Host Microbiome Dictates Infection Risk after Chemotherapy Revealing New Avenues for Antibiotic Stewardship in Oncology

Background: Mucosal barrier injury (MBI) is a recognized risk factor for blood stream infection (BSI) in people undergoing chemotherapy, permitting expansion and translocation of enteric pathobionts; a phenomenon now clinically referred to as MBI-associated laboratory-conrmed BSI. Although recognized to originate from endogenous gastrointestinal sources, MBI-associated BSI continue to be treated with antibiotics; a counterintuitive approach that may increase infection risk by depletion of the host microbiome. While this point has been argued in numerous clinical studies and opinion pieces, there are few data describing mechanistic strategies to decrease MBI-associated BSI, reecting a lack of translationally robust preclinical models. Here, we report on a new translational model of MBI caused by the chemotherapeutic drug, melphalan. We aimed to identify candidate pathways to strengthen the mucosal barrier and prevent infection, prompting new antibiotic stewardship initiatives. Results: Melphalan caused dose-dependent, systemic toxicity characterized by severe neutropenia, self-limiting intestinal injury, inammation and biphasic fever, all of which were clinically and molecularly consistent with the dynamics of melphalan conditioning. The fecal microbiome following melphalan was depleted in richness and commensal taxa, impairing colonization resistance and the microbial metabolome and prompting expansion of enteric pathogens. Breakdown of the intestinal barrier was initiated by melphalan and exacerbated by mucotoxic bile acids. Conclusions: MBI-associated BSI is simply a form of collateral damage resulting from breakdown of the gastrointestinal microenvironment and self-perpetuating injury. Efforts to intervene early in these sequelae are therefore of great clinical signicance to restrict antibiotic use and mitigate their detrimental consequences on treatment outcomes and antibiotic resistance. Our data support interventions targeting the host microbiome and metabolome due to its ubiquitous control of inammation, mucosal injury, bile acid transformation; all of which contribute to infection risk in cancer.

BSI in HSCT has long been considered the result of exogenous introduction of nosocomial pathogens via the central line or venous catheter. However, in 2016 the Centre for Disease Control re-classi ed BSI in HSCT recipients, identifying that not all occur as a result of contamination and instead originate in the gut before translocating across the mucosal barrier (mucosal barrier injury-laboratory con rmed BSI (MBI-LCBI)) (9). MBI-LCBI are now reported to account for approximately 50% of all infections in HSCT recipients, decreasing overall survival and increasing treatment-related mortality (10).
While major progress has been made in preventing central line associated BSI, there are few data describing mechanistic strategies to decrease MBI-LCBIs in HSCT recipients. A major limitation in identify new strategies to prevent MBI-LCBI is the lack of translationally-relevant models in which to study virulence factors (11). The current study therefore aimed to 1) develop and characterize a new preclinical model of HSCT-associated MBI using the conditioning and highly mucotoxic chemotherapeutic drug, melphalan, and 2) identify candidate pathways for interventions targeting the mucosal barrier and gastrointestinal microenvironment.

Materials And Methods
This study is reported using the ARRIVE guidelines for the accurate and reproducible reporting of animal research and the STAR (STAR_RRID:SCR_015899) methods for structured, transparent and accessible reporting. Central Animal Facility) at the University Medical Center Groningen. Rats were housed under 12 h light/dark cycles with ad libitum access to autoclaved AIN93G rodent chow 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.
The study was powered using the primary outcome measure of plasma citrulline, a biomarker of enterocyte mass, based on dynamics con rmed in our model of methotrexate-induced MBI. Our power calculation assumed an alpha of 0.05 and beta of 0.9, resulting in N = 8 animals per group. All analyses (clinical assessments and downstream analyses) were performed in a blinded fashion with all samples identi ed by randomized animal numbers.

Study design
Animal experiments were performed as an initial dose nding study (4-8 mg/kg melphalan, sourced from the Pharmachemie Holding, B.V. The Netherlands), in which body weight, diarrhea, food/water intake and plasma citrulline were used to determine the optimal dose of melphalan in which moderate, self-limiting toxicity was achieved (based on citrulline dynamics). A dose validation study was then performed using the optimized dose of melphalan (5 mg/kg) for which all endpoint analyses were performed (N = 24 vehicle, N = 24 melphalan). In both cases, male Wistar rats weighing between 80-100 g were allowed to acclimatize for 10 days after arriving at the CDP. Rats were then randomized to receive 4 (N = 6), 5 (N = 6 dose nding, N = 24 dose validation), 6 (N = 6) or 8 mg/kg (N = 3) melphalan (10 mg/ml), or a volume equivalent dose of vehicle solution (0.9% NaCl, N = 3 dose nding, N = 24 dose validation). All intravenous injections were performed via the penile vein under anesthetic (3% iso urane) on day 0.

Melphalan treatment and tissue preparation
Rats received a single intravenous injection (via the penile vein) of melphalan (10 mg/ml) or vehicle solution ( Figure S1). For initial dose nding studies, all rats were sacri ced via iso urane anesthesia, cardiac puncture and cervical dislocation at day 10 (recovery). For dose validation studies (5 mg/kg), groups of rats were followed longitudinally until day 10 (N = 8 / group). Subsets of rats were sacri ced on day 4 (N = 4 / group) and day 7 (N = 4 / group) for exploratory investigation. The primary outcome of the study was plasma citrulline, a validated biomarker of mucosal injury (12)(13)(14)(15)(16).

Clinical toxicity assessment
Systemic toxicity was assessed using clinical parameters of body weight, food intake and water intake, as well as routine welfare indicators (reluctance to move, posture, coat condition). Rats were weighed daily, and water/food intake monitored by manual weighing of chow and water bottles. Gut toxicity was determined by diarrhea severity assessed using a validated, semi-quantitative scoring system (0-3) where 0 = no diarrhea, 1 = mild diarrhea with soft stools and perianal staining, 2 = moderate diarrhea with loose stools and perianal staining of fur, 3 = severe diarrhea with watery stools with or without mucous, and fur staining incorporating the hind legs. This assessment scale has been used in both rats (17) and mice(18) treated with chemotherapy.

Plasma citrulline
Plasma citrulline is an indicator of small intestinal enterocyte mass, and a validated biomarker of small intestinal toxicity (12,13). Repeated blood samples (75 µl) were collected from the tail vein into heparinized hematocrit capillary tubes (Fisher Scienti c, Landsmeer, Netherlands) on day 0, 1, 2, 3, 4, 6, 8 and 10 (for dose-nding studies) and days 0, 2, 4, 6, 7, 8 and 10 for dose validation studies (between 8:00 and 10:00 am). Citrulline was determined in 30 µl of plasma (isolated from whole blood via centrifugation at 4000 g for 10 min) using automated ion exchange column chromatography as previously described. 10 Differential blood analysis Whole blood samples (200 µl) were collected into MiniCollect®EDTA tubes (Greiner Bio-One B.V. Alphen aan den Rijn, Netherlands) from the tail vein of all rats at the time of treatment (day 0) and time of termination. Differential morphological analyses were performed using a Hematology Analyzer (Beckman Coulter, Dept Hematology) at the University Medical Centre Groningen.

Body temperature
Core body temperature was used as an indicator of fever and was assessed daily using the Plexx B.V. DAS-7007R handheld reader and IPT programmable transponders (Plexx B.V. Elst, Netherlands).
Transponders were inserted subcutaneously under mild 2% iso urane anesthesia on day − 4 ( Figure S1). Body temperature was assessed once daily between 8:00-10:00 am. Average values from day − 4 to -1 were considered as baseline body temperature.

Intestinal barrier function
Intestinal barrier function was assessed using 4 kDa uorescein isothiocyanate (FITC)-dextran (Sigma-Aldrich, Zwijndrecht, Netherlands) as previously described (19). Brie y, FITC-dextran was prepared at a concentration of 120 mg/ml in sterile 1 X PBS (pH 7.4) and kept on ice, protected from light until administration. FITC-dextran (500 mg/kg) was administered to rats via oral gavage 3 hours prior to termination. Administration was staggered in 15 min intervals to account for the time taken to terminate each animal and collect biospecimens. FITC-dextran concentrations were determined in plasma relative to a standard curve (0.0001-10 µg/ml) using the BioTek Synergy Mx Microplate Reader and Gen5 software (Gen5, RRID:SCR_017317).

Histopathological analysis
Routine hematoxylin and eosin (H&E) staining was performed on jejunum, ileum and colon segments to evaluate intestinal architecture. Brie y, intestinal resections were drop xed in 10% neutral buffered formalin, processed and embedded into para n 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. in which10 well oriented crypts/villi were measured per slide and an average calculated per animal. Intestinal in ammation was determined by quanti cation of in ltrating polymorphic nuclear cells (PMNC) in the jejunum and colon, which were assessed in four regions (500 × 500 µm). Individual counts were averaged over the four regions to generate a single mean value for each specimen.
Blood culture At termination, whole blood was collected via cardiac puncture and 2 ml was immediately dispensed into BD BACTEC™ PEDS Plus™/F plastic blood culture bottles (BD, Vianen, Netherlands). Bottles were stored at room temperature before being transported to the Medical Microbiology laboratory at the University Medical Centre. Culture bottles were incubated in the BD BACTEC™ Automated Blood Culture System for 96 hours. Positive cultures were identi ed using Matrix-Assisted Laser Desorption/Ionization-Time Of Flight (MALDI-TOF) mass spectrometry (MS).

Microbiome analysis
Microbiota analysis was performed on repeated fecal samples collected on day 0, 4, 7 and 10 from N = 8 animals per group. Samples were collected aseptically and stored at -20 o C. 16S rRNA Illumina sequencing was used to determine the composition of the fecal microbiota. All experimental procedures (DNA extraction, PCR ampli cation, PCR product quanti cation and pooling, puri cation, library preparation and sequencing) were performed by Novogene. For full methodology, please see 'Supplementary methods'.

Short chain fatty acid analysis
Short chain fatty acids (SCFAs) were analysed in cecal contents collected at termination. Brie y, the cecum was removed and the apex cut using sterile scissors. The contents were squeezed directly into a sterile 2 ml Eppendorf tube, snap frozen in liquid nitrogen and stored at -80 o C. SCFAs were analysed using gas chromatography adapted from Moreau et al., (2003) (20) with minor adjustments detailed in supplementary methods. Data processing was carried out with MassHunter Workstation Software (MassHunter, Agilent Technologies, Middelburg, Netherlands).

Bile acids and ileal gene expression
Bile acid pro les were determined in plasma isolated from whole blood collected at termination (by cardiac puncture). Plasma bile acids were quanti ed using an Ultra High Performance Liquid Chromatography system (SHIMADZU, Kyoto, Japan), coupled to a SCIEX QTRAP 4500 MD triple quadruple mass spectrometer (SCIEX, Framingham, MA, USA) (UHPLC-MS/MS) as previously described (21,22). Gene expression analysis was performed in on distal ileal segments which were stored in RNAlater at -20 o C. Total RNA was isolated using TRI-reagent (Sigma, St. Louis, MO, USA) and quanti ed by NanoDrop (NanoDrop Technologies, Wilmington, DE, USA). cDNA synthesis was performed from 1 µg of total RNA. Primers were designed with Primer-BLAST(RRID:SCR_003095) and optimized for use with SYBR Green Master Mix (Roche Diagnostics, Mannheim, Germany) (maximum product size 150 nucleotides). Real-time qPCR analysis was performed on a StepOnePlus™ Real-Time PCR System (Applied Biosystems, Thermo Fisher, Darmstadt, Germany). Gene expression levels were normalized to Ubc. Results were quanti ed using the comparative Ct method.

Quanti cation and statistical analyses
All data was analysed using GraphPad Prism version 8.0 (RRID:SCR_002798) with the exception of microbiota (16S rRNA) data. Continuous data were analysed for normality using the D'Agostino and Pearson test and Kolmogorov-Smirnov test. When normality was con rmed, a two-way analysis of variance (ANOVA) or mixed model (when data points were missing) was used to identify statistically signi cant differences. When normality was not con rmed, a Kruskal-Wallis with Dunn's correction for multiple comparisons was used. Where possible, paired or repeated measures were prioritized and indicated using line graphs. In cases where this was not possible, biospecimens collected at termination were used and indicated by grouped data (column/bar graphs). In all cases, P < 0.05 was considered statistically signi cant.

Results
This study has been reported in accordance with the ARRIVE guidelines for transparent preclinical reporting. Baseline characteristics of the dose validation study were compared between control and melphalan groups with no signi cant differences identi ed (Table S1).
In addition to villus blunting and changes in crypt depth, melphalan caused severe architectural derangement in the small and large intestine, characterized by villus fusion and crypt ablation (Fig. 3H), with evidence of gross colonic pathology seen at day 7 (Fig. 3Ii). Histological analysis of a macroscopically evident colonic perforation showed complete destruction of the mucosa, submucosa and muscularis layer (Fig. 3Iii), with pseudomembrane development and visible in ammatory in ltrate (Fig. 3Iiii). In ltration of polymorphic nuclear cells (PMNC) was also elevated in the jejunum of melphalan treated rats (Fig. 4).
Melphalan-induced microbiota disruption is characterized by pathogen expansion and SCFA de cits 16S rRNA-gene analysis was performed in fecal samples collected longitudinally (control and melphalan, N = 8 per group) on day 0, 4, 7 and 10. Melphalan induced signi cant disruption of the fecal microbiota that failed to recover during the 10-day experimental period (Fig. 5). Compositionally, there was a shift towards a Firmicute-dominated microbiota, with expansion of pathogenic taxa largely belonging to the Proteobacteria phylum (Fig. 5A) which was not observed in control animals. This was accompanied by a decrease in the number of operational taxonomic units (OTUs), an indicator of microbial richness, which was signi cantly decreased 7 days post-melphalan compared to controls (P = 0.009, Fig. 5B). There were no signi cant changes in alpha diversity parameters, including Chao1, Simpson index and Shannon's index in both control and melphalan-treated rats (data not shown). Principle component analyses showed no change in beta diversity of control animals throughout the experimental period (Fig. 5C). In contrast, signi cant changes were evident at all time points in melphalan-treated animals compared to baseline (P < 0.0001, Fig. 5D). These changes were signi cant compared to controls at day 4 (P = 0.021) and day 7 (P = 0.015).
While beta diversity and richness were most profoundly affected at day 7, expansion of pathogens was most signi cant at day 4 post-melphalan (Fig. 5E) Orally-administered FITC-dextran was used to assess epithelial barrier permeability at all terminal time points. While no signi cant changes were observed, the highest concentration of plasma FITC-dextran was observed on day 7 post-melphalan treatment ( Figure S2). At this time point we also identi ed one case of positive blood culture (Escherichia coli). All other blood cultures were negative.
Melphalan-induced ileal injury results in bile acid malabsorption and decreased plasma primary to secondary bile acid ratios Plasma bile acids pro les were analyzed at termination on day 4, 7 and 10. Total plasma bile acid concentrations and primary/secondary ratios were decreased in melphalan treated animals on day 4 and 7 but this did not reach statistical signi cance (Fig. 7A-B). Upon analyzing individual bile acid species, the primary conjugated bile acid, taurocholic acid (TCA), was decreased at all evaluated time points postmelphalan treatment (Fig. 7C, P = 0.006, P = 0.0004, P = 0.0001, respectively). The rodent speci c primary bile acid tauro-alpha-muricholic acid (T-α-MCA) was also decreased at all time points (Fig. 7D, P = 0.01, P = 0.003, P = 0.009, respectively), while beta-muricholic acid was unchanged (Fig. 7E). Secondary bile acid deoxycholic acid (DCA) was unchanged by melphalan treatment in its unconjugated form (supplementary data) but was higher as taurine conjugate at day 4 (P = 0.001, Fig. 7F). All other bile acid concentrations are listed in Table S2.
Ileal gene expression analysis was performed at day 4 for target genes involved in bile acid absorption and transport (Fig. 7G). Ileal expression of t-Asbt, Fabp6 and Ostβ was signi cantly lower in melphalan treatment compared to controls, suggesting lower ileal absorption of bile acids (P = 0.03). In contrast, farnesoid X receptor (FXR) target genes Shp and Fgf15 were unchanged and increased (P = 0.03), respectively.

Discussion
The toxic properties of melphalan currently prohibit dose escalations and require intensive antimicrobial prophylaxis to prevent infection (23,24). Together these prevent optimal dosing with curative intent, exacerbate acute and chronic toxicities and contribute to the growing global threat of antimicrobial resistance. Here, we highlight a critical advance in not only modelling melphalan-induced toxicity, but also the mechanisms that drive acute epithelial injury and secondary infectious complications. To our knowledge, this is the only preclinical model that recapitulates the clinically relevant interplay of events including mucosal injury, neutropenia, microbial disruption and fever caused by melphalan, and thus represents a translationally robust approach for the study of HSCT-associated toxicities and evaluation of new interventions.
Intravenous melphalan (5 mg/kg) caused moderate, self-limiting toxicity characterized by a clinically relevant phenotype. Critically, liquid biomarkers (citrulline), microbial dynamics and toxic events mirror previously published clinical human datasets(25) and correlate with gold-standard histopathological criteria used in existing models of gastrointestinal injury (18,26,27). This reinforces the use of plasma citrulline as a minimally invasive liquid biomarker of mucosal barrier injury and its clinical consequences, as well as the applicability of this model for rapid interventional pre-screening.
Importantly, in addition to MBI, our model was characterized by a biphasic fever. An initial spike in body temperature as observed at day + 2, where a rapid loss in citrulline was also observed. This therefore suggests that primary fever responses are re ects of intense mucosal injury, and therefore reinforces the previously described concept of febrile mucositis -the presentation of fever without infectious etiology(13,28). Methods of differentiating febrile mucositis from infectious febrile events would therefore enable more targeted antibiotic use in HSCT recipients. In addition to febrile mucositis, we also effectively model febrile neutropenia, demonstrating expansion of enteric pathogens prior to clinical indicators, aligning with numerous clinical reports (29), (30,31). This is consistent with the proposed sequence of events leading to MBI-LCBI, and suggest that identifying pathogen expansion prior to translocation could facilitate targeted antibiotic use or prompt novel interventions targeting the mucosal barrier.
Characterization of our model also identi ed candidate pathways for intervention consistent with the molecular mechanisms of enteric pathogen expansion and translocation. Of interest, no profound changes were identi ed in diversity and richness indices, suggesting that impaired of colonization resistance (presence of diverse microbiota which out-competes enteric pathobionts) is more likely driven by antibiotics, rather than chemotherapy. However, decreases in some commensals were observed, including S24-7 resulting profound changes in the microbial metabolome indicated by short chain fatty acid (SCFA) loss. SCFA serve dual purposes in the gastrointestinal microenvironment, promoting mucosal barrier integrity and maintaining luminal acidi cation. Maintaining a low luminal pH is critical in controlling enteric pathobionts, in particular Enterobacteriaceae (32,33), with luminal pH regulating oxygen utilization and its expansive capacity. As such, our data support interventions aimed at maintaining or supplementing SCFA after melphalan and suggest they may have dual bene ts for both luminal and mucosal compartments (Fig S3).
Maintaining the mucosal barrier after cytotoxic chemotherapy is a major challenge due to the overlapping mechanisms that govern cell death of tumor cells and intestinal stem cells (34). Identifying unique secondary mechanisms that perpetuate mucosal injury are therefore critical in establishing novel techniques that minimize the duration and intensity of injury. While modulating pro-in ammatory and immune pathways for the prevention of mucosal injury has been investigated thoroughly, it is undermined by the increasingly sophisticated understanding of the innate immune system's contribution to chemoe cacy (34). Here, we provide the rst evidence of bile acid malabsorption after melphalan and propose increased production of mucotoxic secondary bile acids may exacerbate MBI.
Bile acid malabsorption (BAM) has been widely implicated in the development of diarrhea, including diarrhea caused by the tyrosine kinase inhibitor neratinib (35) and in patients with GI-GvHD(36), increasing colonic secretion of water and electrolytes and the induction of propagated contractions (37). We identi ed changes in plasma bile acids and ileal gene expression following melphalan, indicative of BAM. BAM increases the pool of luminal bile acids, with our data supporting microbiota-dependent production of secondary bile acids (38). Importantly, we observed a relative increase of fecal bacteria from the Firmicutes phylum upon melphalan treatment. It is known that Firmicutes, especially certain Clostridia species, are able to 7-alpha-hydroxylate bile acids thus leading to an increased secondary pool as observed in our study (39,40). Secondary bile acids are mucotoxic and pro-in ammatory (41), and as such, their production may serve to amplify mucosal injury. The clinical relevance of this nding is particularly compelling with BAM reported in severe cases of GvHD-related diarrhea and even bowel perforation (36,42,43). Bile acid sequestrants (colesevelam) may offer bene ts in both the prevention of acute diarrhea (caused by motility/secretory mechanisms) as well as the prevention of MBI-LCBI. Coleveselam has already been deemed safe in oncology cohorts and has demonstrated clinical e cacy in treatment neratinib-induced diarrhea (35,44).

Conclusions
Our results reinforce the multi-factorial nature of melphalan toxicity and its symptomology, with death of intestinal stem cells an important initiating factor in establishing various cascades of morbidity. This involves neutrophil recruitment, in ammation, microbial changes and bile acid disruption, each of which are likely to contribute to bacterial translocation and subsequent BSI. Critically, we suggest that this sequence of events only stops when stem cells repopulate the intestinal lining, renewing niches to support the host's commensal microbiota. Replenished SCFA production and normalized enterohepatic circulation then aid in re-establishing a functional barrier. These mechanistic insights highlight that potentially lethal systemic reactions such as fever are simply a form of collateral damage resulting from breakdown of the intestinal mucosal/luminal environments. Efforts to intervene early in these sequelae are therefore of great clinical signi cance to restrict antibiotic use and therefore mitigate their detrimental consequences on treatment outcome and antibiotic resistance. Author contributions: HRW: study conceptualization, ethical approvals, lead laboratory animal work including husbandry, clinical assessments and biospecimen collection, histopathological analysis, data analysis/synthesis and manuscript preparation (lead).

Declarations
CdM: study conceptualization, assisted in laboratory animal work, data analysis and manuscript preparation.
IvdP: bile acid analysis, RT-PCR for ileal gene expression, expert interpretation of bile acid datasets and written description in the manuscript.
RH: animal technician responsible for melphalan administration and sacri ce.
Also contributed to the preparation of the manuscript, with particular emphasis on microbiome elements.
WJET: study conceptualisation, ethical approvals, supervision and infrastructure/funding. Also contributed to the preparation of the manuscript, providing clinical expertise and viewpoint.
NB: study conceptualisation, ethical approvals, supervision and infrastructure/funding. Also contributed to the preparation of the manuscript, providing clinical expertise and viewpoint.
All authors read and approved the nal manuscript.

Acknowledgements:
We would like to thank Hermi Kingma, Martijn Koehorst, Jennifer van der Krogt for their analytical assistance, Mirjam Koster for performing all histological staining, Martijn Geutjes, Alex Ottema and Geetjes Kampinga for assistance in clinical microbiology, Roelof Bekkema for coordinating whole blood analyses and Richard Logan for his pathological advice. We would also like to extend our gratitude to all employees at the University Medical Centre Groningen Central Animal Facility.

Author information:
Our team is highly interdisciplinary, united by the common goal of improving infection control in people with cancer and restricting antibiotic use in the community. The study is led by Dr Hannah Wardill, an early career researcher who has pioneered the eld of oncogastroenterology, de ning dynamics of microbial injury following cancer therapy and developing translational models of gastrointestinal complications caused by chemotherapy. Charlotte de Mooij (MD) and Ana Rita Da Silva Ferreira are active PhD students, with a core interest in developing the next generation of microbial-based therapies to better support people with cancer. Together, they represent a dynamic team with complementary expertise in microbiology, models of mucosal injury and clinical haematology. Importantly, they are supported by a panel of exemplar clinician-scientists, guiding preclinical research activities to address the unmet needs of people with blood cancer. Both Nicole Blijlevens (senior haematology consultant) and Wim Tissing (paediatric oncologist) are actively involved in antibiotic stewardship initiatives in the Netherlands, with Nicole the haematology lead for the Dutch stewardship Committee (SWAB). She has dedicated the bulk of her career to understanding the mechanisms of mucosal injury and its role in fever and infection risk, lobbying for restricted antibiotic use and novel methods to support the microbiome in people with cancer. This project represents a major achievement for our group, developing an accessible platform that re ects the well-de ned dynamics of the clinical environment and identi es new targets for infection control that avoid antibiotics. Figure 1 Dose nding study. Melphalan at 6 and 8 mg/kg caused unaccepted morbidity and mortality (A/B). 5 mg/kg melphalan induced moderate, self-limiting disease. Weight loss at day 4 was dose-dependent (C) and correlated with plasma citrulline (D-F). Hypocitrullinemia in rst 2 days after melphalan was independent of dose and did not correlate with acute weight loss (G). Diarrhea severity was dosedependent (H-L). Non-linear regression analyses with Pearson's correlation analysis was performed for all association plots.    that were unable to resolve (C-D). Panel E shows signi cantly altered taxa (mean ± SEM) at day 4 and median difference between groups (median ± 95% CI).  Bile acid malabsorption is seen in melphalan-treated rats. Bile acids were assessed in plasma isolated from whole blood collected at termination. No signi cant decreases were seen in total bile acid pool (A) or the ratio of primary to secondary bile acids (B). Signi cant decreases were identi ed for taurocholic acid (C) and tauro-alpha-muricholic acid (D) but not for tauro-beta-muricholic acid (E). Taurodeoxycholic