Impact of diet and genes on murine autoimmune pancreatitis

Abstract The impact of environmental factors, such as diet, and the genetic basis of autoimmune pancreatitis (AIP) are largely unknown. Here, we used an experimental murine AIP model to identify the contribution of diet to AIP development, as well as to fine‐map AIP‐associated genes in outbred mice prone to develop the disease. For this purpose, we fed mice of an autoimmune‐prone intercross line (AIL) three different diets (control, calorie‐reduced and western diet) for 6 months, at which point the mice were genotyped and phenotyped for AIP. Overall, 269 out of 734 mice (36.6%) developed AIP with signs of parenchymal destruction, equally affecting mice of both sexes. AIP prevalence and severity were reduced by approximately 50% in mice held under caloric restriction compared to those fed control or western diet. We identified a quantitative trait locus (QTL) on chromosome 4 to be associated with AIP, which is located within a previously reported QTL. This association does not change when considering diet or sex as an additional variable for the mapping. Using whole‐genome sequences of the AIL founder strains, we resolved this QTL to a single candidate gene, namely Map3k7. Expression of Map3k7 was largely restricted to islet cells as well as lymphocytes found in the exocrine pancreas of mice with AIP. Our studies suggest a major impact of diet on AIP. Furthermore, we identify Map3k7 as a novel susceptibility gene for experimental AIP. Both findings warrant clinical translation.

which is frequently associated with inflammatory bowel disease, granulocytic epithelial lesions are the key pathognomonic finding, whereas IgG4-expressing plasma cells are lacking. Common histological features of both subtypes are a periductal fibrosis and the presence of periductal lymphoplasmacytic infiltrates. [1][2][3] shares with other autoimmune diseases basic immunological characteristics, such as presence of autoantibodies and involvement of autoreactive T cells. Furthermore, adoptive transfer of splenic leucocyte subpopulations (specifically, CD44 high memory T cells) from MRL/MpJ mice with AIP into healthy mice induces AIP in the recipients. 4 Environmental and nutritional influences are well-established key factors in the pathogenesis of pancreatitis. Of note, chronic alcohol abuse represents the most common cause of CP in general, and hypertriglyceridemia is associated with an increased risk of acute pancreatitis. 5 The role of such factors in the development of AIP has, however, not been systematically studied yet. Therefore, we here also addressed the contribution of diet, as an every-day environmental factor, on AIP development.
Regarding genetic associations, genome-wide association studies (GWAS) and candidate gene-based approaches have identified several susceptibility loci for human AIP (largely, type 1), including HLA DRB1*04:05-DQB1*04:01, 6 FCRL3, 7 CTLA4 8,9 and KCNA3. 10 Targeted sequencing has suggested additional genetic associations of AIP, specifically CACNA1S, SMAD7, TOP1 11 and CALCB. 12 Noteworthy, specific PRSS1 mutations (PRSS1_IVS 2+56_60 delCCCAG and PRSS1_p.Leu81Met) that were suggested to cause ectopic trypsinogen activation have also been implicated into the pathogenesis of human AIP type 1. 13,14 Further insights into the genetic architecture of AIP have been obtained using the MRL/MpJ mouse AIP model. At an age of approximately 6 months, predominantly female MRL/MpJ mice spontaneously develop AIP that resembles important histopathological features of human AIP. 15,16 Application of polyinosinic:polycytidylic acid (poly I:C) further accelerates and enhances the disease. 17,18 Previously, we generated an autoimmune-prone advanced intercross mouse (AIL) line, to study the impact of genetics and diet on complex (inflammatory disease) traits in the mouse, including AIP. 19 In brief, 3 parental mouse strains that spontaneously develop different autoimmune diseases-MRL/MpJ (AIP), NZM2410/J (lupus) and BXD2/TyJ (arthritis)-and CAST/EIJ mice, which are not prone to any autoimmune disease, were subjected to crossbreeding. 20 These AIL mice have been used for QTL mapping of several complex traits, including murine AIP. 19,20 Regarding AIP, 5 associated QTL were identified, located on chromosomes 2, 4 (two QTLs), 5 and 6. 20 The identified QTL, however, encompassed numerous genes, due to the fact that the study was performed in mice of the 4th generation, where relatively few crossing over had occurred.
To fine map these QTLs, preferably to the single gene level, we here used mice of the 15th, 18th-20th generation. In addition, we took advantage of recently published whole-genome sequences (WGS) of the AIL founder strains 19 to fine map the genetic associations of AIP.

| Animals
MRL/MpJ, NZM2410/J, BXD2/TyJ and CAST/EiJ parental mouse strains were intercrossed at an equal strain and sex distribution as described 4,19,20 to generate an advanced autoimmune-prone intercross line (AIL). As previously reported, AIL mice develop AIP, 20 which is most likely due to the inclusion of the AIP-susceptible MRL/MpJ mice. 16,[18][19][20] For the mapping study herein, mice of the 15th, 18th-20th generation were used. To study the impact of diet on AIP, after weaning, AIL mice were randomly allocated to one of the following diets: control mouse chow, caloric restriction and western diet, as described elsewhere. 19 Regarding randomization: Offspring mice were transferred into separate cages after weaning at 3-4 weeks. Each cage contained mice of either gender and was randomly allocated at a 1:1:1 to one of the three different diets: control mouse chow, caloric restriction and western diet (all mice of one cage received the same diet, which was selected at random). At the age of 6 months, a skin biopsy (for genotyping) and the pancreas (for histological analysis) were obtained ( Figure 1A) Areas, file number 27-2/13) and performed by certified personnel.

| Genotyping and QTL mapping
The tips of the tails of AIL mice were collected during mouse sampling at month 6 and subjected to the isolation of genomic DNA as reported. 19 In brief, the DNeasy Blood & Tissue Kit (Qiagen GmbH, Hilden, Germany) was employed according to the manufacturer's protocol, and DNA samples were stored at −20°C until further use.
DNA samples from 734 mice were analysed by MegaMUGA genotyping array covering 77 800 markers throughout the mouse genome. 21 Genotyping was performed at Neogen/GeneSeek (Lincoln, NE, USA). Non-informative SNPs were filtered out using plink and applying the following criteria: minor allele frequency (maf) >0.05, missing genotype probability <0.1 and common homozygous SNPs among the founders. This approach yielded 55,458 informative SNPs for further downstream analysis. 22 The probabilistic reconstruction of AIL mouse genome in term of four founder strains was performed using the R package of HAPPY 23,24 as described before. 19 Briefly, the posterior probability that each mouse was in one of the four possible genotype states was estimated employing a hidden Markov model (HMM), and probabilities were converted to Rqtl2 object in R.
Subsequently, a chromosome-wise kinship matrix, which represents intra-individual relationship, was calculated using 'calc_kinship' function with leaving out one chromosome at a time (R/qtl2 R package). 25 We estimated heritability of AIP using 'est_herit' function from the same R package, that is R/qtl2. For additive model, LOD scores were calculated by linear mixed model in which AIP score was regressed to posterior probabilities with sex and diet as additive covariate and kinship as a random effect (scan1 function from Rqtl2). For QTL interaction model, covariates diet and sex were considered interactive covariates rather than additive covariate within the same function.
Genome-wide and suggestive thresholds were estimated by traditional permutation (1000)-based approach at 5% and 10% threshold.
The confidence interval for a QTL was described by 1.5 LOD drop within 10 Mb from the peak SNP. To further fine map the QTL, WGS data from founder strains (variant call file format; vcf) were downloaded from database. 19 The annotation of SNPs and Indels within this file was performed using ENSEMBL VEP web server. 26 The genomic region plots for the QTL were visualized and created using R ggplot2 27 and UCSC Genome Browser. 28 This analytical pipeline has been recently validated. 9 F I G U R E 1 Diet modulates the spontaneous development of autoimmune pancreatitis (AIP) in an advanced intercross outbred mouse line (AIL) (A) A total of 734 outbred AIL mice were fed 3 different diets (caloric restriction, control or western diet) for 6 months. Thereafter, mice were genotyped and H&E-stained sections from the pancreata were evaluated for the presence and (if present) severity of AIP. (B) Pancreatic sections were stained with H&E and subjected to the evaluation of pathological changes on a semi-quantitative scale from 0 to 4. Stage 0: healthy; stage 1: small infiltrate of mononuclear cells (arrow); stage 2: large periductal focus of mononuclear cells; beginning parenchymal destruction; stage 3: severe inflammation and more extended parenchymal destruction; and stage 4: organ-wide inflammation; large-scale destruction of acini and partial replacement by adipose and fibrotic tissue. (C) When stratified for sex, no difference in the AIP severity was noted. Graph shows all individual values as dots, whereas the red line indicates the median. Statistical analysis was performed using rank-sum test. (D) By contrast, stratification for the different diets showed that mice fed calorie-reduced diet had a significant lower AIP severity compared to control and western diet-fed mice. Graph shows all individual values as dots, whereas the red line indicates the median. Statistical analysis was performed using one-way ANOVA (P < 0.0001) with Tukey's multiple comparison as post-test

| Histology and immunohistochemistry
Development of spontaneous AIP was assessed based on pancreatic histopathology. Therefore, paraffin-embedded pancreatic sections (4 µm thick) were stained with haematoxylin and eosin (H&E), and subjected to the evaluation of pancreatic lesions on a semi-quantitative scale from 0 to 4 as described before. 16,18,20 Briefly, the stages

| Statistical analysis
Unless stated otherwise, data were analysed employing IBM SPSS Statistics V25.0 and are presented as mean values ± SEM. Statistical tests are detailed in the figure legends. P values of < 0.05 were considered statistically significant. We used Dsquare function from modEvA R package to estimate proportion of variance explained by diet (4.7%) and sex (0.3%) for pancreatitis score.

| Caloric restriction reduces the prevalence and clinical severity of autoimmune pancreatitis
At 6 months, 428 out of 734 mice (58.3%) presented with lymphocytic infiltrates in the exocrine pancreas. A total of 269 animals (36.6%) displayed signs of parenchymal destruction and were classified as AIP stage 2 or greater. Both sexes were similarly affected by AIP ( Figure 1C). Prevalence of AIP stage 2 or greater amounted to 38.5% in female and 33.4% in male mice, and average AIP scores were 1.1 ± 0.4 or 0.9 ± 0.4 for female or male mice, respectively.
AIP scores also showed no differences between both sexes when calculation was performed separately within each dietary group.
When comparing AIP prevalence among the different diets, 71% mice under caloric restriction were free of any AIP, whereas an AIP score of 0 was only observed in 30% and 41% of the mice fed control and western diet, respectively (P < 0.05; chi-squared test).
Corresponding findings were made regarding AIP severity: In mice held at caloric restriction, the histological AIP score was 0.5 ± 0.03; whereas the score reached 1.1 ± 0.03 or 1.2 ± 0.04 in mice fed control or western diet, respectively ( Figure 1D). Strikingly, cases with parenchymal destruction (stages 2-4) were much less common in mice with caloric restriction than in animals on control chow or western diet (Table 1). Taken together, these findings point towards a protective role of caloric restriction on the development and on the progression of murine AIP. The principle effects of diets on the average scores were the same in females and males (data not shown).

| Mitogen-activated protein kinase kinase kinase 7 (Map3k7) is associated with experimental autoimmune pancreatitis
Next, we aimed to fine map previously reported QTLs of AIP, 20 possibly identify so far unreported AIP-associated genes, and evaluate the impact of gene-diet and gene-sex interactions in AIP. For fine mapping, we continuously intercrossed AIL mice leading to smaller  Figure 2A-B), which was located within the previously identified QTL for AIP (Table 2). We next considered diet or sex as interactive covariates for the QTL mapping, allowing to identify gene-diet and gene-sex interaction: The QTL on chromosome 4 was confirmed at the genome-wide significance level (LOD = 7.68, CI = 31.39-34.8 Mb) when diet was taken as an interactive variable.
When using sex as an interactive variable, the QTL was significant at the suggestive level (LOD = 6.89, CI = 31-36.1 Mb, Figure 2B). We to identify polymorphisms (SNPs and Indels) discriminating between BxD2/TyJ and MRL/MpJ strains. We identified 2,994 variants between the two strains among which most of the variants were either intergenic regions (38%), intronic (36%) or located within non-coding transcripts without any predicted severe consequences. For protein-coding genes, we did not identify any non-synonymous variants with severe consequences. However, we detected multiple variants on predicted upstream/downstream region and importantly in 3' UTR region of the full-length protein-coding gene Map3k7, whereas for the partial protein-coding gene Bach2 variants in predicted upstream region were found ( Figure 2D). The importance of variants in 3' UTR has been established previously. Hence, we identify Map3k7, also known as transforming growth factor-beta-activated kinase 1 (Tak1), spans from 31,96 to 32,02 Mb, to be potentially associated with AIP.

| Map3k7 is expressed by infiltrating mononuclear cells of mice with experimental autoimmune pancreatitis
To obtain insights into a potential contribution of Map3k7, we studied its protein expression in the context of murine AIP. In healthy pancreata, MAP3K7 protein expression was largely restricted to pancreatic islets (Figure 3). By contrast, pancreata from mice with AIP showed additional expression by the pancreas-infiltrating mononuclear cells (Figure 3).

| D ISCUSS I ON
Autoimmune disorders are multifactorial diseases that are influenced by the interplay of environmental, for example dietary, and genetic factors. 19 In the case of AIP, both aspects are still relatively poorly understood. We herein addressed this knowledge gap and show that diet has a vast impact on AIP development. More specifically, compared to control or western diet, caloric restriction halved both AIP prevalence and severity. Regarding the genetic AIP association, we fine-mapped Map3k7 as an AIP-associated gene and demonstrate its expression in the leucocyte infiltrate within the pancreas of mice with AIP.
Our genetic studies were based on the report of five, rather large F I G U R E 2 Map3k7 is associated with experimental autoimmune pancreatitis (A) Manhattan plot for mapping of quantitative trait loci (QTLs) for murine autoimmune pancreatitis (AIP9. The plot shows chromosomes on x-axis and LOD (log of odd ratios) for each SNP (grey and black dots) on y-axis. LOD scores were calculated by regressing histological pancreatitis score with the estimated genotype posterior probability. Sex, diet and first principle component of kinship matrix were considered as additive covariate loci. Logarithmic p-values are plotted against genome location. The red line represents the genome-wide threshold (P < 0.05), and the blue line shows the suggestive threshold (P < 0.1) evaluated after 1000 permutations for histological pancreatitis score. (B) The scatter line plot shows LOD scores across chromosome 4 for three models (ie additive, interactive diet and sex). Use of either one of the interactive variables ('diet' or 'sex') considers the variability introduced by either of the variables, allowing to detect gene-diet and gene-sex interactions. The shaded region within the plot described confidence interval in Mb for each model (grey = AIP2, orange = genome-wide and blue = chromosome-wide). (C) The box plot shows ranked haplotype posterior probabilities (derived from happy R package using HMM) for Peak SNP (UNC7008858) across the four founder strains, when stratified for different AIP score. (D) The figure shows the fine-mapped AIP2 loci (using UCSC browser) on chromosome 4 and genes within this loci. The figure also shows all the SNPs and Indels (derived from WGS) different between BxD2/TyJ and MRL/MpJ, as Peak SNP from these strains was negatively and positively correlated with AIP score in previous figure In summary, this investigation has identified a novel susceptibility gene of murine AIP, Map3k7, and a potent protective environmental factor, dietary restriction. As the MAP3K7 protein is in principle druggable 39 and dietary interventions are possible, these findings may shape the future management of AIP patient.

ACK N OWLED G EM ENTS
We gratefully acknowledge the excellent technical assistance of Mrs Katja Bergmann. This work was supported by the Deutsche Forschungsgemeinschaft (DFG JA 819/5-1 and IB 24/9-1) and the

CO N FLI C T O F I NTE R E S T
None to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

O RCI D
Robert Jaster https://orcid.org/0000-0002-8220-4570 F I G U R E 3 MAP3K7 is expressed in the lymphocyte infiltrates of experimental autoimmune pancreatitis Pancreatic tissues from mice with AIP stage 0 (healthy) and stage 3 (severe AIP) were stained with anti-MAP3K7. Arrows point to positively stained pancreatic islets and organ-infiltrating mononuclear cells, respectively. The photographs are representative for n ≥ 5 mice per group