chain reaction (PCR). Each PCR reaction contained 1 µl of 10 µM 341f forward primer, (V3F: 5'aatgatacggcgaccaccgagatct 3') 25 µl Phire HS II Master Mix, 22 µl DNase free water, 1 µl of 10 µM 806r barcoded reverse primer and 1 µl DNA template (100 ng/µl)25. Amplification was confirmed using gel electrophoresis. Size selection and fragment removal was performed using AMPure XP beads as per manufacturer's guidelines, before the final PCR products were normalized to 2mM and pooled to form a single library which was stored at -4oC until sequencing. Sequencing was performed using the MiSeq Benchtop Next Generation Sequencer (Illumina). The paired-end sequencing data received from Illumina software were processed and analyzed using the Qiagen CLC Genomics Workbench v12.0 (kindly provided by Qiagen to HRW). De novo OTU-picking was performed without chimera filtering using Greengenes (RRID:SCR_002830 version 13.5) as a reference database.

Statistics
All data (excluding 16S) were analyzed using GraphPad Prism v8.0 (RRID:SCR_002798). Data were first analyzed for normality using the D'Agostino-Pearson Normality test. Parametric data were subsequently analyzed using one-or two-way analysis of variance (ANOVA) with post-hoc correction for multiple comparisons. For repeated measures, mixed model with Geisser's greenhouse correction was used to identify significant differences. When normality was not confirmed, a Kruskal-Wallis with Dunn's multiple comparison was used to identify significance. Area under the curve (AUC) was used to reduce repeated data sets and simultaneously assess the duration and  Figure 1C).
MTX also impaired weight gain (ΔBW), with significant differences in µBW from day 2-10 (P<0.0001, Figure 1D). This was accompanied by anorexia, with food intake significantly decreased in MTXtreated rats on day 4 (P=0.008, Figure 1E). No changes in water intake were identified (data not shown).
Analysis of the fecal microbiota revealed no significant change in alpha diversity (Shannon's index) following MTX (Figure 1F). Compositionally however, there was a profound increase in the relative abundance of Proteobacteria taxa, especially Enterobacteriales ( Figure 1G, Figure S1), with paralleled decreases in Bifidobacteriales (Figure 1H, Figure S1).  (Figure 2F). It did however increase food intake during peak mucosal injury (3.2 ± 0.7 g/day vs 8.5 ± 1.7 g/day, P=0.007, Figure 2G). There were no significant correlations between peak MBI (defined by citrulline at day 4) and microbial richness or diversity (Figure 2H/I). A significant but mild correlation was identified between citrulline and the relative abundance of Proteobacteria at day 4 (R2=0.24, P=0.002, Figure 2J).

Antibiotic-induced microbial disruption exacerbates MTX-induced gastrointestinal toxicity by impairing mucosal recovery
To further investigate the impact of the microbiota on MTX-induced MBI, we evaluated the effect of antibiotic-induced microbial disruption (ABX) on MTX toxicity in a high-dose model (Figure 3C). MTX was administered at 60 mg/kg (MTX60) to induce a more severe MBI phenotype. ABX, alone or in 78 Chapter 4 combination with MTX60, induced a profound decrease in microbial diversity, while MTX60 alone caused no significant change in alpha diversity ( Figure 3A). Irrespective of diversity, both MTX60 and ABX+MTX60 caused changes in the phylogenic composition of the microbiota characterized by expansion of Proteobacteria (Figure 3B). ABX significantly increased MTX60-induced mortality (P=0.0045; Figure 4A). While acute MBI, indicated by plasma citrulline, was comparable between MTX60 and ABX+MTX60, the ability for the mucosa to recover was significantly impaired in ABX treated rats ( Figure 4B). This resulted in a significantly increased plasma citrulline AUC in ABX+MTX60 treated rats compared to MTX60 alone (583±21 vs 698±11, P<0.0001, Figure 4C). This was accompanied by more profound weight loss in rats treated with ABX+MTX60 compared MTX60 alone ( Figure 4D) irrespective of food intake ( Figure 4E). Histopathological analysis of the jejunum revealed more pronounced villous atrophy in ABX+MTX60 rats compared to all other groups (P<0.0001), including MTX60 (P=0.03; Figure 4F).

FMT mitigates detrimental effects of ABX on MTX-induced diarrhea with its efficacy dictated by colonisation of Muribaculaceae (S24-7)
Disruption of the microbiome with ABX induced a more pronounced toxicity phenotype. As such, we next aimed to restore the microbiome after ABX using autologous FMT. Given high rates of mortality seen in the MTX60 (high-dose) model, MTX was administered as intravenously at 45 mg/kg (MTX45). At this dose, ABX had no significant impact on mortality ( Figure 5A). However, consistent with the high-dose MTX60 model, ABX slowed mucosal recovery, with a significantly lower citrulline value seen at day 8 compared to MTX45 only (P=0.02, Figure 5B). FMT accelerated mucosal recovery, increasing citrulline at day 8 compared to ABX+MTX45 (P=0.04, Figure 5B). These subtle differences did not affect the AUC for citrulline ( Figure 5C). Clinically, weight loss and food intake were not affected by FMT, with significant differences only identified between MTX45 and ABX+MTX45 groups ( P<0.05, Figure 5D/E).

Discussion
Disruption of the intestinal microbiome is increasingly associated with unfavorable treatment outcomes in people with cancer, reported to dictate both the efficacy and toxicity of chemotherapy as well as the risk of relapse 1 . Here we report the first phenotypic analysis of fecal microbiome transplantation (FMT) delivered after cytotoxic chemotherapy (MTX) and antibiotics, identifying Muribaculaceae (S24-7) as a key regulator of mucosal injury and repair. Mucosal barrier injury (mucositis) in the gastrointestinal tract is a major oncological obstacle due to its acute symptomology and systemic consequences. While a strong body of data show microbial injury after chemotherapy drugs, including MTX 26 , it remains hotly debated if these changes are directly mediated by chemotherapy, or simply a form of collateral damage resulting from mucosal injury 27 . Our data support the latter, with FMT delivered after MTX failing to appreciably impact mucosal injury 83 Autologous fecal microbiota transplantation during chemotherapy-induced diarrhea 4 (defined by citrulline) and its clinical manifestations (weight loss, anorexia), despite maintaining a more stable microbiome composition defined by 16S sequencing. The lack of FMT efficacy could also be explained by the hypothesis that MTX-induced mucosal injury creates a luminal environment that is inhospitable to microbes introduced via FMT, or simply the degree of mucosal injury is too profound for the donated microbes to influence. Each interpretation is based on the concept that microbial disruption occurs secondary to mucosal injury; a hypothesis that is consistent with existing data showing chemotherapeutic drugs, including 5-FU and irinotecan, do not directly interfere with the microbiome 28 .
We do however show that disruption of the host microbiome with antibiotics impairs mucosal recovery and exacerbates diarrhea, with FMT having a beneficial impact. This supports the hypothesis that while microbial disruption occurs after mucosal injury, dysbiosis amplifies mucosal injury and establishes self-perpetuating feedback loops that drive clinical manifestations. This strongly supports interventions targeting the microbiome to minimise the depth and duration of MBI, accelerating recovery and controlling clinical symptoms/consequences.
Whilst the dynamics of microbial injury after chemotherapy have been the topic of intense investigation, their correlation with treatment outcomes has been variable. In contrast, the composition of the microbiome at baseline (i.e. at the time of chemotherapy) is increasingly linked to how an individual responds [29][30][31] . This has been most elegantly demonstrated in the case of immunotherapy-induced colitis [32][33][34] , however a growing body of evidence anecdotally suggests that similar mechanisms may dictate an individual's risk of mucosal injury caused by chemotherapy 30. In a recent systemic review of predictors of mucosal injury coordinated by the Multinational Association for Supportive Care in Cancer 11 , antibiotic use in the month leading up to cancer therapy was identified as a modulator of mucosal injury 35,36 . Consistent with these findings, antibiotic administration prior to MTX treatment caused a profound increase in treatment-related mortality (TMR) and diarrhea. Underlying this observation was severe mucosal injury, with antibiotics exacerbating villous atrophy and impairing mucosal recovery. In fact, antibiotic treated rats failed to recover from MTX treatment, with hypocitrullinemia evident at day 10. Given the large relative expansion in Proteobacteria, it may also be the case that expansion of enteric pathogens serve to amplify mucosal injury through interaction with Toll-like receptor 4; an already defined mediator of chemotherapy-induced diarrhea 24,37 . Further research is required to identify the microbial factors that control mucosa recovery in the context of chemotherapy treatment, with short chain fatty acids (SCFA) a promising candidate due to their ability to simultaneously promote luminal acidification (thus controlling pathogen expansion) 38 as well as mucosal proliferation and immunology 39 . Of note, SCFAs have been documented to be deficient after chemotherapy and exert protective effects in chemotherapy-and radiation-induced intestinal injury 40 .

Chapter 4
The detrimental effects we observed when antibiotics were administered prior to MTX combined with the growing number of studies associating increased morbidity, mortality and relapse with antibiotic use 41,42 . This has been explored with the greatest detail in haematopoietic stem cell transplant (HSCT) recipients due to the intensity of chemotherapy they receive and current reliance on antibiotics. Importantly, comparable to our findings, antibiotics and microbial injury have been reported to increase the risk and severity of intestinal graft-versus-host disease [43][44][45] . Mechanistically, this is thought to be caused by increased intestinal permeability and mucin degradation -two welldefined aspects of chemotherapy-induced mucotoxicity and diarrhea 5 -and, as such, may also be mechanisms responsible for our findings.
Microbial injury caused by antibiotic prophylaxis is well documented 41,42 . While the antibiotic schedule used in our study was not intended to reflect clinical practice, when viewed alongside the growing body of literature, our findings raise important implications for antibiotic stewardship.
Specifically, they highlight the need to select appropriate prophylactic anti-microbial drugs that do not damage the microbiome 46 . Ciprofloxacin is reported to have mild effects on the microbiome and may therefore be a potential avenue to restrict antibiotics whilst still providing appropriate infection control 47 . However, in some cases, residual microbial injury may be caused by previous antibiotic exposure, cytotoxic treatment, co-morbidities or other medications. As such, restoring or strengthening the microbiome prior to treatment represents a potentially important supportive care strategy.
We showed that FMT administered after antibiotics but before MTX was able to restore typical dynamics of mucosal recovery, reiterating our findings that suggest a causal contribution of the microbiome to mucosal healing. However, of particular interest was the clear clustering in response to FMT intervention. FMT uptake is documented to be variable both preclinically and clinically, reflecting donor selection and the preparation/status of the recipient 48 . Our data demonstrated a low response rate to FMT intervention, with only 33% of rats responding to FMT. Subgroup analysis suggest response to FMT intervention is mediated through the durability of the FMT, specifically colonization and expansion of Muribaculaceae (S24-7). Muribaculaceae (S24-7) is a typical rodent bacterial family belonging to the Bacteroidetes phylum already documented to regulate FMT efficacy in rodents 49 . It is recognized to increase in abundance following prebiotic intervention with vitamin B2 18 and selenium 50 , both of which have documented benefits in the prevention of mucosal injury caused by chemotherapy 18,51 . Translating this new knowledge is somewhat challenging as Muribaculaceae is absent in humans. However, it is considered to be a core functional microbe of the rodent gut, found in high abundance under healthy circumstances 52 . As such, our findings may simply reflect that restoration of Muribaculaceae is an indicator of successful FMT uptake, and we hypothesise that colonization of other functional microbes relevant to the human gut (e.g. Fecalibacterium or Blautia) would likely dictate clinical FMT efficacy. Importantly, it cannot be 4 excluded that the abundance of Muribaculaceae were comparable between responders and nonresponders prior to FMT, and such, pre-FMT microbiome composition may also be an important factor dictating its efficacy.
While our data strongly support adjunctive FMT strategies to improve treatment outcomes, this study is not without its limitations. It must be noted that due to the mild nature of microbial injury seen at 45 mg/kg MTX, a higher dose was used for the subsequent study (MTX60). This, in combination with antibiotics, induced unacceptable levels of mortality and as such, we reverted to 45 mg/kg for all remaining experiments. Our observations remained largely comparable across both doses, with the exception of citrulline and body weight. As such, we cannot definitively conclude that the exacerbated diarrhea profile is the result of impaired mucosal recovery in the lower dose model. The heightened diarrhea may simply reflect antibiotic-induced dysbiosis. However, we did not see substantial diarrhea in our rats while undergoing antibiotic treatment (6/27 developed grade 1 diarrhea for average of 1.16 days) suggesting it is the combination of antibiotics and MTX that is responsible the exacerbated diarrhea phenotype driven by impaired mucosal recovery.
Nonetheless, given the multifactorial nature of chemotherapy-induced diarrhea, it is most likely that both the dysbiosis and impaired mucosal recovery caused by antibiotics are driving the diarrhea profile.
In conclusion, our data underscore the importance of microbial stability during chemotherapy treatment and provide a rationale to re-evaluate the administration of microbial interventions for the control of mucosal injury. We suggest microbial interventions be used prior to chemotherapy treatment to repopulate or shift the composition of the host microbiome, thus offering a new strategy to improve treatment outcomes. Alternatively, a strong rationale also exists for postchemotherapy intervention aimed at promoting restoration of the mucosal barrier through microbial intervention. This approach would minimize the duration of mucosal injury, decreasing the intensity of symptoms and opportunity for translocation events.