Microbial exposure during early life regulates development of bile duct inflammation

Abstract Objectives The early life microbiome has been linked to inflammatory diseases in adulthood and a role for the microbiome in bile duct inflammation is supported by both human and murine studies. We utilized the NOD.c3c4 mouse model that develops a spontaneous immune-driven biliary disease with a known contribution of the microbiome to evaluate the temporal effects of the early life microbiome. Materials and methods Germ-free (GF) NOD.c3c4 mice were conventionalized into a specific pathogen free environment at birth (conventionally raised, CONV-R) or at weaning (germ-free raised, GF-R) and compared with age and gender-matched GF and conventional (CONV) NOD.c3c4 mice. At 9 weeks of age, liver pathology was assessed by conventional histology and flow cytometry immunophenotyping. Results Neonatal exposure to microbes (CONV-R) increased biliary inflammation to similar levels as regular conventional NOD.c3c4 mice, while delayed exposure to microbes (GF-R) restrained the biliary inflammation. Neutrophil infiltration was increased in all conventionalized mice compared to GF. An immunophenotype in the liver similar to CONV was restored in both CONV-R and GF-R compared to GF mice displaying a proportional increase of B cells and reduction of T cells in the liver. Conclusions Microbial exposure during early life has a temporal impact on biliary tract inflammation in the NOD.c3c4 mouse model suggesting that age-sensitive interaction with commensal microbes have long-lasting effects on biliary immunity that can be of importance for human cholangiopathies. Graphical Abstract


Introduction
Perturbations in the early life microbiome can shape inappropriate immune responses against self and commensal microbes and increase the risk of allergies, autoimmunity and metabolic disease in adulthood [1,2], and thus potentially also affect development of bile duct inflammation.Gut microbes can regulate both local and systemic immunity [3] and modulate innate and adaptive immune responses.
Previously it has been demonstrated that dietary-derived metabolites and microbial antigens can induce immune cells such as regulatory T cells [4,5], and that effects of early life microbial exposure mediated through the gut-liver axis affects unconventional T cells that have both protective and aggravating roles in immune-mediated illnesses and cancer [6][7][8].
The relevance of the early-life microbiome is evident in experimental disease where antibiotic-induced gut dysbiosis during early life leads to increased susceptibility to both spontaneous and chemical-induced colitis driven by upregulation of proinflammatory immune responses in genetic susceptible Il-10 −/− and C57Bl/6 mice [9,10].The NOD.c3c4 mouse model mimicking human autoimmune biliary disease was developed from the non-obese diabetic (NOD) mouse susceptible to autoimmune disease, by introgression of diabetes protective loci on chromosome 3 and 4 from the C57Bl/6 and C57Bl/10 strains [11].Although totally protected from diabetes, NOD.c3c4 mice develop spontaneous bile duct inflammation characterized by peribiliary lymphocyte infiltrates, destruction of smaller hepatic ducts shaping biliary polycystic changes, and common bile duct (CBD) dilatation [12].Additionally, anti-mitochondrial antibodies (anti-PDC-E2) develop at an early age before detectable biliary pathology appears [13].The bile duct inflammation in the NOD.c3c4 mouse model is driven by T cells [13] and transferable by bone-marrow transplantation [14].We have previously shown that the presence of a microbiome in NOD.c3c4 mice is correlated with degree of inflammation and that bacterial depletion by antibiotics ameliorate the biliary inflammatory phenotype [15].These observations are contrasted by the Abcb4 −/− model of cholangiopathy where depletion of gut microbiota increase bile acid synthesis through suppression of the ileal Farnesoid X receptor (FXR), a nuclear bile acid receptor, causing hepatic accumulation of primary bile acids and leading to a more severe biliary phenotype [16].These two studies point to different mechanisms in which the microbiome leads to bile duct inflammation.The importance of the gut microbiome in cholangiopathies is also evident in human studies with microbiota associations in primary sclerosing cholangitis (PSC) [17,18] and primary biliary cholangitis (PBC) [19], two progressive diseases characterized by inflammation in the large [20] and small bile ducts [21], respectively.Moreover, the majority of PSC patients have concomitant inflammatory bowel disease [22] and autoreactive lymphocytes activated in the bowel are suggested as potential key factors in promoting cholestatic liver disease through the gut-liver axis [20].
Whether the root to dysbiosis in hepatobiliary disease is simply the disease state, or a combination of environmental factors potentially starting in the early life combined with host factors as immunity, genetics or bowel transit time, remains unsolved.The early encounters between the host and gut microbes in mucosal tissues tuning the immune responses, is referred to as 'the neonatal window of opportunity' entailing long-term effects on host immunity and health [9,23,24].Gut microbial abundance and composition starts evolving during the first days of life and culminates in a substantial microbial shift characterized by increasing species richness and diversity at weaning [25].To address whether the timing of microbial exposure tune biliary immunity and impacts development of bile duct disease in adulthood, we studied the dynamics of microbiome introduction and development of bile duct disease and accompanying changes in various immune compartments by re-conventionalizing germ-free (GF) NOD.c3c4 mice at different time points representing pivotal immunologic developmental windows.

Models and housing conditions
NOD.c3c4 mice (Jackson Laboratory, Bar Harbor, ME, uSA; Strain: JAX:010971) were rederived as axenic mice at the Core Facility for Germfree Research at Karolinska Institutet, Stockholm, Sweden.Male and female GF NOD.c3c4 mice and conventionalized (CONV) NOD.c3c4 mice were used in the experiments.All mice were housed in the animal facility at Oslo university Hospital, Rikshospitalet, Oslo, Norway where GF mice were maintained in open cages (Eurostandard type II, 11bbB, Tecniplast, Buguggiate, Italy) in flexible-film isolators and regular conventionalized mice were housed in a specific pathogen free (SPF) unit.The group with regular conventionalized mice were maintained for at least two generations before used in experiments.Conventional mice were maintained in individually ventilated cages (IVC, GM500, Tecniplast) or open cages (Eurostandard type III, 1290D, Tecniplast) while animals that were born germ-free and later conventionalized (GF-R or CONV-R) were housed in IVC cages after conventionalization.
All animals were provided ad libitum access to water and autoclaved rodent diet (Labdiet 5021, IPS Products Supplies, Alfreton, Great Britain) and housed with a 12-h light/dark cycle under temperature and humidity-controlled conditions.GF status was verified monthly by aerobic and anaerobic microbial cultivation of collected fecal pellets and mold trap samples from the GF isolator, and yearly serology and PCR testing according to FELASA 2014 recommendations [26].In addition, fecal pellets were cultured from the experimental IVC cages exported from the isolators throughout the experiment to ensure sterility and GF status by bacterial cultivation.
All animal experiments were approved by The Norwegian National Animal Research Authority (project license no.11245/25770) and carried out in full accordance with the Eu Directive on protection of animals used for scientific purposes (2010/63/Eu) and the Norwegian Animal research legislation.The study is reported in accordance with the ARRIVE guidelines.

Study design
Four experimental groups of NOD.c3c4 mice were reared under discrete microbiologic conditions and introduced to microbes at separate time-points (Figure 1A).CONV and GF mice recapitulated the effects of a present or absent microbiota.Time-dependent microbial effects during early-life were characterized in CONV-R mice introduced to microbes within the first day after birth, or in GF-R transitioning from GF to SPF conditions at weaning at 4 weeks of age.All mice were housed for 9 weeks.

Cross-fostering (CONV-R)
To generate mice born from a GF mother and raised in a conventional environment, pregnant GF mice were moved from flexible sterile isolators approximately 14 days after estimated fertilization in IVC cages with autoclaved water, chow and bedding into our SPF unit.All cages remained closed until delivery.After removal of age-matched SPF litters from SPF NOD.c3c4 dams, the newborn GF pups were cross-fostered until weaning at 4 weeks of age and handled as regular SPF mice.

Co-housing and fecal microbiota transplantation (GF-R)
To generate mice that were raised GF before conventionalization, GF mice were exposed to SPF microbiota at four weeks of age by a single fecal microbiota transfer (FMT) followed by co-housing for five weeks.For the FMT procedure, fresh murine cecum contents from 10-15 weeks old donor SPF NOD.c3c4 mice were homogenized in cold phosphate buffered saline (PBS) diluted 1:10, gently vortexed and filtered through a 100 µm cell strainer and administered by oral gavage.The ex-GF pups were then co-housed with age and gender matched SPF pups from weaning until organ sampling.Conventional littermates were ear-marked.

Tissue collection
All animals were euthanized using CO 2 gas at 9 weeks of age and sacrificed at the same time of the day (between 8-12 am.).Tissues were sampled according to recommended guidelines for PSC animal models [27].Blood samples were collected by cardiac puncture and spun at 13500 rcf 4 °C for 10 min following 60 min of clotting at room temperature.Serum samples were stored undiluted at −80 °C and pre-diluted 1:4 with MilliQ water before analysis.Liver tissues were fixed in 4% formalin, paraffin-embedded and sectioned in 4-5 µm sections.Liver sections were stained with hematoxylin and eosin (H&E).First, sections were dried, and paraffin removed before hydration, hematoxylin staining for 6 min and eosin staining for 1 min.All H&E stainings were conducted at the Dept. of Pathology, Institute of Clinical Medicine, university of Oslo.In addition, liver, mesenteric lymph nodes, spleen and thymus were sampled for immune characterization using flow cytometry and processed and stained as described below.

Serum biochemistry
Colorimetric assays were used for the quantitative determination of serum ALT (Atellica CH, Siemens Healthineers, Erlangen, Germany), AST (Atellica CH) and ALP (Atellica CH).All samples were analyzed on a Siemens AtellicaCH Analyzer (Siemens Healthcare Diagnostics Inc., Tarrytown, Ny, uSA) at the Norwegian Veterinary Institute, Ås, Norway.ALT values below the detection range were arbitrary set at the lower detection limit (28 u/L).

Immunohistochemistry
For immunohistochemistry (IHC), liver sections were formalin-fixed and paraffin embedded and stained for CD3 and Ly6G.In brief, heat-induced epitope retrieval was performed in appropriate antigen retrieval buffer (CD3: Envision FLEX Target Retrieval Solution, High pH buffer, DAKO, Glostrup, Denmark.Ly6G: Envision FLEX Target Retrieval Solution, Low pH buffer.DAKO) in a pressure-cooker for 15 min.Endogenous peroxidase activity was blocked by 3% hydrogen peroxide for 20 min.Blocking buffer (CD3e and Ly6G: PBS with 0.5% bovine serum albumin and 0.1% Tween-20) and 2.5% normal goat serum (Ly6G) were used as blocking agents to prevent nonspecific binding to reactive sample sites.Sections were stained for T cells and neutrophils and incubated overnight with primary antibodies for CD3 (Abcam, Cambridge, uK) or Ly6G (BioXCell, Lebanon, NH, uSA) diluted in PBS with 0.5% bovine serum albumin and 0.1% Tween-20.The following day, the slides were incubated for 30 min with the secondary antibody (CD3e: Horse, anti-rabbit IgG HRP.Ly6G: Goat, anti-rat IgG HRP.Vector laboratories, Burlingame, CA, uSA).DAB staining (ImmPACT DAB Peroxidase Substrate kit, Vector laboratories) was performed for chromogenic detection.Hematoxylin (Vector laboratories) was used as nuclear counterstain and finally, all sections were mounted with Eukitt (Sigma-Aldrich/Merck, Darmstadt, Germany).Detailed information about the antibodies is provided in supplementary Table S1.

Image data acquisition and image analysis
High-resolution whole slide images (WSIs) of histological sections were acquired using an automated slide scanner system (Axio Scan Z1, Carl Zeiss Microscopy, Munich, Germany).
All slides were scanned at 20x magnification with a resolution of 0.22 µm/pixel.Quantitation of CD3 + T cells and Ly6G + neutrophils was performed using QuPath machine-learning algorithm (open-source software, version 0.2.3) [28] on the WSIs.First, we trained a segmentation algorithm to detect our regions of interest (ROIs) from background signal and staining artifacts, and then applied the model to the total set of the WSIs.At this point, every WSI was inspected to manually verify the ROIs selected by the algorithm.All cells in the ROIs were then detected, and a second classification algorithm was subsequently trained to detect DAB-stained nucleated cells (positively detected cells).Positive cells were identified using a random forest model and an artificial neural network (ANN) model for classification of Ly6G + and CD3 + cells, respectively, and measured as cells per mm 2 normalized to the measured area.The choice of model was based on feature complexity within the two datasets, whereas the DAB-stained CD3 + cells had a less distinguishable staining phenotype.Machine learning model metrics, performance and validation is provided in the Supplementary methods and schematics for the ANN algorithm is presented in Figure S1.

Cell preparation and staining for flow cytometry
Liver, mesenteric lymph nodes (MLNs), spleen and thymus were subjected to lymphocyte isolation and flow cytometry staining.The MLN located in the colonic mesentery close to the cecum and draining the proximal colon were collected.In brief, liver tissue was mashed, homogenized, and filtered through a 70 µm cell strainer followed by Percoll density gradient (70%/40%) and centrifugation at 700 g.MLN, spleen and thymus were mashed through a 40 µm filter, rinsed with PBS followed by centrifugation at 350 g.Cell suspensions of splenocytes were additionally treated with a lysing reagent (RBC Lysis buffer 10x; BioLegend, San Diego, CA, uSA) to remove erythrocytes.Single cell suspensions enriched with lymphocytes were harvested for subsequent immunofluorescent antibody staining.Thymus and spleen samples were resuspended in flow buffer (PBS with 5% fetal bovine serum) while liver and MLNs were resuspended in PBS.
Lymphocyte samples were incubated with the Fc blocking reagent purified anti-mouse CD16/32 (BioLegend) for 30 min to prevent unspecific antibody binding and then stained with fluorophore conjugated monoclonal antibodies for 30 min.Stained cells were fixed and permeabilized with the FoxP3/ Transcription Factor Staining Buffer Set (eBioscience, San Diego, CA, uSA) followed by a second Fc block incubation with purified anti-mouse CD16/32 overnight.The cell suspensions were then incubated with antibodies for intracellular markers for 45 min.We used a set of lineage markers to identify B cells (CD19), T cell subsets (CD3, CD4, CD8, TCRgd), and innate lymphoid cells as tissue-resident NK cells (CD49a) and conventional NK cells (CD49b).Detailed information about the antibodies used can be found in Table S2.

Flow cytometry sample acquisition and data analysis
All flow cytometry data were acquired on a BD FACSymphony A5 equipped with five lasers (BD Biosciences, Franklin Lakes, NJ, uSA) using BD FACS DiVa software (BD Biosciences).Standard flow cytometry data analysis was performed in FlowJo version 10.8.1 (FlowJo LLC, BD Biosciences, Ashland, OR, uSA).Data was manually curated with removal of events indicative of doublets, dead cells and debris.The applied compensation matrix was automatically generated using the AutoSpill and AutoSpread algorithm [29] in FlowJo with minor manual adjustments.Immune cell populations were manually gated.
T-distributed stochastic neighbor embedding (tSNE) analysis algorithm built-in in FlowJo was used for dimensionality reduction and data visualization.All CD3 positive lymphocytes from each .fcsfile were scaled and concatenated before further downstream analysis using DownSampleV3 plugin in FlowJo and the FlowJo built-in concatenation function.For each sample, the number of events contributing to the final concatenated file was 60.000 events per sample.Next, the FlowJo tSNE function was used by selecting the Opt-SNE algorithm [30] for high-quality visualization and the K-Nearest Neighbor (KNN) algorithm Approximate Random Projection Forest (ANNOy) for similarity measurements.

Statistics
Statistical analyses were performed using GraphPad Prism version 9.3.1 (GraphPad Software, San Diego CA, uSA).Distributions of the data were visually evaluated for normality using Q-Q plots.Normally distributed data were compared using one-way ANOVA followed by Bonferroni correction, while data without a normal distribution were compared using the non-parametric Kruskal-Wallis test followed by Dunn's test for multiple comparisons.Results are reported as median or mean ± standard error of mean (SEM).Corrected P-values < 0.05 were considered statistically significant.

Re-conventionalization of germ-free mice at birth or weaning
To evaluate the temporal impact of microbial introduction on development of bile duct disease in the NOD.c3c4 mouse model, we introduced bacteria at different time-points during early life (Figure 1A).The cecum size is naturally enlarged in GF mice independent of strain or disease model [31] and this feature was also evident in GF NOD.c3c4 mice (Figure 1B).In the two groups of mice that were born GF and later exposed to bacteria, the cecum weights were successfully reversed upon conventionalization independently of whether it was performed at birth (conventionally raised, CONV-R) or weaning (germ-free raised, GF-R) (Figure 1B).As expected, the body weight in GF mice were significantly increased in accordance with the cecum enlargement (Figure 1C).Similar differences were observed in female mice (Figure S2A-B).

Macroscopic liver disease in NOD.c3c4 mice is affected by conventionalization
The NOD.c3c4 phenotype is characterized by an enlarged liver, development of cystic liver lesions located in the peripheral liver lobules and portal inflammation [13].The relative liver weight was significantly reduced under GF conditions while the early conventionalized group attained relative liver weights resembling conventional (CONV) NOD.c3c4 mice (Figure 2A-B).Similar differences were observed in female mice (Figure S2C).No differences in common bile duct dilatation, a parameter known to have considerable variability [32], was observed (Figure S3A-B).

Portal inflammation increased after early-life conventionalization
Next, we examined the presence of liver disease and the degree of biliary inflammation on the cellular level by histological evaluation (Figure 2C-E).The degree of portal inflammation was considerably lower in the GF group and this effect persisted in the GF-R mice after a delayed exposure to a microbial environment.Portal inflammation scores attained similar levels in CONV and CONV-R.Bile duct dilation scores and bile infarct scores were similar between the four groups (Figure 2E).In female mice, a similar trend for portal inflammation was seen, while the other parameters were similar between the groups (Figure S4).Liver transaminases and markers of cholestasis at 9 weeks of age were all within normal range [33] among males and females (Figure S5).These observations are in line with the previous reports where changes in serum transaminases in the NOD.c3c4 mice are normally observed after 9 weeks of age [32].

Immunohistochemical characterization of the immune response
Neutrophil infiltration in liver was increased in all conventionalized groups as evidenced by immunohistochemical staining (Figure 3A-B).Neonatal microbial exposure (CONV and CONV-R) led to elevated levels of infiltrating neutrophils.Similarly, mice encountering microbes for the first time at weaning also had increased neutrophil levels compared to GF mice.In contrast, infiltration of liver T cells during early adulthood was not dependent on microbial status during early life (Figure 3C-D) indicating that this infiltration might be an event taking place at a later age [15].Similar differences were observed in female mice (Figure S6).

Immunophenotyping of liver lymphocytes
After having demonstrated a temporal microbial dependency during early life for development of liver disease in the NOD.c3c4 mouse model, we hypothesized that these changes were governed by alterations in the immune compartment.To address this, we examined the lymphocyte composition in liver, gut-draining mesenteric lymph nodes, spleen and thymus by flow cytometry (gating strategies in Figure S7 and Table S3).Four to five weeks after exposure to microbes, the immune cell phenotype in GF-R mice converged towards a phenotype for the major immune populations of liver lymphocytes as B cells and T cells resembling CONV and CONV-R mice (Figure 4A).Such changes were not seen for liver group 1 innate lymphoid cells (Figure 4A).These changes were not observed in female mice (data not shown) and could be due to known gender associated differences in progression of the inflammatory process in NOD.c3c4 mice [13].Next, we investigated the composition of the overall T cell subsets in NOD.c3c4 livers (Figure 4B-C) illustrating homogenous hepatic T cell compositions across the groups and independent of microbial status or timing of microbial introduction.The immunophenotype of splenocytes and thymocytes of early-conventionalized mice had similarities with CONV while late conventionalized mice resembled a notably GF phenotype (Figure 5).Comparisons of the immune cell populations in liver, MLN, spleen and thymus between CONV and GF NOD.c3c4 mice are shown in Table S4.

Discussion
In this study we demonstrate a clear temporal dependency of the microbiome for the biliary phenotype in the NOD.c3c4 mouse model.We further demonstrate that introduction of a microbiome at specific time-points also influence immune composition in liver and peripheral lymphoid tissues, in a way that could impact the observed time-dependent development of biliary disease.
The severe biliary inflammation in the NOD.c3c4 mouse develops spontaneously but is ameliorated in absence of microbes.This contrasts the reported protective effects of a balanced, homeostatic and diverse microbiome in the digestive tract, airways and on the skin during early life demonstrated in experimental models of colitis [8,9], in airway hyperresponsiveness [8] and allergic contact dermatitis [34].
Biliary disease has been shown to develop in the NOD.c3c4 mouse model independent of an animal facility's microbial status, but the phenotype emerges at different time-points and with variable degrees of severity [11,15,32].Our observations with introduction of the microbiome at specific time-points argue that these previously reported discrepancies could be due to divergent microbiomes.We have previously shown that the bacterial populations comprising the gut microbiota in NOD.c3c4 and NOD mice housed in the same animal facility mice differ [15].It is well known that the microbiological status varies widely between different animal facilities [35,36] and a limitation of our study is that we cannot determine how the microbiota governs the observed inflammation.Further studies are needed to clarify the relative contribution of microbes per se, a specific microbial composition, or the microbiota-dependent immunologic imprinting during early life.
Delayed exposure to microbes by a combination of co-housing and cecal microbiota transfer with donor material from adult NOD.c3c4 mice at 4 weeks of age did not aggravate the histologic evident disease severity to the levels of CONV and CONV-R NOD.c3c4 mice.One limitation of this assessment was that the mice were only followed for 5 weeks after microbiome introduction and a longer follow-up time may have clarified whether the histological features are still restructuring and progressing in the GF-R group towards a similar level of portal inflammation as CONV and CONV-R.Alternatively, the absent host-microbial interplay in GF-R during the neonatal window of immunologic imprinting, might mitigate the progression of biliary disease in the NOD.c3c4 mice with long-term effects.
The adaptive immune system is crucial for disease development in the NOD.c3c4 mouse model as the NOD.c3c4-scid mice, lacking the adaptive immune system, only exhibit minor hepatic histological abnormalities and lack clinical manifestations of biliary disease [37].Second, the pathology is driven by putative autoreactive T cells as anti-CD3 treatment ameliorates disease development [13].These previous findings demonstrate clear evidence of T-cell dependence indicating that effects on the immune system are likely to represent the mechanism for the differences we observe in bile duct inflammation.However, our flow-cytometry assessments demonstrate that changes in different immune compartments on the single cell level not necessarily correspond to the degree of histologic verified bile duct inflammation.
Development of functionally tuned T cells together with other compartments of the adaptive immune system follows a time-dependent microbe-host interplay similar to what we observe for biliary disease [9].Microbial exposure before weaning impacts the tolerance to gut commensals mediated by regulatory T cells [23], reverses accumulation of iNKT cells in lungs and colon [8], establishes MAIT cells in the skin [38] and within other barrier surfaces, and normalizes serum levels of immunoglobulin E [24].Also, maternal factors such as microbial metabolites during gestation impacts the newborn's resilience to biliary inflammation mediated by suppression of immune cell activation [39].
Macrophages are important players driving biliary fibrosis in later stages of chronic human cholangiopathies [40], and the Pkhd1 gene have been pointed out as a regulator of a slowly progressing fibrosis, largely induced by macrophages in mice, on the C57BL/6J genetic background [41].The phenotype differs with the genetic background as NOD.c3c4 mice with a Pkhd1 mutation develops liver immune infiltration accelerated by autoreactive T and B cells [12].
The NOD.c3c4 mouse models two distinct human cholangiopathies with different pathogenetic properties, sex predominance and treatment strategies [21,42], and the intrahepatic bile duct inflammation and autoantibodies characterizes the immune-mediated PBC while the large duct involvement is a feature of PSC [13].We included both male and female NOD.c3c4 mice in our experiments to examine gender as a potential modifier as females are at greater risk of developing human PBC contrasting PSC that has a male predominance [43].Altogether, our male and female data show similar trends although not as strong among females.
The present results clearly demonstrate that the bile duct inflammation in the NOD.c3c4 mouse model is aggravated by microbial presence in a time-dependent manner, which indirectly gives relevance for human cholangiopathies as these diseases often feature a gut microbial imbalance [17][18][19].Whether this dysbiosis is a consequence of chronic inflammatory disease or a contributor to disease development remains unsolved, but nevertheless represents an attractive treatment target.Microbiota modulation may therefore serve as a treatment opportunity in the future as well as a clinical tool to evaluate disease activity and tailor personalized treatment and follow-up in hepatobiliary disease.

Figure 1 .
Figure 1.Germ-free mice were successfully re-conventionalized. a. the figure outlines the study design.Germ-free (Gf) mice were re-conventionalized within the first day after birth or at weaning, at 4 weeks of age, and housed under SPf conditions until sacrifice and sample collection at 9 weeks of age together with age and gender matched untreated conventional (conv) and Gf mice.illustration created with Biorender.com.B. cecum weight as percent of total body weight (conv: n = 14, conv-r: n = 12, Gf-r: n = 11, Gf: n = 16).c.Body weight measured in grams (conv: n = 24, conv-r: n = 12, Gf-r: n = 14, Gf: n = 22).Presented data represents pooled results from two separate experiments with male nod.c3c4 mice and each symbol represents an individual mouse.data are presented as mean ± SeM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by anova with Bonferroni post hoc correction for multiple comparisons.'ns' denotes a comparison as not significant.

Figure 2 .
Figure 2. early-life exposure to microbes increase the severity of bile duct inflammation.a. Macroscopic visualization of liver and intestines immediately after euthanization.B. liver weight as percentage of total body weight (conv: n = 24, conv-r: n = 12, Gf-r: n = 14, Gf: n = 22).the data represents pooled results from two separate experiments.c. representative samples of He stained liver sections by 10x magnification.d. total histological scores including portal inflammation (Pi), intrahepatic bile duct dilatation (iHBd) and bile infarct (Bi) scores (conv: n = 18, conv-r: n = 9, Gf-r: n = 9, Gf: n = 16).e. Segregated histological scores for Pi, iHBd and Bi.Presented data represents results from male nod.c3c4 mice and each symbol represents an individual mouse.data are presented as mean ± SeM. *p < 0.05, **p < 0.01, ***p < 0.001 as determined by anova with Bonferroni post hoc correction for multiple comparisons.'ns' denotes a comparison as not significant.

Figure 3 .
Figure 3. neutrophils are increased in conventional nod.c3c4 mice.a. representative pictures showing neutrophil staining by immunohistochemistry (iHc).B. ly6G positive neutrophils counts (conv: n = 18, conv-r: n = 9, Gf-r: n = 8, Gf: n = 16) reported as counted cells per mm 2 and normalized to the measured area using the random forest algorithm.c. representative slides showing iHc staining of t cells.d. cd3 positive t cell counts (conv: n = 18, conv-r: n = 9, Gf-r: n = 9, Gf: n = 16) reported as counted cells per mm 2 and normalized to the measured area using the artificial neural network algorithm.Presented data represents results from male nod.c3c4 mice.each symbol represents an individual mouse and horizontal lines represents the median value.*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by the non-parametric Kruskal-Wallis test followed by dunn's post hoc correction for multiple comparisons.

Figure 5 .
Figure 5. immunophenotyping of lymphocytes in liver and other lymphoid organs in nod.c3c4 mice with different microbial status and timing of microbial exposure.Heatmaps displaying log2 transformed relative immune subsets normalized to the mean percentage of the conventional group.liver (upper left), Mln (lower left), spleen (upper right) and thymus (lower right).Presented data represents results from male nod.c3c4 mice and each square represents an individual mouse (conv: n = 6, conv-r: n = 3, Gf-r: n = 5, Gf: n = 6).log2 transformed values are displayed on a scale from -2.0 (blue) to 2.0 (red).for gating strategies, see fig.S7. illustrations are created with Biorender.com.