Activated regulatory T-cells promote duodenal bacterial translocation into necrotic areas in severe acute pancreatitis

1355 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas Original research Activated regulatory Tcells promote duodenal bacterial translocation into necrotic areas in severe acutepancreatitis Juliane Glaubitz,1 Anika Wilden,1 Fabian Frost,1 Sabine Ameling,2 Georg Homuth,2 Hala Mazloum,1 Malte Christoph Rühlemann ,3,4 Corinna Bang ,3 Ali A Aghdassi ,1 Christoph Budde,1 Tilmann Pickartz,1 Andre Franke ,3 Barbara M Bröker,5 Uwe Voelker,2 Julia Mayerle ,6 Markus M Lerch,1 FrankUlrich Weiss,1 Matthias Sendler 1 To cite: GlaubitzJ, WildenA, FrostF, etal. Gut 2023;72:1355–1369. ►Additional supplemental material is published online only. To view, please visit the journal online (http:// dx. doi. org/ 10. 1136/ gutjnl2022327448). For numbered affiliations see end of article. Correspondence to Dr Matthias Sendler, Department of Medicine A, University Medicine Greifswald, Greifswald 17475, Germany; matthias. sendler@ unigreifswald. de FUW and MS are last authors. Received 25 March 2022 Accepted 30 December 2022 Published Online First 11January2023 © Author(s) (or their employer(s)) 2023. Reuse permitted under CC BYNC. No commercial reuse. See rights and permissions. Published by BMJ. ABSTRACT Objective In acute pancreatitis (AP), bacterial translocation and subsequent infection of pancreatic necrosis are the main risk factors for severe disease and late death. Understanding how immunological host defence mechanisms fail to protect the intestinal barrier is of great importance in reducing the mortality risk of the disease. Here, we studied the role of the Treg/Th17 balance for maintaining the intestinal barrier function in a mouse model of severe AP. Design AP was induced by partial duct ligation in C57Bl/6 or DEREG mice, in which regulatory Tcells (Treg) were depleted by intraperitoneal injection of diphtheria toxin. By flow cytometry, functional suppression assays and transcriptional profiling we analysed Treg activation and characterised Tcells of the lamina propria as well as intraepithelial lymphocytes (IELs) regarding their activation and differentiation. Microbiota composition was examined in intestinal samples as well as in murine and human pancreatic necrosis by 16S rRNA gene sequencing. Results The prophylactic Tregdepletion enhanced the proinflammatory response in an experimental mouse model of AP but stabilised the intestinal immunological barrier function of Th17 cells and CD8+/γδTCR+ IELs. Treg depleted animals developed less bacterial translocation to the pancreas. Duodenal overgrowth of the facultative pathogenic taxa Escherichia/Shigella which associates with severe disease and infected necrosis was diminished in Treg depleted animals. Conclusion Tregs play a crucial role in the counterbalance against systemic inflammatory response syndrome. In AP, Tregactivation disturbs the duodenal barrier function and permits translocation of commensal bacteria into pancreatic necrosis. Targeting Tregs in AP may help to ameliorate the disease course. Regulatory T cells could represent a therapeutic target for the prevention of infected necrosis during severe acute pancreatitis, which will improve the disease course and outcome. INTRODUCTION Acute pancreatitis (AP) is the most common nonmalignant disease of the gastrointestinal (GI) tract leading to hospital admission with an increasing incidence over the past years in western countries.1 Eighty per cent of pancreatitis cases follow a selflimiting course of disease without complications or longtime hospitalisation. About 20% of patients develop severe AP (SAP), which is associated with systemic complications and increased morbidity and mortality.2 Infected pancreatic necrosis, which is believed to arise from translocation of commensal gut bacteria into the pancreas,3 is regarded to drive persistent organ failure and subsequent mortality.2 Initially bacteria need to overcome the intestinal barrier and evade the immune system to cause infected necrosis. Especially the Treg/Th17balance is crucial for maintaining the intestinal barrier function and regulating tissue homoeostasis.4 WHAT IS ALREADY KNOWN ON THIS TOPIC ⇒Infected necrosis is a severe complication during acute pancreatitis (AP) and is associated with a significant mortality. WHAT THIS STUDY ADDS ⇒In an animal model, the experimental induction of AP associates with significant changes of the intestinal microbiota composition and a marked increase of facultative pathogenic bacteria. The same bacterial taxa are identified in necrotic tissue samples of pancreatitis patients. ⇒During AP, activated regulatory T cells suppress the systemic immune response and impair the immunological intestinal barrier function. ⇒The depletion of Tregs during AP reduced the bacterial translocation of facultative pathogenic strains into the inflamed pancreas. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY ⇒Regulatory T cells could represent a therapeutic target for the prevention of infected necrosis during severe AP which will improve the disease course and outcome. 1356 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas Pancreatitis is a primarily sterile inflammation originating from premature intracellular protease activation. Activated zymogens trigger acinar cell death5 and activate the immune regulating transcription factor NFκB.6 This inflammatory process recruits macrophages and neutrophils to the site of inflammation which further increases the local damage.7–10 Macrophages represent the majority of resident pancreatic immune cells and their activation by damageassociated molecular patterns (DAMPs) initiates a systemic immuneresponse,8 11–14 including the activation of the adaptive immune system through the release of cytokines.12 During an episode of AP resident lymphatic Tcells initiate a suppressive and antiinflammatory response mediated by an increase in FOXP3+/CD25+ regulatory Tcells (Treg) and GATA3+ Th2cells.12 15 This parallel initiation of a systemic inflammatory response syndrome (SIRS) and a compensatory antiinflammatory response syndrome (CARS) is also known from sepsis or critical trauma.16 17 Systemic immunosuppression (CARS), in particular mediated by Tregs, can promote secondary infections18 and thus aggravate disease severity or cause nonpancreatic complications. To date, little is known about the relationship between the gut microbiome and the course of AP.19 Here, we investigated in an animal model of AP how Tregmediated immunosuppression affects the gut microbiome composition and the translocation of bacteria into the inflamed pancreas. RESULTS AP is associated with changes of the intestinal microbiota composition AP was induced by partial duct ligation in C57Bl/6J mice. Duodenal aspirates and faecal samples from distal colon and caecum were collected from AP mice and untreated control animals. 16S rRNA gene sequencing revealed microbiota differences between control and AP mice (figure 1A). Principal coordinate analysis (PCoA) of all GItract samples confirmed microbiota changes, with the most prominent changes seen in duodenal aspirates (figure 1B, online supplemental figure S1A,B). A significantly reduced Shannondiversity and amplicon sequence variant (ASV) richness was observed in the duodenal aspirates of AP animals. Smaller changes of the microbiota composition were observed in caecal but not in distal colonic samples (figure 1C,D). Analysis of the most abundant taxonomic units identified an increase of facultative pathogenic bacteria like Escherichia/Shigella, Enterobacteriaceae diversa, Enterococcus or Staphylococcus. Especially in duodenal samples all these taxa were significantly enriched during AP. In colon and caecum samples, we could also observe an increase of these facultative pathogenic bacteria, but only Escherichia/Shigella showed a significant increase in all parts of the GItract (figure 1E–G). Their high abundance in the duodenal section suggests bacterial overgrowth (online supplemental figure S1C). In contrast, the abundance of beneficial commensal bacteria such as Lachnospiraceae was decreased, and this effect was again most pronounced in the duodenum (figure 1E–G). To verify the results from the model of AP using partial pancreatic duct ligation we investigated the microbiome changes in duodenum and colon samples in a second pancreatitis model. To analyse comparable time points we induced pancreatitis via 8hourly intraperitoneal (i.p.) injections of Caerulein (50 µg/kg/bodyweight) over 3 days. Again, we saw significant changes in the intestinal microbiome composition with the strongest effects in the duodenum (online supplemental figure S2A–C). The disease severity in this model was Figure 1 Acute pancreatitis (AP) is associated with changes of the intestinal microbiota composition. (A)Faecal samples from colon, caecum and duodenum were collected from C57Bl/6 mice with AP (n=24) and from untreated control animals Con (n=10). Isolated DNA was analysed by 16S rRNA gene sequencing. The microbiota composition and APassociated changes of the taxonomic units with the highest abundance are illustrated by a stacked bar graph. (B)Principal coordinate analysis illustrates the changes of gut microbiome between untreated controls and AP mice. (C, D)ShannonDiversity Index (C)and richness of observed species (D)demonstrate significant impact of AP on the duodenal microbiome composition. (E–G)The bar graph illustrates changes of the most abundant taxa in colonic (E),caecal (F)and duodenal samples (G)of AP (red) and control mice (grey). Facultative pathogenic were marked in red, beneficial commensal bacteria were marked in green. Statistical significance was determined by unpaired Student’s ttest and KruskalWallis test followed by a Dunn’s multiple comparisons test to analyse differentially abundant taxa in colon, caecum and duodenum samples. Significance levels of p<0.05 are marked by an asterisk. 1357 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas significantly lower compared with the model induced by partial pancreatic duct ligation (online supplemental figure S2D,E) and we observed less pronounced changes of the intestinal microbiota composition (online supplemental figure S2A–C) and nearly no changes of the systemic immune response (online supplemental figure S2F–H). Intestinal microbiota changes correlate with the severity of AP and the systemic immune response In the next step, we investigated how far the duodenal microbiome changes are influenced by disease severity and through the systemic immune response by analysing splenic lymphocytes. Unexpectedly LY6G+/LY6Clow/CD11b+- cells in spleen of AP mice increased significantly during AP but the concomitant decrease of serum amylase as marker of pancreatic damage suggested an inverse correlation to disease severity (figure 2A). Splenic LY6G+/LY6Clow/CD11b+- cells showed the same inverse correlation with serum lipase activity, histological damage and the serum cytokine level of IL6 and TNFα (online supplemental figure S3A–C). Furthermore, we detected a positive correlation of splenic LY6G+/LY6Clow/CD11b+- cells with the spleen weight, and a negative correlation with the number of splenic CD25+/ FOXP3+ Tregs which are significantly elevated during pancreatitis (online supplemental figure S3D,E). Elevated apoptosis in spleen Figure 2 Intestinal microbiota changes correlate with the severity of AP and the systemic immune response. (A)Dot plot illustrates negative correlation of LY6G+/LY6Clow/CD11b+- cells to the activity of serum amylase of mice with AP (n=50), significance of correlation was tested by Spearman’s rank correlationcoefficient. (B)Disease severity in animals was classified according to the percentage of LY6G+/LY6Clow/CD11b+- cells in spleen (n=5/group). (C)Box plots illustrate the duodenal microbiome species richness and Shannondiversity according to moderate and severe pancreatitis. (D)Principal coordinate analysis illustrates significant differences of duodenal microbiome between these groups (permutational multivariate analysis of variance; control vs severe: p=0.007, R2=59.4%, control vs moderate: p=0.015, R2=24.5% and moderate vs severe: p=0.007, R2=48.4%). (E)Heatmap illustrates 16S rRNA gene sequencing results of the duodenal microbiome analysis, facultative pathogenic were marked in red, beneficial commensal bacteria were marked in green. (F)H&Estaining of pancreatic tissue of healthy and AP mice. (G)Colony forming units (CFU) were counted from necrotic tissue homogenates and exceeded the cutoff level of 1 CFU/mg tissue weight in 5 of 21 (23.8%) animals. (I)From infected (n=5) and noninfected (n=6) groups, we analysed in spleen the percentage of LY6G+/LY6Clow/CD11b+- cells and (H)the percentage of FOXP3+/CD25+/CD4+- cells. Statistical evaluation was done by unpaired Student’s ttest for independent samples and significance levels of p<0.05 are marked by an asterisk. AP, acute pancreatitis. 1358 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas could be responsible for the loss of LY6G+/LY6Clow/CD11b+- cells (online supplemental figure S3F). Based on these findings, we dichotomised the set of animals in a splenic LY6G+/LY6Clow/ CD11b+ high (moderate) or low cell (severe) cohort and investigated severity dependent changes of the duodenal microbiota composition (figure 2B). 16S rRNA gene sequencing analysis confirmed a severitydependent loss of richness and diversity (figure 2C). PCoA of duodenal aspirates from moderate and SAP showed severitydependent significant changes of the microbial composition (figure 2D). A detailed analysis of the most frequent duodenal microbiota taxa is illustrated in the heat map (figure 2E and online supplemental table 1). Facultative pathogenic bacterial clades like Escherichia/Shigella, Enterococcus or Staphylococcus were again detected with increased abundance (online supplemental figure S4A). Infected necrosis is a lifethreatening complication of AP and regarded to originate from translocation of commensal gut bacteria. In our mouse model of AP we examined, bacterial infection of the ligated necrotic part of the pancreas by colony growth assays of the tissue homogenate on agar plates (figure 2F–G). In 5 of 21 animals, we detected a significant bacterial load in pancreatic necrosis with more than 1 colonyforming unit (CFU)/mg pancreas, which is comparable to infection rates seen in patients matched for severity.2 A negative correlation was detected for the rate of infection of pancreatic necrosis and number of splenic LY6G+/LY6Clow/CD11b+- cells (figure 2I and online supplemental figure S4B). To test a functional relationship between LY6G+/LY6Clow/CD11b+- cells and the infection of pancreatic necrosis we depleted LY6G+- cells by antiLY6G antibody treatment before induction of pancreatitis. Treatment with antiLY6G abolished LY6G+/LY6Clow/CD11b+- cells (online supplemental figure S4A–D) and did not affect bacterial translocation into the necrosis (S.4E). Our finding suggested that reduction of LY6G+/LY6Clow/CD11b+- cells in spleen can serve as marker to discriminate between moderate and SAP in mice, but apparently has no functional impact. Since the depletion of LY6G+/LY6Clow/CD11b+- cells did not affect the adaptive immune cells (online supplemental figure S4F) we investigated in further experiments the cells of the adaptive immune response the spleen. Labelling of the IL2 receptor-αchain (CD25) on the surface of CD4+ Tcells (online supplemental figure S5A) revealed a significantly stronger increase in the number of activated CD25+ Tcells in those animals with an infected pancreatic necrosis, indicated by a bacterial load of more than 1 CFU/mg pancreatic tissue. CD69, a second activation marker of CD4+ Tcells, showed the same association (online supplemental figure S5B). We next investigated Tcell differentiation by labelling of the transcription factors TBET, GATA3 and FOXP3. Whereas Th1cell differentiation transcription factor TBET was not affected by pancreatitis (online supplemental figure S5C), GATA3+ Th2cell numbers were found elevated in the spleen, independent of bacterial infection (online supplemental figure S5D). Interestingly, the number of FOXP3+/ CD25+ Tregs was significantly increased in animals suffering from pancreatitis and was highest in those with infected necrosis (figure 2H). Furthermore, we detected a direct correlation between Treg rates with the bacterial load (online supplemental figure S5E). Imbalance of the Th17/Treg ratio in duodenal mucosa during AP To colonise pancreatic necrosis commensal bacteria need to overcome the intestinal barrier. Duodenum histology of AP mice did not show obvious differences (figure 3A). Immunofluorescent labelling of F4/80 or CD68 demonstrated a significant reduction of macrophages within the lamina propria, also sIgA producing plasma cells were significantly reduced (figure 3B,C). The number of Tcells was also affected by AP, CD3+ Tcells within the lamina propria as well as CD8α+ intraepithelial lymphocytes (IELs) were significantly reduced (figure 3D,E). Additionally, we analysed global mRNAprofiles of mouse duodenal tissues (AP vs control) using Affymetrix GeneChiparrays. Analysis of the generated transcriptome data by QIAGEN’s IngenuityPathwayAnalysis (IPA) software identified APspecific differential gene expression in the Tcellresponse, Bcellresponse and other immune systemrelated pathways (figure 3F). Transcript levels of genes encoding Tcell surface markers CD3, CD8α, CD4 and CD28 were significantly reduced, while SOCS3 (suppressor of cytokine signalling 3) specific mRNA was significantly more abundant (figure 3G). Bcellresponse markers like CD79A, TNFS13/proliferationinducing ligand APRIL, TNFSF13B/Bcell activating factor BAFF, or JCHAIN (part of Immunglobulin A) exhibited significantly lower transcript levels in AP mice (figure 3G). In contrast to an apparently attenuated immune response, the mRNA abundance of antimicrobial defence peptides such as DEFA1, REG3A, REG3G, DEFA23, or DEFA21 was significantly higher in the duodenum of AP mice (figure 3G). The expression of genes encoding structural protein components of the intestinal barrier like tight junctions, desmosomes or adherence junctions were only in some cases affected by AP (online supplemental figure S6A). We could observe a significant downregulation of mRNAs encoding for Claudin15 (Cldn15), the Myosin light chain kinase (Mylk), cadherin related family member 2 (Cdhr2) or Pannexin 1 (Panx1), whereas other mRNAs showed a significant upregulation after onset of disease such as Dsp, encoding for Desmoplakin, Cdh5 encoding for the protein Cadherin 5 or the mRNAs encoding for mucins 2, 3 and 3a (Muc2, Muc3 and Muc3a). Cryptcell proliferation, marked by KI67, was not affected by AP (online supplemental figure S6B). Most structural proteins were unaffected by AP like ECadherin (online supplemental figure S6C). On the other hand, AP exhibited significant effects also on several pathways of the immune homoeostasis (cytokines, transmembranereceptors and transcriptionalregulators) in the duodenal system (online supplemental figure S6D–F). To evaluate the effect of disease severity on the intestinal immune response, we isolated and analysed lymphocytes of the lamina propria and duodenal IELs by flow cytometry with respect to disease severity (figure 3H–L). The purity of both cell types was confirmed detecting the CD8α/CD4 ratio (S.7AB). We observed a severitydependent decrease of CD4+ T cells in the lamina propria (online supplemental figure S7C). Within this population of CD4+ Tcells we measured an increase of CD25+/FOXP3+ Tregs during pancreatitis, with the highest numbers observed in SAP (figure 3H). The same result was obtained for the population of IELs, but at a significantly lower ratio (figure 3I). Within the duodenum we observed a severitydependent increase in the number of CD25+/FOXP3+ Tregs (figure 3I,J). Furthermore, effector Tcell (Teff) numbers were significantly elevated during pancreatitis, the induction of CD25+/RORγt+ (Th17) IELs and CD25+/TBET+ (Th1) was not different between moderate or SAP. The ratio of CD25+/ RORγt+ increased to nearly 10%, whereas CD25+/TBET+ did not rise above 1% of total IELs (figure 3K,L). Analysis of GFPexpressing Tregs in DEREGmice (DEpletion of REGulatory Tcells)20 revealed after induction of AP a significant redistribution of GFP+/CD3+ Tcells within the lamina propria, 1359 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas confirming the cytometry analysis (figure 3M). Notably the Teff/ Tregratio was significantly lower in animals suffering from SAP (figure 3N). In summary, these results suggest that AP causes general immunosuppression in the intestinal mucosa and pronounced changes of the intestinal microbiota composition. Immune suppressive function of Tregs during AP To illuminate the role of Tregs, we analysed their pancreatitisassociated suppressive capacity. AP was induced by partial duct ligation in DEREGmice expressing a gene encoding a GFPdiphtheria toxin (DT) receptor fusion protein under the control of a Foxp3promoter.20 GFP flow cytometric analysis confirmed increased Treg numbers after induction of AP (figure 4A). The same increase we could observe in Tcells isolated from the duodenum (figure 4B). Using a fluorescenceactivated cell sorter, GFPpositive Tregs were isolated from the spleen of DEREGmice after APinduction and of control animals. We coincubated Figure 3 Imbalance of the Th17/Treg ratio in duodenal mucosa during AP. (A–E)Histological examination of the duodenum illustrates changes of the intestinal mucosa in C57Bl/6 mice after induction of AP. (A–C)In contrast to only slight changes in H&Estaining we detected differences in immunofluorescent labelling of F4/80, CD68 as marker for macrophages and sIgA as marker for plasma cells (B),all cells were significantly decreased in AP mice (C).(D)CD3 as Tcell marker and CD8α as marker of IELs were labelled in small intestine. (E)Bar graphs illustrate significant changes of Tcells and IELs in duodenum. (F)The bar graph shows the -log10(pBH) values of overrepresented pathways from an IPA analysis which was based on differentially expressed transcriptome data of duodenal tissue from AP and control mice. Heatmap illustrates the zscore which indicates activation (positive zscore, red) or inhibition (negative zscore, blue) of this pathway. (G)Heatmap illustrates fold change differences of upregulated (red) and downregulated (blue) genes in duodenal tissue of AP mice compared with untreated controls. (H)Lymphocytes isolated from lamina propria as well as from the epithelial layer were analysed by flow cytometry. In CD4+ Tcells of the lamina propria the increased levels of CD25 and FOXP3 correlated with disease severity. (I–J)The numbers of CD4+/FOXP3+/CD25+- Tregs (I)and CD4+/CD25+- Tcells (J)from IELs were also increased in a severitydependent manner. (K, L)Th17cells marked by RORγt+/CD25+ (K)and Th1cells marked by TBET+/CD25+ (L)did not show a severity dependent increase. (M)GFPproducing Tregs in the duodenum of DEREGmice were detected by antiGFP and antiCD3 labelling and showed a significant increase after onset of AP. (N)AP induced a shift of the Teff/Treg ratio. Tregs showed a clear increase between moderate and severe AP in contrast to Teff cells (Th1, Th2 and Th17). Statistically significant differences were tested by unpaired Student’s ttest for independent samples and significance levels of p<0.05 are marked by an asterisk. AP, acute pancreatitis. 1360 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas the Tregs at ratios of 1:8, 1:4, 1:2 and 1:1 with naïve CMDillabelled CD3/CD28 activated Tcells in the presence of IL2. Dilution of CMDil fluorescence by cell division was used to measure the suppressive activities of Tregs on Teffcell proliferation and showed a significantly stronger suppression by Tregs from AP animals (figure 4C). In addition, we analysed the transcriptome of the isolated Tregs using Affymetrix GeneChips. Expression of Il10 and Tgfbi were increased, whereas the mRNA levels of Tnf and Stat1 were decreased (figure 4D). Pathway analyses of the transcription data using IPAsoftware indicated APassociated downregulation of cell cycle regulation and of the general immune response pathways in Tregs (online supplemental figure S8A). Upstream regulators included the cytokines IFNγ, IL10, the transcription factors STAT3, FOXP3 and STAT6, and the transmembrane receptors IL10RA, CD28 and CTLA4 (online supplemental figure S8B–D). We further analysed APinduced alterations in the global mRNA profile of isolated Tregs from spleen and from duodenal samples. In both subsets the ‘PD1, PDL1 cancer immunotherapy pathway’ involved in Tregmediated immunosuppression was significantly induced. Analysis of diseaserelated and biofunctionrelated pathways revealed negative activation scores for many immune pathways. Innate and adaptive immune cells seem equally affected and lymphocyte activation was significantly inhibited (online supplemental figure S9A–D). Depletion of Tregs stabilises the homoeostasis of CD4+ T cells in the lamina propria as well as of CD8α+ IELs in the duodenum To verify the immunosuppressive effect of Tregs during AP we depleted Tregs in DEREGmice by i.p. injection of DT before the induction of pancreatitis. The depletion of Tregs in adult mice does not result in autoimmune disease.21 By flow cytometry (figure 5A) and immunofluorescence detection of GFP+ cells in lymph nodes (figure 5B) we confirmed a significant reduction of GFP+/CD25+ Tregs in DTtreated animals, whereas phosphatebuffered saline (PBS)- treated control mice showed the typical APinduced increase of Tregs (figure 5C). The Teff populations in spleen, GATA3+/CD4+ Th2cells and TBET+/CD4+ Th1cells were significantly increased compared with control animals (online supplemental figure S10A). The population of splenic LY6G+/LY6Clow/CD11b+- cells, which inversely correlate with the disease severity were significantly elevated in DTtreated mice, whereas myeloperoxidase activity in lung tissue, a marker of lung injury, was comparable to PBStreated controls (online supplemental figure S10B). We next investigated if the Treg depletion affects duodenal CD4+ Tcells and CD8α+ IELs populations. Immunofluorescence labelling of CD3+ Tcells and CD8α+ IELs showed, that following depletion of Tregs the numbers of CD3+ cells and CD8α+ IELs were significantly higher in DTtreated animals compared with PBStreated mice (figure 5D,E). Subsequently, we isolated leucocytes from the duodenum of DTtreated DEREGmice after induction of AP. PBStreated animals were used as controls. In DTtreated animals GFP+/CD25+ CD4+ Tregs were absent from the lamina propria (figure 5F). Examination of CD4+Teff cells within the lamina propria showed an increase of RORγt+/CD25+ Th17cells (figure 5G), whereas the number of TBET+/CD25+ Th1cells was not changed after Treg depletion (figure 5H). Interestingly, we identified a population of CD4hi cells within the population of IELs which is increased in DTtreated animals (figure 5I). A detailed characterisation of this CD4hi population identified mainly GFP+/CD4hi Tregs and RORγt+/CD4hi Th17cells. Whereas RORγt+/CD4hi Th17cells persisted in DTtreated animals, Tregs were completely abolished (figure 5J,K). We further analysed the population of CD8α+ IELs, which account for the majority of IELs. While immunofluorescence labelling of CD8ɑ+ IELs and of CD3+ Tcells of the lamina propria showed a significant reduction in numbers (figure 5D,E), their ratio was not affected by DTtreatment (figure 5L). Surface markers allow to sort IELs into two different populations: (A) Figure 4 Immune suppressive function of Tregs during AP. AP was induced by partial duct ligation in DEREGand C57Bl/6 mice. (A)The number of splenic GFP producing Tregs was increased in the AP group (n=5) but not in control or sham operated mice (n=3). (B)The same increase of FOXP3+/ CD25+/CD4+ Tregs was observed in the small intestine. (n=5). (C)Suppression assays showed that Tregs from AP mice (red line) have an increased suppressive capacity on Teff cell proliferation. (D)Heat map illustrates fold changes of gene transcription in Tregs from AP mice (n=4) compared with untreated controls (n=4). Statistically significant differences were tested by unpaired Student’s ttest for independent samples and significance levels of p<0.05 are marked by an asterisk, corrected for multiple testing (bonferronicorrection). AP, acute pancreatitis. 1361 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas inducedIELs positive for CD4+/CD8α+ and TCRαβ+ (Tcell receptor), and (B) naturalIELs which are specific for CD8α+/ TCRγδ+ or CD8αβ-/TCRαβ+.22 Analysis of TCRβ and TCRγδ on the surface of CD8α+ IELs in DEREGmice revealed that the population of induced IELs (CD8α+/TCRβ+) was increased in AP, while the number of natural IELs (CD8α+/TCRγδ+) was significantly decreased. This TCRαβ+/TCRγδ+ shift was less pronounced in Tregdepleted mice (figure 5M). Finally, we investigated the presence of the transcription factors TBET and RORγt within the CD8α+ IELs. No differences in the number of TBET synthesising IELs were observed in the absence of Tregs, whereas RORγt producing IELs were significantly increased Figure 5 Depletion of Tregs stabilises the homoeostasis of CD4+ T cells in the lamina propria as well as of CD8α+ IELs in the duodenum. AP pancreatitis was induced in DEREG mice by partial duct ligation, Tregs were depleted by diphtheria toxin (DT) (n=11), controls receive PBS (n=7). (A)The efficiency of Treg depletion with DT was verified by flow cytometry analysis of splenocytes. (B)Immunofluorescence detection of GFPproducing Tregs was performed in lymph nodes of DTtreated and PBStreated mice. (C)Bar graphs show the ratios of CD25+ Tcells and GFP+/CD25+ Tregs analysed by flow cytometry of splenocytes. (D, E)Immunofluorescent labelling of CD3 and CD8α showed a significant persistence of the Tcell and IEL population in the absence of Tregs in DTtreated DEREGmice compared with the PBStreated group, (E).(F–N)Lymphocytes of the lamina propria and the epithelial layer were isolated from duodenum and analysed by flow cytometry. (F)Following Tregdepletion, no GFP+/CD25+ Tregs were detected within the lamina propria. (G, H)Bar graph show the numbers of RORγt+/CD25+ Th17cells (G)and TBET+/CD25+ Th1cells (H).(I)The bar graph shows a higher ratio of CD4hi expressing IELs in DTtreated mice. (J, K)In the CD4hi IELs we observed a shift from GFP+ Tregs (J)to RORγt+ Th17cells (K).(L)The ratio of CD8α+ IELs was not affected by Tregdepletion. (M)Dot plot and bar graphs illustrate the ratio changes of TCRγδ+ IELs and TCRβ+ after DTtreatment. (N)Within the population of CD8α+ IELs we measured an increase in RORγt producing cells whereas TBET was not affected. Statistically significant differences were tested by unpaired student’s ttest for independent samples and significance levels of p<0.05 are marked by an asterisk. AP, acute pancreatitis; IELs, intraepithelial lymphocytes; PBS, phosphatebuffered saline. 1362 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas (figure 5N). In summary, the depletion of Tregs prevented the pancreatitisinduced reduction in Tcells in the lamina propria and of CD8α+ IELs. Flow cytometry data showed a significant increase of RORγt+ Th17cells, as well as a sustained population of CD8α+/TCRγδ+ naturalIELs in DTtreated mice after the onset of AP. Depletion of Tregs attenuates microbial dysbiosis during pancreatitis and prevents bacterial translocation into pancreatic necrosis Tregdepletion resulted in increased Teff numbers thus we analysed whether this increase also affected microbial composition in AP. We performed 16S rRNA gene sequencing based on isolated DNA from duodenal aspirates of DEREGmice. AP induced a pronounced reduction of bacterial taxa within the duodenal aspirates, comparable to what we observed in C57Bl/6 mice. In contrast to PBStreated DEREGcontrols, mice that had received DT before induction of AP retained a more varied microbial composition, indicated by a significantly higher ASV richness and different composition shown by PCoA (figure 6A–C, online supplemental table 2). Overgrowth of facultative bacterial pathogens like Escherichia/Shigella during AP was significantly reduced in the Treg depleted group (S.10C). Quantitative reverse transcriptionquantitative PCR (RTqPCR) confirmed the results from 16S RNA gene sequencing, the abundance of facultative pathogens like Escherichia coli and Enterococcus faecium is significantly increased in mice with AP, but to a minor extend in the DTtreated mice. Beneficial strains like Lachnospiraceae or Lactobacillus remain unchanged (online supplemental figure S10D). We analysed duodenal tissue samples of these mice by RTqPCR analysis and confirmed the loss of Tregs by a DTtreatmentdependent decrease in GFP transcript levels. During AP, we observed significantly higher transcript levels of the genes encoding IL15, IFNγ, TNFα and IL17F in DTtreated mice. Interestingly, transcripts for IL17A, IL22 or IL10 did not differ between PBStreated or DTtreated DEREGmice (figure 6D). In Tregdepleted AP mice, the mRNA amounts of the bacterial defence genes Tnfsf13, Tnfsf13b, Jchain and Defa5 were significantly elevated. On the other hand, the transcript abundances of Defcrrs1 and Reg3g encoding antimicrobial peptides were not significantly increased, while those of the genes encoding the bacteriolytic enzymes Lysozyme 1 and 2, which are normally secreted by Paneth cells, were significantly higher (figure 6E). The expression of genes encoding the physical barrier proteins Occludin and Claudin1 was significantly increased in DTtreated mice whereas changes in other barrier proteins like ECadherin, Claudin5, ZO1 or Mucin2 did not reach a significance level in DTtreated mice (online supplemental figure S10E). Importantly, immunofluorescent labelling of sIgA demonstrated that in Tregdepleted animals the population of sIgA producing cells within the lamina propria remained constant in pancreatitis (figure 6F). Finally, we analysed whether Tregdepletion improves the stability of the intestinal barrier. We performed 16S rRNA gene sequencing in pancreatic tissue DNA of AP mice to detect any bacterial traces. A cutoff of 5000 clean reads was exceeded in 8/25 (32%) of PBStreated mice, indicating infected pancreatic necrosis. In contrast, only 1/14 (7.14%) DTtreated mice reached the cutoff level (online supplemental figure S10F). Statistical analysis by MannWhitney test confirmed a significantly lower bacterial contamination of necrotic tissue in Tregdepleted mice. Facultative pathogens like Escherichia/Shigella, Enterobacteriaceae diversa, Staphylococcus or Enterococcus represented one of the 15th most abundant taxa within the necrotic areas (figure 6G, online supplemental figure S11A,B). A comparison between pancreatic samples and gut samples identified ASVs, which were observed with increased abundance both in duodenal samples (Escherichia/Shigella, Enterococcus and Citrobacter) as well as in necrotic areas of the pancreas (online supplemental figure S11C–E). In a comparative analysis we observed that the very same bacterial taxa that appear or multiply in the duodenal aspirates from APanimals were isolated from murine pancreatic necrosis (figure 6H). Quantitative RTqPCR revealed significantly higher numbers of 16S RNA gene copies (shown as lower dCt value) in the pancreas of PBStreated mice compared with DTtreated mice with AP (figure 6I). Oral administration of an antibiotic cocktail containing 10 mg/mL ampicillin, 10 mg/mL neomycin, 5 mg/mL vancomycin and 10 mg/mL metronidazole, was performed every 12 hours over the experimental period23 and prevented bacterial translocation into the pancreas comparable to the depletion of Tregs by DT (figure 6I). 16S RNA gene sequencing of duodenal samples showed that antibiotic treatment prevented bacterial overgrowth in the duodenum of facultative pathogenic bacteria such as Escherichia/Shigella, Enterococcus, Staphylococcus, Enterobacteriaceae or Klebsiella, whereas beneficial taxa like Lactobacillus remained (online supplemental figure S12A–C). We analysed in DTtreated animals whether the depletion of Tregs or an antibiotic therapy ameliorated the severity of pancreatitis. Enzymatic activities of common severity markers such as serum amylase and serum lipase were significantly lower in DTtreated and antibiotictreated animals (figure 6J). In H&Estained pancreatic sections less organ damage was seen in the Treg depleted mice as well as in antibiotictreated mice. Histological scoring confirmed significantly less damage and reduced pancreatic necrosis in DTtreated and antibiotictreated mice (figure 6K, online supplemental figure S12D). In conclusion, the results suggest that the ameliorating effect of Tregdepletion on disease severity is mainly due to the reduction of bacterial translocation. Depletion of Tregs in combination with antibiotic treatment showed no additional benefit. Bacterial infection of human pancreatic necrosis The bacterial colonisation of human pancreatic necrosis samples was analysed by 16S rRNA gene sequencing. H&Estaining of these samples showed no remaining intact acinar tissue architecture (online supplemental figure S13A,B). For 58 individual patients (online supplemental figure S13C), samples ASVs were inferred from PCRamplified 16S rRNA gene sequences. Samples with >5.000 clean reads (53/58) of representative ASV sequences were subsequently analysed (figure 7A). Firmicutes was found to be the most dominant phylum, representing more than 50% of the total relative abundance, mainly consisting of the genera Enterococcus, Streptococcus and Staphylococcus. The family of Enterobacteriaceae (phylum: Proteobacteria) also showed a high abundance with Escherichia/Shigella as the most frequent taxa (online supplemental figure S13D), the same taxa we observed in murine necrosis. Other phyla like Bacteroidetes, Fusobacteria or Actinobacteria were found with considerably lower abundance. Some patients underwent multiple subsequent necrosectomy procedures and the analysis of followup samples showed that the microbiome in these necrotic samples was rather stable over periods of several weeks (online supplemental figure S14A). Infection of pancreatic necrosis by Enterococcus showed a significant correlation with the length of hospitalisation (figure 7B). None of the other 15 frequent taxa (online supplemental figure S13D) of facultative pathogens or 1363 GlaubitzJ, etal. Gut 2023;72:1355–1369. doi:10.1136/gutjnl-2022-327448 Pancreas the commensal gut bacteria showed this kind of correlation (online supplemental figure S14B). Subgroup analyses regarding AP versus acute episode of CP showed differences, but the high variance in the microbiome composition of pancreatic necrosis and the relative low number of patients did not allow to draw firm conclusions (online supplemental figure S14C). Additional metagenomic sequencing of eight necrosis samples (figure 7C) identified the infecting bacterial species Enterococcus faecalis, Enterococcus faecium, Escherichia coli, Bacteroides spp and Citrobacter spp, these taxa are frequently found in the gut. Figure 6 Depletion of Tregs attenuates microbial dysbiosis during pancreatitis and prevents bacterial translocation into pancreatic necroses. (A)Heatmap illustrates the major microbial taxa in the duodenum of DEREG treated with PBS (n=8) or DT (n=8) after induction of AP. Animals without AP referred as healthy controls (0d). Facultative pathogenic were marked in red, beneficial commensal bacteria were marked in green. (B)Box plots illustrate ShannonDiversity Index and species richness in DTor PBStreated DEREGmice. (C)A principal coordinate analysis illustrates significant differences of the duodenal microbiome between these groups (permutational multivariate analysis of variance; control vs severe: p<0.001, R2=18.2%). (D, E)Box plots show RTqPCR analysis of duodenal transcriptional changes after depletion of Tregs. (F)Immunofluorescent labelling of sIgA producing cells in the duodenal mucosa from control and AP animals showed significantly decreased numbers of sIgA producing cells in AP mice which received PBS. (G)Stakedbar graphs illustrates bacterial taxa which could be identified in all samples with more than 5.000 clean reads from pancreatic necrosis in DEREG APmice (8 PBS treated vs 1 DT treated DEREGmice). (H)The same bacterial taxa that appear or expand in the duodenum of APmice are found in murine pancreatic necrosis samples. The stacked bar graphs show the mean of all mice. (I)We analysed control (con) and ductligated DEREG mice with AP (PBS and DT) by 16S rRNA gene RTqPCR analysis of isolated DNA from pancreatic tissue. Decreased Ct values indicate bacterial infection in the pancreas of mice with AP with a significant greater extent in the PBStreated group compared with Treg depleted mice. Antibiotic treatment (+AB) significantly reduced the copy number of bacterial 16S rRNA gene in the pancreas and therefore prevented bacterial translocation during AP. (J, K)To evaluate how Tregs and microbiome composition affect the disease severity DEREG mice were treated after AP induction with antibiotics (AB), either in presence (+PBS) or absence (+DT) of Tregs. (J)Disease severity was evaluated by analysis of serum amylase and lipase. (K)Pancreatic histology was analysed by H&Estaining of tissue sections, histology score was evaluated by quantification of oedema, necrosis and leucocyte infiltration. The heatmap illustrates the mean result for all groups. Statistically significant differences were tested by unpaired Student’s ttest for independent samples. For more than two groups, statistical significance was determined by ANOVA one way analysis of variance followed by Bonferroni correction for multiple testing, significance levels of p<0.05 are marked by an asterisk. ANOVA, analysis of variance; DT, diphtheria toxin; RTqPCR, reverse transcriptionquantitative PCR; PBS, phosphatebuffered saline.