Bifidobacterium spp. and their metabolite lactate protect against acute pancreatitis via inhibition of pancreatic and systemic inflammatory responses

ABSTRACT Severe acute pancreatitis (SAP) is a critical illness characterized by a severe systemic inflammatory response resulting in persistent multiple organ failure and sepsis. The intestinal microbiome is increasingly appreciated to play a crucial role in modulation of AP disease outcome, but limited information is available about the identity and mechanism of action for specific commensal bacteria involved in AP-associated inflammation. Here we show that Bifidobacteria, particularly B. animalis, can protect against AP by regulating pancreatic and systemic inflammation in germ-free (GF) and oral antibiotic-treated (Abx) mouse models. Colonization by B. animalis and administration of its metabolite lactate protected Abx and GF mice from AP by reducing serum amylase concentration, ameliorating pancreatic lesions and improving survival rate after retrograde injection of sodium taurocholate. B. animalis relieved macrophage-associated local and systemic inflammation of AP in a TLR4/MyD88- and NLRP3/Caspase1-dependent manner through its metabolite lactate. Supporting our findings from the mouse study, clinical AP patients exhibited a decreased fecal abundance of Bifidobacteria that was inversely correlated with the severity of systemic inflammatory responses. These results may shed light on the heterogeneity of clinical outcomes and drive the development of more efficacious therapeutic interventions for AP, and potentially for other inflammatory disorders.


Introduction
Acute pancreatitis (AP) is an inflammatory disorder of the pancreas that is characterized by abnormal activation of trypsinogen within pancreatic acinar cells, resulting in autodigestion of the pancreas and local or even systemic inflammation. 1 Although most AP only results in self-limiting mild disease, a subpopulation of patients progresses to severe systemic inflammatory response syndrome (SIRS) and multiple organ dysfunction syndrome (MODS), which can cause critical illness requiring long-term ICU care. 2 Complications of necrotizing pancreatitis including persistent multiple organ failure (PMOF) and sepsis enormously exacerbate severe acute pancreatitis (SAP), causing mortality at astonishing rates of 30% or even higher. 3 Moreover, despite the global burden of the disease, no effective therapeutic agents are currently available to treat or prevent AP.
Cumulative data strongly support the view that the systemic complications and severity of AP are greatly determined by the intensity of the immune response. 1,2 Injured pancreatic acinar cells release proinflammatory cytokines and chemokines that recruit infiltrating mononuclear-phagocytes and neutrophils to the injury site. 3,4 Subsequently, the cellular contents released from necrotic and injured cells activate monocytes and further propagate pancreatic inflammation, which damages the pancreas and mediates macrophage activation in distant organs, leading to overwhelming SIRS and PMOF in the worst-case scenario. 5,6 Given their pivotal role in activation and amplification of inflammatory responses in AP, the NF-κB, and inflammasome signaling pathways have been key targets for the development of inhibitors looking to cure the disease over the past decade. 4,7,8 Cellular innate immunity sensors including Tolllike receptors (TLRs) such as TLR4 and TLR9, and nucleotide-binding oligomerization domain (NOD)-like receptors (NLRs) such as NOD1 are believed to be major contributors to the progression of AP. [9][10][11] These sensors recognize damageassociated molecular patterns (DAMPs) or pathogen-associated molecular patterns (PAMPs) as the first signals to induce NF-κB complex formation and activation, which upregulates procytokines of pro-IL-1β and pro-IL18 and inflammasome components. 12 A secondary signal, provided by DAMPs released from damaged acinar cells, causes the assembly and activation of inflammasome complexes. 11 Among them, NLR pyrin domaincontaining protein 3 (NLRP3) is the most extensively studied inflammasome sensor protein, and it is a strong determinant of the inflammatory responses and pancreatic tissue damage associated with AP. 13 However, the environmental factors that influence TLR and NLRP3 inflammasome signaling in mononuclear phagocytes and the consequent impact on systemic inflammatory responses in AP remain largely unknown.
Perturbation of the intestinal microbiota plays a role in accelerating the systemic inflammation of AP. 14,15 Accumulating evidence has shown that AP cases can have increased intestinal permeability, and translocation of the intestinal microbiome is thought to be associated with complications such as abdominal infection and sepsis. 16,17 AP induced by caerulein administration in antibiotic (Abx)treated or germ-free (GF) mice exhibited lower pancreatic inflammation and decreased production of systemic proinflammatory cytokines, whereas fecal microbiota transplantation (FMT) worsened the disease, suggesting that the gut commensal bacteria as a whole exacerbates AP. 10,14 On the other hand, probiotics are a promising protective treatment against SAP, considering that specific probiotic strains under the right conditions can alleviate AP. 18,19 Unfortunately, a comprehensive understanding of how the intestinal microbiota is linked to inflammatory signaling is lacking, and specific commensal microorganisms that are capable of limiting inflammatory responses in AP are yet to be identified.
In this study, we discovered that the commensal Bifidobacterium spp. are able to ameliorate both local and systemic inflammation from AP. Using B. animalis as the model organism, we determined that colonization of Abx-treated or GF mice resulted in less severe pancreatic injury and diminished local and systemic inflammatory responses in a TLR4/MyD88-and NLRP3/Caspase1-dependent manner. Exogenous administration of lactate, one of the main metabolites produced by B. animalis, to Abx-treated and GF mice recapitulated the reduction of TLR4-MyD88-and NLRP3 inflammasomemediated inflammation observed with B. animalis monocolonization. Thus, we identified a gutpancreas axis that integrates environmental, metabolic, and immunologic cues to regulate the local and systemic inflammation mediated by NF-κB and NLRP3 inflammasome pathways in AP.

Intensive short-term antibiotics treatment via oral gavage exacerbates caerulein-induced acute pancreatitis
Recent studies demonstrated reduced pancreatic inflammation in Abx-treated or GF mice in a caerulein-induced model of AP, implicating the intestinal microbiota in disease progression. 10,14 Thus, we logically predicted that pancreatic inflammation would be attenuated under conditions of microbiota depletion. We used a more intensive 5-day regimen of highly concentrated Abx by oral gavage, as described previously. 20 MAP or SAP was then induced in these animals by caerulein  (Figure 1a). Surprisingly, mice with orally administered Abx exhibited more severe pancreatic inflammatory lesions and a markedly higher level of serum amylase than PBS-treated controls in both MAP and SAP models. FMT from PBS control mice into Abx-treated mice reversed the amylase and lipase concentration differences in the serum (Figure 1b and Fig. S1A) and prevented pancreatic inflammation (Figure 1c), suggesting that these phenotypes were due to intestinal microbiome alteration. There was no discrepancy in the baseline of serum amylase or pancreatic histology between PBS-and Abx-treated, caerulein-untreated mice (Figure 1c and Fig. S1B). Fecal 16S rDNA copy numbers suggested that FMT reconstituted the enteric bacterial abundance before AP induction (Fig. S1C).
Because our results were contradictory to the study conducted by Tsuji et al., 10 we hypothesized that the distinct phenotype regarding resistance to AP induction could be attributed to the utilization of a different Abx treatment regimen. To confirm this hypothesis, in parallel, we depleted the intestinal microbiota either with a short-term (5-day) oral gavage, or with a longer-term Abx administration (3 weeks) in the drinking water as done in the Tsuji study. Indeed, mice given Abx in the drinking water showed barely detectable pancreatic inflammation and only mild elevation of serum amylase, whereas mice receiving 5-day oral gavage with a more concentrated Abx cocktail developed more severe pancreatic pathology and drastically increased serum amylase concentration compared with PBS control animals (Figure 1d,e).
Based on these observations, we speculated that mice with long-term Abx pretreatment in the drinking water had acquired bacterial communities that conferred resistance to AP, while these critical commensal bacteria had very likely been eliminated in our mouse model after short-term intensive Abx oral gavage. Thus, we analyzed the changes in fecal microbiota composition in PBS control animals before and after induction of MAP or SAP, and also changes in fecal microbiome composition in MAP mice treated with two different Abx regimens. As expected, both Abx treatment regimens substantially reduced the detectable bacterial 16S rDNA readouts to an equally low level (Fig. S1D). However, we observed significantly higher relative abundance of Bifidobacterium, Akkermansia, and Enterococcus spp. in the long-term Abx-treated mice (MAP resistant), but not those treated with high-concentration Abx by oral gavage (MAP susceptible). This finding suggests that Abx administration in the water over 3 weeks somehow resulted in accumulation of distinct commensal bacterial taxa even in the context of global microbiome depletion (Figure 1f,g). Consistent with this, Bifidobacterium was also found to be the most abundant in the PBS control mice, and significantly diminished in the MAP and SAP groups (Figure 1h,i).

Colonization with Bifidobacterium spp. protects the host from MAP and SAP
Since the taxonomic analysis hinted that Bifidobacterium could possibly impact the onset and severity of AP, we performed highthroughput amplicon sequencing of the full-length 16S rDNA from 4 representative fecal samples from MAP mice who had received prolonged Abx in water. Species-level analysis revealed that Bifidobacterium genus in the intestinal microbiome of these mice was dominated by B. longum (~1.2% of total commensal bacteria) (Figure 2a). To further explore the role of Bifidobacterium in AP protection, we chose to use B. pseudocatenulatum (abbreviated as B. pseudo), B. adolescentis (abbreviated as B. adoles), the two commonly identified species in microbiota transplantation (FMT). (b) Serum amylase levels and (c) representative H&E staining of pancreatic sections with pathology scores for SAP or MAP mice at 12 h post-treatment (hpt) with PBS, Abx or FMT (n = 5 each). (d) Serum amylase activity of MAP mice treated with PBS, Abx oral gavage (o.g.) or Abx in drinking water (i.w.) at 12 or 24 hpt (n = 10 each). (e) Pathological sections and pancreatic pathology scores from mice from (d) at 24 hpt. (f) The microbiota composition in fecal samples collected from groups treated with PBS, Abx o.g. or Abx i.w. was analyzed at the genus level by 16S rDNA sequencing (n = 3-5). (g) The relative abundance of Bifidobacterium, Akkermansia and Enterocococcus in fecal samples collected from (f). (h) 16S rDNA sequencing analysis from fecal samples of PBS control, MAP and SAP mice (n = 3-5). (i) Relative abundance of Bifidobacterium from (h). Data are from two independent experiments, and P values were determined by unpaired two-tailed Student's t-test; *: P < .05; **: P < .01; ***: P < .001; ****: P < .0001.   (Figure 2a), and B. animalis, a well-characterized probiotic that has been proven beneficial for a variety of metabolic and infectious diseases. Enterococcus faecalis, a common intestinal commensal bacterium, was included in the protection study in order to determine the role of specific Bifidobacterium species in AP protection.
All three Bifidobacterium isolates were similarly able to repopulate the gut microbiota of Abxtreated mice (Fig. S1E Furthermore, in GF mice with SAP (caerulein plus LPS-induced model), greater serum amylase and more severe pancreatic damage were detected in uncolonized GF mice than in those colonized by B. animalis (Figure 2h,i), suggesting these phenotypes were not due to the residual Abx-resistant bacteria in our Abx treatment model. Additionally, conventionally housed (CNV) B6 mice colonized by B. animalis with SAP (caerulein plus LPS-induced model) displayed not only lower serum amylase but also less pancreatic pathological damage (Figure 2j and Fig. S1I). Collectively, these data indicate that prophylactic administration of B. animalis is capable of ameliorating pancreatitis of different severity, and the suppression of AP by various Bifidobacterium spp. represents a broader mechanism by which this particular bacterial community is able to influence the host inflammatory responses.

B. animalis-associated and exogenous lactate protects host from MAP and SAP
Next, we wanted to test the hypothesis that Bifidobacterium spp. produce specific metabolites that can protect the host from pathogenesis of AP. We performed untargeted metabolomics analysis on serum collected from PBS-or Abx-treated by oral gavage SAP mice with or without B. animalis colonization. The pareto-scaled principal component analysis (PCA) demonstrated significant differences in metabolomics profiles of Abx-treated mice compared with that of PBS control or B. animalis-colonized group (Fig. S2A). KEGG analyses revealed that the differential metabolites in PBS control mice (compared to Abx-treated mice) were mainly enriched in carbohydrate and amino acid metabolism (Fig. S2B). Of particular note, the relative abundance and absolute serum concentration of the carbohydrate metabolite lactate was significantly reduced in Abx mice and restored by B. animalis colonization (Figure 3a  TLR-and inflammasome-mediated inflammation, 8 we prioritized lactate over other upregulated metabolites as our target for further experiments. Consistently, we analyzed the B. animalis cultured supernatants in vitro bacterial culture with LC-MS /MS and identified lactate with an average concentration of 2.4 mM (Figure 3c).
To test whether lactate plays a crucial role in protecting the host from AP induction, groups of Abx-treated mice were given lactate via oral gavage for 3 consecutive days as previously described, 21 and subjected to caerulein-induced MAP (Fig.  S2D). Because butyrate has been determined to mitigate necrotizing pancreatitis, 22 we included it as a positive control. As expected, lactate oral gavage significantly increased the serum lactate concentration in Abx-treated mice by 3 h after the final gavage (Fig. S2E). Lactate exerted a protective effect equivalent to butyrate, with a comparable decrease in serum amylase and similar amelioration of pancreatic histopathology, whereas acetate and propionate pretreatment showed no protective effect (Figure 3d Fig. S2I). Collectively, these findings from several different AP mouse models indicate that lactate, both B. animalis-associated and exogenous, can limit pancreatitis in the context of a perturbed or an intact microbiome.

B. animalis and its metabolite lactate protect against AP via suppression of macrophage-mediated pancreatic and systemic inflammatory responses
Macrophages and neutrophils drawn to the site of pancreatic injury by DAMPs are believed to be the source of systemic inflammatory responses. 23 Thus, we first used flow cytometry to study infiltration of macrophages and neutrophils in the pancreas of mice treated with PBS, Abx, either colonized by B. animalis or pretreated with lactate ( Fig. S3A, B). There was a profound reduction of infiltrating macrophages and neutrophils in the B. animaliscolonized or lactate-pretreated mice compared to those receiving Abx alone ( Figure 4a). IFA revealed significantly more macrophage (CD11b + /F4/80 + ) and neutrophil (CD11b + /Ly6G + ) signal in the pancreatic sections of Abx-treated and E. faecaliscolonized mice compared with the PBS, FMT, B. animalis-colonized or lactate-pretreated Abx mice (   10.5% or 13.5%, respectively) of CD68-positive M1 macrophages in the pancreas of Abx-treated mice during SAP (Figure 4c). Since Colegio et al. demonstrated that lactate produced by aerobic or anaerobic glycolysis of tumor cells can induce expression of vascular endothelial growth factor and M2-like polarization of tumor-associated macrophages, 24 we also measured the CD206-expressing M2 macrophages in the pancreas. B. animalis colonization or lactate pretreatment did not alter the frequency of M2 macrophages compared with mice receiving Abx alone, although AP induction did cause downregulation of M2 macrophages compared to mock mice (Fig. S3D). B. animalis colonization or lactate pretreatment significantly decreased the ratio of M1/M2 macrophages in the pancreas of Abx-treated mice (Figure 4c), indicating that they restrict AP by suppressing M1 polarization of pancreatic macrophages after disease induction. There was also significantly lower expression of pancreatic proinflammatory cytokine genes including Il1b, Il6, Il10 and Tnfa in B. animalis-colonized or lactate-pretreated Abx mice (Figure 4d), suggesting an important role for B. animalis/lactate in dampening pancreatic inflammatory responses.
During SAP, a vicious cycle is driven by activated monocytes in circulation and macrophages in distant organs. 23 B. animalis colonization or lactate administration decreased the frequency of monocytes in peripheral blood mononuclear cells of Abx mice by 1.6-fold or 2-fold, respectively, and reversed the differences in serum cytokine profile (Figure 4e, f).
Consistent with the results of these studies, we discovered that B. animalis colonization or lactate administration significantly impaired the upregulation of bulk and activated M1 macrophages, greatly lowering the M1/M2 ratio and decreasing expression of Il1b, Il6, Il10, and Tnfa in the spleen in Abxtreated mice at 12 hours following SAP induction (Figure 4g,h and Fig. S3E, F). In summary, B. animalis and microbial metabolite lactate both protected mice from AP by suppressing the local and systemic inflammatory responses related to macrophage stimulation.

Prevention of macrophage-mediated inflammation during AP by B. animalis colonization or lactate administration requires TLR4-MyD88
As lactate can restrict AP with or without intestinal microbiota, we hypothesized that lactate may suppress the inflammatory response in macrophages in a cell-intrinsic manner. Thus, we extracted peritoneal macrophages from PBS or Abx-treated mice, pretreated them with lactic acid and stimulated with LPS ex vivo. Indeed, the expression of Il1b, Il6, and Tnfa transcripts was significantly reduced by a similar magnitude (Fig. S4A), indicating that the inflammatory response in macrophages was impaired in a cell-intrinsic manner following lactate administration.
A previous study reported that the NF-κB pathway was activated when bone marrowderived macrophages were co-incubated with acini, and transcriptome analysis indicated that upstream TLR signaling pathways were activated by DAMP signals arising from necrotic acinar cells. 7 When we evaluated the expression of inflammatory response genes associated with the TLR and NF-κB signaling pathways in our SAP mouse model, we discovered that transcripts from Tlr4, Myd88, Traf6, Ikkb, P50 and P65, but not Tlr1 or Tlr9 were significantly   in the pancreas and spleen of B. animalis-colonized or lactate-treated Abx mice compared to mice receiving Abx alone (Fig. S5). Reduced protein expression was confirmed by western blot (Figure 5a), indicating that B. animalis colonization or lactate administration diminished TLR-and NF-κB-associated inflammatory responses locally and systemically during SAP.
To study the functional role of TLR signaling in the B. animalis-or lactate-mediated protection against AP, we established a caerulein-induced SAP model in PBSor Abx-treated Tlr4 −/− or Myd88 −/− mice and their respective wild-type (WT) parents C57BL/10JGpt (referred to as B10/WT) and B6/WT mice. Consistent with our other AP models, serum amylase concentration and pancreatic histopathological changes were markedly augmented in B10/WT and B6/WT mice treated with Abx alone. This pathology was reduced to PBS control levels by B. animalis colonization or lactate pretreatment. However, there were no signs of pathology induced by caerulein in the B10/Tlr4 −/− or B6/Myd88 −/− mice (Figure 5b-d). Correspondingly, pancreatic infiltration by macrophages and neutrophils and proinflammatory cytokine expression were equivalent among B10/Tlr4 −/− or B6/Myd88 −/− mice, regardless of Abx treatment, B. animalis colonization or lactate pretreatment (Fig. S6).
Furthermore, we observed that secretion of IL-1β, IL-6, and TNF-α was greatly diminished in lactic acid-pretreated peritoneal macrophages extracted from B10/WT and B6/WT mice after LPS stimulation, a phenotype not evident in peritoneal macrophages obtained from B10/Tlr4 −/− and B6/Myd88 −/− mice (Figure 5e,f and Fig. S4B). Overall, these data demonstrate the suppression of macrophage-associated pancreatic and systemic inflammatory responses by B. animalis and lactate directly correlate with activation of TLR4 and MyD88-dependent signaling during AP.

Suppression of NLRP3 inflammasome-mediated inflammatory responses is required for B. animalisor lactate-dependent AP protection
Growing evidence suggests that NLRP3 inflammasome activation is crucial for systemic immune responses during SAP. 7,8,11,25 Therefore, we next studied the transcriptional profile of the NLRP3 inflammasome signaling pathway in the pancreas and spleen of Abx-treated mice either colonized with B. animalis or pretreated with lactate. In general, both treatments reversed the greater splenic and pancreatic transcription of Nlrp3, Asc, and Casp1 observed in the mice who received Abx alone (Fig. S7). These findings were perfectly consistent with the study conducted by Hoque et al, which demonstrated that lactate supplementation suppresses NLRP3 inflammasome-mediated inflammation. 11 We further corroborated the activation of NLRP3 inflammasomes by evaluating the protein level of NLRP3, ASC, pro-Caspase1, Caspase1 mature fragment (P10), pro-IL-1β and mature IL-1β in the pancreas and spleen of mock or SAP mice treated with PBS or Abx, and Abx mice colonized by B. animalis or treated with lactate in SAP. Indeed, depletion of the microbiota significantly augmented the NLRP3 inflammasome components and secretion of IL-1β, whereas colonization of B. animalis or lactate gavage greatly reduced expression of these proteins (Figure 6a,b).
To further investigate the role of the NLRP3 inflammasome in B. animalis-or lactateassociated protection from AP, we next repeated the experiments as previously described in Nlrp3 −/− and Caspase1 −/− mice. B. animalis colonization or lactate administration prevented an increase in serum amylase concentration and mitigated pancreatic tissue damage in microbiota-depleted B6/WT AP mice but not in mice lacking Nlrp3 or Caspase1 (Figure 6c,d). The   (Fig. S8). Similar to the experiments performed on Tlr4 −/− and Myd88 −/− animals, lactic acid pretreatment of peritoneal macrophages from B6/WT mice resulted in decreased expression of IL-1β, IL-6, and TNF-α after LPS stimulation, which was not seen in peritoneal macrophages from Nlrp3 −/− or Caspase1 −/− mice (Figure 6e). These data suggest that the suppression of inflammatory responses in macrophages mediated by B. animalis and lactate depends on NLRP3 inflammasome signaling.

Pancreatitis patients exhibit a decreased fecal abundance of Bifidobacterium that was inversely correlated with the severity of their systemic inflammatory responses
To understand the clinical relevance of our findings, we analyzed the dissimilarity of fecal microbiota composition in a total of 63 hospitalized patients (32 with MAP, 9 with moderately severe AP [MSAP] and 22 with SAP) and 30 healthy control volunteers (HC) by 16S rDNA sequencing. Remarkable differences in relative abundance of commensal bacteria were found between HC and AP patients, with a significantly lower abundance of Bifidobacterium observed in the AP patients (Figure 7a,b). Moreover, the abundance of the Bifidobacterium genus was inversely correlated with the severity of AP among the three disease groups (Figure 7b). Consistent with these results, serum lactate levels in MAP and MSAP patients were significantly higher than in SAP patients although there was no difference in serum lactate concentration between MAP and MSAP patients (Figure 7c). Compared with HC, AP patients showed markedly higher serum expression of the inflammatory cytokines IL-1β and IL-6, and their concentrations positively correlated with disease severity in general (Figure 7d,e). There was trend (not statistically significant) for higher IL-1β in MSAP vs MAP patients, and a similar trend for higher IL-6 between SAP and MSAP patients (Figure 7d,e). Furthermore, the serum concentrations of IL-1β and IL-6 inversely correlated with the relative abundance of Bifidobacterium in AP patients (Figure 7f,g). Taken together, these data support a role for Bifidobacterium in the regulation of systemic inflammatory responses during AP.

Discussion
Here, we describe a novel role for the Bifidobacterium genus in limiting immunopathogenic macrophage functions in AP and potentially other immunological disorders. Using B. animalis as a representative species and combining our mouse model and clinical data, we have demonstrated that: 1) Bifidobacterium spp. protect the host from AP by ameliorating pancreatic and systemic inflammation and immunopathological damages; 2) the Bifidobacterium-derived metabolite, lactate, dampens macrophage-associated inflammatory responses both locally and systemically in AP; 3) lactate limits inflammatory responses in macrophages by suppressing NF-κB and NLRP3 inflammasome activation in a TLR4-MyD88-and NLRP3-Caspase1-dependent manner during AP (Fig. S9).
Until now, most studies based on the caeruleininduced AP mouse model have reported that the microbiota promotes AP. 10,14,26 These investigations, without exception, utilized a similar treatment regimen for depletion of intestinal microbes, one based on long-term supplementation of broadspectrum Abx in drinking water. In the current study, we used a short-term (5-day) oral gavage  with a cocktail of high-dose Abx which resulted in an equally efficient microbiota depletion (Fig. S1D). Amazingly, despite their resultant microbiota composition only differing by a few bacterial taxa (Figure 1f), mice on the two Abx regimens had remarkably disparate susceptibility to caeruleininduced AP (Figure 1d,e). The limited distinctions in microbiota composition between the two Abx regimens allowed us to target the beneficial bacterial genus Bifidobacterium for mechanistic experiments. Importantly, the severe disease phenotype in our caerulein induction AP model provided us the frame of reference to evaluate the protective effect exerted by bacterial colonization and related metabolites. The enrichment of Bifidobacterium in the long-term drinking water Abx regiment suggests that the genus is more resistant to the Abx cocktail delivered in drinking water and became one of the dominant microbial communities after the Abx-susceptible microbiota had been depleted.
Our functional studies in both Abx-treated and GF mouse models demonstrated that colonization of Bifidobacterium spp., particularly B. animalis, reduced the serum amylase concentration and ameliorated pancreatic tissue damages in caeruleininduced MAP and SAP models, and improved survival rates in a retrograde injection SAP model ( Figure 2). Although subspecies of B. animalis have been intensively studied and recently shown to have beneficial effects for obesity, metabolic health and microbiota dysbiosis, 27,28 our study is the first to reveal its crucial role in regulation of the pathogenesis of AP.
Metabolic profile screening revealed greatly diminished carbohydrate fermentation and amino acid metabolism in Abx-treated mice, and that B. animalis colonization restored serum lactate to control levels (Figure 3a and Fig. S2B, C). Exogenous lactate administration protected against AP in Abx and GF mice, as evidenced by lower serum amylase concentration, fewer histopathological changes in the pancreas and lower mortality after retrograde injection of sodium taurocholate ( Figure 3). These data are in agreement with the literature that suggests lactate reduces pancreatic injury in AP via suppression of TLR-and inflammasome-mediated inflammation. 8 Future studies with B. animalis mutants lacking the ability to ferment carbohydrates into lactate are needed to definitively confirm lactate's role in Bifidobacteriumdependent anti-inflammatory responses.
The current study showed that B. animalis colonization or lactate administration significantly decreased the infiltration of neutrophils into the pancreas of Abx-treated and GF mice (Figure 4a and Fig. S3C), implicating that lactate may interfere with signals derived from injured pancreatic tissue at early phases of AP. Rapid neutrophil infiltration is usually followed by macrophage recruitment in AP. 29 We did observe the infiltration of macrophages upon microbiota depletion, especially proinflammatory, activated M1 macrophages within the necrotic areas of the pancreas, though B. animalis colonization or lactate administration restored this to control levels (Figure 4a-c). Similar patterns were observed in GF mice following B. animalis colonization or lactate administration (Figure 4b), although AP induction caused less severe pancreatic infiltration by neutrophils and macrophages in GF mice compared with CNV mice (data not shown), supporting that GF mice are truly less sensitive to caerulein-induced AP, as previously described. 14 Furthermore, the upregulation of peripheral monocytes, serum cytokine expression and splenic activated bulk and M1-macrophages in Abx-treated mice were reverted by B. animalis colonization or lactate pretreatment (Figure 4e-h), indicating that SIRS exacerbated by macrophage activation in distal organs during SAP can be efficiently mitigated by lactate. Our findings are consistent with a recent publication that demonstrated how infiltrating macrophages within areas of pancreatic necrosis secrete proinflammatory cytokines to develop SIRS, and at the same time induce a parallel compensatory anti-inflammatory response syndrome (CARS) of the adaptive immune system, which drives M2-macrophage polarization. 7 Our study utilized a panel of knockout mice genetically deficient in key sensors and mediators, identifying a mechanism by which Bifidobacterium and the microbial metabolite lactate suppress caerulein-induced pancreatic and systemic inflammatory responses through a pathway that requires TLR4 and MyD88 signaling as well as an NLRP3and Caspase1-dependent inflammasome pathway in macrophages (Figure 5e,f and Figure 6e). Our model perfectly matches the study conducted by Hoque et al., which determined a crucial role for lactate in suppressing TLR4-and NLRP3 inflammasome-mediated inflammation in AP via G iprotein-coupled receptor 81 (GPR81). 8 Thus, additional studies are needed to evaluate whether or not GPR81 also mediates the protective antiinflammatory effects of Bifidobacterium and lactate under AP in vivo.
Finally, our clinical data was consistent with the mouse models, showing an inverse relationship between the relative abundance of gut Bifidobacterium and AP severity (Figure 1h,i and Figure 7b), suggesting a potential role for Bifidobacterium as a prognostic microbial marker for AP outcome. Furthermore, increased serum IL-1β and IL-6 were found to be inversely related to abundance of intestinal Bifidobacterium, matching the findings reported by Tan et al., and suggesting that during the development of AP, dysbiosis of the gastrointestinal tract may contribute to an increase in severity of SIRS. 15 One limitation of the clinical study is that the large number of clinical variables in patients with AP might affect their microbiome composition, which represents a major hurdle to the interpretation of disease causality for human microbiota alteration. Moreover, caution must be taken when substituting probiotics such as Bifidobacterium in AP, as Pushalkar et al., suggested that microbiota reconstitution with B. pseudologum accelerated oncogenesis in Abx-treated KPC mice, which express mutant intra-pancreatic Kras and p53. 30 Further, considering the cell surface lactic acid receptor GPR81 is crucial to the survival of cancer cells, it is necessary to assess the appropriate, effective dose of lactate to be used, which is usually administrated in the form of lactated Ringer's solution. 23 In conclusion, our detailed functional study demonstrates how macrophages and inflammatory signaling pathways are modulated by the commensal Bifidobacterium to limit inflammatory responses locally and systemically, reducing subsequent immunopathology. The finding that B. animalis and its associated metabolite lactate can ameliorate AP may shed light on the heterogeneity of clinical outcomes and drive the development of more efficacious therapeutic interventions for pancreatitis, and potentially other inflammatory disorders.

Patients
A total of 63 AP patients ranging from 18 to 90 years old presenting within 7 days of onset of symptoms were enrolled from the Department of Intensive Care Unit, Sir Run Run Shaw Hospital affiliated with Medical College of Zhejiang University. All patients met the criteria for AP according to the revised Atlanta classification, and they were stratified into three groups by clinical severity: MAP, MSAP, and SAP. Exclusion criteria included those patients with chronic pancreatitis or metabolic, liver, and immunosuppressive diseases as well as cancer. Age, gender and body mass index (BMI)matched 30 HC individuals were composed of volunteers after a routine physical examination. The HC group had the following inclusion criteria: free of chronic metabolic, cardiovascular or gastrointestinal diseases; not pregnant; no medical treatment influencing intestinal function. Both patients and healthy subjects who had received antibiotics or probiotics in the 4 weeks prior to sample collection were excluded from the study. Written informed consent was obtained from each participant following protocols approved by the institutional review boards of Sir Run Run Shaw Hospital affiliated to Medical College of Zhejiang University. Demographics and clinical variables were collected during clinic visits.
For human studies, written informed consent was obtained from all subjects before the study protocol. All human experiments were conducted in accordance with a protocol (#20200224-31) approved by the Medical Ethics Committee of Sir Run Run Shaw Hospital. Table S1 describes the baseline demographic and clinical characteristics of AP patients.

Isolation and treatment of splenocytes, pancreatic cells, peritoneal macrophages and PBMCs in mice
Spleens were dissected, and single-cell suspensions of splenocytes were generated by grinding through a 70-μm strainer. Erythrocytes were lysed with red blood cell lysis buffer (RCLB; HyClone), and remaining cells were resuspended in PBS supplemented with 2% fetal bovine setum and 1 mM EDTA.
Samples of pancreas were dissected, minced and washed twice with 10 mL Hank's Balanced Salt Solution containing 10% FBS. Tissue was resuspended in 10 mL HBSS containing 10% FCS and 2 mg/mL collagenase (type IV; Sigma Chemical Co. St. Louis, MO), shaken continuously in a 37°C water bath (100 cycles/min) and vortexed for 20 seconds. Cellular suspensions were centrifuged and resuspended in 0.5 mL pancreas ACK Lysis Buffer to remove red blood cells. Resultant cells were immediately centrifuged, washed 3 times with HBSS, and passed through a 70-µm filter to produce a single-cell suspension of pancreatic cells.
Peritoneal macrophages were isolated as previously described. 11 PBMCs were isolated by density gradient centrifugation using a PBMC isolation kit (TBD, Tianjin, China).

Abx treatment, FMT, and bacterial colonization
Mice were given an Abx cocktail composed of ampicillin 33.2 mg, neomycin 33.2 mg, metronidazole 33.2 mg, and vancomycin 16.7 mg daily for 5 days via oral gavage. After the 5th day of oral gavage, Abx were added to the drinking water at a concentration of 1 g/L each for ampicillin, neomycin, and metronidazole and 500 mg/L for vancomycin. Fecal samples collected from microbiota-depleted mice at 5 days post-treatment were homogenized, plated on BHI agar with 10% sheep blood, and cultured under anaerobic conditions at 37°C for 2 days followed by incubation under aerobic conditions at 37°C for 1 day to confirm efficient microbial depletion. Animals were maintained on Abx-or PBScontaining water for the duration of the experiment.
For FMT experiments, feces were collected under 20% O 2 condition and were used immediately after fresh harvest. 200 mg of pooled fecal pellets were homogenized with sterile silica beads in 1.5 mL PBS at 45 Hz for 1 min and filtered with 70-μm strainers. Mice were orally administered Abx as described above, except Abx were discontinued on day 6, and mice were individually subjected to gavage with 150 μL filtered stool homogenates.
For bacterial colonization experiments, mice were subjected to gavage with 10 10 CFU of E. faecalis, B. pseudocatenulatum, B. animalis or B. adolescentis in 150 μL PBS after 5 consecutive days of oral Abx administration. Two days after FMT or bacterial colonization (3 days after final oral Abx gavage), stool samples were collected to determine the efficiency of colonization, and colonized mice were processed for pancreatitis modeling.

qRT-PCR
RNA was isolated from tissues or cells using TRIzol reagent (Invitrogen) per the manufacturer's instructions. Gene expression levels were determined via qRT-PCR (primers used for the assay are listed in Table S2) and normalized to GAPDH expression. Results are presented as fold change of gene expression in AP mice relative to that in mock animals (2 −ΔΔCT ).

Histopathology processing, scoring, and immunofluorescence assays
The pancreas was treated with 4% paraformaldehyde (PFA) and embedded in paraffin. 5-µm sections were cut and stained with hematoxylin and eosin (H&E) and semiquantitatively scored using Schmidt's criteria by two board-certified veterinary pathologists in a double-blinded manner. 32 The final pathology score expressed is the average of these two values.

Flow cytometry analysis
Splenocytes, pancreatic cells, and PBMCs were harvested for the analysis of different surface antigens in cell subsets following blockade of Fcγ receptors with anti-CD16/32 (eBioscience). Fluorescent antibodies used included those specific to CD11b, CD68, CD206, Ly6G and F4/80 (BioLegend).

Western blots
Tissues were homogenized in ice-cold buffer containing 20 mM HEPES (pH 7.0), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, Proteinase Inhibitor Cocktail (Roche Applied Sciences), and Halt Phosphatase Inhibitor Cocktail (Thermo Scientific) for 30 s. The samples were sonicated through the Cell Ultrasound Cruizer (Thermo Scientific) for 10 s and centrifuged at 15,000 rpm for 10 min, and total proteins were recovered in the supernatants. The supernatants were collected for BCA protein kit (Beyotime Institute of Biotechnology, Nanjing, China). Samples (40 μg protein per lane) were loaded on 10% SDS-PAGE gels. After electrophoresis, the proteins were transferred to PVDF membranes (Millipore, Boston, MA, USA). Then, the membranes were blocked with 5% BSA and incubated at 4°C overnight with the appropriate primary antibodies. The membranes were washed four times for 10 min each in TBS-Tween-20, followed by incubation with secondary Ab (Fude Institute of Biotechnology, Hangzhou, China) for 60 min at room temperature. Then the membranes were detected by exposure to high ECL assay. Size estimates for proteins were obtained using m.w. standards from Bio-Rad. All experiments were performed in triplicate and carried out three times independently.

Sodium acetate, propionate, and butyrate supplementation
Briefly, 200 mM each of sodium acetate, sodium propionate, and sodium butyrate (Sigma-Aldrich) were dissolved in drinking water of mice and replaced every 3 days. Mice received drinking water supplemented with acetate, propionate, and butyrate for 4 weeks prior to AP modeling.

Lactate treatment
For in vivo experiment, mice were perfused with 0.24 g/kg body weight sodium lactate by oral gavage three days before AP modeling. For in vitro studies, peritoneal macrophages were pretreated with 15 mM lactic acid for 15 minutes before LPS (200 ng/mL) stimulation and tested by qRT-PCR.

Serum amylase, lipase, and lactate measurement
Serum samples collected from animal experiments were assessed for lactate concentration by a commercial kit (ab65330; Abcam). Serum amylase concentration was determined by the Alphaamylase Determination Kit (BIOSINO, Beijing, China). Serum lipase concentration was determined by the Lipase Activity Assay Kit (BC2340; Solarbio).

Induction of different AP mouse models
For the MAP model, groups of mice were given 7 hourly intraperitoneal (i.p.) injections with 25 μg/ kg of caerulein (Sigma). The caerulein plus LPSinduced SAP model was established by 10 hourly i. p. injections of caerulein (25 μg/kg) plus a single i.p. injection of LPS (7.5 mg/kg). A surgical SAP model was established by retrograde injection of 125 μg/kg sodium taurocholate (5%) into the pancreatic-bile duct, and survival kinetics were observed for 72 hours as described previously. 23

Germ-free experiments
Male C57BL/6 GF mice at 6 to 7-weeks-of-age were provided by the Experimental Animal Center of Huazhong Agricultural University, Wuhan, China. GF mice were fed in a flexible film plastic isolator. All conditions were kept sterile, and the feces and skin of GF mice were tested to verify compliance with the Chinese laboratory animal microbiological standards and monitoring (GB 14922.22-2011).

DNA extraction, 16S rDNA sequencing, deep sequencing, and data analyses
Genomic DNA in fecal samples was extracted using the ALFA-SEQ Advanced Stool DNA Kit (Magen). The quality and quantity of DNA were measured using a NanoDrop One instrument (Thermo Fisher Scientific). Subsequently, bar-coded PCR primers targeting the V3-V4 region of bacterial 16S rDNA (forward primer, ACTCCTACGGGAGGCAGCA; reverse primer, GGACTACHVGGGTWTCTAAT) were used to generate amplicons, and multiplex sequencing of amplicons with sample-specific e2127456-20 barcodes was performed using an Illumina Novaseq 6000 platform (paired-end 2 × 250-nucleotide reads; Guangdong Magigene Biotechnology Co., Ltd., Guangzhou, China).
Raw sequence data were filtered using fastp (version 0.14.1, https://github.com/OpenGene/fastp) with the parameters -W4 -M20 and further processed with cutadapt (https://github.com/marcelm/ cutadapt/) to remove the primer sequences and generate the paired-end clean data. Subsequently, the usearch-fastq_mergepairs tool (version 10; http://www.drive5.com/usearch/) was utilized to merge the raw tags, which were later trimmed by fastp to get clean tags. The operational taxonomic units (OTUs) were then clustered using UPARSE software, with a cutoff value of 97% similarity. Each representative sequence of these OTUs was assigned using the SILVA database to annotate taxonomic information. The richness of certain commensal bacteria taxa was calculated using the usearch-alpha_div (version 10; http://www.drive5. com/usearch/) according to the OTU abundance. Based on the relative abundance of species at each classification level in otu_table, R software was used to draw the histogram, heat map, and ternary-phase diagram.

Metabolomics profiling
LC-MS/MS analysis was conducted using an ultrahigh performance liquid chromatography (1290 Infinity LC, Agilent Technologies) and quadrupole time-of-flight (AB SCIEX QQQ 5500). 5500 QTRAP (AB SCIEX) was performed in positive and negative switch mode. Peak chromatographic area and retention time were analyzed with Multiquant software.

Statistical analysis
Statistical analyses were performed with Prism GraphPad software v 8.0. Error bars represent standard errors of the means in all figures and P values (the cutoff for statistical significance was P ≤ .05) were determined by unpaired, two-tailed Student's t-test. A log-rank test was used for survival curves. One-way ANOVA was used for correlation analysis.

Study approval
All animal experiments were conducted in accordance with a protocol (#117113) approved by the Institutional Animal Care and Use Committee of Zhejiang University. For human studies, written informed consents were obtained from all subjects before the study protocol. All human experiments were conducted in accordance with a protocol (#20200224-31) approved by the Medical Ethics Committee of Sir Run Run Shaw Hospital, affiliated with Medical College of Zhejiang University.