Apyrase-mediated ampliﬁcation of secretory IgA promotes intestinal homeostasis

SUMMARY Secretory immunoglobulin A (SIgA) interaction with commensal bacteria conditions microbiota composition and function. However, mechanisms regulating reciprocal control of microbiota and SIgA are not deﬁned. Bac-teria-derived adenosine triphosphate (ATP) limits T follicular helper (Tfh) cells in the Peyer’s patches (PPs) via P2X7 receptor (P2X7R) and thereby SIgA generation. Here we show that hydrolysis of extracellular ATP (eATP) by apyrase results in ampliﬁcation of the SIgA repertoire. The enhanced breadth of SIgA in mice colonized with apyrase-releasing Escherichia coli inﬂuences topographical distribution of bacteria and expression of genes involved in metabolic versus immune functions in the intestinal epithelium. SIgA-mediated conditioning of bacteria and enterocyte function is reﬂected by differences in nutrient absorption in mice colonized with apyrase-expressing bacteria. Apyrase-induced SIgA improves intestinal homeostasis and attenuates barrier impairment and susceptibility to infection by enteric pathogens in antibiotic-induced dysbiosis. Therefore, ampliﬁcation of SIgA by apyrase can be leveraged to restore intestinal ﬁtness in dysbiotic conditions

Correspondence fabio.grassi@irb.usi.ch In brief Secretory IgA plays pleiotropic function in ensuring the integrity of the intestinal ecosystem. Perruzza et al. show that amplification of gut SIgA repertoire by hydrolysis of endoluminal extracellular ATP can condition enterocyte transcriptional activity and promote intestinal homeostasis and colonization resistance in dysbiosis.

INTRODUCTION
The intestine ensures the digestion and absorption of nutrients and, concomitantly, the establishment and maintenance of a beneficial microbial community (Chow et al., 2011). The gut microbiota is essential for intestinal and immune system differentiation, tissues homeostasis, and systemic metabolism (Sommer and Backhed, 2013). Alterations in the microbial community structure have been associated to susceptibility to diseases (Blaser, 2006), and dysbiosis to an increasing number of medical conditions, including metabolic disorders (e.g., diabetes, obesity), blood pressure alteration, and autoimmunity (Cerf-Bensussan and Gaboriau-Routhiau, 2010;Dicksved et al., 2008;Holmes et al., 2008;Larsen et al., 2010;Scher et al., 2013;Turnbaugh et al., 2009;Wu et al., 2010). Many factors contribute to the shaping of the gut microbiota, but specific mechanisms responsible for host microbiota mutualism are not thoroughly understood.
The glycan-rich gut mucous layer constitutes an essential niche for symbionts by providing nutrients and a scaffold for growth (Backhed et al., 2005;Li et al., 2015). Secretory immunoglobulin A (SIgA) may enhance commensal colonization of this microbial niche by promoting adhesion and/or nutrient utilization of bacteria within the colonic mucus (Johansson et al., 2008;Rogier et al., 2014). In fact, immunoglobulin A (IgA)-coated bacteria contribute to host physiology and metabolism  and are important for the preservation of commensals diversity and community networks in the human gut (Fadlallah et al., 2018). Recently, a regulatory system where IgA fosters mucosal colonization of the human intestinal commensal Bacteroides fragilis was described (Donaldson et al., 2018;Lee et al., 2013). Furthermore, glycan-dependent, epitope-independent IgA coating of Bacteroides thetaiotaomicron, a prominent gut symbiont of the phylum Bacteroidetes, regulated gene transcription and metabolism by inducing the expression of polysaccharide utilization loci, which in turn promoted symbiosis with other members of the gut microbiota and colonic homeostasis (Nakajima et al., 2018). Because the immune system and the gut microbiota start developing together at birth, it has been hypothesized that their Germ-free (GF) C57BL/6 mice were colonized with 10 10 CFUs of E. coli pBAD28 or E. coli pApyr . Twenty-eight days after colonization, Ig repertoire analysis was performed on IgA + plasma cells isolated from small-intestine lamina propria. (A and B) Hierarchical clustering (A) and significance analysis (B) of pairwise clonal (CDR3aa) overlap are shown. co-evolution selects and maintains mutualistic microorganisms within the gastrointestinal (GI) habitats. Adenosine triphosphate (ATP) is a ubiquitous extracellular messenger, which activates plasma membrane receptors for extracellular nucleotides termed P2 receptors (Burnstock, 2006). Peyer's patches (PPs) are the secondary lymphoid organs within the ileal mucosa, where T cell-dependent IgA responses originate. Most lymphocytes localized in PPs inhabit germinal centers (GCs), where T follicular helper (Tfh) cells interact with B cells and facilitate B cell proliferation, Ig class-switch recombination (CSR), somatic hypermutation (SHM), and affinity maturation (Crotty, 2011). Because Tfh cells in PPs are essential for GC reactions and IgA affinity maturation, they play a critical role in the modulation of the structure and function of intestinal microbial communities (Kawamoto et al., 2014). Tfh cells express high levels of the ionotropic P2X7 receptor (P2X7R) in the plasma membrane; in the PPs, they are exposed to micromolar concentrations of extracellular ATP (eATP) released by the microbiota that permeates the intestinal epithelium. Bacterial eATP limits Tfh cell abundance in PPs via stimulation of P2X7R, thus promoting the generation of a beneficial microbiota via modulation of SIgA response (Proietti et al., 2014). We previously demonstrated that Escherichia coli transformants expressing the phon2 gene from Shigella flexneri (E. coli pApyr ) encoding for the highly active ATP-diphosphohydrolase apyrase efficiently hydrolyses intestinal ATP in both monocolonized and specific pathogen-free (SPF) mice (Perruzza et al., 2017;Proietti et al., 2019); hydrolysis of eATP results in Tfh cell expansion and enhanced B cell help (Perruzza et al., 2017). Here, we investigated whether abrogation of signaling by bacterial eATP affected SIgA repertoire structure and intestinal homeostasis.

RESULTS
Regulation of the SIgA response and repertoire diversity by eATP in the small intestine In C57BL/6 germ-free (GF) mice colonized with E. coli pApyr , hydrolysis of intestinal eATP resulted in enhanced coating of fecal bacteria by SIgA as compared with mice colonized with transformants bearing the empty pBAD28 vector (E. coli pBAD28 ) (Figure S1A); colonization with E. coli pApyr resulted in the increase of SIgA in the small intestine, cecum, and colon ( Figure S1B). Accordingly, we observed a significant increase of IgA-secreting plasma cells (PCs) by ELISPOT in the intestinal lamina propria (LP) of these mice ( Figure S1C). To investigate whether reduction of bacteria-derived ATP could have an impact on the SIgA repertoire, we performed high-throughput sequencing of Ig V H regions in IgA + PCs isolated from the LP of gnotobiotic mice colonized with E. coli pApyr or E. coli pBAD28 . Hierarchical clustering analysis of the percentage of IgA CDRH3 amino acid (aa) overlap in PCs from the two groups of animals showed significantly higher overlap among mice colonized with E. coli pBAD28 than E. coli pApyr . The control group of non-colonized GF mice showed the highest CDR3aa overlap ( Figures 1A and 1B). These data suggest that colonization with E. coli pApyr was associated with more diversified SIgA repertoires between different animals. To further define the architecture of the SIgA repertoire, we performed clonal sequence similarity network analysis (Ben-Hamo and Efroni, 2011;Miho et al., 2019). The number of clones isolated from mice colonized with E. coli pApyr or E. coli pBAD28 did not differ significantly between samples (94-132 clones), as well as the extent of SHM (3.16-4.90 aa mutations). However, E. coli pBAD28 mice showed a higher degree of clonal expansion of public CDR3 sequences  with respect to E. coli pApyr mice, as indicated by a higher abundance of clones with similar sequences within one repertoire; conversely, repertoires isolated from E. coli pApyr mice showed a more diverse sequence similarity distribution with less clustering, similarly to the repertoire architecture observed in healthy unimmunized mice . The sequence similarity within each repertoire was quantified by determining the repertoire's median clonal degree (Greiff et al., 2015) ( Figures 1C and 1D). Altogether, these data indicate that reduction of bacteria-derived ATP by apyrase is associated with the generation of a more diversified SIgA repertoire in monocolonized mice.
The breadth of the anti-E. coli SIgA response conditions the epithelial transcriptional activity and function in monocolonized mice We investigated if the enhancement of SIgA coating affected the topography of E. coli in the proximal colon. We used SPF C57BL/ 6 mice expressing E-cadherin mCFP treated with broad-spectrum antibiotics (ABXs) before colonization with E. coli as a proxy for gnotobiotic mice. Colonization of these animals with GFP-expressing E. coli pBAD28 or E. coli pApyr showed the distribution of E. coli pBAD28/pGFP in closer proximity to the intestinal epithelium as compared with E. coli pApyr/pGFP (Figure 2A). No differences were observed in the thickness of the inner mucous layer between differently colonized mice. We detected similar mucous thickness independently of apyrase also in Igh-J À/À mice, which carry a deletion in the J segment of the Ig heavy-chain locus and therefore are devoid of Igs ( Figure S2B). These data suggest that apyrase does not influence mucous thickness in the colon either in the presence or absence of Igs, and that abrogation of SIgA modulation by eATP modifies E. coli interaction with the mucous layer and intestinal epithelium.
Next, we addressed whether apyrase-mediated enhancement of anti-E. coli SIgA and modified topography of bacteria had an impact on the transcriptional regulation in intestinal epithelial cells (IECs). We performed genome-wide expression profiling of small intestinal epithelium from GF mice colonized with E. coli pApyr or E. coli pBAD28 . Hierarchical clustering segregated differentially expressed genes in three clusters corresponding to GF, E. coli pApyr , and E. coli pBAD28 gnotobiotic mice ( Figure 2B;  Table S2), strikingly, minimal differences in gene expression were detected between the corresponding Igh-J À/À mice ( Figure 2D). These data indicate that endoluminal ATP does not substantially affect per se the transcriptional activity of ileal IECs in gnotobiotic mice. In contrast, eATP-mediated shaping of the SIgA response against a commensal bacterium can condition IEC function. Gene Ontology (GO) enrichment analysis revealed that most of the genes downregulated in IECs from WT mice colonized with E. coli pApyr were involved in immunity, including genes related to defense responses against bacteria and cell-cycle regulation ( Figures 2E and S2A). In particular, Paneth cell-derived antimicrobial peptide angiogenin 4 (Ang4), Lyz1 and Lyz2 encoding two isoforms of lysozyme, 14 members of Defa family encoding for defensin a isoforms, and their processing enzyme matrix metalloproteinase 7 (Mmp7) were all significantly downregulated in IECs conditioned by E. coli pApyr with respect to E. coli pBAD28 (Figure S2A). Moreover, genes involved in the regulation of cell cycle (Cdk1, Cdk4, and Cdk20) and cell division (Cdc20, Bub1b, and different members of the kinesin family) were downregulated in mice monocolonized with E. coli pApyr with respect to E. coli pBAD28 .
To understand which cell types were responsible for this differential transcriptional activity, we applied t-statistic stochastic neighbor embedding (t-SNE) overlay of regulated genes on publicly available gene signatures of individual epithelial cells isolated from the small intestine and organoids (Haber et al., 2017) ( Figure S2C). The t-SNE overlay revealed the enrichment of the apyrase-induced signature in the mature enterocyte populations ( Figure S2D). In particular, the majority of genes that were upregulated in IECs from WT mice colonized with E. coli pApyr were expressed in absorptive mature enterocytes (Figures S2E and S2F), while the downregulated genes encoding for anti-microbial peptides and cell-cycle regulators were expressed in particular by Paneth cells and transit-amplifying stem cells, respectively (Figures S2G and S2H). Overall, these data suggest that hydrolysis of eATP by apyrase in the small intestine conditions the symbiotic relationship between the host and microbiota via amplification of the SIgA response.
The breadth of the anti-E. coli SIgA response conditions metabolites absorption in the intestine of monocolonized mice To test if the transcriptional regulation induced in IECs by E. coli pApyr had an impact on nutrient absorption, we performed targeted metabolomics analysis of portal vein serum from WT and Igh-J À/À gnotobiotic mice. Colonization with E. coli irrespective of apyrase expression resulted in a significantly smaller diversity of the metabolic profiles of WT compared with Igh-J À/À mice ( Figure 4A; Marti Anderson's PERMDISP2 [Permutational Analyses Of Multivariate Dispersions] procedure + permutation test, p = 0.001), suggesting SIgA could mediate ''metabolic speciation'' of the gut ecosystem on microbial colonization. Ig secretion appeared particularly important in regulating lipid (A) Representative images and statistical analysis of bacterial interaction with proximal colon mucosa measured as distance of E. coli cells from the epithelial surface in C57BL/6 SPF ABX mice colonized with E. coli pApyr/pGFP and E. coli pBAD28/pGFP . Blue: E-cadherin coupled tog monomeric cyan fluorescent protein (mCFP) in E-cadherin mCFP mice; green: GFP expressing E. coli pApyr or E. coli pBAD28 ; red: autofluorescence signal of the mucus obtained by spectral unmixing of the CFP and GFP channels. Two-tailed Mann-Whitney U test was used. ****p < 0.0001. Representative images are displayed as the maximum intensity projection (MIP) of a full-depth, 10-mm-wide portion of the xz view of the stacks. Data shown are representative of two pooled experiments (n = 7-9). (B-E) GF C57BL/6 WT and Igh-J À/À mice were colonized with 10 10 CFUs of E. coli pBAD28 or E. coli pApyr . Twenty-eight days after colonization, genome-wide expression profiling was performed on intestinal epithelial cells. (B) Hierarchical clustering of gene expression profiles in the intestinal epithelium of C57BL/6 GF mice and mice monocolonized with E. coli pApyr or E. coli pBAD28 (FDR-corrected p < 0.05 and log 2 FC > |1.5|) (see Table S1). (C) Volcano plot descriptive of differential gene expression in C57BL/6 mice colonized with E. coli pApyr versus E. coli pBAD28 . (D) Volcano plot descriptive of differential gene expression in Igh-J À/À mice colonized with E. coli pApyr versus E. coli pBAD28 . Volcano plots show for each gene (dots) the differential expression (log 2 fold-change [log 2 FC]) and its associated statistical significance (log 10 p value). The red dots indicate those genes with an FDR-corrected p < 0.05 and log 2 FC > |1|. We detected 395 upregulated and 411 downregulated genes in mice monocolonized with E. coli pApyr as compared with E. coli pBAD28 . The names of strongly downregulated and upregulated genes (FDR-corrected p < 10 À5 and log 2 FC > |1.5|) are also reported (see Table S2). (E) Gene Ontology (GO) analysis of the differentially expressed genes (FDR-corrected p < 0.05 and log 2 FC > |1|) in E. coli pApyr versus E. coli pBAD28 monocolonized C57BL/6 WT mice visualized as GOCircle plot. The inner ring is a bar plot where the height of the bar indicates the significance of the GO term (log 10 FDRcorrected p value), and color corresponds to the Z score, i.e., the number of genes upregulated minus the number of genes downregulated divided for the square root of the total number of genes analyzed: z = ðn upÀ n downÞ ffiffiffiffiffiffiffi ffi n tot p Article ll OPEN ACCESS digestive and absorptive pathways because sphingolipids and glycerophospholipids were enriched in portal vein serum from gnotobiotic WT with respect to Igh-J À/À mice ( Figure S3A). Next, we focused on the possible role of eATP-mediated shaping of SIgA repertoire structure in conditioning nutrient absorption. E. coli colonization had a dominant effect on serum metabolites composition, because colonization with E. coli, irrespective of apyrase expression, resulted in rather homogeneous metabolic profiles that were distinct from GF mice ( Figure 4B). Nevertheless, the metabolic profiles between the two groups of gnotobiotic animals (i.e., E. coli pApyr and E. coli pBAD28 ) showed few but relevant differences. The univariate analysis performed on the full set of metabolites showed that the significant differences between WT mice colonized with E. coli pApyr and E. coli pBAD28 were concentrated in specific metabolic classes. Indeed, 8 of 13 aa showed a significantly higher abundance in the portal vein serum of E. coli pApyr mice, as well as three biogenic amines and glycerophospholipids. In contrast, acylcarnitines were enriched in serum collected from E. coli pBAD28 mice ( Figure 4C). These differences were not detected in differently colonized Igh-J À/À mice ( Figure S3B), suggesting that the alterations in metabolites composition detected in mice conditioned by apyrase were mediated by shaping of the SIgA response against E. coli.
Administration of apyrase-expressing bacteria promotes maintenance of gut microbial homeostasis in dysbiosis SIgA has multiple functions in regulating microbiota composition and gut homeostasis (Weis and Round, 2021). To investigate whether SIgA enhancement by apyrase could beneficially affect gut microbial homeostasis during an environmental perturbation, we used a mouse model of ABX-mediated dysbiosis. A mix of ABXs (vancomycin 1.25 mg, ampicillin 2.5 mg, and metronidazole 1.25 mg) was administered via orogastric gavage to WT mice for 4 consecutive days. After the ABX treatment, mice were orally gavaged for 4 days with PBS (control) or 10 10 colony-forming units (CFUs) of E. coli pApyr or E. coli expressing a loss-of-function isoform of the apyrase enzyme with R192P amino acid substitution (E. coli pHND19 ) (Scribano et al., 2014) (Figure 5A). Quantification of eATP in E. coli pHND19 and E. coli pBAD28 cultures showed no significant differences, suggesting that both strains released similar amounts of ATP ( Figure S1D). The analysis of metabolic and immunological parameters in ABX-treated mice colonized with E. coli pHND19 and E. coli pBAD28 revealed superimposable values, suggesting that both strains produced equivalent effects on gut homeostasis following ABXs administration (Figures S1E-S1O).
Mice treated with ABX alone or in combination with E. coli pHND19 showed a strong reduction of bacterial taxonomic richness in the cecum as reflected by a-diversity (Shannon index, observed features, and Faith's phylogenetic diversity) (Figure 5B), indicating strong dysbiosis. Interestingly, treatment with E. coli pApyr resulted in a significant improvement of this parameter. To determine similarity in bacterial composition in the different experimental groups, we analyzed b-diversity by principal-coordinate analysis (PCoA) derived from unweighted and weighted UniFrac. Despite each ABX treatment group clustered separately from the untreated control (PERMANOVA <0.001), E. coli pApyr -treated mice clustered closer to the control group, indicating an improved recovery of the physiological microbiota composition ( Figure 5C). Relative abundance analysis of differentially represented amplicon sequence variants (ASVs) revealed that that E. coli pApyr administration resulted in the selective preservation of 41 species belonging to the orders of Bacteroidales, Clostridiales, Lactobacillales, and Burkholderiales ( Figure 5D). Among Bacteroidales, Muribaculum intestinale was detected by multiple ASVs. The reduction of this bacterial species was shown to correlate with higher susceptibility to ileitis (Dobranowski et al., 2019). E. coli pApyr administration favored the preservation of Clostridium scindens, a bacterium that was shown to protect from Clostridioides difficile infection through the generation of secondary bile acids deoxycholic acid (DCA) and lithocholic acid (LCA). Reconstitution with C. scindens alone or within a bacterial consortium protected ABX-treated mice from C. difficile intestinal colonization (Buffie et al., 2015). Different species belonging to Lactobacillales order, in particular Lactobacillus johnsonii and Lactobacillus reuteri, were also significantly enriched in E. coli pApyr -treated mice. These results suggest that apyrase enzymatic function can positively influence gut microbial homeostasis in ABX-mediated dysbiosis. Consistent with SIgA function in mediating the effect of apyrase, the improved recovery in microbiota composition was associated to increased coating of fecal microbiota by SIgA in mice gavaged with E. coli pApyr ( Figure 5E).
Administration of apyrase-expressing bacteria attenuates intestinal barrier impairment and glucose homeostasis perturbation in ABX-mediated dysbiosis ABXs strongly affect microbial diversity and intestinal barrier function, leading to bacterial translocation of live commensal bacteria to the mesenteric lymph nodes (MLNs) (Knoop et al., 2016). A characteristic feature of reduced bacterial load after ABX treatment is cecum enlargement, which characterizes also GF animals (Devkota et al., 2012;Poteres et al., 2020). Analysis of cecum weight 4 days after recovery from ABX treatment revealed a pronounced increase of this parameter, as expected. However, mice treated with E. coli pApyr showed significantly reduced cecum weight as compared with mice treated with (F) Perigonadal white adipose tissue (WAT) weight. (G-J) GF Igh-J À/À mice were colonized with 10 10 CFUs of E. coli pBAD28 or E. coli pApyr . Twenty-eight days after colonization, different metabolic parameters were evaluated in the two groups of animals. Data points represent single mice. One experiment representative of two is shown.  P C ae C 40 :5 P C a e C 4 0 :6 P C a e C 4 2 :1 P C a e C 4 2 :2 P C a e C 4 2 :3 P C a e C 4 2 :4 P C a e C 4 4 :3 P C a e C 4 4 :4 P C a e C  (legend continued on next page)  Figure 5F). Quantification of CFUs from MLN, both in aerobic and anaerobic conditions, revealed a significant increase of bacterial recovery in mice treated with ABX or ABX and E. coli pHND19 compared with untreated animals, indicating gut barrier integrity was compromised. However, mice treated with the combination of ABX and E. coli pApyr showed a number of CFUs in the MLN that was not significantly different from untreated animals ( Figures 5G and 5H). These data indicate that E. coli pApyr administration mitigates the effects of ABXinduced dysbiosis.
The gut microbiota encodes a more versatile metabolome than the host, and a healthy microbiota is a necessary requirement for stable functional metabolic interactions with the host. To investigate the effect of apyrase on host metabolism perturbation caused by dysbiosis induced by ABXs, we analyzed blood glucose 4 days after recovery from the ABX treatment. Administration of ABXs resulted in a pronounced decrease in blood glucose. Mice treated with apyrase-expressing bacteria showed higher serum glucose levels compared with mice treated with ABX or ABX and E. coli pHND19 ( Figure 5I). Quantification of white adipose tissue (WAT) revealed a significant reduction of WAT in ABX-treated mice both as ABX alone treatment or in association with E. coli pHND19 . This reduction was significantly attenuated by administration of E. coli pApyr ( Figure 5J). No improvements in blood glucose levels ( Figure S4A) and WAT deposition (Figure S4B) after ABX treatment were observed in Igh-J À/À mice treated with E. coli pApyr as compared with the counterparts treated with ABX or the combination of ABX and E. coli pHND19 . These results are consistent with the function of apyrase in attenuating metabolic consequences of ABX-mediated dysbiosis via enhancement of SIgA.
Intestinal conditioning by apyrase promotes resistance to infection by enteric pathogens following ABX treatments Microbial cells in the GI tract confer colonization resistance against intestinal pathogens. ABX-mediated depletion of endogenous microbes results in increased susceptibility to a number of opportunistic and pathogenic enteric infections (Blaser, 2011;Preidis and Versalovic, 2009). Enterohemorrhagic E. coli (EHEC), enteropathogenic E. coli (EPEC), and Citrobacter rodentium are Enterobacteriaceae that belong to the family of attaching and effacing (A/E) lesion-forming bacteria. EHEC and EPEC can cause severe intestinal inflammation and diarrhea. In addition, EHEC strains expressing the highly potent Shiga toxin (Stx) cause nephrotoxicity, resulting in severe cases in the death of infected individuals. Because the human pathogens EHEC and EPEC induce only modest pathogenicity in ABX-treated adult mice, C. rodentium is used to mimic these infections in mice (Collins et al., 2014). We investigated whether apyrase could promote resistance to the challenge of ABX-treated mice with 10 8 CFUs of C. rodentium ( Figure 6A). Mice gavaged with E. coli pApyr after ABXs showed reduced body weight loss on infection as compared with the groups treated with ABX alone or followed by E. coli pHND19 ( Figure 6B). Inflammatory monocytes (CD45 + CD11b + Ly6c + Ly6g À cells) ( Figure 6C) and neutrophils (CD45 + CD11b + Ly6c + Ly6g + cells) in the cecum LP ( Figure 6D), C. rodentium CFUs in both the spleen ( Figure 6E) and liver (Figure 6F), and fecal and serum lipocalin-2 (LCN-2) levels ( Figures 6G and 6H), which are linked to epithelial damage and neutrophil infiltration, were all significantly reduced in mice treated with E. coli pApyr as compared with the groups treated with ABX alone or in combination with E. coli pHND19 . Notably, Iga À/À mice did not show any improvement in intestinal inflammation and control of the infection by combining E. coli pApyr administration to the ABX treatment ( Figures S4C-S4I), indicating that SIgAs are instrumental in promoting colonization resistance by apyrase.
We further tested the beneficial effect of apyrase in the enteric infection by C. difficile. C. difficile is a major cause of ABX-associated diarrhea dependent on reduced bacterial community diversity and depletion of key taxa within the intestinal microbiota (Seekatz and Young, 2014). To investigate whether the microbiota community structure induced by apyrase-expressing bacteria could counteract intestinal invasion by C. difficile, we administered ABX to WT mice for 4 days to induce microbiota depletion. Thereafter, mice were orally gavaged with E. coli transformants for 4 days, infected with 10 5 C. difficile VPI 10463 spores, and analyzed 72 h postinfection to evaluate intestinal inflammation ( Figure 6I). The loss of body weight ( Figure 6J) and the clinical score ( Figure 6K) following C. difficile infection were both significantly ameliorated by E. coli pApyr administration. In mice treated with ABX followed by E. coli pApyr , colon length, an important pathological parameter in colitis, was similar to non-infected mice, whereas in mice treated with standalone ABX or in combination with E. coli pHND19 , it was drastically reduced (Figure 6L). Finally, fecal and serum LCN-2 concentrations were significantly reduced in mice treated with ABX and E. coli pApyr as compared with mice treated with ABX alone or in combination with E. coli pHND19 (Figures 6M and 6N). These data further support the notion that treatment with apyrase-expressing bacteria mitigates ABX-induced dysbiosis and promotes resistance to C. difficile infection.

DISCUSSION
The diversity of the SIgA repertoire generated by intestinal B cells does not reflect the complexity of the antigenic universe borne by the microbiota. This phenomenon is consistent with the coevolution of the gut immune system and microorganisms, and contrasts with the diversification of the systemic B cell repertoire that emerges on exposure to microbial antigens . GCs in PPs from different mice expand public B cell receptor (BCR) clonotypes, some of which are dependent on bacterial antigens, while others are not (Chen et al., 2020). A substantial percentage (5%-10%) of GCs from SPF mice contain highly    Article ll OPEN ACCESS dominant B cell clones that are selected by antigens derived from commensal bacteria. The antigen-driven selection of public clonotypes specific for bacterial and non-bacterial antigens appears to be tunable by the presence and composition of the microbiota (Nowosad et al., 2020). The identification of mechanisms responsible for the regulation of the SIgA diversity would allow to intervene for enhancing the efficacy of mucosal vaccination and counteracting disruption of microbiota homeostasis in pathophysiological conditions. Microbiota-derived eATP limits Tfh cells activity in the PPs via the ATP-gated P2X7R and thereby BCR affinity maturation (Proietti et al., 2014). Hydrolysis of eATP by apyrase delivery to the small intestine resulted in increased SIgA production and higher B cell clonal diversity that was reflected by enhanced coating of intestinal bacteria. SIgA is key in determining a non-inflammatory relationship between the host and microbiota (Peterson et al., 2007). SIgA coating accelerates the small-intestinal transit of bacteria by limiting bacterial motility, by reducing adherence, or by exclusion effects (Uchimura et al., 2018). We found that apyrase-mediated conditioning of SIgA repertoire modified the topographical distribution of bacteria in the mucus, suggesting SIgA shaping by eATP influences bacterial interaction with the epithelium. Accordingly, SIgA generated by an apyrase-bearing live attenuated oral vaccine conferred enhanced protection from the invasiveness of the virulent bacteria (Proietti et al., 2019). Along this line, SIgA generated in mice treated with ABXs and gavaged with E. coli pApyr significantly limited both aerobic and anaerobic bacteria translocation to MLN (Figures 5G and 5H).
SIgA plays a central role in conditioning transcriptional activity of the intestinal epithelium. Shulzhenko et al. (2011) have shown that lack of SIgA results in upregulation of immune response genes and concomitant repression of genes correlated with metabolic functions. This shift in intestinal function led to lipid malabsorption and decreased deposition of body fat (Shulzhenko et al., 2011). The enhanced SIgA production by colonization of GF mice with E. coli pApyr induced the downregulation of genes involved in immune response against bacteria and upregulation of genes involved in metabolic processes. This effect mediated by apyrase-conditioned SIgA prompted us to characterize metabolites absorption in the portal vein of E. coli pApyr monocolonized mice. First, the targeted metabolomic analysis revealed that sphingolipids and glycerophospholipids were particularly enriched in portal vein serum from gnotobiotic WT with respect to antibody-deficient Igh-J À/À mice. Second, the upregulation in the intestinal epithelium of genes connected to lipid metabolism and uptake of amino acids that we observed in mice colonized with E. coli pApyr with respect to control E. coli transformants devoid of apyrase activity correlated with enhanced absorption of glycerophospholipids and essential amino acids. Mucosal conditioning by apyrase induced a significant reduction of serotonin absorption, a biogenic amine indole derivate with pleiotropic function in immune system regulation that was shown to be increased in the small intestine of antibody-deficient mice (Uchimura et al., 2018). Moreover, in E. coli pApyr monocolonized mice, we detected an overall reduction of plasma acylcarnitines that are connected with inflammation and insurgence of metabolic disorders (Makrecka-Kuka et al., 2017;Rutkowsky et al., 2014). Therefore, apyrase-mediated enhancement of SIgA diversity and bacterial coating promotes a shift from immune to metabolic functions in the intestinal epithelium.
In SPF mice, modulation of GC reaction in PPs by bacteriaderived eATP via P2X7R in Tfh cells promotes a proficient microbiota for metabolic homeostasis (Perruzza et al., 2017); chronic deregulated SIgA coating of commensal bacteria results in altered systemic metabolism . Conversely, the increase in SIgA diversity in E. coli pApyr colonized mice partially compensated the lack of metabolic fitness of GF and monocolonized mice by ameliorating host glucose homeostasis. SIgA coating alters gene expression of mucous-associated bacteria and modulates interphylum interaction (Nakajima et al., 2018). The diversity of antigen targeting by SIgA affects bacterial function and metabolism (Rollenske et al., 2021). Therefore, the apyrase-mediated increase in SIgA diversity could improve the metabolic fitness of E. coli pApyr .
SIgA possesses two antithetic functions acting either in preventing or promoting bacterial colonization (Kubinak and Round, 2016). This dichotomy is at least in part driven by characteristic Figure 5. E. coli pApyr administration improves gut microbiome recovery, intestinal barrier integrity, and metabolic homeostasis in ABXinduced dysbiosis (A-I) Dysbiosis was induced by daily oral gavage of a mix of ABXs for 4 days. After the ABX treatment, during the recovery phase, mice were orally gavaged for 4 days with PBS (control) or 10 10 CFUs of E. coli pApyr or E. coli pHND19 . (A) Experimental layout of ABX-induced dysbiosis and recovery phase. (B) Bacterial a-diversity calculated by Shannon index, observed features, and Faith's phylogenetic diversity. Two-tailed Mann-Whitney U test was used. **p < 0.01. Data points represent single mice. (C) Bacterial b-diversity. The PCoA plots of microbial b-diversity were generated using unweighted and weighted UniFrac algorithms. PERMANOVA was used. p < 0.001. Data points represent single mice. (D) Heatmap showing bacterial amplicon sequence variants (ASVs) in cecal microbiota that discriminate the different experimental groups. ASVs were selected according to p < 0.05 with Wald test using FDR p value correction following DESeq2 read counts normalization. Each line represents one ASV, and each column represents an individual mouse. Mean relative abundances (log10) of ASVs detected in the different experimental groups are shown. (E) Statistical analysis of fecal IgA-coated microbiota at the end of the recovery phase (day 4). Microbiota from Iga À/À mice was used as negative control for the secondary antibody. (F) Percentage of cecum weight normalized by total body weight. (G and H) CFU quantification of aerobic (G) and anaerobic (H) bacteria recovered from the MLN (pooled data from two independent experiments). (I) Blood glucose variation between day À4 (start of ABX treatment) and day +4 (end of recovery phase). (J) Percentage of WAT deposition at the end of recovery phase (day 4). Data points represent single mice. Except where indicated, one experiment representative of three is shown (n = 4-7/experiment). Means ± SEM are shown. Two-tailed Mann-Whitney U test (E-I) was used. *p < 0.05, **p < 0.01. See also Figures S1D-S1O, S4A,  features of the bound bacteria, such as the replicative activity, which induces enchained growth and clearance of enteropathogens (Moor et al., 2017). Conversely, SIgA coating of Bacteroides fragilis capsular polysaccharides helps bacteria occupy a defined mucosal niche, resulting in exclusion of possibly pathogenic competitors. This phenomenon was shown to be spread across different commensal microbes (Donaldson et al., 2018). ABX treatment results in loss of intestinal microbiota diversity, which requires several weeks to be restored (Palleja et al., 2018), exposing the host to possible colonization by exogenous enteropathogens. Four-day course of an ABX mix renders SPF mice susceptible to enteric infections, including C. rodentium and C. difficile, two murine models of ABX-mediated human diseases (Becattini et al., 2016). Daily gavage immediately after ABX treatment with E. coli pApyr , but not E. coli pHND19 , expressing a loss-of-function mutant of apyrase, promoted the colonization of different bacterial species belonging to the orders of Bacteroidales, Clostridiales, Lactobacillales, and Burkholderiales that ensured resistance to the infection by C. rodentium and C. difficile. Among Bacteroidales, Muribaculum intestinale, the reduction of which was correlated with higher susceptibility to ileitis (Dobranowski et al., 2019), was detected by multiple ASVs. Notably, E. coli pApyr gavaging resulted in the enrichment of C. scindens that was previously shown to protect mice from C. difficile infection through the generation of the secondary bile acids DCA and LCA (Buffie et al., 2015). Among the Lactobacillales order, L. johnsonii and L. reuteri were significantly enriched in E. coli pApyr -treated mice. These two species are commonly used as probiotics and were shown to confer protection against C. rodentium (Mackos et al., 2013) and Campylobacter jejuni (Bereswill et al., 2017) infections. Finally, ABX-mediated alteration of glucose metabolism (Zarrinpar et al., 2018) was improved by apyrase-mediated SIgA enhancement. We originally proposed bacteria-derived eATP as an important mediator of gut ecosystem homeostasis by modulating SIgA response (Perruzza et al., 2017;Proietti et al., 2014). The observations reported in the present study suggest that abrogation of eATP signaling and expansion of SIgA diversity by apyrase can be used to restore intestinal microbiota fitness in dysbiotic conditions, such as provoked by ABX treatment.

Limitations of the study
Although our experiments establish a function for apyrasemediated degradation of luminal ATP in promoting the amplification of the SIgA repertoire in the intestine, the following limitations of the study have to be considered. To analyze the impact of apyrase activity on SIgA structure in a controlled intestinal environment, we used monocolonized mice. In physiological conditions, the complexity of the microbiota and microbiota-derived metabolites might further influence the SIgA repertoire structure and/or bias the response specificity in the absence of eATP. An analogous limitation can be considered for the transcriptional regulation in the epithelial component of the intestine and metabolites absorption by SIgA amplification. The picture we provide, albeit quite explanatory for the role of eATP in regulating intestinal function by shaping the commensal-specific SIgA response, is limited by monocolonization that cannot provide a comprehensive definition of the transcriptional landscape and metabolites absorption of a normally colonized organism. We demonstrated that enhancing the SIgA coating of the microbiota could eventually lead to the attenuation of pathological consequences of dysbiosis induced by ABX treatment. The maintenance of gut homeostasis correlated with the SIgA-dependent preservation of a beneficial bacterial community in mice exposed to apyrase. These experiments were performed in mice housed in an SPF environment; additional studies are warranted to establish the effectiveness of SIgA amplification by apyrase in correcting dysbiosis in non-SPF conditions and in other organisms.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We thank Sara Maffei and the staff of the animal facility of the Institute for Research in Biomedicine and the staff of the clean mouse facility of the University of Bern for excellent mouse husbandry, David Jarossay (Institute for Research in Biomedicine) for cell sorting, and Prof. Siegfried Hapfelmeier (University of Bern) for providing the tetracycline-resistant strain of C. rodentium used in this study. This work was supported by the Swiss National Science Foundation (SNSF 310030_192531), the Swiss Cancer League (KFS-5033-02-2020), the Fondazione Ticinese per la Ricerca sul Cancro, the Fondazione San Salvatore, and the Novartis Foundation for Medical-Biological Research (to F.G.).  Igh-J -/-(B6.129P2-Igh-J tm1Cgn /J) Jackson Labs (Gu et al., 1993) JAX stock # 002438 E-cadherin mCFP (B6.129P2(Cg)-Cdh1 tm1Cle /J) Jackson Labs (Snippert et al., 2010) JAX stock # 016933

AUTHOR CONTRIBUTIONS
Iga -/-(Igha tm1Grh ) SPF Vivarium LASC Schlieren (Harriman et al., 1999)  Article ll de-multiplexing and trimming of Illumina adaptor residuals using Illumina's real time analysis software included in the MiSeq reporter software v2.6 (no further refinement or selection). The quality of the reads was checked with the software FastQC version 0.11.8. The sequences were analyzed through the Qiime2 virtual environment (Bolyen et al., 2019). The raw sequences were in total 4'896'770 (median = 71'942, mean = 72'011.3, SD = 15'891.2). The trimming step on the first 7 and the last 25 bases and the reads filtration have allowed to obtain excellent quality sequences (Phred > 30). A denoising algorithm (Callahan et al., 2016) was implemented on these sequences. The overlapping regions R1 and R2 were joined and the chimeric reads discarded. The reads that resulted from trimming, filtering and joining steps were in total 1'145'671 (median = 16'277, mean= 16'848.1, SD= 3'897.6). The taxonomic assignment was performed by BLAST feature-classifier. It performs BLAST+ local alignment between query and reference reads. Then, it assigns consensus taxonomy to each query sequence on the last database version of Greengenes (gg_12_10). A rooted tree was constructed based on IQ-TREE stochastic algorithm that allows maximum likelihood analysis of large phylogenetic data (Nguyen et al., 2015). Sequence reads from 16S rRNA gene profiling have been deposited in the European Nucleotide Archive (ENA) of the European Bioinformatics Institute under the accession number of the study: PRJEB49686.

QUANTIFICATION AND STATISTICAL ANALYSIS
All statistical analyses were performed using the statistical programming environment R version 4.0.3 (Team, 2017) or GraphPad Prism v7.04 (GraphPad Software, La Jolla, CA, USA). Alpha diversity was calculated using the main indexes to allow an exploration of data in term of richness and evenness. Alpha-diversity estimates were computed using the phyloseq R package (McMurdie and Holmes, 2013). Statistically significant changes in the alpha diversity were determined through the Mann-Whitney signed-rank test. The microbial community comparison was calculated using PERMANOVA on Weighted distance metrics performed by UniFrac algorithm (Lozupone et al., 2011). Statistically significant differences in the relative abundance of ASVs between groups were performed by Wald test using FDR p value correction following DESeq2 read counts normalization. Statistical significance was set at p % 0.05 (*, 0.01 % p % 0.05; **, 0.001 % p % 0.01; ***, p % 0.001). the mean differences with 0.05 < p % 0.10 were accepted as trends.