EBioMedicine 48 (2019) 630–641 Contents lists available at ScienceDirect EBioMedicine journal homepage: www.elsevier.com/locate/ebiom A single faecal microbiota transplantation modulates the microbiome and improves clinical manifestations in a rat model of colitis Marina Lleal a,1, Guillaume Sarrabayrouse a,1, Joseane Willamil a, Alba Santiago a, Marta Pozuelo a, Chaysavanh Manichanh a,b,∗ a Department of Gastroenterology, Vall d’Hebron Research Institute, Barcelona, Spain b CIBERehd, Instituto de Salud Carlos III, Madrid, Spain a r t i c l e i n f o a b s t r a c t Article history: Background: Faecal microbiota transplantation (FMT) is a novel potential therapy for inflammatory bowel Received 16 July 2019 diseases, but it is poorly characterised. Revised 1 October 2019 Methods: We evaluated the performance of the mouse and rat as a pre-clinical model for human micro- Accepted 1 October 2019 biota engraftment. We then characterised the effect of a single human stool transfer (HST) on a human- Available online 15 October 2019 ised model of DSS-induced colitis. Colonic and faecal microbial communities were analysed using the 16S Keywords: rRNA approach and clinical manifestations were assessed in a longitudinal setting. Inflammatory bowel disease Findings: The microbial community of rats showed greater similarity to that of humans, while the micro- Faecal microbiota transplantation biome of mice showed less similarity to that of humans. Moreover, rats captured more human microbial Rat model of colitis species than mice after a single HST. Using the rat model, we showed that HST compensated faecal dys- biosis by restoring alpha-diversity and by increasing the relative abundance of health-related microbial genera. To some extent, HST also modulated the microbial composition of colonic tissue. These faecal and colonic microbial communities alterations led to a relative restoration of colon length, and a significant decrease in both epithelium damage and disease severity. Remarkably, stopping inflammation by remov- ing DSS before HST caused a faster and greater recovery of both microbiome and clinical manifestation features. Interpretation: Our results indicate that the rat outperforms the mouse as a model for human microbiota engraftment and show that the efficacy of HST can be enhanced when inflammation stimulation is with- drawn. Finally, our findings support a new therapeutic strategy based on the use FMT combined with anti-inflammatory drugs. © 2019 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/) Research in context mid/long-term effect of this treatment on the evolution of the host microbiome, immune response and clinical outcomes has not been Evidence before this study fully investigated through clinical trials nor through proper animal models. As a result, the efficiency of current FMT protocols is low Inflammatory bowel disease (IBD) is a chronic incurable disease in IBD. affecting approximately 0.5% of the general population in West- ern countries. Its aetiology remains unknown and its prevalence Added value of this study is expected to increase exponentially over the next decade mak- ing it a growing healthcare burden. Faecal microbiota transplan- In the present study, we first show that a conventional rat tation (FMT), an efficacious treatment for Clostridium difficile in- model, more reflective of human microbiota, could be an alterna- fection, has emerged as a potential therapy for IBD. However, the tive to a mouse model to study microbe-host relationships in hu- man disease. Then, using the humanised rat model of IBD (coli- ∗ tis induced by inflammatory agents), we demonstrate that a sin- Corresponding author at: Department of Gastroenterology, Vall d’Hebron Re- search Institute, Pg Vall d’Hebron, 08035 Barcelona 119-129, Spain. gle faecal transplantation from a healthy human donor could com- E-mail address: cmanicha@gmail.com (C. Manichanh). pensate dysbiosis by restoring alpha-diversity and by increasing 1 Authors contributed equally. the relative abundance of health-related microbial genera. Further- https://doi.org/10.1016/j.ebiom.2019.10.002 2352-3964/© 2019 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/) M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 631 more, we show that the faecal transplantation also corrected, to over, rats captured more human bacterial species than mice af- some extent, colonic tissue alterations, leading to a relative restora- ter a single human stool transfer. We then showed that a single tion of colon length, and a significant decrease in epithelium dam- stool transfer, obtained from a previously selected healthy human age and disease severity. Finally, we show that stopping the in- donor, modulated the dysbiotic microbiota, which was associated flammation by withdrawing the inflammatory agent before faecal with improvements in clinical and biological manifestations. transplant results into a faster and greater recovery of both micro- biome and clinical manifestation features. 2. Materials and methods Implications of all the available evidence 2.1. Ethical statement Our findings lead to promote the use of a rat model, instead The animal study was approved by the Ethical Committee of mouse, as the more appropriate pre-clinical animal model for of Animal Experimentation at the Vall d’Hebron Research Insti- FMT and IBD. Our first longitudinal study, using a humanised rat tute (Barcelona, Spain) and the Department of Agriculture, Live- model of colitis, proves the efficacy of a single faecal microbiota stock and Fisheries of the Generalitat de Catalunya (Departament transplantation and encourages the use of anti-inflammatory drugs d’Agricultura, Ramaderia i Pesca). Animal care complied with the during FMT to maintain a remission state in future clinical trials. criteria outlined in the Guide for the Care and Use of Laboratory Animals Institute for Laboratory Animal Research, Division on Earth 1. Introduction and Life Studies, Washington, D.C., USA. All experiments were con- ducted in the animal facilities of the Vall d’Hebron Research Insti- Inflammatory bowel disease (IBD) is a complex and chronic in- tute (Barcelona, Spain). testinal disorder associated with an exacerbation of the host im- For human stool donors, approval was provided by the Univer- mune response, a disruption of the intestinal barrier, and alter- sity Hospital of Vall d’Hebron Ethics Committee and informed con- ations in the luminal and mucosal microbial communities [1,2]. IBD sent was given in all cases. includes two chronic intestinal inflammatory forms, ulcerative col- itis (UC), which affects the rectum and colon, and Crohn’s disease, 2.2. Animals which may involve the entire gastrointestinal tract, but is most common in the colon and terminal ileum [3]. Patients with IBD Male Sprague-Dawley rats and BALB/c mice were purchased experience episodes of relapse alternating with remission. from Charles River (France). Rats weighed about 200 g and mice Dysbiosis in IBD patients is often characterised by a significant about 25 g and were 6 weeks of age. After 7 days of quarantine, and global alteration of the composition and structure of the mi- each animal was isolated in a conventional cage in order to pre- crobial community and is associated with a lower microbial alpha- vent microbial transmission. They were maintained at a constant diversity compared to healthy controls [4–7]. room temperature of 21 °C and were exposed to a 12:12 h light- Faecal microbiota transplantation (FMT) has been used for sev- dark cycle. All rats drank sterilised water and were fed an auto- eral years to treat recurrent Clostridium difficile infection (CDI) and claved standard chow diet. No contact was allowed between rats has recently proven effective in randomised clinical trials [8,9]. In- either through water, chow or faeces. terest in using FMT to restore a healthy gut microbiota and thereby attenuate inflammatory responses in IBD patients is growing. How- 2.3. Bowel preparation for stool transfer ever, many aspects of this therapeutic approach remain unan- swered, including questions regarding patient stratification, donor Omeprazole was diluted in water, filtered, and administered via selection and methods to administer the FMT [10–14]. While sev- oral gavage at a dosage of 50 mg/kg/day. All animal groups, except eral meta-analyses suggest that FMT is safe, this approach shows group A-CTL, received the omeprazole solution on day 2 (9 a.m.), variable efficacy in IBD patients [12,15,16]. In this regard, more day 3 (12 a.m.) and day 4 (8 a.m.). To empty the gastrointestinal studies are required in order to establish optimal and standardised tract, on day 3, rats underwent a 24-h fast and mice a 12-h fast. FMT protocols [17]. Concurrently, many experimental models, in- Animals were placed individually in a cage to prevent coprophagy cluding in vivo animal models such as mouse TNBS or DSS-induced and they have access to water. Then, 0.2 ml of Citrafleet®, prepared colitis [18,19] and T cell transfer colitis in severe combined immun- following the manufacturer’s instructions, was administered twice odeficient (SCID) mice, have been widely used to study IBD and (9 a.m. and 9 p.m. of day 3) via oral gavage. Citrafleet® (Casen- the effect of FMT in this disease [20]. For instance, using a mouse Fleet, Zaragoza, Spain) was administered two times via oral gav- model of colitis and a short-term follow-up study of 8 days, Zhou age: 1 ml, 24 h before the stool transfer, and 2 ml for rats (1 ml for et al. [20] demonstrated that FMT decreases the severity of colitis. mice), 12 h before the transfer [21]. A laparotomy was performed This effect was associated with significantly decreased myeloperox- to evaluate the efficiency of Citrafleet® and to collect colon tis- idase activity, reduced levels of TNF-α and IL-1β , and an increased sues for microbiome analysis. The use of cages with grids pre- level of IL-10 in colon tissues. However, these authors did not ad- vented coprophagy during the cleansing procedure, which lasted dress the impact of FMT on the microbiome community. In this about 24 h. The rats and mice were then given via oral gavage a context, to the best of our knowledge, the present study is the first dose of omeprazole (at the concentration of 50 mg/kg/day). to evaluate the performance of mouse and rat models for human microbiota transplantation. 2.4. Human stool transfer (HST) in mice and in rats The aims of this study were to: (a) Identify which animal species (rat or mouse) was the most appropriate to receive and The healthy stool donor was selected in a previous project maintain a single human stool transfer; and (b) to examine the (Sarrabayrouse et al., unpublished). Briefly, ten healthy volunteers long-term effect of a single human stool transfer on a humanised (≥ 18 years of age) provided self-collected stool samples. Two and chemically-induced colitis murine model at both microbiome stools with one month interval were collected and immediately and clinical manifestation level. frozen at home and brought to the laboratory without break- We demonstrated that the microbiome of rats has a similar ing the cold chain, where they were stored at −80 °C. Stool and composition and structure to that of humans, while the micro- serology screenings were performed for bacterial, parasitic and biome of mice shows less similarity to that of humans. More- viral pathogens (Clostridium difficile, Salmonella, Shigella, Yersinia, 632 M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 Aeromonas, Campylobacter, Vibrio, E. coli O157, Rotavirus, Adenovirus, flow cytometry using an LSRFortessa cytometer (BD) and then Atrovirus, Norovirus, and Giardia intestinalis). Potential donors did analysed with FlowJo software (FlowJo LLC). not take antibiotics in the eight weeks preceding the screening. One of the donors was selected on the basis of the stability and 2.10. Microbiome analysis high diversity of the stool microbial community and the lack of genera potentially involved in inflammation, such as Escherichia or 2.10.1. 16S rRNA gene amplification Fusobacteria, and the presence of Faecalibacterium. The transfer of Genomic DNA was extracted using the protocols previously de- human stool to rodents was performed once through an oral gav- scribed by Pascal et al. [7] for stool samples and by Santiago et al. age using 100 mg of a fresh stool sample diluted in 2 ml of PBS for [25] for tissue samples. An equivalent of 1 mg of each sample rats (in 1 ml for mice). A second HST containing 100 mg of faeces was used for DNA quantification using a Nanodrop ND-1000 Spec- diluted in 2 ml of PBS was performed 22 days after the induction trophotometer (Nucliber). DNA integrity was examined by micro- of colitis. Faecal samples were collected at various time points. capillary electrophoresis using an Agilent 2100 Bioanalyzer with the DNA 12,000 kit. 2.5. Induction and assessment of colitis by administration of DSS in For profiling microbiome composition, the hyper-variable region rats (V4) of the bacterial and archaeal 16S rRNA gene was amplified by PCR using universal primers, as previously described by Pascal Rats were provided with drinking water containing 5% (wt/vol) et al. Standard PCR using 0.75 units of Taq polymerase (Roche) and dextran sodium sulphate (DSS) (mol wt 40,000; TdB Consultancy, 20 pmol/μL of the forward and reverse primers was run in a Mas- Uppsala, Sweden) ad libitum for 5 days in order to induce acute tercycler gradient (Eppendorf) at 94 °C for 3 min, followed by 35 experimental colitis. Chronic colitis was then induced by reduc- cycles of 94 °C for 45 sec, 56 °C for 60 sec, 72 °C for 90 s, and a final ing the concentration of DSS to 2% for 17 (group Dw-DSS) or 57 cycle of 72 °C for 10 min. Amplicons were first purified using the (group D) days depending on the study group, as previously de- QIAquick PCR Purification Kit (Qiagen, Barcelona, Spain), quantified scribed by Oishi et al. [22]. Colitis was assessed daily on the basis using a Nanodrop ND-1000 Spectrophotometer (Nucliber) and then of the combined scores of body weight, stool consistency and pres- pooled in equal concentration. For colon samples, a second purifi- ence of macroscopic blood in stools to calculate disease activity in- cation step was applied to the pool of amplicons, as described in dex (DAI), as shown in Supplementary Table S1. The DAI score was Santiago et al. [25]. The pooled amplicons (2 nM) were then sub- graded on a scale of 0–4, as previously described by Murthy et al. jected to sequencing using Illumina MiSeq technology at the tech- [23]. nical support unit of the Autonomous University of Barcelona (UAB, Spain), following standard Illumina platform protocols. In order to 2.6. Laparotomy and euthanasia identify possible contamination in colonic tissues, considered as low-biomass samples, and subtract the sequences of the potentially On the last day of the study, laparotomy was performed by in- contaminated DNA generated, negative controls (blanks) were in- jecting 10 mg/kg xylazine (Xilagesic®; Calier, Barcelona, Spain) and troduced during two technical steps: genomic DNA extraction and 75 mg/kg ketamine (Ketolar®; Parke-Dawis, Madrid, Spain). The 16S rRNA gene PCR amplifications. whole intestine was removed and the colon was washed with PBS and cut into small pieces. Some of the colon fragments were sam- 2.10.2. Upstream 16S rRNA sequence analysis pled for histological evaluation while others were stored directly at The upstream sequence analysis was performed with QIIME −80 °C for microbiome analyses. (v.1.9.0) using the guidelines proposed by Navas-Molina [26]. First, we generated a mapping file containing sample identifiers, bar- 2.7. Histological analysis codes, primer sequences, time points, symptoms, treatments and any other additional information of the samples needed for the For histology, tissues were fixed in Methanol-Carnoy’s solution analyses and validated its compatibility with the QIIME tool. Then, (60% methanol, 30% chloroform, 10% glacial acetic acid) and main- we demultiplexed and performed sequence quality filtering of the tained at room temperature for at least 3 h. Small pieces of colon obtained fastq files with a minimum acceptable Phred score of 20. tissue were placed into cassettes and dehydrated in ethanol baths. From a total of 531 rat faecal samples, we obtained a total of 23.8 They were then immersed in xylene, a clearing agent, to remove millions of high-quality sequences with a number of reads rang- ethanol. Next, colon tissues were embedded in molten paraffin and ing from 1 to 85,679 per sample with an average of 44,201 reads. sectioned (5 μm). Slices were stained with hematoxylin and eosin. From a total of 36 rat colon samples, we obtained a total of 4.6 Finally, the degree of intestinal inflammation was determined fol- millions of high-quality sequences with a number of reads ranging lowing the guide described by Erben et al. [24]. from 3711 to 460,161 per sample with an average of 127,934 reads. From a total of 60 mice faecal samples, we obtained a total of 2.8 2.8. TNF-α measurement millions of high-quality sequences with a number of reads ranging from 22,663 to 69,572 per sample with an average of 46,230 reads. TNF-α was determined in tissue samples by enzyme-linked im- We clustered similar filtered sequences into Operational Tax- munosorbent assays (ELISAs), following the manufacturer’s proto- onomic Units (OTUs) with the USEARCH algorithm (v8.1.1861) cols (eBiosciences). Results are expressed as the ratio of picograms [27] using a 97% similarity threshold. Although the QIIME devel- of TNF-α per milligram of total protein. Limits of detection were opers recommend the open-reference method that first groups se- 30 pg/ml. quences using a reference database and then by sequence iden- tity, we preferred, in terms of time consumption and discovery of 2.9. Lymphocyte isolation and staining from spleen novel diversity, to use the de novo approach, based only on se- quence similarity that focuses on finding new bacteria. In this step, Single cell suspensions were generated from spleens, and ery- we also identified and removed chimeric sequences using UCHIME throcytes were removed from splenic samples by Ficoll gradi- [27]. ent centrifugation. 2.5 × 10^5 splenocytes were stained with FITC- We then performed a taxonomical assignment step using the conjugated anti-CD3 (Miltenyi Biotec), PB-conjugated anti-CD4 and Basic Local Alignment Search Tool (BLAST) to map each representa- PE-conjugated anti-CD8 antibodies. Stained cells were analysed by tive sequence against a combined database of Greengenes (gg_13_8 M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 633 release) and PATRIC databases. Since each OTU could comprise to compute correlations between the microbiota and clinical or ex- many related sequences, we picked a representative sequence from perimental variables, we used non-parametric Spearman correla- each OTU and these representative sequences were aligned using tions. For the statistical analysis of clinical longitudinal data, we PyNAST against Greengenes template alignment. After this step, we used a non-parametric method of the R package nparLD (https: did a phylogenetic construction with FastTree. Finally, we gener- //CRAN.R-project.org/package=nparLD). ated a table containing the predicted taxonomy and abundance of all the OTUs for each sample. 2.10.8. Microbial source tracking Microbial source tracking was performed using SourceTracker 2.10.3. Removal of unknown sequences not corresponding to the 16S (v1.0.1) [32], a Bayesian approach that estimates the proportion rRNA gene of OTUs in a given community that come from possible source In the case of colon (low biomass) samples, many of the se- environments. We used the default parameters (burn-in = 100, quences could correspond to unknown taxa, which could be due to restart = 10, alpha and beta Dirichlet parameters = 0.001 and 0.01, the presence of Eukaryotic host sequences. We identified those se- respectively), except for the rarefaction depth, which was set to quences assigned as unknown by mapping them against GenBank 7512, as in the analyses involving human and rat faecal samples. using BLAST and we removed from the OTU table all sequences To analyse the effect of HST in the two rodent models, mouse and that did not correspond to the 16S rRNA gene. rat faecal samples from the first day of the trial and HST donor samples were defined as mouse, rat and human sources, respec- tively. To determine the proportions of healthy rat microbiota ver- 2.10.4. Removal of bacterial contaminants sus microbiota associated with the administration of DSS, rat fae- After removal of non-bacterial DNA, we identified bacterial con- cal samples from the day after DSS administration and from af- tamination in colon samples and removed it. We used the ’preva- ter 3 weeks of DSS treatment were defined as healthy and DSS- lence’ method of the open-source R package decontam [28] with associated sources and used to train the model. The rest of samples both extraction and PCR negative controls and with the default were then assigned some percentages either from these sources or threshold (P = 0.1). We decontaminated each sequencing run sep- an unknown source. arately. 2.10.9. Data deposition and accession numbers for raw data 2.10.5. Downstream sequence analysis Sequence data and metadata have been deposited in the Gen- We performed downstream analyses, including diversity analy- Bank database with the following access number: PRJNA549436. ses and statistical tests with QIIME (v1.9.1) and R software (v3.4.4). We used the biom program to convert tables between biom and 3. Results txt formats. We assigned arbitrary numbers to unknown species to avoid collapsing possible novel species into one. Finally, we sum- 3.1. Choice of rodent model marised the OTU table by distinct taxonomic levels from phylum (L2) to species (L7). 3.1.1. Rat microbiome compared to mouse and human microbiome Conventional mice are often used to study human diseases and, 2.10.6. Diversity in particular, to examine the effect of alterations in the micro- We performed rarefaction (random selection of the same num- bial community in human diseases. To study whether conventional ber of reads per sample) in the OTU table in order to account for rats outperform mice as recipients for human microbiota, we first uneven library sizes. For rat faecal samples, we rarefied the OTU compared the faecal microbiome of rats (n = 52), mice (n = 16) and table at 10,912 sequences per sample, which allowed us to keep healthy human individuals (n = 72). Human sequence data were re- 521 samples and 16.42 million reads for further analyses. For the covered from our previous study [7] and data for rats and mice analyses involving human and rat faecal samples, we rarefied the were newly generated. OTU table at 7512 sequences per sample (in order to keep most of Firmicutes and Bacteroidetes accounted for 93%, 98% and 93% of the samples) and retained 524 samples and 16.46 million reads for the sequence data in the human, mouse and rat microbiome, re- further analyses. For rat colon samples, we rarefied the OTU table spectively (Fig. 1(a)). The microbial faecal community of the three at 2538 sequences per sample, which maintained all 36 samples species was significantly different for the most predominant phyla and 2.81 million reads for further analyses. Regarding mice faecal and genera (Mann Whitney pairwise test, FDR < 0.05). Alpha- samples, we rarefied the OTU table at 7512 sequences per sam- diversity analysis using Chao1 estimates showed significant pair- ple, which retained all 60 samples and 2.8 million reads for further wise differences between the three groups (Fig. 1(b)). Shannon in- analyses. dexes revealed significant pairwise differences between all groups, To estimate the alpha-diversity (within samples), we calculated except between mice and humans (Fig. 1(c)). Altogether, mice pre- the Chao1 and Shannon indexes [29,30]. Beta diversity (between sented a higher alpha-diversity than rats and humans. Surprisingly, samples) was computed using weighted and unweighted UniFrac beta-diversity analysis, based on weighted and unweighted UniFrac [31] to produce distance matrices, which were later used for indexes, revealed that the distance between rats and humans was grouping samples into hierarchical cluster trees with unweighted smaller than that between mice and humans (Fig. 1(d)–(g)). These Pair Group Method with Arithmetic mean (UPGMA) and Principal findings confirmed that the faecal microbial community of rats is Coordinate Analyses representations (PCoA). more similar to that of humans than that of mice to humans. 2.10.7. Differential abundance and statistics 3.1.2. Human stool transfer (HST) in rats and mice To assess differences in the microbiota composition between Human microbiota-associated (HMA) animal models are becom- groups based on treatment and/or time point, we used the non- ing a standard tool through which to study the relation between parametric statistical Kruskal-Wallis test (more than two groups, gut microorganisms and human diseases. We performed a HST into unpaired), Wilcoxon signed rank test (pairwise paired compar- conventional rat and mouse by gavage to evaluate how each ro- isons) and Mann-Whitney U test (pairwise unpaired comparisons). dent shaped the human microbiota over time (Fig. 1(h)). Omepra- To account for multiple comparisons of all taxa, we used FDR cor- zole, a proton pump inhibitor, was used to suppress stomach acid rection and considered the results significant when P < 0.1. Then, secretion and therefore to increase bacterial survival. CitraFleet® 634 M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 Fig. 1. The microbiome of the rat showed greater similarity to that of humans than the mouse microbiome to humans. (a) Taxonomic profile at the phylum level for human (n = 72), rat (n = 52) and mouse (n = 16) faecal samples. Alpha-diversity analysis using the Chao1 (b) and the Shannon (c) indexes showed differences between humans, mice and rats (FDR < 0.05). Pairwise distances between rats and humans were lower than between mice and humans in both unweighted (d) and weighted (e) UniFrac (FDR < 0.05). Unweighted (f) and weighted (g) UniFrac Principal Coordinate Analysis representation of the three species, highlighting the HST donor sample. (h) Experimental design to assess the engraftment of the human microbiota in mouse and rat. (i) Proportion of predicted microbiota from a human source in rats and mice allocated to different groups as a function of receiving a HST. Significant differences were found between rats receiving a HST or not, and also between rats and mice receiving a HST (Mann-Whitney U test; p < 0.05). NM = No HST mouse; HM = HST mouse; NR = No HST rat; HR = HST rat. (sodium picosulfate), a stimulant laxative, was used to empty the Given that the rat microbiome was more similar to the hu- gastrointestinal tract before the HST. The healthy human donor man microbiome, in terms of composition and structure, than stool was previously selected from ten volunteers on the basis of the mouse microbiome to humans and that it showed a better its high microbial diversity and stability, and tested for the absence and long-lasting engraftment of this human microbiome, the rat of potential common pathogens (see Methods section). Addition- emerges as a better model for studying the relationship between ally, the healthy donor stool was selected for having the lowest the microbiome and human intestinal diseases. relative abundance of Escherichia and Fusobacteria and a high abun- dance of Faecalibacterium, characteristics of a healthy stool as com- 3.2. Modulation of the rat gut microbiota pared with an IBD stool [7] (see Methods section). To monitor the evolution of the transferred human microbiome in the rodents, we 3.2.1. Experimental design used SourceTracker, a Bayesian, OTU-based algorithm, to estimate To assess the effect of a HST in a rat model of chronic intesti- the proportion of the human microbial community remaining in nal inflammation, a total of 50 animals were randomly placed into the animals over time, up to 8 weeks (60 days) for mice and 16 seven experimental groups (Fig. 2) as follows: a control group re- weeks (112 days) for rats. To train the model, we used rat and ceiving a standard diet and water without treatment (group A- mouse faecal samples collected immediately after quarantine, and CTL, same group as that described in Fig. 1(h) as NR); a group the healthy human donor sample. Surprisingly, the proportion of receiving omeprazole and Citrafleet® (group B-OME-CIT), to evalu- the human microbiome present in the rats was significantly higher ate the cleansing effect; a group receiving omeprazole, Citrafleet® than in mice over the 8-week follow-up (Fig. 1(i), day 60), thereby and a human stool (group C–HST1, same group as that described suggesting that rats captured human microbiota more efficiently in Fig. 1(h) as HR), to evaluate the humanization protocol; a group than mice. The engraftment effect lasted for at least 38 days and receiving omeprazole, Citrafleet®, human stool and then dextran was lost at 60 days. sodium sulfate (DSS) until the end of the study (group D), to M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 635 terella, Prevotella, Bacteroides, Odoribacter, and three unknown gen- era from Rikenellaceae, Desulfvibrionaceae and S24-7 families, and a decrease or disappearance of Lachnospira and Parabacteroides (Fig. 3(c)). Changes in the taxonomic profile were associated with a significant increase in alpha-diversity, as shown by the Chao1 and Shannon indexes and the number of observed species from D0 to D112, which increased from 161 to 289 (Fig. 3(d)–(f)). 3.2.3. Short-term effect of bowel cleansing The short-term effect of omeprazole alone on the gut microbial community was not significant (animal group B-OME-CIT). How- ever, omeprazole combined with CitraFleet® led to an increase in Verrucomicrobia and Proteobacteria but a decrease in Firmi- cutes (Supplementary Fig. S1a). At the genus level, omeprazole and CitraFleet® led to increases in several genera, such as Parabac- Fig. 2. Experimental design to evaluate the effect of a human stool transfer on a teroides, Morganella and Citrobacter and deceases in others, includ- rat model of chronic intestinal inflammation. Seven groups were distributed as fol- ing Anaerostipes, unknown Clostridiales, and Lactobacillus, among lows: a control group receiving a standard diet and water without treatment (group A-CTL, n = 10); a group receiving omeprazole and Citrafleet® (group B-OME-CIT, others (Supplementary Fig. S1b). To the best of our knowledge, this n = 8); a group receiving omeprazole, Citrafleet® and a human stool (group C–HST1, is the first study to report this combined effect of omeprazole and n = 8); a group receiving omeprazole, Citrafleet®, human stool, and DSS until day CitraFleet® on the composition of the gut microbial community. 39, which was replaced by water (w) until the end of the study (group Dw-DSS, n = 7); a group receiving omeprazole, Citrafleet®, human stool and then DSS un- til the end of the study (group d-DSS, n = 7); a group receiving omeprazole, Cit- 3.2.4. Inflammation and dysbiosis induction rafleet®, human stool, and DSS until day 39, which was replaced by water (w) until To simulate human IBD, we induced inflammation using dex- the end of the study, and a second human stool (group Ew-HST2, n = 7); a group receiving omeprazole, Citrafleet®, human stool, DSS until the end of the study, and tran sodium sulfate (DSS) in two additional animal groups, d-DSS a second human stool (group E-HST2, n = 7). HST: Human stool transfer; DSS: dex- and Dw-DSS. DSS, a negatively charged sulphated polysaccharide, tran sodium sulfate. The “n” in the right side of the figure represents the number was given in drinking water at 5% (from day 17) and then at 2% of animals for which colonic tissues were collected and analysed for microbiome (from day 22) in order to achieve a model of chronic intestinal in- composition. flammation [22]. The Dw-DSS group was used to assess resilience after withdrawal of DSS treatment on day 39 (Fig. 3). DSS re- sults in the death of epithelial cells, leading to the disruption of evaluate dysbiosis after inflammation induction; a group receiv- the epithelial monolayer lining. This disruption allows the entry ing omeprazole, Citrafleet®, human stool, DSS until the end of the of luminal bacteria and associated antigens into the mucosa and study and a second human stool (HST2) on day 39 (group E-HST2), causes an altered mucosal immune response [19]. Small intestine to assess the effect of microbiome modulation in response to the and colon lengths, like the histological scores, were measured on transfer of a healthy human stool. the day each animal was sacrificed. A blinded histological evalua- Two additional groups, Dw-DSS and Ew-HST2, were added. tion of colonic tissues was performed by two researchers on the These were similar to groups d-DSS and E-HST2, except that DSS day of animal sacrifice. was replaced by water at day 39 before HST2, in order to study Significant alterations in the faecal microbial communities were the resilience of the mucosal layer and the microbiota after injury observed at the genus level when comparing the two groups of to the tissue was stopped. As specified above, omeprazole and Cit- rats with (D-DSS) and without (C–HST1) DSS treatment at day 79. raFleet® were used to increase microbial survival and to empty the As shown in Fig. 4(a), changes in the microbial community induced gastrointestinal tract, respectively, before stool transfer. DSS was by DSS included genera that were significantly affected over time used to induce chronic intestinal inflammation, which occurred af- (Fig. 3(b)), such as Bacteroides, Suterella and an unknown mem- ter 5 days of DSS treatment at 5% and then at 2% (group d-DSS vs. ber of the S24-7 family. Bacteroides and Suterella increased both Dw-DSS and group E-HST2 vs. Ew-HST2). over time and upon DSS treatment. Conversely, the unknown S24-7 The first HST was performed to generate a human microbiota- group increased over time but decreased in response to DSS. These associated animal model and the second to assess the correction findings suggest that the S24-7 group is involved in maintaining of disease severity (group E-HST2 vs. d-DSS and group Ew-HST2vs. homeostasis in the gut microbiota. No significant correlations were Dw-DSS). Given that human microbiota differs significantly from found in colonic tissues—an observation that could be attributed that of rodents, the transfer of human microbiota was performed to the low number of samples involved. Notably, the microbiome to ensure a human microbiota context before inducing inflamma- returned to a baseline and healthier composition when DSS treat- tion. ment was withdrawn, as shown by SourceTracker (Fig. 4(b)). In- In order to calculate the disease activity index (DAI), animals deed, no significant differences were found at the genus level be- were monitored daily for body weight and for the presence of tween groups C–HST1 and Dw-DSS. blood in stool. Histological, immunological and colonic microbial To determine whether the dysbiosis in a rat model of DSS- community analyses were performed on colonic samples obtained induced colitis was similar to ulcerative colitis or to Crohn’s dis- on the day of sacrifice. Faecal microbial community and DAI scores ease in humans, we compared the microbiome data of rats from were evaluated at various time points. group d-DSS collected at day 116 with published sequence data [7] of adult patients with ulcerative colitis (UC, n = 28) and adult 3.2.2. Rat intestinal microbial community viewed across time patients with Crohn’s disease (CD, n = 32), using the UniFrac dis- To study the effect of time on the faecal microbial composi- tances. Our results showed that dysbiosis in the rat model of tion, we analysed changes in the taxonomic profile over 16 weeks DSS-induced colitis was closer to dysbiosis in patients with UC (112 days, animal group A-CTL) of follow-up (Fig. 3(a)). The rel- than in those with CD, when taking into account only microbial ative abundance of six phyla (Fig. 3(b)) and ten genera changed composition (unweighted UniFrac distances) (non-parametric test, significantly over 16 weeks, with an increase in Adlercreutzia, Sut- p < 0.01) (Fig. 4(c) and (d)). This result validates the description of 636 M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 Fig. 3. Evolution of the rat microbiome over time (112 days). (a) Taxonomic profile at the phylum level. Significant differences (marked with asterisks) were found in most of the phyla over time (Kruskal-Wallis test, FDR < 0.05). Each bar represents an individual faecal sample. (b) Taxonomic profile at the genus level, only significant changes are shown (Kruskal-Wallis test, FDR < 0.05). Each bar represents an individual faecal sample. Alpha-diversity was calculated on the basis of the species observed (c), the Chao1 index (d), and the Shannon index (e). Significant differences were found in all three indexes, indicating a progressive increase in microbial diversity over time (FDR < 0.05; ∗∗ : FDR < 0.01; I: FDR < 0.1). A total of 24 faecal samples were analysed, with n = 4 or n = 6 at each time point. several clinical characteristics of the rat model, such as colon lo- the peripheral immune response of the animals (Fig. 5(f)). On the calization and the lack of transmural inflammation, as reported in other hand, DSS significantly increased the frequency of these cells, this study and others [33]. which is in agreement with an increase in the inflammation in- As shown in Fig. 5(a), small intestine length was not affected duced by this treatment. When DSS was maintained until the end by DSS treatment. In contrast, colon length was significantly re- of the study, no effect on the production of TNF-α in colonic tis- duced in all animal groups compared to the control group C–HST1, sues was observed; however, a significant decrease in this param- as reported by others (Fig. 5(b)) [19,23,34]. The mucosal layer of eter was detected when DSS was stopped on day 39 (Fig. 5(g)). group d-DSS was highly damaged compared to group C–HST1, as All together, these results demonstrate that, as expected, DSS shown by an average histological score of 4.4 (SD = 1.1) vs. 1.7 induced a significant reduction in colon length and mucosal dam- (SD = 0.8), respectively (Fig. 5(c) and (d)). When DSS treatment age. Moreover, it exacerbated the immune response and caused an was withdrawn from group Dw-DSS, which implied that mucosal increase in disease severity. Furthermore, cessation of DSS after a tissues were not exposed to DSS for 40 days, we observed a sig- daily treatment of 22 days revealed an improvement of all these nificant recovery from tissue damage, as shown by an average his- clinical and biological manifestations. tological score of 2.9 (SD = 0.8). Furthermore, disease activity in- dexes (DAI), which were calculated on the basis of animal weight, 3.2.5. Correction of dysbiosis and disease severity by HST presence of blood in stool and stool consistency scores for each an- To evaluate the therapeutic potential of HST to correct DSS- imal at different time points, significantly increased in group d-DSS induced inflammation, we added two animal groups, E-HST2 and compared to Dw-DSS (mixed ANOVA, P < 0.05) when DSS admin- Ew-HST2, in which we performed a second HST (HST2) on day istration was stopped on day 39 (Fig. 5(e)). 39 (22 days after the start of DSS-induced inflammation). The Ew- Moreover, to evaluate the immune response associated with HST2 group was added to assess the resilience of rats after with- DSS treatment, we studied the frequency of CD4 splenic T lympho- drawal of the DSS treatment (see experimental design in Fig. 2). cytes in spleen and measured TNF-α in colonic tissues. To this end, Analysis of group E-HST2 revealed a significant reversal of ep- splenocytes were isolated and analysed by flow cytometry, as de- ithelial damage as compared with group d-DSS, as shown by a rel- scribed in the Methods section. On the one hand, we observed that ative reduction of the histological score (P = 0.07; Mann-Whitney humanization of the rats decreased the frequency of these cells, U test) and a significant reduction of the DAI score (P = 0.003; thereby suggesting that the human microbial community affected non-parametric mixed ANOVA-like test). Rats did not present dif- M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 637 Fig. 4. DSS administration causes dysbiosis in rats. (a) Significant differences in the relative abundance of several microbial genera between group C–HST1 and D (no DSS vs. 9 weeks DSS) at the end of the trial (at day 79). (b) Proportion of predicted microbiota from a healthy rat source in rat groups C, D and Dw-DSS, using the SourceTracker software. The vertical, dotted line separates the time points of the 3 weeks in which rats in group Dw-DSS received DSS (left) from the rest of time points in which DSS treatment was withdrawn (right). Significant differences were found between DSS-treated rats and untreated rats (group C–HST1 vs. D) but not between control rats and rats after DSS treatment was withdrawn (group C–HST1 vs. Dw-DSS, at day 60) (Mann-Whitney U test; p < 0.05). (c) Unweighted UniFrac distances between rats in group d-DSS (9 weeks DSS) and UC and CD patients. Significant differences were found, UC patients being closer to rats with DSS-induced colitis than CD patients to this rat model (non-parametric test, p < 0.01). (d) Weighted UniFrac distances between rats with DSS-induced colitis in group d-DSS (9 weeks DSS) and human UC and CD patients. No significant differences were found. ferences in small intestine length with (E-HST2) or without (Ew- Finally, we also found several correlations between faecal mi- HST2) the maintenance of colitis induction. However, colon length crobiome and clinical manifestation. Among them, an unknown tended to increase in group E-HST2 compared to D (Mann-Whitney Ruminococcaceae and Ruminococcus were positively correlated U test, P = 0.08) (Fig. 5(a) and (b)). Of note, HST2 significantly de- with colon length and negatively with histological score. The S24- creased the frequency of CD4 splenic T lymphocytes (group d-DSS 7 group was positively correlated only with colon length. Con- vs. E-HST2), reducing it to basal level (group C–HST1). This find- versely, Bacteroides was negatively correlated with colon length but ing therefore indicates the efficacy of this treatment on one of the positively correlated with histological score. The longitudinal set- markers of inflammation. At the faecal microbiome level, HST per- ting allowed us to assess the effect of the different treatments on formed in group E-HST2 increased alpha-diversity (Chao1 indexes) the relative abundance of various microbial genera, which evolved as compared with group d-DSS (FDR = 0.057, on the day of sacri- as expected from their correlations with clinical manifestations fice (Fig. 6(a) and (b)). (Fig. 6(f)). HST2 in group Ew-HST2 caused a significant decrease in dis- ease severity through a reduction in both the DAI scores (P = 1e- 4. Discussion 13; non-parametric mixed ANOVA-like test) and histological scores (P = 0.005; Mann-Whitney U test) compared to the group that re- Here we first studied the performance of the rat and mouse for ceived DSS (D) (Fig. 5(c) and (d)). Simultaneously, at the micro- HST. We then evaluated the therapeutic potential of a stool trans- biome level, we observed a recovery of alpha-diversity (Fig. 6(a) fer to treat chronic inflammation induced by a colitogenic chemi- and (b)) and a greater proportion of healthier gut microbiota, us- cal. To this end, we analysed clinical manifestations and immuno- ing the SourceTracker method, in both faecal samples and colonic logical markers, as well as faecal and colonic microbiota in a lon- tissues (Supplementary Fig. S2). gitudinal setting (up to about 100 days) using a conventional an- Furthermore, the Spearman correlation test revealed correla- imal model. Our experimental design, which included several ani- tions between colonic microbiome composition and clinical man- mal groups, also allowed us to gather information on the natural ifestations such as DAI scores, histological scores and colon length. evolution of the animal gut microbiome over a period of 112 days, Indeed, several microbial phyla and genera in the colonic tissues the long-term effect of colitis induction, and the resilience of the showed a correlation with colon length and DAI scores (Fig. 6(d) microbiome when this induction was stopped. and (e)). Ruminococcus, Lactobacillus and an unknown Ruminococ- Most studies on IBD using DSS-induced colitis models have caceae were positively correlated with colon length, and an un- been performed on mice [20,33–35]. Comparison of rat, mouse known Erysipelotrichaceae and Bacteroides were negatively corre- and human genomes has revealed that the rat genome contains lated with this parameter. In contrast, an unknown Rikenellaceae about the same number of genes as that of humans and mice was positively correlated with DAI scores, and Oscillospira, an un- [36]. A recent study has provided a gene catalogue of the Sprague- known Christensenellaceae and an unknown genus of Mollicutes Dawley rat gut metagenome, and comparison of rat with human class were negatively correlated with these scores. or mouse has revealed a higher pairwise overlap between rats and 638 M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 Fig. 5. HST ameliorated clinical manifestations during DSS-induced colitis. Small intestine (a) and colon length (b) measured on the day of sacrifice. (c) Histological scores. (d) Haematoxylin and eosin staining of colonic tissue. (e) Disease activity indexes (DAI) were calculated on the basis of animal weight, presence of blood in stool and stool consistency scores for each animal. Significant differences were found using a non-parametric test equivalent to a mixed ANOVA (P < 0.05, the different letters (a, b and c) indicate which groups were different). (f) Lymphocyte isolation from spleen. (g) TNF-α measurement in colonic tissues. Mann Whitney test, ∗ : P < 0.05; ∗ ∗ : P < 0.01. humans (2.47%) than between mice and humans (1.19%) at the of FMT in treating IBD and other human disorders. Future studies gene level [37]. Our study, using 16S rRNA sequence data, has com- should test the effects of transplanting stools with diverse micro- plemented this previous work, showing that the overall composi- bial compositions from various human donors and evaluate a range tion and structure of the faecal microbiome of rat has a greater of conditions to improve the engraftment of human microbiota. similarity to that of the human than the microbiome of the mouse For instance, using a mouse model, a recent paper by Staley et al. to human. Furthermore, the rat appeared to capture the human [38] has shown that an appropriate antibiotic treatment protocol stool microbiome more efficiently than the mouse. Overall, on the followed by a single gavage of human microbiota provides a useful, basis of these findings, the rat emerges as a potential alternative complimentary human microbiota-associated model to that estab- to the mouse for evaluating a human microbiota-associated ro- lished in germ-free facilities. A similar protocol could be applied dent, thus highlighting the potential use of rats to study the effect to the rat model, although in the case of a colitis model, the use M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 639 Fig. 6. Effect of HST on DSS-induced dysbiosis. (a) Alpha-diversity in stool samples as measured by the Chao1 indexes in a longitudinal setting. Alpha-diversity in stool (b) and colon (c) samples as measured by the Chao1 indexes on the day of animal sacrifice. There was a significant reduction in faecal microbiome diversity only in group d-DSS compared to control group C–HST1 (n = 5–8 per group, non-parametric test, FDR < 0.1). No significant differences were found in the colonic microbiome (n = 4–6 per group). Correlations between microbiome composition in colonic tissues and clinical manifestations as measured by the Spearman correlation test (P < 0.05). In colon samples, genera were significantly correlated with colon length (d) and with DAI score (e). (f) In stool samples, the evolution of several bacterial genera selected based on their correlation with clinical manifestations are plotted according to the longitudinal setting. 640 M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 of an antibiotic cocktail could cause a bloom of Escherichia (Pro- restoration of colon length after withdrawal of DSS treatment. The teobacteria), which may sustain inflammation. This protocol would discrepancy between studies could be attributable to the dose of therefore need further evaluation on a model of colitis. DSS used (administration of 5% DSS for 5 days before a treatment Our study presented several limitations. We used a 16S rRNA of 2% for 17 days) in our work. Instead, in Bento et al., DSS was gene approach to characterise changes in the microbial community. used at only 2% for 20 days. Furthermore, the transfer of a healthy Indeed, to evaluate the effect of treatments, more comprehensive human stool allowed the amelioration of clinical manifestations, approaches could have been applied. These include DNA and RNA leading to a significant decrease in the DAI scores, which were shotgun sequencing, proteomic or metabolomic analysis. However, correlated with the relative abundance of several microbial gen- these approaches are also much more expensive and may not be era in colonic samples. Interestingly, an unknown Christensenel- worth this project using an animal model. Another limitation of laceae genus presented one of the highest negative correlations the present work is the use of only one colitis model, which may with DAI scores, and thus with disease severity. This genus was not reproduce all the characteristics of human chronic UC. Another previously detected in a higher relative abundance in healthy indi- in vivo model such as TNBS-induced colitis might have provided viduals compared to IBD patients [7] and in normal weight individ- different outcomes. However, DSS is used in many studies for its uals compared to overweight subjects [41], thereby validating its simplicity (given in drinking water and not by enema administra- role in maintaining a healthy status. The S24-7 family, another re- tion as for TNBS) and a relative short time to induce chronic coli- markable group of bacteria, was depleted upon DSS treatment but tis. Moreover, the evaluation of the combined effect of Citrafleet® increased after DSS withdrawal. Moreover, it was found to be pos- and omeprazole unexpectedly uncovered a dramatic alteration of itively correlated with colon length. This result is consistent with the microbiome composition. However, we did not assess whether a previous study showing that S24-7 was depleted in a model of this treatment could influence donor microbiome engraftment or type 1 diabetes [42], thereby suggesting that this bacterial group dysbiosis caused by DSS. Future studies should provide details of may be a good candidate for maintaining homeostasis. the long-term effect of these drugs on engraftment and include an- Moreover, HST led to a significant restoration of microbiome other control group without this treatment. Finally, we performed alpha-diversity in stool samples but not in colonic tissues unless HST using a single faecal sample collected from only one healthy DSS treatment was withdrawn. This observation indicates that HST donor. This approach did not therefore allow testing of whether a alone was not sufficient to correct dysbiosis in mucosal tissues specific microbial community composition affects the efficiency of when inflammation was maintained. To the best of our knowledge, colonization and recipient immune response. Future studies could there are currently only four randomised published clinical trials also test the effects of stool transfers at several time points on the on FMT in patients with UC [10,43–45]. All of them showed low ef- efficiency of this procedure. ficacy of the treatment (between 20 and 30% of success rate), with- Our longitudinal study has provided valuable information re- out a clear understanding of the clinical events because of the ob- garding the natural evolution of the rat faecal microbiome over vious limitations of working with patients. Our study has pointed 112 days, demonstrating a continuous increase in alpha-diversity out possible reasons why FMT alone was not sufficient to treat and changes in composition. These results suggest that, although IBD and recommends evaluation of the effect of anti-inflammatory rats reach sexual maturity at six weeks of age, their gut micro- drugs on FMT treatment efficacy. Moreover, pre-clinical models, bial community may still be developing until at least five months such as the present one, allow the analysis of a wide range of clin- of age, the age that these rats reached after the 112-day follow- ical features, including intestinal length, and also mucosal tissue up. We propose that future studies determine the age at which harvesting and a long-term follow-up, which would not be easily rats should be tested in order to achieve greater stool transfer effi- performed with patients. ciency. Moreover, a longitudinal setting has allowed us to evaluate Our study shows for the first time that a conventional rat the resilience of the rat microbiome after HST and to monitor the model, more reflective of human microbiota, is a potential alter- disease severity index after withdrawal of DSS treatment. native to a mouse model to study microbe-host relationships in Our study not only confirmed previous studies regarding the human disease. We demonstrate that a healthy stool transfer com- increase in DAI and histological scores and the decrease in colon pensates faecal dysbiosis, with a restoration of alpha-diversity and length upon DSS treatment but also revealed a lower microbial di- an increase in the relative abundance of health-related genera. In versity in stools, as well as in colonic tissues. These characteristics addition, we show that a single faecal transplantation also cor- are also observed in both UC and CD patients [6,39] and are more rects, to some extent, colonic tissue alterations, leading to a rela- marked in the latter [7]. Many studies have shown that rat models tive restoration of colon length, and a decrease in epithelium dam- of DSS-induced colitis share more features with UC than CD pa- age and disease severity. Finally, our data suggest that new strate- tients [19,40], from a structural, clinical and ultrastructural point gies based on the combination of anti-inflammatory drugs with of view. Our comparison at microbiome level also confirmed that FMT may allow a better clinical outcome in IBD patients. the rat model of DSS-induced colitis was relatively closer to UC than to CD patients. 5. Authors’ contributions In an attempt to understand the mechanism underlying a fae- cal microbiota transplantation as a potential treatment for IBD, we CM: Study concept and design; AS, GS and JW performed the transferred a single stool collected from a healthy human donor to animal experiments and acquired biological samples; AS, GS and a humanised DSS-induced colitis rat model. Our findings revealed JW processed samples for microbiome and immune response anal- an improvement in several clinical and microbial variables. The rel- yses; ML and MP analysed all the data; GS, ML and CM interpreted ative recovery of colon length in response to HST confirmed an ef- the data; CM drafted the manuscript; all authors provided critical fect (direct or indirect) of an unaltered stool microbiota on modi- revision of the manuscript for intellectual content; CM: obtained fying the structure of the colon. This effect overpowered the action funding. ML and GS contributed equally. of DSS, as its treatment in group E-HST2 was maintained until the end of the study. However, a complete restoration of colon length was not seen even after withdrawal of DSS treatment 40 days be- 6. Funding fore the end of the study. This result is not consistent with the finding of a previous study by Bento et al. using a mouse model Study funded by the Instituto de Salud Carlos III/FEDER of DSS-induced colitis [34]. Those authors demonstrated complete (PI17/00614), a government agency. The funder had no role in M. Lleal, G. Sarrabayrouse and J. Willamil et al. / EBioMedicine 48 (2019) 630–641 641 study design, data collection, data analysis, interpretation or writ- [19] Gaudio E, Taddei G, Vetuschi A, Sferra R, Frieri G, Ricciardi G, et al. Dextran ing of the report. sulfate sodium (DSS) colitis in rats: clinical, structural, and ultrastructural as- pects. Dig Dis Sci 1999;44:1458–75. [20] Zhou J, Zhou Z, Ji P, Ma M, Guo J, Jiang S. Effect of fecal microbiota transplan- Declaration of Competing Interest tation on experimental colitis in mice. Exp Ther Med 2019;17:2581–6. [21] El-Salhy M, Wendelbo IH, Gundersen D, Hatlebakk JG, Hausken T. Colonoscopy with mucosal biopsies in young rats: a model for experimental gastroenterol- None declared. ogy. Mol Med Rep 2013;7:1757–60. [22] Oishi M, Tokuhara K, Miki H, Tanaka Y, Yamaki S, Kaibori M, et al. Tempo- Acknowledgments ral and spatial dependence of inflammatory biomarkers and suppression by fluvastatin in dextran sodium sulfate-induced rat colitis model. Dig Dis Sci 2014;59:2126–35. We thank Dr. Anna Barceló for her technical support at the Il- [23] Murthy SN, Cooper HS, Shim H, Shah RS, Ibrahim SA, Sedergran DJ. Treatment lumina sequencing platform (Servei de Genòmica i Bioinformàtica, of dextran sulfate sodium-induced murine colitis by intracolonic cyclosporin. Dig Dis Sci 1993;38:1722–34. UAB, Barcelona). [24] Erben U, Loddenkemper C, Doerfel K, Spieckermann S, Haller D, Heime- saat MM, et al. A guide to histomorphological evaluation of intestinal inflam- Supplementary material mation in mouse models. Int J Clin Exp Pathol 2014;7:4557–76. [25] Santiago A, Sanchez E, Clark A, Pozuelo M, Calvo M, Yanez F, et al. Sequen- tial changes in the mesenteric lymph node microbiome and immune response Supplementary material associated with this article can be during cirrhosis induction in rats. mSystems 2019;4. found, in the online version, at doi:10.1016/j.ebiom.2019.10.002. [26] Navas-Molina JA, Peralta-Sanchez JM, Gonzalez A, McMurdie PJ, Vazquez-Baeza Y, Xu Z, et al. Advancing our understanding of the human References microbiome using qiime. Methods Enzymol 2013;531:371–444. [27] Edgar RC. Search and clustering orders of magnitude faster than blast. Bioin- formatics 2010;26:2460–1. [1] Fiocchi C. Inflammatory bowel disease: etiology and pathogenesis. Gastroen- [28] Davis NM, Proctor DM, Holmes SP, Relman DA, Callahan BJ. Simple statisti- terology 1998;115:182–205. cal identification and removal of contaminant sequences in marker-gene and [2] Guarner F, Malagelada JR. Gut flora in health and disease. Lancet metagenomics data. Microbiome 2018;6:226. 2003;361:512–19. [29] Chao A, Chazdon RL, Colwell RK, Shen T-J. Abundance-based similarity indices [3] Baumgart DC, Sandborn WJ. Inflammatory bowel disease: clinical aspects and and their estimation when there are unseen species in samples. Biometrics established and evolving therapies. Lancet 2007;369:1641–57. 2006;62:361–71. [4] Frank DN, St Amand AL, Feldman RA, Boedeker EC, Harpaz N, Pace NR. Molec- [30] Shannon CE. A mathematical theory of communication. Bell System Technical ular-phylogenetic characterization of microbial community imbalances in hu- Journal 1948;27:379–423. man inflammatory bowel diseases. Proc Natl Acad Sci U S A 2007;104:13780–5. [31] Lozupone C, Knight R. UniFrac: a new phylogenetic method for comparing mi- [5] Manichanh C, Rigottier-Gois L, Bonnaud E, Gloux K, Pelletier E, Frangeul L, crobial communities. Appl Environ Microbiol 2005;71:8228–35. et al. Reduced diversity of faecal microbiota in Crohn’s disease revealed by a [32] Knights D, Costello EK, Knight R. Supervised classification of human micro- metagenomic approach. Gut 2006;55:205–11. biota. FEMS Microbiol Rev 2011;35:343–59. [6] Ott SJ, Musfeldt M, Wenderoth DF, Hampe J, Brant O, Folsch UR, et al. Re- [33] Chassaing B, Aitken JD, Malleshappa M, Vijay-Kumar M. Dextran sulfate duction in diversity of the colonic mucosa associated bacterial microflora in sodium (DSS)-induced colitis in mice. Curr Protoc Immunol 2014;104:15–25. patients with active inflammatory bowel disease. Gut 2004;53:685–93. [34] Bento AF, Leite DF, Marcon R, Claudino RF, Dutra RC, Cola M, et al. Evalua- [7] Pascal V, Pozuelo M, Borruel N, Casellas F, Campos D, Santiago A, et al. A mi- tion of chemical mediators and cellular response during acute and chronic gut crobial signature for Crohn’s disease. Gut 2017;66:813–22. inflammatory response induced by dextran sodium sulfate in mice. Biochem [8] van Nood E, Vrieze A, Nieuwdorp M, Fuentes S, Zoetendal EG, de Vos WM, Pharmacol 2012;84:1459–69. et al. Duodenal infusion of donor feces for recurrent clostridium difficile. N [35] Park YH, Kim N, Shim YK, Choi YJ, Nam RH, Choi YJ, et al. Adequate dextran Engl J Med 2013;368:407–15. sodium sulfate-induced colitis model in mice and effective outcome measure- [9] Ianiro G, Masucci L, Quaranta G, Simonelli C, Lopetuso LR, Sanguinetti M, ment method. J Cancer Prev 2015;20:260–7. et al. Randomised clinical trial: faecal microbiota transplantation by [36] Gibbs RA, Weinstock GM, Metzker ML, Muzny DM, Sodergren EJ, Scherer S, colonoscopy plus vancomycin for the treatment of severe refractory clostrid- et al. Genome sequence of the brown Norway rat yields insights into mam- ium difficile infection-single versus multiple infusions. Aliment Pharmacol malian evolution. Nature 2004;428:493–521. Ther 2018;48:152–9. [37] Pan H, Guo R, Zhu J, Wang Q, Ju Y, Xie Y, et al. A gene catalogue of the [10] Moayyedi P, Surette MG, Kim PT, Libertucci J, Wolfe M, Onischi C, et al. Fecal Sprague-Dawley rat gut metagenome. Gigascience 2018;7. microbiota transplantation induces remission in patients with active ulcerative [38] Staley C, Kaiser T, Beura LK, Hamilton MJ, Weingarden AR, Bobr A, et al. Stable colitis in a randomized controlled trial. Gastroenterology 2015;149:102–9 e6. engraftment of human microbiota into mice with a single oral gavage follow- [11] Paramsothy S, Kamm MA, Kaakoush NO, Walsh AJ, van den Bogaerde J, ing antibiotic conditioning. Microbiome 2017;5:87. Samuel D, et al. Multidonor intensive faecal microbiota transplantation [39] Sokol H, Leducq V, Aschard H, Pham HP, Jegou S, Landman C, et al. Fungal for active ulcerative colitis: a randomised placebo-controlled trial. Lancet microbiota dysbiosis in ibd. Gut 2017;66:1039–48. 2017;389:1218–28. [40] Okayasu I, Hatakeyama S, Yamada M, Ohkusa T, Inagaki Y, Nakaya R. A novel [12] Paramsothy S, Paramsothy R, Rubin DT, Kamm MA, Kaakoush NO, Mitchell HM, method in the induction of reliable experimental acute and chronic ulcerative et al. Faecal microbiota transplantation for inflammatory bowel disease: a sys- colitis in mice. Gastroenterology 1990;98:694–702. tematic review and meta-analysis. J Crohns Colitis 2017;11:1180–99. [41] Garcia-Mantrana I, Selma-Royo M, Alcantara C, Collado MC. Shifts on gut mi- [13] Sartor RB, Wu GD. Roles for intestinal bacteria, viruses, and fungi in patho- crobiota associated to mediterranean diet adherence and specific dietary in- genesis of inflammatory bowel diseases and therapeutic approaches. Gastroen- takes on general adult population. Front Microbiol 2018;9:890. terology 2017;152:327–39 e4. [42] Krych L, Nielsen DS, Hansen AK, Hansen CH. Gut microbial markers are asso- [14] Vermeire S, Joossens M, Verbeke K, Wang J, Machiels K, Sabino J, et al. Donor ciated with diabetes onset, regulatory imbalance, and IFN-gamma level in nod species richness determines faecal microbiota transplantation success in in- mice. Gut Microbes 2015;6:101–9. flammatory bowel disease. J Crohns Colitis 2016;10:387–94. [43] Costello SP, Hughes PA, Waters O, Bryant RV, Vincent AD, Blatchford P, [15] Colman RJ, Rubin DT. Fecal microbiota transplantation as therapy for inflam- et al. Effect of fecal microbiota transplantation on 8-week remission in pa- matory bowel disease: a systematic review and meta-analysis. J Crohns Colitis tients with ulcerative colitis: a randomized clinical trial. JAMA 2019;321:156– 2014;8:1569–81. 164. [16] Fang H, Fu L, Wang J. Protocol for fecal microbiota transplantation in inflam- [44] Paramsothy S, Nielsen S, Kamm MA, Deshpande NP, Faith JJ, Clemente JC, matory bowel disease: a systematic review and meta-analysis. Biomed Res Int et al. Specific bacteria and metabolites associated with response to fecal mi- 2018;2018:8941340. crobiota transplantation in patients with ulcerative colitis. Gastroenterology [17] Pigneur B, Sokol H. Fecal microbiota transplantation in inflammatory bowel 2019;156:1440–54 e2. disease: the quest for the holy grail. Mucosal Immunol 2016;9:1360–5. [45] Rossen NG, Fuentes S, van der Spek MJ, Tijssen JG, Hartman JH, Duflou A, [18] Antoniou E, Margonis GA, Angelou A, Pikouli A, Argiri P, Karavokyros I, et al. Findings from a randomized controlled trial of fecal transplantation for et al. The TNBS-induced colitis animal model: an overview. Ann Med Surg patients with ulcerative colitis. Gastroenterology 2015;149:110–18 e4. (Lond) 2016;11:9–15.