ATP released by intestinal bacteria limits the generation of protective IgA against enteropathogens

T cell dependent secretory IgA (SIgA) generated in the Peyer’s patches (PPs) of the small intestine shapes a broadly diverse microbiota that is crucial for host physiology. The mutualistic co-evolution of host and microbes led to the relative tolerance of host’s immune system towards commensal microorganisms. The ATP-gated ionotropic P2X7 receptor limits T follicular helper (Tfh) cells expansion and germinal center (GC) reaction in the PPs. Here we show that transient depletion of intestinal ATP can dramatically improve high-affinity IgA response against both live and inactivated oral vaccines. Ectopic expression of Shigella flexneri periplasmic ATP-diphosphohydrolase (apyrase) abolishes ATP release by bacteria and improves the specific IgA response against live oral vaccines. Antibody responses primed in the absence of intestinal extracellular ATP (eATP) also provide superior protection from enteropathogenic infection. Thus, modulation of eATP in the small intestine can affect high-affinity IgA response against gut colonizing bacteria.

E nteric pathogens such as enteropathogenic E. coli and nontyphoidal Salmonella are a major health burden in both humans and animals. The rapid spread of antibiotic resistance in these species highlights the need for better disease prophylaxis. Protection from infection is most effective when strong mucosal immune responses have been induced, either by prior infection or by oral vaccination 1,2 . High-affinity secretory IgA (SIgA) promotes enteropathogen enchainment and aggregation to disable and clear potentially invasive species from the intestinal lumen 3 . However, balancing safety of the vaccination strain with sufficient immune stimulation has proved challenging 4 .
T follicular helper (Tfh) cells express high levels of the ATPgated P2X7 receptor, a non-selective cationic channel that opens to form a cytolytic pore when exposed to micromolar concentrations of extracellular ATP (eATP). P2X7 activity therefore controls Tfh cells abundance in Peyer's patches (PPs): Resistance of Tfh cells to ATP-mediated cell death by deletion of P2X7 enhances germinal center (GC) reactions 5 . As eATP is produced in large quantities by the intestinal microbiota, this directly suppresses commensal-specific IgA responses primed in the gutdraining lymphoid tissues and affects microbiota composition 6 . This study is based on the hypothesis that similar effects may dampen immunity against enteric pathogens and oral vaccines.
We show that ATP released by intestinal bacteria permeates the intestinal epithelium and can be found at high concentrations in hepatic portal blood. Eliminating this eATP, via administration of apyrase, dramatically improves the induction of specific IgA in response to either Salmonella infection or an inactivated oral vaccine. We could not measure any adverse effects of altered antimicrobiota immunity secondary to oral apyrase administration, suggesting that apyrase application is safe. Moreover, these enhanced immune responses provide superior protection from secondary infection.

ATP released by microbiota affects Tfh cells in PPs via P2X7.
In the small intestine and portal vein of specific pathogen free (SPF) mice, we measured micromolar concentrations of eATP that was detected at much lower levels in germ-free (GF) mice or in other circulatory districts (Fig. 1a, b). To address the contribution of the epithelium to eATP in the small intestine, we induced epithelial regeneration in the ileum by starvation and refeeding, as described 7 . In the presence of bacteria, the variations in epithelial turnover by starvation and re-feeding corresponded to undistinguishable concentrations of ileal eATP. In the absence of bacteria, starvation did not affect the percentage of proliferating cells 8 . However, the concentration of ileal ATP was dramatically reduced with respect to SPF mice with comparable amount of proliferating epithelial cells, suggesting that the great majority of eATP measured in the ileal lumen is of bacterial origin ( Supplementary Figure 1a, b). Therefore the microbiota generates high levels of eATP that can penetrate into the intestinal epithelium and draining blood. We cannot exclude that fungi, archaea, and protozoa might also contribute to the eATP present in the intestinal lumen. Consistent with other reports 9,10 ; eATP was detectable in cultures from different bacterial strains isolated from ilea of our mouse colony (Fig. 1c) and could be acutely exacerbated by vancomycin/ampicillin/metronidazole (VAM) treatment (Fig. 1d, e). In vivo VAM administration resulted in an acute significant increase of eATP in the ileum and portal vein blood (Fig. 1f). In wild type (WT), but not P2rx7 −/− mice, VAM administration-induced enhanced phosphatidylserine (PS) exposure in Tfh cells from PPs ( Fig. 1g and Supplementary Figure 1c, d), suggesting bacteria-derived eATP can modulate high-affinity SIgA response. Antibiotic treatment can be contraindicated in acute bacterial gastrointestinal infections due to negative effects on microbiota recovery. This data further suggest that antibiotic treatment may negatively effect the induction of T-cell dependent intestinal immunity in these infections.
Enhanced SIgA response by depletion of eATP. The IgA response to E. coli is dependent on Tfh cells in PPs 11 and is significantly enhanced in P2rx7 −/− mice 5 , suggesting that P2X7 activity can affect the T-cell dependent SIgA response. To address whether depleting bacteria-derived ATP could influence T celldependent IgA responses via P2X7, we used a recombinant E. coli strain (E. coli pApyr ) carrying an expression plasmid for Shigella flexneri's periplasmic ATP-diphosphohydrolase (apyrase) (Supplementary Figure 2a-d) 12,13 . The supernatant of E. coli pApyr cultures showed ATP-degrading activity that was absent in E. coli pBAD28 (Fig. 2b); its fractionation resulted in the recovery of apyrase activity within outer membrane vesicles (OMVs), suggesting the enzyme was released in the extracellular space (Fig. 2a,  c). Notably, eATP 14,15 was undetectable in cultures of E. coli pApyr , indicating that apyrase efficiently degraded ATP (Fig. 2d). To address whether bacteria-derived ATP could selectively limit SIgA responses in the small intestine of normally colonized animals, we administered E. coli pApyr or E. coli pBAD28 as control, to SPF mice by orogastric gavage (Supplementary Figure 3b). Administration of E. coli pApyr resulted in the increase of Tfh cells in the PPs concomitant to reduced Annexin V staining in flow cytometry, suggesting degradation of bacterial ATP reduced Tfh cell death via P2X7 receptor (Fig. 3a). As expected, we observed a poor SIgA response to E. coli in mice gavaged with E. coli p-BAD2816 . However, anti-E. coli IgA was significantly increased in mice gavaged with E. coli pApyr (Fig. 3b, c), despite identical intestinal E. coli load (Supplementary Figure 3a), indicating that abrogation of ATP release by bacteria results in the development of high-affinity IgA responses. The analysis of IgA in intestinal washes from mice gavaged with the two E. coli transformants on different bacterial species revealed lack of detectable reactivity (Supplementary Figure 4a). Moreover, intestinal IgA from untreated and immunized mice stained an analogous percentage of commensals from WT mice (Supplementary Figure 4b), indicating the absence of epitope-spreading to resident microbiota members.
To further address the role of ATP released by bacteria in modulating the SIgA response, we monitored endoluminal ATP after orogastric administration of E. coli pBAD28 and E. coli pApyr in mice maintained with Chloramphenicol and Ampicillin (CA) (a bactericidal mix active on endogenous flora but not on CAresistant E. coli pBAD28 and E. coli pApyr ) or Penicillin/Streptomycin/Vancomycin (PSV) (bactericidal on both endogenous flora as well as E. coli transformants) in drinking water (Fig. 3d). Oral gavaging with E. coli pBAD28 in mice maintained in PSV as compared to CA resulted in a significant acute increase of endoluminal ATP because of bacterial lysis (Fig. 3e). Notably, the analysis of anti-E. coli IgA after multiple gavaging in this setting showed that the increase in eATP concomitant to E. coli pBAD28 gavaging in the presence of PSV correlated with reduced anti-E. coli IgA with respect to the group treated with non-bactericidal CA (Fig. 3f). In contrast, in mice colonized with E. coli pApyr , ATP degradation by apyrase in both treatment groups (Fig. 3e) resulted in undistinguishable anti-E.coli IgA response (Fig. 3f). These data further show that an increased release of ATP by bacteria corresponds to a reduced generation of specific IgA. To address whether apyrase expression in live-attenuated S.Tm could increase the specific SIgA response and confer enhanced protection from infection by a virulent strain, we generated an attenuated S.Tm strain (ATCC 53648) carrying either pBAD28 (S.Tm pBAD28 ) or apyrase-bearing pHND10 (S.Tm pApyr ) (Supplementary Table 1). As observed with E. coli pApyr , ATP was undetectable in culture medium of apyrase-expressing S.Tm pApyr (Fig. 4a). Notably, Tfh and GC B cells as well as plasma cells secreting IgA specific for Salmonella LPS were all significantly increased in mice immunized with S.Tm pApyr (Fig. 4b, c and Supplementary Figure 5b-d), as was the concentration of anti-Salmonella IgA in intestinal wash (Fig. 4d).
Effective protection from Salmonella infection by S.Tm pApyr . SIgA protects the host from invasion by S.Tm or other enteropathogens by limiting the interaction of bacteria with the gut epithelium 19,20 . In the non-typhoidal salmonellosis model, both enchained growth and classical agglutination, requiring highaffinity IgA to cross-link dividing and colliding bacteria, are the main protective effects 3   The boxplot is overlaid with the visualization of single observations. Two-tailed Mann-Whitney U-tests. **p < 0.01. One representative experiment out of at least three is shown NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-018-08156-z ARTICLE approach the intestinal epithelium and invade into host tissues, this has the effect of hugely reducing the infectious burden in the intestine. We first addressed whether the enhanced IgA responses observed after vaccination with S.Tm pApyr corresponded with enhanced aggregation of Salmonella in the gut lumen. Vaccinated mice were therefore orally infected with 1:1 mixture of GFPtagged and mCherry-tagged attenuated S.Tm (S.Tm att , used to avoid confounding inflammation in the control animals) 21 and cecal content was analysed at 5 h post infection by confocal microscopy. In mice immunized with S.Tm pApyr , we observed a significant decrease in planktonic S.Tm att as compared to mice immunized with S.Tm pBAD28 (Fig. 4e), and fewer S. Typhimurium swimming in the cecal crypts (Fig. 4f). Correspondingly, when challenge infections were carried out with fully virulent S. Typhimurium (S.Tm WT ) in S.Tm pApyr vaccinated mice, disease parameters, including histopathological score, fecal Lipocalin 2 (LCN2) and GALT infection, were all significantly decreased as compared to controls or S.Tm pBAD28vaccinated mice (Fig. 5a-c). Salmonella colonizing systemic compartments (e.g. spleen and liver) follows a GALTindependent route of infection (presumably via blood circulation) 22 . Accordingly, infection with S.Tm WT determines increased permeability of the gut-vascular barrier that is reflected by translocation of FITC-dextran from the intestinal lumen into the bloodstream and liver 23 . Vaccinated mice were significantly more resistant to blood absorption of FITC-dextran administered via orogastric route with respect to non-vaccinated animals and immunization with S.Tm pApyr conferred enhanced protection (Fig. 5d). Moreover, S.Tm WT CFUs were significantly reduced in the liver and spleen of these mice (Fig. 5e). These results indicate that immunization with apyrase-expressing bacteria confers improved protection from S.Tm systemic spreading.

Protection by S.Tm pApyr depends on T-cell-dependent SIgA.
To control for effects of immunization that occur independently of adaptive immunity, we immunized recombinase-1 deficient Improved response to inactivated oral vaccines by apyrase. We tested the ability of apyrase to enhance the induction of IgA by inactivated oral vaccines 3,24 (Supplementary Figure 8a). We generated these vaccines by treating pure-cultured bacteria with 1% peracetic acid (PA)-a strong oxidizing agent. Notably, no increase of extracellular ATP was detected in the intestine of GF mice upon oral administration of inactivated bacteria (Supplementary Figure 8b). The oxidative treatment also abolished the function of the apyrase enzyme, resulting in identical IgA priming to a vaccine constructed from the empty-vector-carrying strain (Fig. 6a). Correspondingly, similar protection from infection with virulent S.Tm was observed with both vaccines (Fig. 6b, c). We therefore administered inactivated S. Typhimurium with or without crudely-purified apyrase extracted from E. coli that significantly decreases eATP in the small intestine (Supplementary Figure 8c). Inclusion of apyrase in the vaccine preparation greatly improved the IgA titer induced, similarly to that induced by infection with the live-attenuated apyrase-expressing strain (Figs. 4d, 6d), and generated enhanced protection from challenge with virulent S.Tm (Fig. 6e, f). It was previously observed that the a b  The boxplot is overlaid with the visualization of single observations. Kruskal-Wallis with Dunn's post-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. One representative experiment out of at least three is shown effectiveness of inactivated vaccines was dependent on the extent of colonization 24 . These results suggest that a dominant inhibitory effect of the microbiota is eATP production, as antibody titers induced by the inactivated vaccine in GF mice are similar to those induced by the inactivated vaccine plus apyrase in colonized animals (Fig. 6d, g). This secretory response correlates with effective protection from subsequent infection (Fig. 6e, f, h). Therefore, inclusion of recombinant apyrase into inactivated oral vaccines greatly improves the responsiveness to these preparations without any major side-effects.

Discussion
The purinergic signaling system, which uses ATP and related nucleotides as signaling molecules, plays pleiotropic roles in regulating physiological and pathological responses in virtually all mammalian tissues 25   The boxplot is overlaid with the visualization of single observations. Kruskal-Wallis with Dunn's post-test. *p < 0.05, **p < 0.01, ***p < 0.001. One representative experiment out of two is shown modality emerged very early in evolution. In the endosymbiotic relationship between α-proteobacteria and the archeon, from which the eukaryotic cell originated, ATP released by mitochondria (i.e. α-proteobacteria) evolved also as a signaling molecule to communicate mitochondrial fitness 27 . Bacteria release ATP via mechanosensitive channels 28,29 . Proteobacteria, in particular, were shown to secrete elevated amounts of ATP that were modulated by adherence to different surfaces 30 . Indeed, ATP release as well as ATP breakdown enzymatic systems are present in all kingdoms of life 26 (Supplementary Figure 2a). The observation that ATP released by bacteria limits Tfh cells abundance via P2X7 receptor and ensures controlled Tdependent SIgA responses in the small intestine indicates eATP can act as an inter-kingdom signalling molecule. This regulatory pathway plays a crucial role in shaping a beneficial gut ecosystem for host metabolism 6 . On the other hand, it contributes to the relative resistance of the intestinal adaptive immune system to generating high-affinity SIgA upon oral immunization. Overcoming this insensitivity has typically required the use of liveattenuated oral vaccines, which encompass significant safety risks. Transient subversion of bacterial ATP-mediated control of Tfh cells by apyrase could be exploited to enhance T cell-dependent IgA induction by inactivated oral vaccines or to limit the harmful potential of live-attenuated oral vaccines by specifically enhancing SIgA response and limiting intestinal inflammation, at the same time eliciting more effective and protective SIgA against enteropathogens. Our experiments demonstrate that this strategy is safe and generates potent protective immune responses. Determination of the ileal volume. To calculate the ileal volume, the last 5 cm of the terminal ileum were excised from 8-week-old-female mice and fixed in neutral buffered formalin (16 h at 4°C). After fixation samples were dehydrated (70% ethanol, two changes, 1 h each; 80% ethanol, one change, 1 h; 95% ethanol, one change, 1 h; 100% ethanol, three changes, 1.5 h each; xylene, three changes, 1.5 h each), embedded in paraffin and then cut at microtome to the desired thickness (6 μm). Ten sections spanning the 5 cm were obtained and stained with hematoxylin and eosin (H&E). The internal areas of sections were calculated by ImageJ and the mean value (base) multiplied for 5 cm (height) 31 (Supplementary Figure 9).

Methods
Quantification of ATP. For quantification of ileal ATP, intestinal content was collected by lavage with 10 ml of intestinal wash buffer (PBS, 0.5 M EDTA, Soybean trypsin inhibitor, PMSF), spun and filtered (0.22 μm) to remove any bacteria-sized contaminants and immediately frozen in dry ice. ATP concentration in the intestinal washes was multiplied for the dilution factor to obtain the actual endoluminal ATP concentration (Supplementary Figure 9). Bile and urine were collected from gallbladder and bladder through puncture with a 34G needle. For quantification of ATP secreted by commensal bacteria in culture, intestinal content was plated on BHI agar and cultured for 16 h at 37°C. Single colonies were picked and cultured in BHI broth or LB. To quantify the ATP production during bacterial growth, the bacterial culture supernatant was collected at different O.D., centrifuged and filtered (0.22 μm). For quantification of ATP in serum, inferior caval, jugular and portal veins, and hearth were exposed and blood collected through puncture with a 34G needle. Emolysed sera were discharged. The extracellular ATP concentration was evaluated by bioluminescence assay with recombinant firefly luciferase and its substrate D-luciferin according to the manufacturerʼs protocol (Life Technologies Europe B.V.).
OMVs isolation. OMVs were isolated from 250 ml of LB cultures 32 . E. coli pBAD28 and E. coli pApyr were grown to late-exponential phase (OD 600~0 .8-1.0) and removed from culture supernatants by centrifugation. The collected supernatants were filtered (0.22 µm) and concentrated using the Vivaspin 20 concentrators, molecular weight cutoff 50-kDa (GE Healthcare), to eliminate free apyrase (~27 kDa) from the medium. The collected concentrates were then centrifuged at 270,000×g for 3 h at 4°C to yield crude OMVs preparations that were resuspended in PBS and tested for apyrase activity.
Apyrase activity test. To test the apyrase activity in bacterial supernatants, OMVs and intestinal washes, samples were incubated with 50 µM ATP for 30 min at room temperature. ATP concentration was evaluated by a bioluminescence assay with recombinant firefly luciferase and its substrate D-luciferin according to the manufacturer's protocol (Life Technologies Europe B.V.). The ATPase activity of the samples was expressed as the percentage of non-degraded ATP.
Antibodies and flow cytometry. Bacterial strains and growth conditions. Full length phoN2::HA fusion, encoding periplasmic ATP-diphosphohydrolase (apyrase) of Shigella flexneri, was cloned into the polylinker site of plasmid pBAD28 (ATCC 8739387402), under the control of the P BAD L-arabinose inducible promoter, generating plasmid pHND10 12 . Bacteria transformed with pBAD28 or pHND10 were grown in LB medium supplemented with L-arabinose (0.03%) and ampicillin (100 μg ml −1 ). For infection experiments, S.Tm WT (SL1344 wild-type clone SB300) or the respective mutants were cultured in LB containing the appropriate antibiotics for 12 h at 37°C, diluted 1:20 and sub-cultured for 3 h in 0.3 M NaCl supplemented LB without antibiotics. The bacterial strains used in the study are listed in Supplementary Table 1.
Determination of specific antibody titers by flow cytometry. Specific antibody titers in mouse intestinal washes were measured by flow cytometry 34 . Intestinal contents were collected by lavages with 5 ml of intestinal wash buffer (PBS, 0.5 M EDTA, Soybean trypsin inhibitor, PMSF), spun and filtered (0.22 μm) to remove any bacteria-sized contaminants. Bacterial targets were resuspended at a density of 10 7 bacteria ml -1 . Intestinal washes 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 1 h 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 intestinal wash sample used for analysis in flow cytometry. Median fluorescence intensities (MFI) were plotted against antibody 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 abovebackground signal. The concentration of total antibody titer required to achieve a given MFI (for example = 500) was calculated by re-arrangement of the fitted 4parameter logistic equation for each samples. 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 34 . Treatment of bacterial cultures with antibiotics. Ampicillin (2.5 μg ml −1 ), vancomycin (1 μg ml −1 ), metronidazole (1 μg ml −1 ) were added to intestinal bacterial culture when OD 600 reached 0.5 value. At different times after addition of antibiotics, bacterial cultures were spun and supernatants collected in a sterile tube. ATP concentration was evaluated by bioluminescence assay in filtered supernatants (see above).
Production of peracetic acid killed vaccines. To produce PA killed vaccines, bacteria grown for 16 h to late stationary phase were collected by centrifugation and resuspended at a density of 10 9 -10 10 ml −1 in sterile PBS. Peracetic acid (Sigma-Aldrich) was added to a final concentration of 1% and after vigorous mixing, the suspension was incubated for 60 min at room temperature. After extensive washing, the final pellet was resuspended at a concentration of 10 11 particles ml −1 in sterile PBS 24 . Each batch of vaccine was tested before use in order to confirm absence of live bacteria.
Preparation of periplasmic extract. E. coli pBAD28 and E. coli pApyr were grown as described above and collected by centrifugation. After washing, bacteria were resuspended (10 10 CFU ml −1 ) in PBS with 30 mM Tris-HCl (pH 8.0), 4 mM EDTA, 1 mM PMSF, 20% sucrose and 0.5 mg ml −1 lysozyme and incubated 2 min at 30°C. MgCl 2 (10 mM final) was added to the bacterial solution and incubation was continued for 1 h at 30°C. At the end of the incubation period bacterial suspensions were centrifuged at 11,000×g for 10 min at 4°C and supernatants were stored (periplasmic extract).
Oral vaccination protocols. For vaccination with E. coli transformants, E. coli p-BAD28 and E. coli pApyr were collected by centrifugation, washed in sterile PBS and 10 10 CFUs administered to mice by orogastric gavage. The procedure was repeated every 3 days for 3 weeks and mice were sacrificed at day 22 or 28 (Supplementary Figure 3b). For vaccination with S.Tm transformants, S.Tm pBAD28 and S.Tm pApyr were collected by centrifugation, washed in sterile PBS and 5 × 10 9 CFUs administered to mice by orogastric gavage. The procedure was repeated every 3 days for three times (Supplementary Figure 5b). The pBAD promoter is constantly active in the gut lumen (E.S. and W.D.H., unpublished observations), however, to ensure optimal apyrase expression by administered S.Tm transformants, mice were maintained with 0.3% arabinose in drinking water. On day 28, mice were used for infection experiments. For oral vaccination with PA killed S.Tm pBAD28 and S. Tm pApyr , mice received 10 10 particles of the respective transformants in PBS every 3 days for three times. When periplasmic extracts from E. coli pBAD28 (mock extract) or E. coli pApyr (APY extract) were used, 100 μl of the extract was orally gavaged every 12 h starting three days before the first day of immunization until day 10. On day 28, mice were used for infection experiments (Supplementary Figure 8a).
Challenge infections with S. Typhimurium. Mice were pretreated with 1 g kg −1 streptomycin sulfate in sterile PBS by gavage. Twenty-four hours later, S.Tm wt (10 8 CFUs 0.1 ml −1 PBS) were gavaged into the stomach. For determination of total bacterial loads, homogenates of PPs, mLN, spleen and liver collected at 24 and 48 h after infection, were plated on MacConkey agar plates containing 50 μg ml −1 streptomycin.
Live confocal microscopy of cecal content. Vaccinated or control mice were pretreated with 0.8 g kg −1 ampicillin sodium salt in sterile PBS by gavage. Twentyfour hours later, mice received 10 7 CFUs of a 1:1 mix of mCherry-(pFPV25.1) and GFP-(pM965) expressing avirulent S.Tm. For imaging, cecum content was gently diluted to 1:10 w/v in sterile PBS containing 6 μg ml −1 chloramphenicol, avoiding heavy mixing 3 . The suspension (200 μl) was transferred to 35 mm dish, 14 mm glass diameter, poly-D-lysine coated Petri dish (MatTek Corporation) and imaged using a Leica TCS SP5 confocal microscope with a ×100/1.44 NA oil immersion objective (HCX PL APO CD ×100/1.44 oil). Individual bacteria were visually scored as planktonic, whereas aggregates of equal or more than three bacteria were scored as clumps.
Multiphoton microscopy and analysis. Vaccinated or control E-cadherin-mCFP mice were pretreated with 0.8 g kg −1 ampicillin sodium salt in sterile PBS by gavage. After 24 h, mice received 10 7 CFUs of S.Tm att expressing GFP constitutively. After 18 h, the infected animals were sacrificed and the whole-ceca collected for 2-photon analysis. Deep tissue imaging was performed on a customized two-photon platform (TrimScope, LaVision BioTec) 35 . The objective used was a Nikon Apo LWD ×25/1.10W IR Corrected. The fluorescence signal has been separated using a custom configuration of detecting PMTs equipped with a set of dichroic mirror and selective bandpass filters for the Blue, Green, and Red channels (respectively detecting the fluorescence in the range of 450 nm-495 nm, 500 nm-550 nm, 600 nm-635 nm). 3D reconstructions of the whole crypts shown in Fig. 4f and Supplementary Figure 10 were performed by acquiring a z-stack with zstep between slices of 3 μm, for a total depth of 120 μm. Images were analysed using FIJI software 36 with a custom-developed macro to automate image processing: the internal area of crypts was segmented by applying a threshold on the intensity of the red autofluorescence (intensity above I = 1500 a.u.). Within the identified ROI (in the red channel) the green particles above the threshold (intensity above I = 5000 a.u. in the green channel) were detected and total fluorescence was measured by summing the single particles fluorescence intensity along the entire z-stack. The obtained total intensity was normalized by the total internal area of the crypts (obtained by summing up the single slice areas found in the red channel). 3D renderings shown in Fig. 4f and Supplementary Figure 10 were made with Clear Volume plugin of FIJI software 37 .
Histological evaluation of Salmonella-induced typhlitis. Ceca from all animals were examined at necropsy, fixed in 10% neutral buffered formalin for at least 48 h prior to embedding in paraffin and stained with H&E. Pathological scores were determined in a blinded manner using a scoring scheme, which takes into consideration the severity of submucosal edema (scores 0-3), neutrophilic infiltration into the lamina propria (scores 0-4), loss of goblet cells (scores 0-3), and epithelial damage (scores 0-3) 17,38 .
Phylogenetic tree of apyrases. Protein sequences were retrieved from GenBank by searching for the terms "apyrase" and "ectonucleoside triphosphate diphosphohydrolase". Putative and partial sequences were excluded from the analysis. Protein sequences were aligned using MUSCLE 39 . The phylogenetic trees were inferred by using FastTree2 40 .