Specific microbiota enhances intestinal IgA levels by inducing TGF‐β in T follicular helper cells of Peyer's patches in mice

In humans and mice, mucosal immune responses are dominated by IgA antibodies and the cytokine TGF‐β, suppressing unwanted immune reactions but also targeting Ig class switching to IgA. It had been suggested that eosinophils promote the generation and maintenance of mucosal IgA‐expressing plasma cells. Here, we demonstrate that not eosinophils, but specific bacteria determine mucosal IgA production. Co‐housing of eosinophil‐deficient mice with mice having high intestinal IgA levels, as well as the intentional microbiota transfer induces TGF‐β expression in intestinal T follicular helper cells, thereby promoting IgA class switching in Peyer's patches, enhancing IgA+ plasma cell numbers in the small intestinal lamina propria and levels of mucosal IgA. We show that bacteria highly enriched for the genus Anaeroplasma are sufficient to induce these changes and enhance IgA levels when adoptively transferred. Thus, specific members of the intestinal microbiota and not the microbiota as such regulate gut homeostasis, by promoting the expression of immune‐regulatory TGF‐β and of mucosal IgA.


Introduction
Antibodies of the IgA isotype are dominant at mucosal surfaces in humans and mice. IgA antibodies are an integral part of the mucosal barrier, controlling the growth and attachment of specific bacteria [1]. Ig class switching to IgA occurs mainly in GC reactions of Peyer's patches (PP) [2], but can also occur in mesenteric lymph nodes, in isolated lymphoid follicles and in the small intestinal lamina propria (siLP), where B cells are activated without the help of T cells [3][4][5]. Class switch recombination is targeted to IgA by TGF-β [6,7], which induces IgA switch transcripts [8,9], an essential prerequisite for IgA antibody class switch recombination [10]. In PP, the class switch to IgA is regulated by T follicular helper (Tfh) cells expressing TGF-β and IL-21 [11,12] and by dendritic cells activating latent TGF-β [13]. An important role in the promotion of mucosal IgA expression has been attributed to eosinophils [14,15]. It has been reported that dblGATA-1 mice, which lack eosinophils [16], have reduced numbers of IgA + GC B cells in their PP, reduced numbers of IgA-expressing plasma cells in their siLP, and lower levels of mucosal IgA than WT mice [14,15]. How eosinophils that are not present in PP can contribute to IgA class switching in PP had remained enigmatic. In the study of Chu and colleagues, dblGATA-1 mice and WT control mice had been derived from different animal facilities, with different composition of the microbiota. Here, we report that allowing the exchange of microbiota by cohousing equalizes their levels of mucosal IgA.
It has been known for some time that the expression of mucosal IgA is also dependent on microbiota. Germ-free mice show reduced production of intestinal IgA, unless they are repopulated with bacteria [17,18]. So far, members of the intestinal microbiota that enhance IgA expression specifically and selectively have not yet been identified. Here, we report the serendipitous observation when comparing the microbiome of dblGATA-1 mice and WT control mice that only the presence of bacteria of the genus Anaeroplasma in the microbiota correlated with increased TGF-β expression in Tfh cells of PP, enhanced class switch recombination to IgA in GC B cells of the PP, increased numbers of IgA + plasma cells in the siLP, and enhanced expression of mucosal IgA. Adoptive transfer of bacteria highly enriched in Anaeroplasma and depleted for most other bacteria into secondary abiotic mice resulted in the generic induction of high intestinal IgA levels, supporting the notion that Anaeroplasma specifically and not bacteria as such are dictating TGF-β and IgA expression in the small intestine. Anaeroplasma could thus have the potential as prime candidates for effective anti-inflammatory probiotics.

Results
The intestinal microbiota determine mucosal IgA levels independently of eosinophils It had been previously described that dblGATA-1 mice, lacking eosinophils, also have reduced mucosal IgA, suggesting that eosinophils are regulating mucosal IgA production [14,15]. In view of recent findings, that the phenotypic manifestation of a par-ticular genotype can be dependent on the microbiota (reviewed in [19]), we cohoused dblGATA-1 mice bred and maintained at the German Rheumatism Research Center (DRFZ) with BALB/c mice purchased from Charles River for 3 weeks to allow for the exchange of microbiota by coprophagy. Cohousing of dblGATA-1 mice and BALB/c mice equalized their expression of mucosal IgA. Non-cohoused dblGATA-1 mice expressed 0.035 ± 0.003 μg/ mg of fecal IgA, and non-cohoused BALB/c mice about four times more (0.166 ± 0.020 μg/mg; Fig. 1A). The absolute number of IgA + B220 − plasma cells (1.28 ± 0.34 × 10 5 ) was more than 10fold lower in the siLP of non-cohoused dblGATA-1 mice when compared to BALB/c mice (15.11 ± 2.30 × 10 5 ; Fig. 1B and C), and in the PP of non-cohoused dblGATA-1 mice the numbers of IgA + PNA hi B220 + GC B cells were more than 20-fold lower than in non-cohoused BALB/c mice (6.50 ± 1.29 × 10 3 vs. 145.1 ± 30.3 × 10 3 ; Fig. 1D, E, and G). Interestingly, while the frequency of IgG1 + GC B cells was increased in the PP of dblGATA-1 (Fig. 1D), absolute numbers of switched IgG1 + GC B cells did not differ between non-cohoused mice of the two strains (5.05 ± 1.75 × 10 3 vs. 3.43 ± 0.96 × 10 3 ; Fig. 1F). Fecal IgA concentrations were not significantly different between cohoused dblGATA-1 mice and BALB/c mice (0.163 ± 0.046 μg/mL vs. 0.145 ± 0.039 μg/mL; Fig. 1A). Absolute numbers of IgA + plasma cells in the siLP (19.32 ± 4 × 10 5 ) and of IgA + GC B cells in the PP of dblGATA-1 mice (60.92 ± 21.65 × 10 3 ) were no longer significantly different from those of BALB/c mice (21.96 ± 3.79 × 10 5 and 189.4 ± 64.8 × 10 3 ; Fig. 1B-E and G). Numbers of IgG1 + GC B cells in the PP were not affected by cohousing ( Fig. 1D and F). Taken together, these data show that eosinophils are dispensable for the regulation of mucosal IgA expression, and more importantly, that distinct components of the murine microbiota, rather than the microbiota as such, regulate mucosal IgA expression. In line with this notion, mice from any strain of the DRFZ breeding facility had consistently lower titers of fecal IgA (<0.05 μg IgA/mg feces) than mice of the strains we obtained from the Charles River breeding facility (>0.08 μg IgA/mg feces; Supporting Information Fig. 1).
The fecal microbiome composition of BALB/c and dblGATA-1 mice under non-cohoused and cohoused conditions was compared by next-generation 16S rDNA sequencing of V3/V4 regions. The fecal microbiome was substantially different between dblGATA-1 from the DRFZ and BALB/c mice from Charles River (Fig. 1H). However, we did not observe substantial differences in the number of bacterial genera detected, or in Shannon and Simpson diversity indices of the fecal microbiome of BALB/c and dblGATA-1 mice (Supporting Information Fig. 2A-C). Following 3 weeks of cohousing, Bray-Curtis similarity clustering on the family level revealed that the microbiomes between BALB/c and dblGATA-1 can no longer be distinguished ( Fig. 1I and J), indicating an exchange of the microbiota by coprophagy.

The IgA high phenotype is transferred in a dominant fashion by the microbiota
To confirm that (1) the IgA phenotype is conferred by the microbiota and (2) is independent of the dblGATA-1 genotype, we C 2020 The Authors. European Journal of Immunology published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Specific members of the microbiota enhance mucosal IgA
To elucidate which bacterial species participate in inducing the IgA high phenotype, we compared the microbiome of BALB/c and dblGATA-1 mice filtering for bacterial OTUs that are present in BALB/c and non-detectable in dblGATA-1 ( Fig. 3A; Supporting Information Fig. 3) and that can be detected in dblGATA-1 mice after cohousing with BALB/c mice (Fig. 3B). Of 100 bacterial OTUs detected in total (Supporting Information Fig. 2A), 19 OTUs were exclusively detected in the microbiome of BALB/c mice (Fig. 3A) and only two, Anaeroplasma and Erysipelotrichaceae incertae sedis, were consistently detected in dblGATA-1 mice following cohousing (Fig. 3B), thus qualifying them as prime candidates for the induction of IgA.
To analyze whether Anaeroplasma (Phylum Tenericutes) and/or Erysipelotrichaceae incertae sedis (Phylum Firmicutes) can induce IgA, we separated them using a 0.45-μm pore size filter. Tenericutes are 0.2-0.5 μm in size [20], smaller than most bacteria of the intestinal microbiota. Anaeroplasma were the only representatives of the phylum Tenericutes within the BALB/c intestinal microbiomes. By filtration of the fecal microbiota of BALB/c mice, Tenericutes, i.e., Anaeroplasma, could be separated from Erysipelotrichaceae incertae sedis and from all other genera, which were identified before to be specific for the microbiome of our BALB/c mice (Fig. 3C). Tenericutes could be enriched by several orders of magnitude by filtration of the fecal microbiota of BALB/c mice (Fig. 3D). The Anaeroplasma-enriched filtrate was then transferred into antibiotics-treated, microbiota-ablated BALB/c mice (Fig. 3E). The filtrate was supplemented with Anaeroplasmanegative dblGATA-1 microbiota, since it has been suggested that some Anaeroplasma strains are auxotrophs lacking sterol synthesizing capability [20]. Transfer of Anaeroplasma-enriched filtrate enhanced expression of fecal IgA from 0.255 ± 0.017 to 0.469 ± 0.055 μg/mg, comparable to the transfer of the entire BALB/c microbiota (from 0.307 ± 0.014 μg/mg feces to 0.612 ± 0.051 μg/mg feces), strongly suggesting that Anaeroplasma are indeed IgA inducing bacteria (Fig. 3F). Mice not having received any microbiota after antibiotic treatment showed no increase in their fecal IgA concentration (Supporting Information Fig. 4A). The serum IgA concentration was not affected by the presence of Anaeroplasma in the transferred microbiota (Supporting Information Fig. 4B). Numbers of IgA + plasma cells in the siLP were significantly increased in recipients of Anaeroplasma-enriched microbiota in comparison to recipients of microbiota lacking Anaeroplasma ( Fig. 3G and H). Since filtration of BALB/c microbiota through a 0.1-μm filter did not augment IgA levels (Supporting Information Fig. 4C), the presence of the bacteria, rather than any secreted product, is the main drivers of IgA induction. By FISH using an Anaeroplasma-specific 16S rDNA probe, we could localize Anaeroplasma in the small intestine of BALB/c mice (Fig. 3I). Anaeroplasma could not be detected by FISH in the large intestine of BALB/c mice or the small intestine of dblGATA-1 mice (Fig. 3J).

Generic induction of IgA by Anaeroplasma-containing microbiota
To determine whether the presence of Anaeroplasma results in the general, generic enhancement of IgA levels or rather induces IgA of distinct specificities, e.g., against Anaeroplasma itself, we stained fecal bacteria of Rag1 −/− mice, lacking own IgA, with fecal IgA of BALB/c mice having received Anaeroplasma (−) or (+) microbiota (Supporting Information Fig. 5A). Staining of the IgAcoated bacteria was then evaluated by high-resolution microbiota flow cytometry [21]. Microbiota stained with IgA of Anaeroplasma (−) and (+) mice showed the same cytometric profile (Supporting Information Fig. 5B and C), and sequencing of IgA-bound bacteria (IgA-Seq [22]) confirmed that the composition of IgA-bound bacteria was indistinguishable between the two groups (Supporting Information Fig. 5E-G). However, the Rag1 −/− microbiota were stained two to four times brighter (MFI) by IgA of Anaeroplasma (+) mice (Supporting Information Fig. 5D), corresponding to the difference in fecal IgA levels observed between Anaeroplasma (+) and (−) mice (Fig. 3F). Thus, Anaeroplasma appear to enhance mucosal IgA responses, irrespective of the antigen targeted, and qualify as an IgA enhancing adjuvant. It may be also effective in humans, since fecal IgA levels of healthy human donors correlate to the amount of Anaeroplasma detected by quantitative PCR in these feces (Supporting Information Fig. 6).

Anaeroplasma induce TGF-β expression in T follicular helper cells of Peyer's patches
The selective and dominant effect of Anaeroplasma-enriched microbiota on absolute numbers of IgA + PNA hi B220 + GC B lymphocytes in the PP (Figs. 1E and 2G) suggests that Anaeroplasma directly influence the regulation of antibody class switch in B lymphocytes. TGF-β is the cytokine targeting antibody class switch recombination to IgA [6][7][8][9], and in PP, it is expressed by Tfh controlling activation and differentiation of B lymphocytes [12]. We isolated CD3 + CD4 + CXCR5 + PD-1 + Tfh and CD3 + CD4 + CXCR5 − PD-1 − non-Tfh cells from PP of BALB/c mice, dblGATA-1 mice, and of dblGATA-1 mice cohoused with BALB/c mice for 3 weeks (Supporting Information Fig. 7A). Expression of Tgfb1 mRNA was determined in these cells by quantitative RT-PCR directly ex vivo. Tfh cells of Anaeroplasma-deficient dblGATA-1 mice express significantly less Tgfb1 than Tfh cells of Anaeroplasma-sufficient BALB/c mice and those of co-housed dblGATA-1 mice (Fig. 4A). This effect was selective for TGF-β, since Il4 mRNA expression did not differ between Tfh of the various mice (Fig. 4B). It was also selective for Tfh lymphocytes, since in MHC class II-positive cells of the siLP and in non-Tfh of PP Tgfb1 expression did not differ between Anaeroplasma-sufficient and -deficient mice (Supporting Information Fig. 7B and C). In intestinal epithelial cells, Tgfb1 transcript was not detectable (data not shown). In vitro, a 0.45-μm filtrate of BALB/c microbiota containing Anaeroplasma could significantly enhance Tgfb1 expression by a factor of 3 in splenocytes activated with anti-CD3 and anti-CD28 antibodies, as compared a 0.45-μm filtrate of dblGATA-1 microbiota, not containing Anaeroplasma (Fig. 4C), and control cultures with no filtrate added. Apparently, the induction of TGFβ expression in Tfh of PP is the mechanism by which Anaeroplasma enhance mucosal IgA expression.

Discussion
IgA antibodies [23] and the cytokine TGF-β [24] dominate mucosal immunity. It is less clear how TGF-β and IgA are targeted in mucosal immune reactions. Here, we report identification of a distinct member of the intestinal microbiota, which generically enhances mucosal IgA expression in a dominant fashion, probably by inducing in mucosal Tfh the expression of TGF-β, the cytokine targeting antibody class switch recombination to IgA [8,9]. So far, the species-specific induction of IgA immune reactions by bacteria of the microbiota has been described [17,18,22], but it has been unclear whether such bacteria induce mucosal immune reactions predisposed to TGF-β/IgA, or whether they themselves target TGF-β and/or IgA expression.
Eosinophil-deficient dblGATA-1 mice had been described to be impaired for mucosal IgA [14,15], however, the role of differences in the microbiota between dblGATA-1 and their WT BALB/c counterpart were not considered in these studies. We cohoused eosinophil-deficient dblGATA-1 bred and maintained at the animal facility of the DRFZ with BALB/c mice obtained from Charles River for 3 weeks and could show, that dblGATA-1 mice increase expression of mucosal IgA, numbers of IgA + GC B cells in their PP, and numbers of IgA + plasma cells in their siLP to the same level as BALB/c mice. The assumption that the enhancement of mucosal IgA expression in dblGATA-1 mice had been due to an uptake of BALB/c microbiota by coprophagy was confirmed by oral gavage of BALB/c microbiota into antibioticstreated, secondary abiotic BALB/c mice. In these mice, only the microbiota of BALB/c mice but not that of dblGATA-1 mice reconstituted mucosal IgA levels. Below the line, these data show that the microbiota of BALB/c mice from the Charles River animal facility contains bacteria that are able to enhance mucosal IgA expression.
By comparing the microbiomes of the dblGATA-1 with the BALB/c mice using 16S rDNA sequencing, we could identify only two bacterial genera correlated to the enhanced expression of IgA and were transferred by cohousing from BALB/c mice into dblGATA-1 mice, namely Anaeroplasma, and Erysipelotrichaceae incertae sedis. Anaeroplasma and Erysipelotrichaceae incertae sedis could be physically separated by filtration, and only Anaeroplasma enhanced mucosal IgA expression in antibiotics-treated, secondary abiotic BALB/c mice. Presumably, Anaeroplasma do so in a unique way, namely by enhancing expression of TGF-β in Tfh of the PP. TGF-β-receptor signaling in B cells targets antibody class switch recombination to IgA [6,25,26] inducing Iα germline transcripts [27][28][29]. Several cell types of the mucosa have been shown to be able to produce TGF-β: Dendritic cells of PP [30][31][32], intestinal epithelial cells (IEC) [33], and PP follicular dendritic cells [34]. These cells express TGF-β upon stimulation of TLR signaling. This  may be a mechanism to sense microbiota as such and provide a basic level of expression of TGF-β and IgA. It should be noted that Anaeroplasma, like all members of the constituting phylum Tenericutes, do not have a cell wall [20] and could potentially escape notice of inflammatory cell wall-detecting pattern recognition receptors. Obviously, Anaeroplasma, enhancing this basic IgA expression two-to fivefold, is not using this pathway, since we did not observe any difference in TGF-β expression in MHC class II + cells of PP or non-Tfh cells of PP and no expression of TGF-β in IEC of Anaeroplasma-deficient and -sufficient mice. However, Tfh of PP differed significantly in their transcription of the Tgfb1 gene. Tfh cells regulate T-dependent (TD) mucosal IgA responses [12,35]. In vitro, Anaeroplasma also induce Tgfb1 transcription in activated splenocytes. So far, the mechanism is unclear, but it is C 2020 The Authors. European Journal of Immunology published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.
www.eji-journal.eu obviously mediated by the bacterium itself and not a secreted product, since the 0.1 μm filtrate of Anaeroplasma-containing microbiota does not enhance mucosal IgA expression. By FISH, we localized Anaeroplasma to the small intestine of BALB/c but not dblGATA1 mice, closely associated with the epithelium but not in PP, suggesting that in vivo Anaeroplasma induce Tgfb1 expression in Tfh cells indirectly. In summary, Anaeroplasma qualify as a key anti-inflammatory component of the microbiota, since both, TGF-β, and IgA have been shown to protect from intestinal inflammation [36,37].
Bacteria that affect IgA levels in the gut have been described previously. Mice monocolonized with "segmented filamentous bacteria" (SFB) show increased IgA levels compared to mice colonized with Escherichia coli. This has been attributed to the vigorous induction of germinal centers in secondary and tertiary lymphoid structures in the intestine [38]. In the BALB/c and dblGATA-1 mice analyzed here, SFBs were present in both microbiomes. Bacterial metabolites, in particular short-chain fatty acids, have been shown to promote IgA responses in the gut. They apparently do so by boosting the differentiation of B cells into antibodysecreting plasma cells. The levels of systemic IgG increase, too [39]. Moon et al. have described a Sutterella-containing murine microbiota, which degraded secretory IgA [40]. The Sutterellamicrobiota induced IgA low phenotype is dominant. In contrast, Anaeroplasma-containing microbiota induces a dominant IgA high phenotype. Obviously, Anaeroplasma are special among IgA regulating bacteria of the intestinal microbiota, in that they enhance TGF-β and IgA expression in TD mucosal immune reactions generically and impressively. We could show that also in humans, levels of fecal IgA correlate to levels of fecal Anaeroplasma. Anaeroplasma may thus qualify as a robust adjuvant or probiotic to enhance mucosal immunity. Although the cultivation of Anaeroplasma has been reported previously [41], we have not been able to reproduce these findings so far. Nevertheless, our data based on the isolation of Anaeroplasma by physical means provide compelling evidence that Anaeroplasma are an IgA and TGF-β inducing component of the intestinal microbiota.
Apart from the identification of Anaeroplasma as IgA and TGF-β inducing bacteria, the present experiments also show that eosinophils are dispensable for the regulation of mucosal IgA. In the presence of Anaeroplasma, dblGATA-1 mice, deficient for eosinophils, show the same numbers of IgA + GC B cells in PP, IgA + plasma cells in siLP, and fecal IgA, as WT BALB/c mice.

Materials and methods
Mice BALB/c and Rag1 −/− mice were purchased from Charles River Laboratories, Sulzfeld, Germany. dblGATA-1 [16] mice were bred under specific pathogen-free conditions in the DRFZ breeding facility at the Bundesinstitut für Risikobewertung, Berlin. All animal experiments were approved by the regulatory office "Landesamt für Gesundheit und Soziales" in Berlin, Germany (permit number G0045/17).

Cohousing and microbiota transplantation
For cohousing experiments, BALB/c and dblGATA-1 females were kept in the same cage at a ratio of 1:1 for 3 weeks. For the depletion of the native intestinal microbiota, BALB/c mice were treated with ampicillin (1 mg/mL ad libitum) via the drinking water and with additional gavage feedings b.i.d. of neomycin 5 mg/mL, metronidazole 10 mg/mL and vancomycin 5 mg/mL (all purchased from Sigma, Darmstadt) for 14 days. Twenty-four hours after the last antibiotics administration the corresponding fecal microbiota was transferred via gavage on three consecutive days. For this, fresh fecal pellets of BALB/c (IgA high ) or dblGATA-1 (IgA low ) mice were collected and resuspended in PBS (one pellet in 1 mL). Following filtration through a 30 μm cell strainer (Partec, Görlitz), 200 μL of the suspension were transferred via gavage feeding. The mice were analyzed 3 weeks after the last microbiota administration.

Determination of soluble IgA levels
Fresh feces were collected, weighed, and resuspended in PBS (1 mg feces in 10 μL PBS). The levels of fecal IgA in supernatants as well as in blood sera were determined by ELISA (ELISA antibodies purchased from Southern Biotech via BIOZOL Diagnostica, Eching).

Immunofluorescence of PP tissue sections
PP of non-cohoused and cohoused BALB/c and dblGATA-1 mice were excised from the intestines, washed in ice-cold PBS, embedded in O.C.T. compound (Sakura Finetek; Alpen aan den Rijn, NL), and frozen. The frozen tissue was cut into 7 μm sections, which were fixed in 100% acetone (Carl Roth; Karlsruhe) for 20 min. For the immunofluorescence staining, the sections were blocked with anti-FcγR antibody (clone 2.4G2; DRFZ), following by the staining with the goat anti-mouse IgA-FITC (Southern Biotech via BIOZOL Diagnostica, Eching) and anti-mouse IgG1-PE (RMG1-1; Biolegend; Fell). The images acquired using a laser scanning microscope LSM 710 (Carl Zeiss) and ZEN2011 software.

Cell isolation
For the isolation of siLP cells, PP were removed and small intestines were longitudinally opened and washed in cold PBS. After cutting into short pieces, intestines were incubated twice in RPMI 1640 medium containing 25 mM EDTA, 10% FCS, 100 U/mL penicillin, and 100 μg/mL streptomycin (P/S) for 15 min C 2020 The Authors. European Journal of Immunology published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim. at 37°C followed by vigorous shaking to remove the epithelium. Intestinal pieces were washed in PBS, homogenized using scalpel, and digested twice with RPMI 1640 containing 10% FCS, P/S, 1 mg/mL Collagenase D (Roche, Mannheim), 0,1 mg/mL DNase I, and 1 mg/mL Dispase II (Sigma, Darmstadt) for 20 min at 37°C in a shaker at 200 rpm. Cell suspensions were passed through an 18 gauge needle and filtered through a 70 μm cell strainer (BD Biosciences, Heidelberg). After washing with RPMI1640 medium (10% FCS, P/S), cells were purified by 30% Percoll (GE Healthcare, Hamburg) solution in RPMI 1640 medium (10% FCS, P/S). Cells were then washed and resuspended in RPMI 1640 medium (10% FCS, P/S). To isolate cells from PP, tissue was removed from the small intestine, meshed through a 70 μm cell strainer (BD Biosciences, Heidelberg) and resuspended in RPMI 1640, 10% FCS, P/S medium.

Analysis of fecal bacterial DNA
Bacterial DNA was isolated from frozen murine fecal pellets using ZR Fecal DNA MiniPrep Kit (Zymo Research, Freiburg) according to the manufacturer's protocol. The relative amount of specific bacterial phyla was quantified using a real-time PCR (95°C 30 s; 55°C 30 s, 72°C 30 s; 55 cycles) and 16S rRNA gene-targeted phylum specific primers [43]. The total amount of Eubacteria [44] was used as an internal reference. The deep sequencing analysis was performed by LGC Genomics, Berlin, Germany using Illumina MiSeq V3 Chemistry (300 bp paired-end read) and Bacteria 16S (341F-785R) amplicon. The preparation of 16S rDNA templates, construction of Illumina libraries as well as the processing of reads occurred as described previously [21]. Combined reads were classified using "classifier.jar" from the Ribosomal Database Project [45] with a confidence cutoff of 50% and agglomerated at Genus level using RDPUtils [Quensen J. RDPutils: R Utilities for Processing RDPTool Output. R package version 1.2]. Frequencies of bacterial genera were estimated using the copy number-adjusted output and total bacterial counts. Heatmaps represent confidentially classified bacterial genera, with frequencies above 0.01% in noncohoused BALB/c and below in non-cohoused dblGATA-1 mice. The alpha and beta diversity estimation were performed without removing unclassified taxa and resampling of samples to equal sizes by the "vegan" (Oksanen, J. et al. Vegan: Community Ecology Package) and "phyloseq" [46] packages. The Shannon as well as the Simpson index were computed using the "estimate richness" function. The non-metric multidimensional scaling was performed by the "ordinate" function in default parameter settings using the Bray-Curtis distance. Raw sequence data were deposited at the GEO database under the accession number PRJNA607006.

Preparation of microbiota filtrates
Fresh feces from BALB/c mice were collected and resuspended in ice-cold PBS. The fecal suspension was subsequently filtered through a 70 μm (Corning), 30 μm (Partec, Görlitz) cell strainer, 5 μm (Merck via Sigma, Darmstadt) and 0.45 μm (Sarstedt, Nümbrecht) bacterial syringe filters. After centrifugation, the pellet was added to the microbiota suspension of dblGATA1 mice. Two hundred microliters of the final product was administered per mouse by gavage.

Fluorescent in situ hybridization
Intestinal mouse tissue biopsies were embedded in Tissue-Tek R O.C.T. TM Compound. Four micrometer thick sections were placed on SuperFrost plus slides, fixed with Carnoy's solution on ice for 20 min, and left to dry. Hybridization was performed in hybridization buffer (900 mM NaCl, 100 mM Tris-HCl (pH 8.0), 6.7 mM formamide, 346.7 mM SDS, 2 ng/μL DNA probe) at 50°C for 90 min. The slides were rinsed with de-ionized water, washed with washing buffer (900 mM NaCl, 100 mM Tris-HCl (pH 8.0), 346.7 mM SDS) for 5 min at 50°C, and rinsed again with deionized water. After drying, the sections were stained with DAPI solution (3.6 μM DAPI, 900 mM NaCl, 20 mM Tris-HCl) at room temperature for 5 min and rinsed with de-ionized water. Hybridization was performed in the a-Hyb hybridization station (Miltenyi Biotec). Dry sections were mounted and examined by fluorescence microscopy (Keyence Biorevo BZ-9000). The following probe conjugated with AlexaFluor 647 was used: AnaeroF1: 5'-ATGTAAAGTTCTTTTATCAG-3'.

Statistical analysis
Statistical analysis was done with GraphPad Prism 5.04. The corresponding tests are indicated in figure legends.
Nancy-Ann Erickson, Anna Pfeffer for taking expert care of our animals. We thank Dr. Alexander Swidsinski and Petra Schlien for the introduction to FISH. We also thank Dr. Sergei Nedospasov and Dr. Randall Lindquist for critical reading of the manuscript.