Abstract
We previously reported that NOD.c3c4 mice develop spontaneous autoimmune biliary disease (ABD) with anti-mitochondrial Abs, histopathological lesions, and autoimmune T lymphocytes similar to human primary biliary cholangitis. In this article, we demonstrate that ABD in NOD.c3c4 and related NOD ABD strains is caused by a chromosome 1 region that includes a novel mutation in polycystic kidney and hepatic disease 1 (Pkhd1). We show that a long terminal repeat element inserted into intron 35 exposes an alternative polyadenylation site, resulting in a truncated Pkhd1 transcript. A novel NOD congenic mouse expressing aberrant Pkhd1, but lacking the c3 and c4 chromosomal regions (NOD.Abd3), reproduces the immunopathological features of NOD ABD. RNA sequencing of NOD.Abd3 common bile duct early in disease demonstrates upregulation of genes involved in cholangiocyte injury/morphology and downregulation of immunoregulatory genes. Consistent with this, bone marrow chimera studies show that aberrant Pkhd1 must be expressed in the target tissue (cholangiocytes) and the immune system (bone marrow). Mutations of Pkhd1 produce biliary abnormalities in mice but have not been previously associated with autoimmunity. In this study, we eliminate clinical biliary disease by backcrossing this Pkhd1 mutation onto the C57BL/6 genetic background; thus, the NOD genetic background (which promotes autoimmunity) is essential for disease. We propose that loss of functional Pkhd1 on the NOD background produces early bile duct abnormalities, initiating a break in tolerance that leads to autoimmune cholangitis in NOD.Abd3 congenic mice. This model is important for understanding loss of tolerance to cholangiocytes and is relevant to the pathogenesis of several human cholangiopathies.
Introduction
Autoimmune biliary disease (ABD) in humans includes primary biliary cholangitis (PBC) and primary sclerosing cholangitis in adults (1–3) and biliary atresia in children (4, 5). The biliary epithelial cell (cholangiocyte) is the main autoimmune target in these diseases (6, 7). Several animal models of ABD have been established. NOD.c3c4 (8, 9), NOD.ABD (10), and dnTGFβRII mice (11) develop spontaneous ABD similar to PBC. Infection of neonatal BALB/c mice produces an autoimmune reaction very similar to biliary atresia (7, 12). The NOD.c3c4 strain arose in a project to refine Idd loci by introgressing B6 and B10 Idd regions onto the NOD genetic background (8). We showed that NOD.c3c4 mice are completely protected from diabetes but develop ABD that is characterized by hepatosplenomegaly, wasting, abdominal swelling, liver function abnormalities, and, eventually, death from obstructive liver disease. Histologically, their livers show substantial lymphocytic infiltration, nonsuppurative destructive cholangitis, and macrophage aggregation in the bile ducts, all features similar to human PBC. In addition, NOD.c3c4 mice spontaneously develop anti–pyruvate dehydrogenase E2 autoantibodies (anti-mitochondrial Abs), which are highly specific for human PBC. In contrast to human PBC, NOD.c3c4 mice also develop common bile duct (CBD) dilation and inflammation, which more closely resemble primary sclerosing cholangitis or biliary atresia. Finally, they develop extensive proliferation of intrahepatic bile ductules, far exceeding the ductule proliferation seen in stage II human PBC and more resembling polycystic liver disease (8, 9, 13). However, transfer of splenocytes from NOD.ABD donors with severe disease resulted in overwhelming inflammation, nonsuppurative destructive cholangitis, high titer anti–pyruvate dehydrogenase E2 Abs, and severe illness in NOD.c3c4-scid recipients, in the absence of any additional significant ductular proliferation (10). We concluded that the NOD.c3c4 mouse strain was a useful model for understanding the mechanisms of ABD.
We have previously identified several immune mechanisms of NOD ABD. First, NOD.c3c4-scid mice do not develop clinical disease and have much diminished hepatic histological abnormalities, indicating that the adaptive immune system is critical to disease pathogenesis (10). Second, CD8 T cells from NOD.ABD donors, but not NOD donors, transferred disease into NOD.c3c4-scid recipients (10). Finally, NOD-scid recipients did not develop ABD upon adoptive transfer, even when receiving 20 million splenocytes from NOD.ABD donors that rapidly caused overwhelming ABD in NOD.c3c4-scid recipients (10). These results showed that the genetic background of the target tissue was critical to disease pathogenesis; however, the meaning of the requirement of B6/B10 genetic components in the adaptive immune system was unclear. An alternate explanation for our results was that NOD splenocytes did not cause disease, because autoreactive cholangiocyte-directed T cells were not expanded as a result of T cell repertoire development in the absence of the cholangiocyte autoantigen(s). This issue is addressed in the current study by constructing bone marrow chimeric mice.
We used a congenic mapping approach to define the genetic origin of ABD in the NOD.c3c4 model (10). Although regions on chromosomes 3 and 4 were initially linked to disease, a genome-wide 5K single nucleotide polymorphism (SNP) chip analysis showed a small non-NOD region on chromosome 1 in all strains that developed disease. To assess the role of this region, we constructed a new chromosome 1–congenic strain (NOD.Abd3). The NOD.Abd3 strain develops highly penetrant ABD, similar to the previously described strains, in the absence of the original chromosome 3 and 4 congenic regions present in the NOD.c3c4 or NOD.ABD models. We undertook the present studies to understand how the chromosome 1 genetic region mediates biliary disease and to further characterize the role of genetic background and the chromosome 1 region in the target tissue (biliary epithelium) and hematopoietic system. From these investigations, we discovered that the congenic region on chromosome 1 is itself unlikely to cause ABD; rather, the highly likely cause is a mutation altering the expression of a gene closely linked (2.5 Mb) to the congenic region, polycystic kidney and hepatic disease 1 (Pkhd1).
Materials and Methods
Animals and genotyping
NOD, NOD.ABD (10), NOD.CD45.2 (14), B10.H2g7, B6.PL-Thy1a/CyJ (B6.PL), and C57BL/6 (B6) mice and mice derived therefrom during this study were bred and housed under specific pathogen–free conditions, and all procedures were conducted according to approved protocols of the University of Cincinnati College of Medicine or the Merck Research Laboratories Animal Care and Use Committees. We discovered during quality control screening of congenic strains in our colony, using a 5K mouse single nucleotide polymorphism (SNP) chip (ParAllele Bioscience, South San Francisco, CA), that the NOD.c3c4 (9) and NOD.ABD strains included a non-NOD–derived congenic region (derived from the B6.PL or B10 mouse strains) on chromosome 1 (Supplemental Table I). This region was isolated by selective backcrossing to NOD as line 7825 (N11) (Fig. 1). Line 7825 was used to generate additional chromosome 1 recombinants to isolate smaller congenic segments, resulting in lines NOD.Abd3 (line 8706, N13), 8727 (N11), 8728 (N11), and 8730 (N11) (Fig. 2A). Novel primers developed to screen for recombination events and to define recombination points in detail are listed in Supplemental Table II. To assess whether the NOD MHC contributed to ABD, line 7825 mice were bred to NOD.B10-H2b (NOD.H2b) mice (15) and then backcrossed to line 7825 mice. Mice homozygous for the chromosome 1 congenic region and heterozygous for the H2b MHC were intercrossed to fix both regions as homozygous for the non-NOD haplotype. Line 7825 mice were crossed to NOD, and the F1 mice were intercrossed to assess whether the Abd3 region was recessive and sufficient to cause ABD on the NOD background in (7825 × NOD) F2 mice. A cohort of line 7825 mice was compared with a cohort of NOD mice for diabetes development, as described (16). NOD.Abd3 mice were crossed once with B6 mice and then the F1 offspring were intercrossed to produce F2 mice, or they were crossed twice to B6 mice to generate Backcross 2 mice, which were then further intercrossed to generate the backcross 2 intercrossed (BC2i) generation. NOD.Abd3 mice backcrossed to B6 mice were genotyped as follows using markers outside the B6 congenic region, which genotype as NOD for NOD.Abd3 mice. The proximal marker is rs13475762 (Fig. 1A) with the forward primer rs13475762_F = 5′-TTCCCCCTTTTAATATTTTGCAT-3′ and reverse primer rs13475762_R = 5′-CAGGGAGGCAGTGATTTAGC-3′. The distal marker is rs32040516 with forward primer rs32040516_F = 5′-TGAGCCATCTGACAGACCAG-3′ and reverse primer rs32040516_R = 5′-TGGATGGCCATGACAAAAA-3′. PCR product was digested with restriction enzyme ACC1 (for proximal marker) or MNL1 (for distal marker) (New England Biolabs) at 37°C for 2 h, and the PCR products were run on a 4% agarose gel with 1× Tris/Borate/EDTA buffer. The size of the digested PCR product of proximal markers is 87 bp for NOD, and the size of the digested PCR product of distal markers is 182 bp. Comparison of NOD and B6 genomic sequences in the congenic portion of the Abd3 region was performed as described (17).
Histopathology and disease scoring
Livers and/or CBDs were scored for visible/clinical liver and CBD pathology, as previously described (8–10) (see below for methods), isolated from mice, immediately fixed in 10% formalin, and embedded in paraffin. Samples were stained with H&E, and histology was scored blindly, using microscopy, for duct epithelial hyperplasia and lymphocytic infiltration.
CBD and liver gross pathological scoring.
CBD dilation and the quantitation of liver abnormalities were examined and scored separately. The clinical liver score, performed at the University of California at Davis, was assigned based on tissue induration (hardness) and the presence of cysts on the surface of liver as follows: 0: soft and no cysts; 1: mild induration, no cysts; 2: indurated with cysts on one lobe surface; 3: indurated with cysts on two to three lobes; and 4: indurated and cysts on all four lobes. The clinical CBD score was identical to the diameter (in millimeters) of the maximum CBD width; if CBD measured <1 mm (“normal” CBD), the score was assigned as 0. The clinical biliary duct score performed at Merck (on 7825 × NOD F2 mice) was assigned as follows: CBD, 0: normal; 1: size just above normal; 2: CBD clearly enlarged but <2 mm; 3: CBD ∼ 2 mm; 4: CBD ∼ 2–4 mm; and 5: CBD > 4 mm. The clinical liver score was assigned as follows: liver: 0: normal; 1: very few cysts in the liver; 2: cysts in liver are easily seen; 3: many cysts easily seen; 4: many cysts in most or all lobes; and 5: liver completely full of cysts.
Histological liver score: duct epithelial hyperplasia.
The histological liver duct score was calculated as follows: 0: 0–3 ductules per portal triad across the section; 1: at least two times the normal diameter of ductule lumen in 10–25% of the entire section and no fewer than three small ductules around one triad in 25–50% of the entire section; 2: enlarging ductules across 25–50% of the entire section, and the average diameter of a ductule lumen section is ∼2–4 times the normal ductule diameter; 3: larger ductules across >50% of the entire section, and the average diameter of ductule lumen section is ∼4–5.5 times of the normal ductule diameter; and 4: dilated, diffuse, torturous ducts (“cysts”) in much of the section (this is the end stage disease), and the average diameter of ductule lumen section is >5.5 times the normal ductule diameter.
Histological score: leukocytic infiltration of portal areas.
The histological liver leukocyte infiltration score was calculated as follows: 0: none or a few cells in <25% of entire section; 1: small numbers of cells in multiple areas (25–50%), including some clusters of immune cells; 2: small numbers of cells (cell infiltrate 2–3 cells deep) diffusely (>50% of entire section) or patchy moderate infiltrates (25–50% of entire section); 3: moderate numbers of cells diffusely (>50% of entire section) or patchy large infiltrates (larger cluster than for “2,” 25–50% of entire section); and 4: diffuse or significant sections of large cellular infiltrates.
Bone marrow chimeras
Two- to three-month-old NOD (CD45.1) or NOD.Abd3 (CD45.1) mice were irradiated (1200 rad). Fifteen to twenty million bone marrow cells from 4-mo-old NOD.CD45.2 or 2–3-mo-old NOD.Abd3 (CD45.1) donors were extracted without RBC lysis. Mature CD4, CD8, and CD90 cells were removed using magnetic beads (Miltenyi Biotec), and the bone marrow was injected i.v. into the irradiated recipient mice. Recipient mice were given water treated with antibiotics (neomycin trisulfate salt hydrate) for 2 wk after transfer. The recipient mice were sacrificed 120 d post–bone marrow transfer or when they developed abdominal swelling indicative of severe hepatobiliary disease (∼2.5 mo for NOD.Abd3 recipients of NOD.Abd3 bone marrow). Tissues were analyzed for anatomical and histological biliary/liver disease. CD45.1 or CD45.2 quantification was done for the recipients of different donor cells, and the bone marrow reconstitution rate ranged from 87.6 to 98.9%.
Quantitative real-time PCR of mRNA
Extrahepatic CBDs were isolated from 2-wk-old NOD.Abd3 and NOD mice and stored in RNAlater solution (QIAGEN). For RNA extraction from CBD, the tissue was removed from RNAlater solution and disrupted in lysing/binding buffer provided in a mirVana miRNA Isolation Kit (Life Technologies, CA) using a homogenizer (QIAGEN), and RNA was extracted from the tissues using a mirVana miRNA Isolation Kit and converted to cDNA by random primers (High-Capacity cDNA Reverse Transcription Kit; Applied Biosystems). Relative gene expression of Pkhd1 was performed using selective mouse TaqMan gene-expression assays (Mm00467747_m1 for region spanning exons 35 and 36, Mm00467728_m1 for region spanning exons 15 and 16, Mm01233737_m1 for regions spanning exon 60 and 61) and an endogenous control assay Rn18S (Mm03928990_g1; all from Life Technologies).
RNA sequencing profiling of CBDs and data analysis
Total CBD RNA was extracted from 2-wk-old NOD or NOD.Abd3 mice using a mirVana miRNA Isolation Kit (Life Technologies). Two samples were sequenced for each strain; each of the two NOD.Abd3 samples had RNA from three individual CBDs, and the two NOD samples had RNA from three to five individual CBDs. RNA concentration, purity (28S:18S rRNA ratio), and quality (RNA integrity number) were assessed using a 2100 Bioanalyzer (Agilent, Santa Clara, CA) at the DNA Sequencing and Genotyping Core, Cincinnati Children’s Hospital Medical Center. The RNA samples had an RNA integrity score (RNA integrity number) ≥ 7.9 on the Agilent Bioanalyzer (scale of 1–10, with 10 being completely intact).
The RNA sequencing (RNA-seq) library was generated with an Illumina TruSeq RNA preparation kit and subsequently sequenced on an Illumina HiSeq 2000, using single-end 50-bp read specifications with a read depth ≥ 10 million (Illumina, San Diego, CA).
RNA-seq analysis was carried out using Bowtie (18) and TopHat2 (19). RNA-seq BAM files generated using the Bowtie–TopHat2 pipeline were analyzed for the expression of known and unknown genes/transcripts using the Cufflinks2 pipeline (20) and included removal of reads that did not map uniquely to the mm9 Ensembl genome or ENSEMBL-annotated genes. Workflows included filtering to remove duplicate reads and those with postaligned read metrics mapping quality < 40. Our samples showed normal 3ʹ/5′ read distribution ratios compared with hundreds of other samples run in the Cincinnati Children’s Hospital Medical Center genomics core. Transcript/isoform and gene-summarized expression tables were filtered to identify entities whose expression was at least five fragments per kilobase of exon per million fragments mapped (Cufflinks) or five reads per kilobase per million reads mapped (GeneSpring) in at least one sample. Differentially expressed gene signatures were identified using the Audic–Claverie test (p < 0.05) and the Student t test (false discovery rate < 0.05), followed by a two-fold change requirement. Gene ontology and other enrichment and biological network analyses were carried out using ToppGene (http://toppgene.cchmc.org) (21) and ToppCluster (http://toppcluster.cchmc.org) (22). ToppCluster output of xgmml data files were network analyzed in Cytoscape, yielding similar and complementary results compared with the ToppFun functional enrichment analysis.
PCR and sequencing of Pkhd1 exons and intron 35
For different exons of Pkhd1, primer pairs were designed using Primer3 (http://biotools.umassmed.edu/bioapps/primer3_www.cgi) to amplify representative exon sequences, using genomic DNA as template. PCR primer sequences are detailed in Supplemental Table III.
Primer N7 forward primer and primer 3 reverse primer were used with PrimeSTAR GXL DNA Polymerase (Clontech Laboratories) for long-range PCR to identify the inserted DNA fragment in Pkhd1 intron 35 in NOD.Abd3 mice. Long-range PCR products from NOD and NOD.Abd3 mice were then sent for next-generation sequencing at Cincinnati Children’s Hospital Medical Center DNA Sequencing and Genotyping Core.
Statistical analysis
GraphPad Prism 6 (GraphPad, San Diego, CA) software package was used to perform unpaired two-tailed Student t tests for comparison of the expression level of genes, as well as the Mann–Whitney U test to estimate significance levels of histology scores and clinical scores.
Results
A region on chromosome 1 is necessary for spontaneous autoimmune cholangitis in NOD congenic mice
We previously reported that NOD.c3c4 and NOD.ABD mice, with B6/B10 genetic regions on chromosomes 3 and 4 placed on the NOD background, develop a genetically determined spontaneous autoimmune cholangitis; however, the causative genetic regions were not clear (8–10). Further analysis of these strains by a 5K SNP chip analysis revealed a region of B6/B10-derived DNA on chromosome 1 in the NOD.c3c4 and NOD.ABD strains (Supplemental Table I) (10). To assess the significance of this region, we developed, from the NOD.ABD strain, a novel congenic mouse that had the chromosome 1 region but lacked the chromosome 3 and 4 regions (line 7825 in Fig. 1A). This chromosome 1 congenic mouse has highly penetrant ABD that is identical in kinetics, penetrance, and phenotype to the previously published NOD.c3c4 strain. CBD dilation score, which we previously reported as a sensitive marker of ABD in this strain (8), at 120 d is identical between strain 7825 and NOD.c3c4 (strain 1112) (Fig. 1B) as are liver and histology scores (data not shown). Therefore, we designated the chromosome 1 region as Abd3 [chromosome 3 and 4 regions previously described in NOD.ABD mice (10) were termed Abd2 and Abd1, respectively].
To test whether the Abd3 region is fully recessive and is the only segregating genetic region contributing to the ABD phenotype in line 7825, we examined (line 7825 × NOD) F2 mice. Thirty-four F2 mice were scored at 120 d of age in a blinded manner for CBD and liver disease and genotyped at Abd3. All mice genotyped as homozygous for the NOD allele at Abd3 (n = 6) had clinically normal livers and CBDs, whereas all mice genotyped as homozygous for the B6/B10 allele at Abd3 (n = 10) had macroscopic disease evident in liver and bile duct (Fig. 1C). All mice genotyped as heterozygous at Abd3 (n = 18) were also phenotypically normal. This confirmed that ABD caused by the B6/B10-derived chromosome 1 region is recessive and that the chromosome 1 region is sufficient to cause disease in the context of the NOD genetic background; no other gene influencing ABD to a detectable degree is segregating outside the Abd3 region. The effect of the chromosome 1 region on the development of autoimmune diabetes was also examined, and the frequency of diabetes in line 7825 was equivalent to that observed in the NOD parental strain, indicating that Abd3 had no effect on type 1 diabetes (Supplemental Fig. 1).
To refine the Abd3 region and aid in the discovery of the causal gene, several additional congenic strains were developed by screening for recombination events in the congenic segment present in line 7825 (Fig. 2A). Line 8706 has the smallest congenic interval, a maximal size of 1.0003 Mb (proximal boundary defined by rs13475762 and the distal boundary defined by rs32040516); we designated this strain NOD.Abd3. Like strain 7825, NOD.Abd3 mice developed liver histological abnormalities (lymphocytic infiltrates and biliary ductule epithelial hyperplasia) identical in appearance and kinetics to NOD.c3c4 mice (9) (Fig. 2B). (NOD.Abd3 × NOD) F1 mice lacked clinical disease and had no liver histological abnormalities compared with NOD.Abd3 mice (Fig. 2C), clearly demonstrating that Abd3 mediates a recessive trait, even at a microscopic level.
There are only a few genes in the 1-Mb Abd3 region, including two microRNAs (miR-30a and miR-30c-2), Ogfrl1, a predicted gene with unknown function (Gm6420), and partial sequence of B3gat2 (Fig. 2D). However, none of these genes showed significant expression differences in target tissue or CD8 T cells isolated from NOD.Abd3 mice compared with NOD mice (data not shown). A comparison of the NOD and B6/B10 sequence also did not reveal any functional polymorphisms, such as amino acid–changing SNPs or SNPs affecting splice sites (data not shown). Notably, strains developed from proximal recombination events (lines 8727 and 8730, Fig. 2A) did not develop ABD. This led us to consider the possibility that a proximal gene outside, but closely linked to, the congenic interval could be mutated and that the congenic interval was carried along with the mutation due to linkage disequilibrium.
Bile duct abnormality in NOD.Abd3 mice associated with aberrant expression of Pkhd1
All recombinant strains that did not develop disease had recombination events removing the proximal portion of the congenic region present in line 7825 (Fig. 2A, lines 8727 and 8730), suggesting there could be a genetic mutation proximal to the Abd3 region that was causing disease (because a recombination event removing the proximal portion of the 7825 congenic region would also replace a large portion of the DNA upstream of the congenic region with NOD DNA derived from the backcross partner). To assess this, we performed RNA-seq analysis of CBD from NOD.Abd3 and NOD mice. We then compared the RNA-seq reads from NOD and NOD.Abd3 CBDs, focusing on 0–24 Mb (ending at the proximal border of the Abd3 region) on chromosome 1, and found a single major genetic difference between these strains: Pkhd1 was aberrantly expressed in NOD.Abd3 CBD samples compared with NOD CBD samples (Fig. 3). Pkhd1 is located ∼2.5 Mb upstream from the proximal border of the Abd3 congenic region. In human and murine models, Pkhd1 mutations mediating biliary-hepatic disease (in the absence of renal pathology) produce biliary duct–proliferation abnormalities similar to NOD.Abd3 mice (but largely lacking significant cholangitis/inflammation) (23–26). RNA-seq reads aligned to the reference genome showed that NOD.Abd3 CBD lacked expression of exon 36 to exon 67 of Pkhd1 compared with NOD controls (Fig. 3A). Moreover, the read density aligned to each exon of Pkhd1 indicated that the expression level of exons 1–35 of Pkhd1 in NOD.Abd3 CBDs at 2 wk of age is higher than that in NOD CBDs; in contrast, the read density of all 67 Pkhd1 exons in NOD CBDs was relatively constant (Fig. 3B).
To rule out the possibility that truncated expression of Pkhd1 was due to genomic DNA deletion of exons 36–67, we designed primers to amplify genomic DNA for proximal and distal exons before and after the observed transcript cutoff site at exons 35–36). PCR showed product bands for all representative exons in NOD.Abd3 comparable to NOD and B6 (Fig. 3C), clearly demonstrating that missing expression of Pkhd1 exons 36–67 in NOD.Abd3 was not caused by a genomic DNA deletion event. We then tested Pkhd1 expression level using quantitative RT-PCR (qRT-PCR). Expression of exons 15/16 in CBD from 2-wk-old NOD.Abd3 mice was nearly two times that in NOD and B6 mice, and the expression of exons 35/36 and 60/61 is dramatically diminished (Fig. 3D), consistent with the RNA-seq results (Fig. 3B). There was no significant difference in the expression of these exons between NOD and B6 mice (Fig. 3D), and the expression level within NOD or B6 mice was relatively consistent (data not shown). Given these results, we defined Abd3 as including the B6 genetic region and Pkhd1 (Figs. 1A, 2A).
Abd3-mediated clinical ABD requires the NOD genetic background but not the NOD MHC region
To examine the influence of non-MHC NOD genes on ABD, we crossed NOD.Abd3 and B6 mice and intercrossed the resulting F1 mice. This F2 generation allows for the examination of mice homozygous for the Abd3 region but, on average, having 50% NOD and 50% B6 genetic contributions to the remainder of the genome. To reduce the proportion of NOD-derived genes even further, F1 mice were backcrossed twice to B6 mice and in each generation breeders were selected to retain the disease-associated Abd3 region and then intercrossed to produce BC2i mice having, on average, only 12.5% NOD-derived alleles. The pathological CBD enlargement and anatomically visible liver abnormalities seen in Abd3 homozygous mice are greatly reduced and absent, respectively, in F2 and BC2i offspring that are homozygous at Abd3 but have with a reduced NOD background (Fig. 4A, 4B). Therefore, progressively reducing the NOD genetic background and increasing the B6 genetic background ameliorates clinical manifestations of ABD in Abd3 homozygous animals, strongly emphasizing the importance of the NOD genetic background for the presence of clinical ABD. Notably, because clinical disease is present in a substantial proportion of the F2 mice homozygous for Abd3, NOD and B6 alleles contributing to disease penetrance could be mapped in future studies using a large cohort of such mice.
We performed histological analysis of the same F2 and BC2i crosses. Mice lacking the Abd3 allele, or heterozygous for the region in the F2 and BC2i cohorts, had no histological lesions, whereas homozygous animals had reduced histology scores for duct proliferation and leukocytic infiltration compared with NOD.Abd3 mice (Fig. 4C, 4D). Notably, histological disease was not different between F2 and BC2i mice (Fig. 4C, 4D), in contrast to the clinical manifestations of disease (Fig. 4A, 4B). The fact that histology scores were similar between F2 and BC2i Abd3 homozygous mice suggests that the B6 immune system does respond, albeit less aggressively, to the aberrant regulation of Pkhd1 (i.e., the NOD genetic background does not seem to be necessary for initiation of disease [histological lesions] but is necessary for clinical progression of disease).
The development of autoimmune diabetes in NOD mice is dependent on having at least one copy of the NOD MHC, as well as a large number of other NOD background genes (27). To understand the interaction between the Abd3 region and the NOD MHC in the context of the NOD background on the development of ABD, we constructed NOD.H2b Abd3 mice. NOD.H2b Abd3 mice had a small, but statistically significant, decrease in CBD score, but their liver score (Fig. 4F) and histological severity of ABD (data not shown, n = 9 NOD.Abd3 mice and n = 24 NOD.H2b Abd3 mice, age 120 d) were not significantly different from NOD.Abd3 mice. This suggests that, unlike type 1 diabetes in the NOD model, biliary disease was not dependent on the presence of the NOD MHC.
Early cholangiocyte damage response and abnormal immune activation in NOD.Abd3 CBD
In the NOD mouse, islet inflammation begins as early as 3 wk (27) and is associated with developmental processes of the pancreatic islets. We examined the CBD of 2-wk-old mice and found disrupted biliary epithelial architecture in NOD.Abd3 CBD compared with the smooth lumen in the CBD from NOD and NOD.c3c4-scid mice (Fig. 4E). No significant histological difference was observed in the liver between NOD.Abd3 and NOD control at this very early time point (data not shown). These findings are consistent with our previously published results showing that hepatic inflammation was not detectable until ∼8 wk in NOD.ABD mice and that NOD.c3c4-scid mice had minimal hepatic disease (8, 10). To further understand the striking difference between NOD and NOD.Abd3 CBD architecture at 2 wk, we performed RNA-seq on CBD from 2-wk-old mice. Two-hundred and fifty-five genes were significantly differentially expressed between the strains (Fig. 5A). Cluster enrichment network analysis of these 255 genes, based on biological processes, molecular functions, mouse-knockout phenotypes, and biological pathways, showed that many of the genes upregulated in 2-wk-old NOD.Abd3 CBD were extensively involved in cholangiocyte injury (reflecting the tissue target cell in ABD), whereas genes involved in the immune response (adaptive and innate) were downregulated in NOD.Abd3 CBD (Fig. 5B). Biological process–enrichment analysis shows that many of the upregulated genes in NOD.Abd3 mice are involved in epithelial cell differentiation, morphology, and morphogenesis (all genes associated with these processes were upregulated), whereas the downregulated genes are largely involved in abnormal inflammatory response/inflammation, leukocyte transendothelial migration, hepatobiliary system and gland physiology, and cellular secretion (all genes associated with these processes were downregulated) (Table I). A few processes involved upregulated and downregulated genes, including cell adhesion and epithelial cell proliferation (Table I). These results establish that a very early tissue abnormality interacting with an abnormal immune response drives the development of NOD.Abd3 biliary disease.
Biological process enrichment analysis shows the significantly differentially expressed genes in CBD from 2-wk-old mice (from Fig. 5A) associated with biological processes characterized by by upregulated genes (in red) and downregulated genes (in green). The biological processes characterized by all upregulated or downregulated genes are in green and red, respectively. The processes involving some upregulated and some downregulated genes are in black and the individual genes are in green or red depending on whether they are upregulated or downregulated.
Bone marrow chimera studies demonstrate that expression of Abd3 in target tissue and immune cells is necessary for disease
We next dissected the level at which the NOD background was acting in the autoimmune process. We had previously shown that NOD.ABD splenocytes or CD8 T cells could not transfer disease into NOD.scid mice and, conversely, that NOD splenocytes or CD8 cells could not transfer disease into NOD.c3c4-scid mice. These results suggested that the Abd3 locus acted on the target tissue and the immune system. An alternate explanation, however, was that the NOD CD8 T cell repertoire has the potential to cause ABD but that the disease-mediating T cell clones were not expanded because they were never exposed to their tissue autoantigen (i.e., NOD mice lack abnormal biliary epithelial cells). To distinguish between these possibilities, we performed bone marrow chimera studies. NOD.Abd3 irradiated recipients were reconstituted with NOD or NOD.Abd3 bone marrow; conversely, NOD recipients were reconstituted with NOD.Abd3 bone marrow. The results clearly reproduced the transfer study findings: the recipients must express NOD.Abd3 in biliary tissue to develop disease. NOD.Abd3→NOD recipients had no clinical abnormality in CBD or liver (Fig. 6A, bars 7 and 8, significantly different from NOD.Abd3→NOD.Abd3 recipients) and no duct epithelial hyperplasia and minimal lymphocytic infiltrates (Fig. 6B, bars 7 and 8, significantly different from NOD.Abd3→NOD.Abd3 recipients). In addition, NOD bone marrow did not cause significant disease in NOD.Abd3 recipients: the clinical gross liver score and histological duct epithelial hyperplasia score at 120 d post–bone marrow transplant were significantly less than in NOD.Abd3→NOD.Abd3 recipients (Fig. 6). NOD.Abd3 recipients already had some disease at the time of irradiation (Fig. 6, bars 1 and 2). In the NOD.Abd3→NOD.Abd3 transfer, NOD.Abd3 bone marrow repopulated the mouse and caused the disease to progress. In NOD→NOD.Abd3 mice, NOD bone marrow did not accelerate the disease, and it actually reduced the histological score compared with 10-wk-old NOD.Abd3 mice (Fig. 6B, bars 1 and 2 compared with bars 5 and 6). Therefore, there is clearly a difference in the ability of the hematopoietic system to cause disease between the NOD and NOD.Abd3 strains; the Abd3 mutant Pkhd1 allele must be expressed in the hematopoietic system and target tissue for ABD to develop.
Although the expression of Pkhd1 in immune cells has never been reported previously, to our knowledge, the Ensembl database shows that RNA-seq reads from spleen are aligned to Pkhd1 (28). We hypothesized that a deficient exon expression pattern of Pkhd1 could be detected in CBD, the target tissue, as well as in immune cells, explaining why both systems were necessary for ABD. However, we were unable to detect Pkhd1 expression in unactivated splenocytes, intrahepatic immune cells, or MACS-sorted CD4 or CD8 T cells from spleen and liver using exons 15/16, 35/36, and 60/61 (data not shown).
Details of the Pkhd1 genetic mutation in NOD.Abd3 and potential mechanism of truncated transcription
Pkhd1 is ∼2.5 Mb upstream from the Abd3 allele (Fig. 7A). The RNA-seq data, as well as the qRT-PCR results, revealed a novel loss of Pkhd1 transcription after exon 35, and we sought to understand how this transcription deregulation occurs in NOD.Abd3 mice. The alignment of the RNA-seq reads from 2-wk-old NOD.Abd3 CBD and NOD CBD identified a unique transcribed intronic region immediately after exon 35 (the second longest Pkhd1 intron) in NOD.Abd3 mice but not in NOD mice (Fig. 7B). We found no mutations at the splice donor site (GT) after exon 35 or the splice acceptor site (AG) before the start of exon 36. In contrast, there were multiple microsatellite repeats in the uniquely transcribed intron region and poly(A) signal (AAUAAA), which suggested that it could be a 3′ untranslated region (UTR)-like region (29) in Pkhd1.
To evaluate further the unique 3ʹ UTR–like region in intron 35 of NOD.Abd3 Pkhd1 mice, we designed three primer pairs (primers 1–3 in Table II, Supplemental Table III) to perform tiled PCR starting from the end of the 3′ UTR–like region in intron 35 toward exon 35 (Fig. 8A). The first two pairs of tiled PCR primers at the end of the novel 3′ UTR–like region of intron 35 amplified in NOD mice but not in NOD.Abd3 mice (Fig. 8B, Table II). We then designed seven primer pairs (every 5 kb) from exon 36 toward the end of the novel 3′ UTR–like region in intron 35 (N1–N7, Table II, Supplemental Table III). Each of these primer pairs amplified in NOD and NOD.Abd3 mice (Fig. 8C). Lastly, using the forward primer of primer pair N7 and the reverse primer of primer pair 3 (Fig. 8A, Table II), we found that a PCR product size in NOD mice was ∼3 kb (similar to reference target sequence). In striking contrast, the PCR product size in NOD.Abd3 mice was ∼9 kb (Fig. 8D). Therefore, the mutation within Pkhd1 was a DNA insertion of ∼6 kb.
Primer Set or Exon Position . | Starting Positiona . | Ending Positiona . | Amplification of NOD.Abd3 . | Amplification of NOD . |
---|---|---|---|---|
Exon 36 | 20,440,302 | 20,440,458 | NA | NA |
Primer N1 | 20,440,373 | 20,440,692 | Yes | Yes |
Primer N2 | 20,445,607 | 20,445,925 | Yes | Yes |
Primer N3 | 20,450,397 | 20,450,733 | Yes | Yes |
Primer N4 | 20,455,542 | 20,455,846 | Yes | Yes |
Primer N5 | 20,460,303 | 20,460,626 | Yes | Yes |
Primer N6 | 20,465,591 | 20,465,989 | Yes | Yes |
Primer N7 | 20,470,616 | 20,470,976 | Yes | Yes |
Primer 1 | 20,472,908 | 20,473,208 | No | Yes |
Primer 2 | 20,473,134 | 20,473,447 | No | Yes |
Primer 3 | 20,473,238 | 20,473,540 | Yes | Yes |
Exon 35 | 20,493,023 | 20,493,173 | NA | NA |
Primer Set or Exon Position . | Starting Positiona . | Ending Positiona . | Amplification of NOD.Abd3 . | Amplification of NOD . |
---|---|---|---|---|
Exon 36 | 20,440,302 | 20,440,458 | NA | NA |
Primer N1 | 20,440,373 | 20,440,692 | Yes | Yes |
Primer N2 | 20,445,607 | 20,445,925 | Yes | Yes |
Primer N3 | 20,450,397 | 20,450,733 | Yes | Yes |
Primer N4 | 20,455,542 | 20,455,846 | Yes | Yes |
Primer N5 | 20,460,303 | 20,460,626 | Yes | Yes |
Primer N6 | 20,465,591 | 20,465,989 | Yes | Yes |
Primer N7 | 20,470,616 | 20,470,976 | Yes | Yes |
Primer 1 | 20,472,908 | 20,473,208 | No | Yes |
Primer 2 | 20,473,134 | 20,473,447 | No | Yes |
Primer 3 | 20,473,238 | 20,473,540 | Yes | Yes |
Exon 35 | 20,493,023 | 20,493,173 | NA | NA |
Genomic coordinates refer to chromosome 1. Genomic assembly: NCBI37/mm9.
NA: no PCR was performed; No: no band shown on the gel; Yes: single band with expected PCR product size.
Using next-generation sequencing technology, we sequenced the PCR products through de novo reads assembly. The assembled PCR product sequence of NOD mice could be fully aligned with the reference genome, whereas only part of the three nodes of the assembly sequence of NOD.Abd3 mice matched with the NOD sequence. The node 2–assembled sequence in NOD.Abd3 Pkhd1 intron 35 was analyzed in RepeatMasker and had 99.73% identity with long terminal repeat elements of four class II endogenous retroviruses (ERVs) (Fig. 8E).
Given these results of the genetic characterization of Pkhd1 in NOD.Abd3 mice, we propose the following potential mechanism: the 6-kb long terminal repeat ERV DNA insertion in intron 35 of NOD.Abd3 Pkhd1 mice disrupts normal recognition of splicing sites in this intron and forces the exposure of an alternative polyadenylation site within the intron, thus giving rise to a novel shortened transcript that ends after exon 35 (Fig. 8F). Given that qRT-PCR and RNA-seq show an upregulation of the first 35 exons of Pkhd1 in NOD.Abd3 CBD, this alternative polyadenylation site must predominate over the normal polyadenylation site (after exon 67) in NOD.Abd3 mice.
In summary, we have shown that ABD in NOD congenic mouse strains is associated with a novel Pkhd1 mutation. Uniquely, on the NOD background, this mutation results in an autoimmune disease that depends on expression of mutated Pkhd1 in the target tissue and hematopoietic system. Early abnormalities in cholangiocyte morphology and development are accompanied by dysregulation of the immune system, resulting in a genetically controlled autoimmune disease.
Discussion
In this article, we prove conclusively that NOD ABD is caused by a recessively acting region on chromosome 1. This region, designated Abd3, is necessary and sufficient for ABD when on the NOD background. The only genetic difference that we found in the Abd3 region of diseased NOD.Abd3 mice compared with NOD mice was aberrant expression of Pkhd1. Pkhd1 is a large gene with 67 exons (4059 aa) encoding fibrocystin, a membrane-associated receptor-like protein with a single transmembrane domain, a small intracellular domain, and a large extracellular domain containing multiple repeats of Ig-like plexin-transcription factor domains and two G8 domains (predicted to contain 10 β strands and an α helix) (30, 31). We anticipate that the mutated Pkhd1 protein in NOD.Abd3 mice, if it is expressed, has lost the G8 domains, as well as the transmembrane and intracellular domains. Fibrocystin has domains that are similar to those found in plexin and hepatocyte growth factor families; these domains function in cellular adhesion and proliferation (24). Pkhd1 is extensively alternatively spliced and is spliced differently in different cell types, but a role for alternately spliced products has not been demonstrated (32, 33). The human and mouse protein sequences share 73% identity (30, 34). During embryogenesis, fibrocystin is widely expressed in epithelial derivatives, including neural tubules, gut, pulmonary bronchi, renal cells, and hepatic cells. By embryonic day 15, fibrocystin is strongly expressed in bile duct cholangiocyte cilia, as well as in the epithelial cells in the kidney-collecting tubes (32, 35, 36). mAbs also detect fibrocystin in human pancreatic duct, islets, testis, and adrenal gland (37).
NOD.Abd3 mice have normal renal structure throughout life. Consistent with this, prior mouse models with engineered deletions of Pkhd1 have no renal abnormalities. Mice with targeted deletion of Pkhd1 exon 3, 4, or 40 had no kidney abnormality, but they demonstrated intrahepatic bile duct proliferation with progressive cyst formation (similar to our model) and associated periportal fibrosis (not seen in NOD.Abd3 mice) (23, 24, 33). Pkhd1ex4del mice develop splenomegaly, which is also found in our model. Pkhd1ex4del mice develop a syndrome of congenital hepatic fibrosis, and Pkhd1ex4del cholangiocytes secrete chemokines, including CXCL1, CXCL2, and CSCL12, which aberrantly recruit macrophages (38). Macrophage depletion in Pkhd1ex4del mice decreases fibrosis and cystic disease. The similarity of the biliary epithelial abnormalities and, in some cases, the presence of hematopoietic cell infiltrates in these other models of Pkhd1 mutation, leads us to conclude that mutated Pkhd1 is the primary cause of ABD in our model. Our model is unique, however, because none of the previous models were shown to involve an autoimmune response nor did they require, as far as we know, Pkhd1 expression in hematopoietic cells.
Disease expression in other Pkhd1 models was also apparently unaffected by the genetic background. Thus, the role of the genetic background in our model is also unique; we show decreased histological abnormalities (and reduced or no clinical disease depending on the amount of residual NOD genetic material remaining) when the Pkhd1 mutation is expressed on a genetic background that is predominantly B6. Notably, however, there was no significant difference in histological disease scores between F2 and BC2i mice (Fig. 4C, 4D), suggesting that disease initiation (manifest as histological lesions) does not require the NOD genetic background. In contrast, clinical disease (liver and CBD scores in Fig. 4A, 4B) requires the NOD genetic background. Although outside the scope of the current study, mapping the NOD background regions required for clinical disease, in conjunction with Abd3 homozygosity, would be achievable by genetic analysis of (NOD.Abd3 × B6) F2 mice having two doses of the Abd3 region, because disease was observed to be present in a portion of these mice (Fig. 4A, 4B). We note that introgression of genetic regions derived from B6 or B10 mice into the NOD strain that prevent type 1 diabetes [i.e., introgression of the MHC region and large segments of chromosomes 3 and 4 (9, 15)] does not prevent clinical or histological ABD in the presence of Abd3 homozygosity (Fig. 4F) (9). This suggests that NOD genes distinct from those critical for the development of type 1 diabetes in the NOD model may be essential for the clinical manifestation of ABD.
At this point, we cannot fully explain our findings, because we have been unable to verify Pkhd1 expression in the immune cells that we have tested. It is possible that the upregulated Pkhd1 exons 1–35 play a role in immune cells; another possibility is that mutant Pkhd1 affects alternative splicing in an immune cell subset (e.g., macrophages) and that we need to use different primers to detect the immune cell–specific transcript. Pkhd1 expression in the thymus could affect the immune repertoire. Finally, we cannot exclude that the linked B6-derived congenic region present in the NOD.Abd3 strain plays some role in the pathogenesis of these mice. Although we have not shown any differentially expressed genes from this region, we will ultimately only be able to formally disprove a role for the 1.0-Mb B6-derived congenic segment within the Abd3 region in future studies by selecting a recombination event between the congenic region and Pkhd1 and testing the effect of the mutant Pkhd1 gene in the absence of the congenic region (on the NOD background) on disease.
Bone marrow– and splenocyte-transfer studies show that the Abd3 allele/mutated Pkhd1 must be expressed in the immune system and target tissue to develop cholangitis. The question remains how an alteration of a single gene (Pkhd1) involved in cilia function and epithelial integrity in cholangiocytes can ultimately lead to an autoimmune disease. Interestingly, the biological process–enrichment analysis of RNA-seq data from NOD.Abd3 CBD reveals a striking pattern, with the downregulated genes extensively involved in acute immune/inflammatory response (Table I). Many of these genes can negatively regulate the immune response, thus mediating a protective role against, for example, overexuberant immune responses to microbial infection. For example, although there is a large number of reports on its proinflammatory effect, S100A9 (Table I) also has anti-inflammatory effects through inhibition of neutrophil oxidative metabolism (39, 40) and neutrophil chemorepulsion. Serpinc1 can induce the anti-inflammatory cytokine PGI2 from cultured HUVECs (41), Spdef negatively regulates TLR signaling in airway epithelial cells (42), Vip executes protective humoral responses against pathogens (43), and CORO1A is linked to inhibition of neutrophil apoptosis in chronic inflammation, which is essential to resolve inflammation (44).
Downregulation of several of these genes results in increased inflammation and autoimmune disease [e.g., ablation of Syt7 in mice resulted in inflammatory myopathy (45); C2GnT2-knockout mice have decreased mucosal barrier function in the digestive tract, reduced levels of circulating IgGs and fecal IgA, and increased susceptibility to experimental colitis (46); and complement deficiency is associated with lupus (47)]. We speculate that the downregulation of these anti-inflammatory/regulatory genes in the innate immune system or in cholangiocytes makes it difficult to resolve inflammatory responses, which may facilitate the generation of neoantigen in the target tissue. Generation of neoantigens, in the presence of altered bile composition and abnormal duct physiology, could escalate the immune response and progression to an adaptive immune response to self-targets.
Mutation of Pkhd1 likely results in a cholangiociliopathy with impaired primary cholangiocyte cilia. Defective cilia structure can compromise the integrated sensory/transducing functions, resulting in increased cholangiocyte cAMP and decreased [Ca2+] (48), causing cholangiocyte hyperproliferation, changes in metabolic processes, abnormal cell–matrix interactions, and altered fluid secretion/absorption (water and bile modification), as reflected by the gene cluster analysis (Fig. 5B, Table I). Therefore, the entire cholangiocyte environment is changed, and the outcome of the change in one gene (Pkhd1) leads to altered morphogenesis and altered immune-related responses.
The results of the bone marrow–transfer studies clearly demonstrate that NOD.Abd3 immune cells are pathogenic, whereas NOD immune cells are not. NOD T cells are unable to cause disease, even when they develop in an Abd3+ host with abnormal cholangiocytes. Taken together with a similar result in our transfer studies (10), this suggests that the Pkhd1 mutation must be expressed in the target tissue and in the hematopoietic tissue. In contrast, requirement of the NOD genetic background (which is highly disposed to autoimmunity) for disease manifestation supports the idea that the immune system perpetuates and enhances cholangitis in NOD.Abd3 mice. Future studies will be directed to discover the critical immune subsets and the mechanism by which the Pkhd1 mutation affects the immune system.
In summary, we have identified a novel mutation in Pkhd1 as a key genetic factor causing the biliary tissue abnormality and activating the adaptive immune system in the context of the NOD genetic background. Although we have clearly demonstrated that the target tissue (cholangiocytes) and the immune system must express the Pkhd1 mutation, we do not yet understand the role of the mutation in each system or how the systems interact. Abnormalities involving hundreds of tissue-specific (cholangiocyte specific) and immune-origin genes are affected by 2 wk of age, indicating a complex orchestrated event producing this organ-specific autoimmune disease. Thus, our mouse model of spontaneously mutated Pkhd1 is valuable for exploring several cholangiopathies, including primary biliary cholangitis, primary sclerosing cholangitis, and polycystic hepatobiliary disease.
Acknowledgements
We thank Mehdi Keddache and Satwica Yerneni for assistance with RNA-seq.
Footnotes
This work was supported by the Biomedical Laboratory Research and Development Service of the Veterans Affairs Office of Research and Development, Veterans Affairs Merit Grant BX000827-01A1 (to W.M.R.). L.S.W. and D.B.R. were supported by Wellcome Trust Grant 107212/Z/15/Z and Juvenile Diabetes Research Foundation Grant 4-SRA-2017-473-A-N. The Wellcome Trust Center for Human Genetics is supported by a core award from the Wellcome Trust (203141/Z/16/Z). W.M.R., L.S.W., and M.E.G. were supported by National Institutes of Health Grant R01DK074768.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- ABD
autoimmune biliary disease
- B6
C57BL/6
- BC2i
backcross 2 intercrossed
- B6.PL
B6.PL-Thy1a/CyJ
- CBD
common bile duct
- ERV
endogenous retrovirus
- NOD.H2b
NOD.B10-H2b
- PBC
primary biliary cholangitis
- Pkhd1
polycystic kidney and hepatic disease 1
- qRT-PCR
quantitative RT-PCR
- RNA-seq
RNA sequencing
- SNP
single nucleotide polymorphism
- UTR
untranslated region.
References
Disclosures
The authors have no financial conflicts of interest.