Enrichment of intestinal Lactobacillus by enhanced secretory IgA coating alters glucose homeostasis in P2rx7−/− mice

The secretory immunoglobulin A (SIgA) in mammalian gut protects the organism from infections and contributes to host physiology by shaping microbiota composition. The mechanisms regulating the adaptive SIgA response towards gut microbes are poorly defined. Deletion of P2rx7, encoding for the ATP-gated ionotropic P2X7 receptor, leads to T follicular helper (Tfh) cells expansion in the Peyer’s patches (PPs) of the small intestine, enhanced germinal centre (GC) reaction and IgA secretion; the resulting alterations of the gut microbiota in turn affects host metabolism. Here, we define gut microbiota modifications that correlate with deregulated SIgA secretion and metabolic alterations in P2rx7−/− mice. In particular, Lactobacillus shows enhanced SIgA coating in P2rx7−/− with respect to wild-type (WT) mice. The abundance of SIgA-coated lactobacilli positively correlates with Tfh cells number and body weight, suggesting Lactobacillus-specific SIgA response conditions host metabolism. Accordingly, oral administration of intestinal Lactobacillus isolates from P2rx7−/− mice to WT animals results in altered glucose homeostasis and fat deposition. Thus, enhanced SIgA production by P2X7 insufficiency promotes Lactobacillus colonization that interferes with systemic metabolic homeostasis. These data indicate that P2X7 receptor-mediated regulation of commensals coating by SIgA is important in tuning the selection of bacterial taxa, which condition host metabolism.

colonic homeostasis 9 . The vast majority of SIgA-coated bacteria resides in the small intestine and is targeted by T cell-independent antibodies; only a minority of commensals would be responsible for eliciting T cell-dependent SIgA responses 10 . T follicular helper (Tfh) cells in PPs are essential for SIgA affinity maturation that in turn modulates the structure and function of the intestinal microbiota 11 . Adenosine triphosphate (ATP) is an ubiquitous extracellular messenger, which activates purinergic receptors in the plasma membrane of eukaryotic cells termed P2X and P2Y receptors 12 . The ATP-gated ionotropic receptor P2X7 is a signature gene of effector T cell subsets 13,14 and is selectively upregulated in Tfh cells of PPs. In mice with deletion of the P2rx7 gene, Tfh cells are expanded in PPs because of resistance to cell death induced by extracellular ATP (eATP). The altered control of Tfh cells by defective sensing of microbiota derived ATP leads to enhanced secretion of T cell dependent IgA and increased frequency of replacement mutations in the IgV H 1 family's complementarity determining region (CDR) 2 suggesting enhanced affinity maturation of IgA responses 15 . Therefore, eATP modulates adaptive IgA responses to ensure physiological mucosal colonization. Furthermore, the alteration of the gut microbiota due to the lack of P2X7 mediated control of Tfh cells results in dysregulated metabolic homeostasis, consistent with the central role of SIgA in regulating host-microbiota interactions and host physiology 16 . A number of studies in mice and humans have demonstrated that obesity is associated with alterations of the gut microbiota. Intestinal dysbiosis has been suggested to play a causal role in the development of insulin resistance 17 as well as inflammation and macrophage accumulation in adipose tissue 18 . Recently, a genome-wide association study has shown the association of hypo-functioning P2X7 variants with impaired glucose homeostasis and obesity in humans 19 . Here, the characterization of the faecal microbiota targeted by SIgA in P2rx7 −/− mice allowed us to identify the enhanced SIgA coating of Lactobacillus as a possible mechanism contributing to the observed metabolic disturbance.

Results
P2rx7 −/− mice show altered metabolic parameters and enhanced Tfh cells activity. P2rx7 −/− mice are characterized by altered fat distribution 20 . In these mice, dysregulated Tfh cells activity with consequent enhanced GC reactions and secretion of high affinity IgA affects microbiota composition resulting in altered glucose homeostasis and fat deposition 15,16 . White adipose tissue (WAT), body weight and blood glucose were increased in P2rx7 −/− mice with respect to WT littermates (p < 0.01, Wilcoxon rank-sum test; Fig. 1a-c). Food and WT (blue) mice. Box plots are defined by the 25th and 75th percentiles. Centre line represents the median (50th percentile). Whiskers are defined as 1.5 times the interquartile range from the 25th or 75th percentiles. (g) Dynamic pattern of respiratory exchange ratio (RER) and relative mean RER values during the light and dark phase in WT and P2rx7 −/− mice (the displayed experiment is representative of three). *p < 0.05, **p < 0.01, ***p < 0.001, **** p < 0.0001, Wilcoxon rank-sum test; n = 10 per group. consumption and energy harvesting were not different between the two strains of mice. However, the daily profile of fuel metabolism, as measured by the respiratory exchange ratio (RER), the ratio of consumed oxygen to produced carbon dioxide, showed differences between the two groups of mice. P2rx7 −/− mice showed lower RER values during the inactive (light) phase with respect to WT littermates, suggesting lower energy expenditure might at least partially contribute to the body weight increase in these mice (Fig. 1g). As expected, the proportions of Tfh and GC B cells were significantly higher in P2rx7 −/− mice than in WT littermates (p < 0.01, Wilcoxon ranksum test; Fig. 1d,e) as well as the percentage of IgA-coated bacteria (p < 0.0001, Wilcoxon rank-sum test; Fig. 1f).
P2rx7 −/− mice harbour an altered gut microbiota and enhanced IgA response towards bacteria residing in the small intestine. In order to define possible differences in bacterial IgA coating between P2rx7 −/− mice and WT littermates, we characterized the IgA + and IgA − fractions of the faecal microbiota through high-throughput sequencing of the V5-V6 region of the 16S rRNA gene (IgA-SEQ) (Fig. S1). The analysis of alpha-diversity (i.e. the within samples diversity) revealed no significant differences between the IgA + or IgA − fractions of P2rx7 −/− and WT mice. Nevertheless, we observed a reduction close to statistical significance, of alpha-diversity in the presorted faecal samples from P2rx7 −/− mice compared to WT controls (p = 0.055 on the Inverse Simpson index; Wilcoxon rank-sum test), in agreement with previous observations on the gut microbiota of obese mice and humans 21,22 . We then characterized the microbial community structure of the IgA + , IgA − and presorted faecal microbiota through beta-diversity analysis on the unweighted UniFrac distance and Bray-Curtis dissimilarity. The faecal microbiota of P2rx7 −/− mice clustered apart from that of WT littermates as well as the IgA + fraction (p < 0.05, PERMANOVA on the Bray-Curtis dissimilarity, Table 1 and Fig. 2a), suggesting that enhanced IgA secretion due to lack of Tfh cells control via P2X7 has a significant effect on the composition of the gut microbiota (Figs 2b and S2a). Phylum level analysis showed a significant increase in the Firmicutes/Bacteroidetes ratio in the P2rx7 −/− IgA + microbiota (p = 0.03, Wilcoxon rank-sum test, Fig. 2c) due to the significantly higher relative abundance of Firmicutes (mean relative abundance, 65.6% in P2rx7 −/− , 27.8% in WT), as confirmed by LEfSe analysis (Fig. 3a). On the contrary, we did not detect significant differences in the Firmicutes/Bacteroidetes ratio of the IgA − microbiota (Fig. 2d) and presorted faecal microbiota of P2rx7 −/− and WT mice (Fig. S2b). The Firmicutes/Bacteroidetes ratio is a rough estimator of intestinal dysbiosis and its increase has been associated to obesity and metabolic abnormalities in humans and mice 21,22 .
Correlation of gut microbes conditioned by enhanced sIgA response with metabolic and immunological parameters in P2rx7 −/− mice. The relationship between metabolic disorders and gut microbiota has been widely established 1 as well as the role of the immune system and Tfh cells activity in selecting a beneficial microbiota for host metabolism 16 . The enhanced Tfh cells activity in P2rx7 −/− mice was accountable for gut microbiota alterations in both IgA + and IgA − fractions. To evaluate which bacterial taxa might be important for energy metabolism and regulation of mucosal immunity via P2X7, we correlated metabolic and immunologic parameters with the most abundant bacterial genera retrieved by IgA-SEQ. The genus Lactobacillus, within the IgA + microbiota of P2rx7 −/− and WT mice, positively correlated with body weight and abundance of Tfh cells in PPs (Fig. 4a,b). On the other hand, we observed negative correlations of Tfh cells, GC B cells, blood glucose, body and WAT weight with the genera Prevotella, Bacteroides and Barnesiella (Fig. 4a,b). Thus, modified SIgA targeting of these genera by deregulated T follicular help in P2rx7 −/− mice could contribute to host metabolic alterations. Within the IgA − microbiota of P2rx7 −/− and WT animals, the relative abundance of unclassified Lachnospiraceae www.nature.com/scientificreports www.nature.com/scientificreports/ positively correlated with body weight, blood glucose and Tfh cells while Prevotella and Bacteroides negatively correlated with body weight and % of GC B cells (Fig. 4c,d), consistent with previous observations on the high relative abundance of Lachnospiraceae, and low abundance of Bacteroides and Prevotella in obese individuals and mice 25,28,29 . Moreover, significantly increased Lachnospiraceae were found in the caecal microbiota of P2rx7 −/− mice 16 . Finally, different taxa (i.e. Alistepes, Oscillibacter, Mucispirillum, Clostridium XIVb, unclassified genera of Clostridiales and Ruminococcaceae) in the WT and P2rx7 −/− faecal microbiota negatively correlated with body and WAT weight, blood glucose and GC B cells (Fig. S2d,e). Altogether, these data suggest that P2X7 activity in Tfh cells conditions microbiota composition and host metabolism via regulated SIgA targeting of selected bacterial genera that in turn might affect metabolically relevant taxa independently of SIgA targeting.
Intestinal Lactobacillus isolates from P2rx7 −/− mice alter glucose metabolism in wild-type animals. The genus Lactobacillus was significantly enriched in the IgA + fraction of the P2rx7 −/− microbiota ( Fig. 3a,c) and correlated with the metabolic and immune phenotype of P2rx7 −/− mice (Fig. 4a,b). Different species of lactobacilli have been associated with body weight gain 30 and juvenile growth rate 31 through the increase of dietary protein digestion and amino acid intake by the host 32 . Quantification of the genus Lactobacillus by qPCR in samples from small intestine, caecum and faeces confirmed the significant enrichment of lactobacilli in the gut of P2rx7 −/− mice (Fig. 5a). To investigate whether lactobacilli may contribute to the metabolic alterations induced by non-functional P2X7 receptor 16 , we recovered from the gastrointestinal tract of P2rx7 −/− mice different isolates that were all belonging to the species L. murinus and L. reuteri. Accordingly, we detected increased titres of faecal IgA specific for these Lactobacillus species in P2rx7 −/− as compared to WT mice (Fig. S3c). Conversely, IgA coating of small intestine microbes by faecal IgA derived from either WT or P2rx7 −/− mice was undistinguishable (Fig. S3a,b), suggesting SIgA response in the small intestine of P2rx7 −/− mice is skewed toward lactobacilli.

Discussion
Intestinal homeostasis requires a balanced microbiota 1 , which is also shaped in structure and functions by secreted IgA 33 . IgA coating identifies bacterial taxa with the potential ability to interact with the host and colonize the intestinal mucosa; in addition, it can influence bacterial gene expression, metabolism and ability to colonize different intestinal ecological niches 8,9 . Since P2X7 deficiency leads to enhanced secretion of intestinal IgA and alterations of both gut microbiota and host metabolism 15,16 , the P2rx7 −/− mouse represents a unique model for the study of the role of SIgA in the remodelling of gut microbiota and metabolic homeostasis. In fact, the enhanced production of SIgA resulted in increased SIgA coating of bacteria typically residing in the small intestine, especially Lactobacillus, Enterococcus and Enterobacteriaceae 23 .
The P2rx7 −/− SIgA + microbiota was characterized by a significant increase of the Firmicutes/Bacteroidetes ratio, a common feature of obese mice and humans 21,22 , suggesting that enhanced SIgA-coating could enrich bacterial taxa contributing to metabolic alterations. The genus Lactobacillus, belonging to the phylum Firmicutes, has been associated with body weight gain 30 , obesity 29 and modulation of SIgA production 34 , although a consensus regarding its role in health and disease has not been fully achieved 35 . In P2rx7 −/− mice, we observed a  www.nature.com/scientificreports www.nature.com/scientificreports/ or Igh-J −/− mice reproduced the impaired glucose metabolism observed in P2rx7 −/− mice. These experiments suggest that purinergic regulation of adaptive SIgA response in GALT can modulate intestinal colonization by commensals, which affect host physiology.
A physiological bacterial IgA coating regulated by T follicular regulatory (Tfr) cells and P2X7 proficient Tfh cells, contributes to the maintenance of a well-balanced intestinal microbial community within different ecological niches 11,16 . Specific changes in the IgA + and IgA − microbiota of P2rx7 −/− mice correlate with dysmetabolic features of these animals. How SIgA controls the diversification and balance of the gut microbiota is not yet clearly understood; our work sheds light on the importance of the regulation of T cell dependent SIgA via the eATP/ P2X7 axis in controlling the abundance of bacterial taxa, such as Lactobacillus, that can affect host metabolic homeostasis.
In conclusion, by analysing mice deficient in the ATP-gated ionotropic P2X7 receptor, which limits Tfh cells in the PPs and adaptive SIgA production, we positively correlated Tfh cells number and body weight with increased SIgA coating and enrichment of lactobacilli. We hypothesize the eATP/P2X7 axis constitutes a crucial www.nature.com/scientificreports www.nature.com/scientificreports/ regulatory pathway in Tfh cells to ensure controlled SIgA coating and abundance of commensals which affect host metabolism.

Materials and Methods
Mice and in vivo experiments. C57BL/6J, P2rx7 −/− (B6.129P2-P2rx7tm1Gab/J) and Igh-J −/− (B6.129P2-Igh-Jtm1Cgn/J) mice from Jackson Lab were bred in the specific pathogen-free (SPF) facility at the Institute for Research in Biomedicine, Bellinzona, Switzerland. The colonies of C57BL/6J, P2rx7 −/− and Igh-J −/− were maintained onsite with heterozygous breeders and littermates kept in the same cages until weaning at 4 week of age. Animals were housed in ventilated cages in a 12 h light/dark cycle, with free access to water and standard autoclaved chow. Food intake was measured by using metabolic cages. For the in vivo Lactobacillus administration experiments, 4 weeks old C57BL/6J and Igh-J −/− animals were treated with an antibiotic mixture containing Vancomycin (1.25 mg), Ampicillin (2.5 mg) and Metronidazole (1.25 mg) (VAM) in 200 µl water per mouse by oral gavage for 7 days to promote a more efficient bacterial colonization 36 . Later, these animals were given 5 * 10 9 CFU of Lactobacillus reuteri, Lactobacillus murinus or E. coli DH10B by oral gavage in 200 µl PBS for 21 days. Glucose tolerance test was performed as follow: animals were fasted for 12 h and then received an intraperitoneal injection of glucose (2 g/kg of body weight). Blood glucose was measured using a glucometer (Healthpro-X1, Axapharm) on samples collected from tail vein. For RER measurement, mice were transferred to single housing in Phenomaster System (TSE Systems Gmbh, Bad Homburg, Germany) one day before the study start for acclimatization, followed by two days of continued measurements. During the study period, air flow, temperature, oxygen and carbon dioxide content, oxygen uptake (VO 2 ), carbon dioxide production (VCO 2 ) were measured simultaneously using standard indirect calorimetry analysis. Respiratory exchange ratio was calculated automatically from VO 2 and VCO 2 . Data were collected in TSE Phenomaster software and exported to excel. For ex vivo experiments, mice were euthanized by CO 2 inhalation and Peyer's patches, white perigonadal adipose tissue and faeces, small intestine and caecal contents were collected. All animal experiments were performed in accordance with the Swiss Federal Veterinary Office guidelines and authorized by the relevant institutional committee (Commissione cantonale per gli esperimenti sugli animali) of the Cantonal Veterinary with authorization numbers TI44/18 and TI22/16. Faecal IgA flow cytometry and sorting of IgA + and IgA − bacteria. For analysis of IgA coated bacteria in flow cytometry, fresh faecal pellets were collected into sterile 2 mL Eppendorf tubes and homogenized in PBS (0.1 g/ml). The homogenized samples were centrifuged at 400 × g for 5 min to remove larger particles from bacteria. Supernatants were centrifuged at 8,000 × g for 10 min to remove unbound IgAs. Bacterial pellets were resuspended in PBS 5% goat serum (Jackson Immunoresearch, West Grove, PA, USA), incubated 15 min on ice, centrifuged and resuspended in PBS 1% BSA for staining with APC conjugated rabbit anti-mouse IgA antibodies (Cat.#: SAB1186; Brookwood Biomedical, Birmingham, AL, USA). After 30 min incubation, bacteria were washed twice and resuspended in 2% paraformaldehyde in PBS for acquisition at LSRFortessa. Both for analysis and sorting of the IgA + and IgA − fractions at FACSAria, forward and side scatter parameters were used in logarithmic mode. SYTO BC was added to identify bacteria-sized particles containing nucleic acids. Rag1 −/− mice were used as control for absence of Igs-coated bacteria.

Cells isolation and flow cytometry.
Determination of binding of faecal IgA to small intestine microbiota and titers of faecal IgA specific for lactobacilli. Binding of faecal IgA to small intestine microbiota and titers of lactobacilli specific IgA in faecal samples were measured by flow cytometry. To detect the binding of faecal IgA to small intestine microbiota, the intestinal content of C57BL/6 and P2rx7 −/− mice was collected and homogenized in PBS (0.1 g/ml). The homogenized samples were centrifuged at 400 × g for 5 min to remove larger particles from bacteria. Supernatants were then centrifuged at 20,000 × g for 10 min to remove unbound IgAs. The pellet was resuspended in 1 ml PBS and 10 μl of bacterial suspension were incubated with 25 μl of fecal IgA from C57BL/6 or P2rx7 −/− mice at 4 °C for 1 h. After two washes, bacteria were incubated for 30 min with biotinylated anti-mouse IgA mAb followed by SYTO-BC and Alexa Fluor 405-labeled streptavidin. The samples were resuspended in 2% paraformaldehyde in PBS for acquisition on a FACSCanto using FSC and SSC parameters in logarithmic mode.
To determine the titer of lactobacilli specific IgA in faecal samples, L. reuterii and L. murinus were resuspended at www.nature.com/scientificreports www.nature.com/scientificreports/ a density of 10 7 bacteria ml −1 . Fresh faecal samples were collected and carefully resuspended in PBS (0.01 g/ml). The obtained suspension was centrifuged two times at 20,000 × g and the supernatant collected to determine the titer of IgA specific for lactobacilli. Faecal samples were serially diluted and 25 μl of each dilution were incubated with 25 μl of bacterial targets suspension at 4 °C for 1 h. After two washes, bacteria were incubated for 30 min with monoclonal FITC anti-mouse IgA and then resuspended in 2% paraformaldehyde in PBS for acquisition on a FACSCanto using FSC and SSC parameters in logarithmic mode. ELISA was used to determine the total IgA concentration in an undiluted aliquot of the same faecal sample used for analysis in flow cytometry. Median fluorescence intensities (MFI) were plotted against IgA concentrations for each sample and 4-parameter logistic curves fitted using Prism (Graphpad, La Jolla, CA). Titers were calculated from these curves as the inverse of the antibody concentration giving an above-background signal. The concentration of total IgA titer required to achieve a given MFI (for example 200) was calculated by re-arrangement of the fitted 4-parameter logistic equation for each sample. As this value is low where a strong antibody response is present, the inverse of this value was plotted. Thus, titers are calculated as the inverse total antibody concentration required to achieve a given MFI. The y-axis value chosen as "above background" necessarily varies between experiments due to the flow cytometer settings, but is constant within any one analysis 37 .
16S rRNA gene sequencing and data analysis. DNA was extracted using the ZR faecal DNA Miniprep kit (Zymo Research, Irvine, CA, USA) following manufacturer's instructions. A primer set specific for the V5-V6 hypervariable regions was used for the amplification of the bacterial 16S rRNA gene (Fw: 5′-ATTAGATACCCYGGTAGTCC-3′ and Rev: 5′-ACGAGCTGACGACARCCATG-3′) 38 . The 16S rRNA gene amplicons were then purified and pair-end sequenced on an Illumina MiSeq platform as previously described 39 Table S1. Reads were pre-processed using the MICCA pipeline (v1.5.0) (http://www.micca.org) 40 . The overlapping 2 × 250 paired-end reads were merged using micca mergepairs 41 . Forward and reverse primer trimming and quality filtering were performed using micca trim and micca filter, respectively. De novo greedy clustering and chimera filtering were performed by using micca otu: operational taxonomic units (OTUs) were assigned by clustering the sequences with a threshold of 97% pairwise identity, and their representative sequences were taxonomically classified using micca classify with the RDP classifier version 2.11 42 . Singleton OTUs and OTUs present only in the sorted IgA + and IgA − fractions but not in the pre-sorted faecal samples were discarded from the final OTU table. Multiple sequence alignment was performed using the Nearest Alignment Space Termination (NAST) 43 algorithm implemented in micca msa with the template alignment clustered at 97% similarity of the Greengenes database 44 (release 13_05). The phylogenetic tree was inferred using micca tree 45 . Sampling heterogeneity was reduced rarefying samples at the depth of the less abundant sample (56,444 sequences). Alpha-(within-sample richness) and beta-diversity (between-sample dissimilarity) estimates were computed using the phyloseq R package 46 . Permutational MANOVA (PERMANOVA) was performed on the unweighted UniFrac distance and Bray-Curtis dissimilarity using the adonis() function of the vegan R package with 999 permutations. The identification of taxa differentially distributed in the groups of study was obtained by using the linear discriminant effect size analysis (LEfSe) 47 . LEfSe ranks features by effect size, putting at the top features that explain most of the biological difference. LEfSe combines Kruskal-Wallis and Wilcoxon rank-sum tests with linear discriminant analysis (LDA). LEfSe was performed under the following conditions: α value for the statistical test equal to 0.05 and threshold on the logarithmic LDA score for discriminative features equal to 2.0. Spearman's correlation tests were computed using the psych R package 48 . All statistical analyses were performed using R 49 and GraphPad Prism v7.04 (GraphPad Software, La Jolla, CA, USA). A p-value < 0.05 was considered significant in all cases.  50 and 20 ng of gDNA as template. A seven point standard curve consisting in tenfold serial dilutions of gDNA extracted from a Lactobacillus pure culture at known concentration was used for absolute quantification. Amplification specificity was checked by melting curve analysis, efficiency and reliability of PCR amplifications were also calculated.