Quantitative Trait Locus Analysis Implicates CD4+/CD44high Memory T Cells in the Pathogenesis of Murine Autoimmune Pancreatitis

The mouse strain MRL/MpJ is prone to spontaneously develop autoimmune pancreatitis (AIP). To elucidate the genetic control towards the development of the phenotype and to characterize contributions of immunocompetent cell types, MRL/MpJ mice were interbred with three additional strains (BXD2/TYJ, NZM2410/J, CAST/EIJ) for four generations in an advanced intercross line. Cellular phenotypes were determined by flow cytometric quantification of splenic leukocytes and complemented by the histopathological evaluation of pancreatic lesions. An Illumina SNP array was used for genotyping. QTL analyses were performed with the R implementation of HAPPY. Out of 41 leukocyte subpopulations (B cells, T cells and dendritic cells), only three were significantly associated with AIP: While CD4+/CD44high memory T cells and CD4+/CD69+ T helper (Th) cells correlated positively with the disease, the cytotoxic T cell phenotype CD8+/CD44low showed a negative correlation. A QTL for AIP on chromosome 2 overlapped with QTLs for CD4+/CD44high and CD8+/CD44high memory T cells, FoxP3+/CD4+ and FoxP3+/CD8+ regulatory T cells (Tregs), and CD8+/CD69+ cytotoxic T cells. On chromosome 6, overlapping QTLs for AIP and CD4+/IL17+ Th17 cells and again FoxP3+/CD8+ Tregs were observed. In conclusion, CD4+/CD44high memory T cells are the only leukocyte subtype that could be linked to AIP both by correlation studies and from observed overlapping QTL. The potential role of this cell type in the pathogenesis of AIP warrants further investigations.


Introduction
Given the problems of obtaining large sizes of human samples, studies in animal models of AIP may provide valuable additional insights into the genetic basis of the disease. We have previously established an advanced intercross line originating from MRL/MpJ parental mice and three other mouse strains: CAST/EIJ (healthy controls), BXD2/TYJ (susceptible to collageninduced arthritis) and NZM2410/J (a model of Lupus erythematodes) [26]. The idea beyond this concept was to map both general autoimmune disease-associated loci and AIP-specific quantitative traits. Therefore, generation 4 of outbread intercross mice was characterized phenotypically by scoring histopathological changes of the pancreas and genotyped employing SNP arrays. By this approach, five quantitative trait loci (QTL), located on chromosomes 2, 4 (n = 2), 5 and 6, were mapped [26]. Similarly, we identified QTLs controlling arthritis and skin inflammation [27,28]. Taking sex as a covariate, we have now used the same approach to study genomic loci that control immune cell phenotypes in the spleen and to determine their overlap with the QTLs for AIP. Out of several leukocyte subtypes, only CD4 + /CD44 high memory T cells where not only controlled by such an overlapping QTL, but also showed a significant correlation of their relative frequency with the appearance of AIP.

Animal Model and Experimental AIP
The establishment of the 4-way autoimmunity advanced intercross line has been described before [26]. Briefly, MRL/MpJ, NZM2410/J, BXD2/TyJ and CAST/EiJ parental mouse strains were intercrossed at an equal strain and sex distribution. To maintain an equal distribution of original strains in subsequent generations, parental origin of offspring mice of the predecessor generation was considered. For each generation of mice, at least 50 breeding pairs were used as parentals. As previously described, MRL/MpJ mice, but no individuals of the other parental strains, developed AIP in an age and gender specific manner [26,29].
Development of spontaneous AIP in parental strains and in intercross generation 4 (156 males and 175 females) was assessed in 6-months-old mice by evaluating the severity of pancreatic lesions. Therefore, paraffin-embedded pancreatic sections were stained with hematoxylin and eosin (H&E), applying standard protocols. Pathological changes were graded on a semi-quantitative scale from 0 to 4 [26]. The stages were defined as follows: 0, no pathological changes; 1, minimal infiltration of periductal tissue with mononuclear cells but no parenchymal destruction; 2, moderate periductal infiltration with mononuclear cells associated with beginning parenchymal destruction; 3, severe periductal inflammation and/or more extended parenchymal destruction; 4, diffuse mononuclear cell infiltrates, destruction of acini and (partial) replacement by adipose tissue. All samples were assessed by two independent investigators and blinded before evaluation. AIP stages were determined by microscopic analysis of at least two tissue sections per sample. Mice with pancreatic lesions that scored ! 2 were defined as positive for AIP.
Animals were kept under specific pathogen-free conditions at a 12 h light/dark cycle with food and water ad libitum. All procedures were performed with adherence to the EU Directive 2010/63/EU for animal experiments and approved by the local governmental administrations (Landesamt für Landwirtschaft, Lebensmittelsicherheit und Fischerei Mecklenburg-Vorpommern).

Immunohistochemical Analysis
Cryostat sections of pancreatic tissue (6 μm) were fixed by incubation in ice-cold methanol for 1 min at 4°C and washed three times with PBS. Subsequently, they were stained using the Vectastain ABC staining kit (Vector Laboratories, Burlingame, CA, USA) according to the manufacturer's instructions. For the detection of CD4 and CD44, mouse-specific primary rat antibodies were employed (anti-CD4, Immunotools, Friesoythe, Germany and anti-CD44, eBioscience, San Diego, CA, USA, respectively). The sections were counterstained with hemalaun and examined by light microscopy (Axioskop 40, Zeiss, Oberkochen, Germany).

Analysis of Leukocyte Subtypes by Flow Cytometry
Splenocytes were isolated from the spleen of G4 mice using a cell strainer (70 μm). Red blood cells were lysed applying RBC lysis buffer (eBioscience) according to the manufacturer's instructions. After washing and centrifugation steps, 1x10 6 cells per stain were subjected to subsequent analysis. Prior to staining, Fc receptors on splenocytes were blocked by preincubation with anti-CD16/CD32 antibodies (BD Biosciences, Heidelberg, Germany) for 5-10 minutes on ice. Surface staining was performed by incubating the cells with fluorochrome-conjugated specific antibodies (listed in the supplement, S1 Table) for at least 20 min in dark on ice. After washing and centrifugation steps, stained cells were fixed with 1% paraformaldehyde for 10-20 min at 4°C and subjected to flow cytometry.
For staining of intracellular cytokines, single cell suspensions of splenocytes were fixed in 4% paraformaldehyde for 10-20 min at 4°C and permeabilized employing Saponin (0.3%; Sigma-Aldrich, Deisenhofen, Germany) for 10 min. Afterwards, optimized concentrations of fluorochrome-conjugated anti-cytokine antibodies were applied at 4°C for 30 min in the dark, followed by washing steps and flow cytometry.
Details regarding marker combinations and the corresponding leukocyte subpopulations are given in Table 1. The table also indicates the averaged results for each phenotype, obtained from the analysis of 331 G4 mice. Data are expressed as relative frequency of the investigated leukocyte subset, with 100% corresponding to the total number of splenocytes. Representative dot plots of the experimental data are shown in the supplement (S1 Fig).

Correlation Analysis
Correlations between AIP appearance and sex, as well as between AIP and leukocyte subpopulations, were analyzed using Spearman's correlation coefficient (Spearman's rho). Significance of correlations was calculated using a correlation test in R. Correction for multiple testing was done by controlling the false discovery rate using the Benjamini-Hochberg approach.

Pancreatic Histopathology and Immune Cell Phenotypes of G4 Mice
Generation 4 of the 4-way autoimmune-prone intercross mouse line was employed, at an age of 6 months, to assess pancreatic histopathology and relative frequencies of different leukocyte subsets in the spleen. Therefore, 331 mice (156 females and 175 males) were included into the investigations. The mice represent a subset of the previously analyzed 351 animals [26], which was chosen based on the availability of flow cytometry data. Scoring of AIP-typical pancreatic lesions, such as presence of lymphocytic foci and parenchymal destruction, revealed an AIP stage 2 or 3 in 44 mice (32 females and 12 males). Stage 4 was not detected, but has been observed in mice of the advanced intercross line outside of this study (R.J., unpublished data). These numbers correspond to 18.3% of the females but only 7.7% of the males (13.3% of all mice). Details are given in Table 2. Exemplary tissue stains are shown in Fig 1. The data are in line with our previous report [26], indicating that the 331 mice are representative of the entire cohort. They are also in agreement with the fact that AIP of the parental MRL/MpJ strain is largely restricted to females (at an age of at least 6 months) [11,29]. Gender dependency of AIP was also confirmed by correlation analysis (Spearman's rho = -0.16; p = 0.0090).
To study immune cell phenotypes in the context of AIP, splenocytes were phenotyped employing flow cytometry. We focused on 41 subsets of leukocytes (B cells, T cells and DCs), covering a large range of immunocompetent cells that have been implicated in the pathogenesis of various autoimmune diseases [32,33]. Special attention was paid to B cells, cytotoxic and helper T cells, Tregs and DCs. Table 1 in the Materials and methods section shows the relative frequencies of all these leukocyte subtypes as averaged values of the 331 G4 mice.

Correlations between AIP and Leukocyte Abundance
We next studied correlations between appearance of AIP and the relative frequencies of the 41 leukocyte subsets (Table 3).
A significant correlation (corrected p<0.05) was found for three populations of lymphocytes: activated Th cells (CD4 + /CD69 + ), CD4 + /CD44 high memory T cells, and cytotoxic T cells of the phenotype CD8 + /CD44 low . While the first two leukocyte subsets were positively correlated with the disease, CD8 + /CD44 low cells displayed a negative correlation. For three further lymphocyte populations, p values <0.05 were determined only prior to correction for multiple testing. These cell types are activated cytotoxic T cells (CD8 + /CD69 + ; positive correlation), and Table 2. Pancreatic histopathology. Pancreatic sections of 331 G4 mice were stained with H&E, and evaluated applying a scoring system from 0 to 4 as further described in the methods section. The data are a subset of previously published data obtained from 351 mice [26]. Th cells with the phenotypes CD4 + /CD62L -(positive correlation) and CD4 + /CD44 low (negative correlation).

QTL Analysis
We have previously employed Illumina murine HD arrays to genotype 351 G4 mice of the 4-way autoimmune-prone intercross mouse line [26]. Here, we have re-analyzed the data subset of the 331 phenotyped mice (see above) to perform QTL analyses with sex as covariate. As shown in Table 4, a total of four QTLs could be mapped to the chromosomes 2, 4 and 6 (global p<0.05). The-log10(p) values varied from 4.2 to 5.5. To avoid any confusion with the previously identified QTLs, these new QTLs were termed AIP s1-s4. The mean confidence interval (CI) of each QTL was roughly 19 Mb. Next, the same genotyping data subset was used to determine QTLs of all 41 leukocyte subpopulations (Table 5; S2 Table). Most significant QTLs (global p<0.05) were found for B cells (n = 80), T helper cells (n = 68) and cytotoxic T cells (n = 59). For conventional and plasmacytoid DCs, 8 and 7 QTLs, respectively, could be mapped, while for Tregs 25 loci were detected. Only 4 leukocyte subsets (CD19 + /CD86 + , CD8 + /CD62L -, CD4 + /CD69 + and PDCA1 + /CD11ccells) were not influenced by any QTL, whereas for CD19 + /CD69 -, CD19 + /MHC IIand CD8 + / CD69 + lymphocytes more than 20 QTLs were discovered. The mean CI of each QTL was about 13.6 Mb. -log10(p) values varied from 3.8 to 32.5 (S2 Table). QTLs were located on all chromosomes, but most were found on chromosomes 2, 6, 7, 9, 12 and 20 (Table 5). For B cells, most QTLs were mapped on chromosomes 8 and 9, for cytotoxic T cells on chromosome 20 and for T helper cells on chromosome 17 (Table 5).

Overlapping QTLs for Immune Cell Phenotypes and AIP
The significant QTLs for AIP with sex as a covariate are located on chromosomes 2, 4 and 6 ( Table 4). Of these, chromosomes 2 and 6 also harbor significant QTLs for B cells, cytotoxic T cells, T helper cells and Tregs. On chromosome 4, we identified QTLs for the same subpopulations, except of Tregs (Table 5). No QTLs for conventional or plasmacytoid DCs are located on these disease-associated chromosomes.

Detection of CD44 on Lymphocytes in Pancreatic Tissue
CD4 + /CD44 high lymphocytes were the only leukocyte subtype that could be linked to AIP both by the correlation studies (Table 3) and from observed overlapping QTL (Table 6). We therefore asked if cells with this phenotype were also present in the inflamed pancreatic tissue. Using serial sections, we found that cells expressing CD4 and CD44, respectively, could be detected in overlapping regions of lymphocytic foci (Fig 3).   Table 6. doi:10.1371/journal.pone.0136298.g002

Discussion
As other autoimmune diseases, AIP is considered to develop after environmental triggering of genetically susceptible individuals [1,[15][16][17][18]32,33]. Except of a few HLA serotypes and SNP polymorphisms in non-HLA genes, however, no AIP-associated genetic factors have been identified in humans so far. This is largely due to the relative rarity of the disease (estimated prevalence in the general population: <1/100,000), which hampers the buildup of large cohorts of patients.
Employing an autoimmune-prone intercross mouse line, we previously identified five QTLs of murine AIP that may provide additional insights into the pathogenesis of the disease [26]. In our current work, we expanded these investigations by analyzing AIP-associated immune cell phenotypes and their genetic control. Since AIP of the parental MRL/MpJ mouse strain is more frequent in females, sex was considered as a covariate in all our investigations. Taking this approach, we confirmed the susceptibility loci on mouse chromosomes 2 (CI 56.3-81.9 Mb), 4 (CI 81.3-120.7 Mb) and 6 (CI 111.2-133.9 Mb) ( Table 4). Since our previous analysis [26] was performed without sex as covariate, these QTLs can now be assigned as Table 6. Overlapping QTLs for AIP and leukocyte subtypes with distances between peaks below 10 Mbp. QTLs are listed with chromosome (Chr), peak position, confidence interval (CI) and -log10(p).  sex-independent. Neither here nor in our previous study, we could identify any X-linked QTL, suggesting that the effect of X-linked genes is too small to be verified by our approach. The reason for the higher susceptibility of the female mice to AIP, therefore, remains to be deciphered. Recently, Okada et al. reported a pathogenetic role of mag, an autoimmune susceptibility locus encoded by the telomeric region of MRL/MpJ mouse chromosome 1, in murine AIP [34]. Interestingly, this susceptibility locus is not preserved in our intercross line since we did not map significant genetic markers on chromosome 1.
In a comprehensive effort, we phenotyped a variety of B cell, T cell and dendritic cell subtypes by determining their relative frequencies in the spleen of G4 mice, and systematically searched for genetic loci that influence these phenotypes. We came up with no less than 273 QTLs that control 37 different splenocyte subtypes (Table 5). This original set of loci was considerably narrowed down when an overlap with one of the AIP QTLs was introduced as an additional criterion: Of the immune cell-associated QTLs, only five overlapped with the QTL AIP s1, and one each with AIP s3 and AIP s4 ( Table 6). The corresponding immune cell phenotypes consist of activated cytotoxic T cells (CD8 + /CD69 + ), Th17 cells (CD4 + /IL17 + ), two types of T cells expressing the memory marker CD44 (CD4 + /CD44 high and CD8 + /CD44 high ), and also two types of regulatory T cells (FoxP3 + /CD4 + and FoxP3 + /CD8 + ; with the latter being influenced by two distinct loci).
So far, we had identified a small subset of QTLs that influence both the appearance of AIP and control a total of six different immune cell phenotypes in the spleen. These data, however, did not address the question of a direct association between the two phenotypes (AIP and frequency of the respective immune cells). Therefore, a correlation analysis ( Table 2) was performed which revealed a statistically significant negative correlation between AIP and cytotoxic T cells of the phenotype CD8 + /CD44 low , and a positive such for AIP and activated T helper cells (CD4 + /CD69 + ) as well as CD4 + /CD44 high cells. For the first two T cell subtypes, no QTLs overlapping with QTLs for AIP had been found. CD4 + cells expressing the memory marker CD44 are therefore the only cell type which could be linked to AIP by both approaches. Both cell surface markers could also be detected on immune cells in pancreatic lesions (Fig 3). The latter findings complement our previous studies regarding the composition of the immune cell infiltrate in MRL/MpJ mice, which had shown a predominance of CD4 cells over CD8 cells infiltrating the pancreas as well as the presence of plasma cells, regulatory T cells and macrophages [13,29].
To the best of our knowledge, this is the first experimental hint for a potential involvement of memory T cells in the pathogenesis of AIP. Apart from AIP, however, memory T cells (and here in particular self-antigen-reactive CD4 + effector memory T cells) have been suggested to drive the progression of autoimmune diseases because of their ready effector functionality and relative longevity [35]. Along these lines, persistent antigen increases the pool of effector memory T cells, which may in turn trigger the progression of the autoimmune disease through the potent production of inflammatory cytokines, such as interferon-γ [36,37]. Indeed, we have previously observed high mRNA levels of this cytokine (as well as of interleukin-2 and interleukin-6) in pancreatic tissue of MRL/Mp mice with advanced AIP and also shown that injections of interferon-γ accelerate and aggravate the disease [12,29].
Of course, immune cell phenotypes that fulfilled only one criterion (existence of overlapping QTLs with AIP or correlation with the appearance of AIP) may nevertheless be important in the progression of the disease, and their QTLs demand further investigation let alone as a controlling factor in immunity. Thus, activated CD4 + and CD8 + lymphocytes, Th17 cells and Tregs all have already been implicated in the pathogenesis of AIP [reviewed in 1]. Moreover, employing MRL/MpJ mice we have recently shown (based on immunohistochemical studies and a flow cytometric analysis of splenocytes) that the immunosuppressant drug rapamycin significantly reduces pancreatic damage by expanding Tregs of the phenotype CTLA4 + /CD4 + / FoxP3 + and a subsequent reduction of the effector T cell response [13]. We consider the fact that this part of our observations is in line with previous investigations also as supportive for our novel concept regarding the role of memory T cells in the development of murine AIP.
The overlapping QTL region for AIP and CD4 + /CD44 high lymphocytes, which is located on chromosome 2 [ Table 6], still spans roughly 11 Mb and contains more than 70 genes. Noteworthy, the interval contains the genes Sjogren syndrome antigen B (Ssb) and Ubiquitin protein ligase E3 component n-recognin 3 (Ubr3), which we have previously described as part of the QTL AIP1 [26]. The encoded proteins might be disease-relevant autoantigens and therefore deserve special attention in future studies. A full list of genes in the overlapping QTL region is provided in the S3 Table. Taken together, we hypothesize that CD4 + /CD44 high memory T cells play a previously unrecognized role in murine AIP, and suggest that these cells are an important link between genetic susceptibility and development of the disease (Fig 4). Currently, fine-mapping studies are underway to narrow down AIP loci and overlapping QTLs for immune cell phenotypes, and to identify candidate genes that control the respective phenotypes.