Genome-wide analyses have shown that the MHC class II region is the principal locus that confers susceptibility to a number of human autoimmune diseases. Due to the high degree of linkage disequilibrium across the MHC, it has been difficult to dissect the contribution of individual genes to disease susceptibility. As a result, intensive efforts have been made to generate mice transgenic for human class II molecules as models of autoimmune disease. However, in every case, additional manipulations—such as immunization with Ag in adjuvant, expression of immunostimulants on target tissues, or coexpression of TCR transgenes—have been required to induce disease. In this study, we show that expression of the human HLA-DQ8 (DQA1*0301/DQB1*0302) molecule alone in three lines of transgenic nonobese diabetic murine class II-deficient (mII−/−) mice results in the spontaneous development of autoimmune myocarditis. The disease shares key features of human myocarditis and was characterized by lymphocytic infiltrates in the myocardium and cardiac myocyte destruction, circulating IgG autoantibodies against cardiac myosin heavy chain, and premature death due to heart failure. We demonstrate that myocarditis could be transferred into healthy HLA-DQ8+RAG-1−/−mII−/− nonobese diabetic recipients with lymphocytes, but not sera. It has been widely thought that autoimmune myocarditis is of infectious etiology, with the immune responses arising secondary to cardiac damage from pathogens. These studies provide direct experimental evidence that spontaneous autoimmune myocarditis can occur in the absence of infection and that expression of HLA-DQ8 confers susceptibility to this organ-specific autoimmune disease.
Myocarditis is a major cause of sudden death in children and young adults (1, 2, 3) and is a frequent precursor of dilated cardiomyopathy, a common indication for cardiac transplantation (4). Myocarditis presents a challenging clinical problem because the etiology is heterogeneous, the diagnosis is difficult to establish, and the treatment options are limited (4, 5). Although cardiotropic viruses are considered to be the most common causative agent, in the majority of cases it has not been possible to isolate viral genome from the myocardium, and the etiology of the disease is therefore unknown (4).
There is increasing evidence suggesting that subgroups of patients with myocarditis may have disease of autoimmune origin. Support for this theory has come from several observations. Patients with myocarditis have an increased prevalence of other autoimmune disorders such as celiac disease (6), an inflammatory disease of the small bowel that is associated with the structurally similar HLA class II molecules DQ8 and DQ2 (7, 8, 9). Circulating autoantibodies against cardiac Ags have been found in patients with dilated cardiomyopathy and their first-degree relatives, suggesting familial clustering characteristic of autoimmune diseases (9). Familial dilated cardiomyopathy is associated with the HLA-DR4/DQ8 haplotype, which is prevalent in other autoimmune diseases (10). In addition, recent studies have shown that patients with myocarditis who have circulating cardiac autoantibodies, but no evidence of viral genome in the myocardium, may benefit from immunosuppression (11). Because myocarditis is difficult to study in humans and the diagnosis is most often made postmortem (4), much of our understanding of the disease pathogenesis has come from animal models. However, to date, all models of autoimmune myocarditis have involved experimental induction of disease, either by infection with cardiotropic viruses or by immunization with adjuvants containing cardiac myosin (12).
In this study, we describe a spontaneous animal model of autoimmune myocarditis. This discovery arose unexpectedly during the course of other studies to make a mouse model to investigate the role of HLA-DQ8 in the development of type 1 diabetes. Genome-wide analyses in humans and mice have shown that susceptibility to organ-specific autoimmune diseases is strongly linked to the MHC class II region. This is exemplified by type 1 diabetes, in which the MHC class II region represents the most important genetic susceptibility locus with a logarithm of odds score of 65.8 (13). However, because of the high degree of linkage disequilibrium within the MHC complex, it has been difficult to analyze the contribution of individual loci to disease susceptibility. For this reason, several groups have generated mice that are transgenic for various HLA class II molecules, including HLA-DQ8, to model human autoimmune diseases. However, in no instance did a spontaneous autoimmune disease phenotype arise that could reproduce a human disease. For example, in an HLA-DQ8 type 1 diabetes model, both the initiation and adoptive transfer of disease were dependent upon the expression of the costimulatory molecule, B7.1, on the target tissues that do not physiologically express costimulatory molecules (14). In HLA-DR and -DQ transgenic models of arthritis, immunization with cartilage Ag in adjuvant was required to induce disease (15, 16, 17, 18). More recently, HLA-DQ transgenic mice have also been evaluated as a potential model for polychondritis (19). In an HLA-DR2 model of multiple sclerosis, coexpression of a myelin basic protein-specific, human TCR transgene was required for spontaneous disease (20). In this study, we show that the expression of a human class II molecule alone—in the absence of immunization, adjuvants, or other transgenes—can serve as a restriction element for murine CD4+ T cells in generating a full-blown spontaneous autoimmune disease syndrome that closely resembles the corresponding human disease.
Materials and Methods
Generation of the HLA-DQ8 transgenic mice
The HLA-DQ8 (DQA1*0301/DQB1*0302) gene was excised from the cosmid clone (pDCβ2 (21); a gift from J. Strominger (Harvard Medical School, Boston, MA)) by digestion with EcoRV/KpnI. The 2.6- and 10.8-kb fragments of DQA1*0301 and the 13-kb ClaI/KpnI and 6.9-kb KpnI fragments of DQB1*0302 were subcloned into separate pCR-Script (Stratagene, La Jolla, CA) vectors. The 13.4-kb DQA1*0301 and the 19.9-kb DQB1*0302 genes were then excised with SrfI/SacII, and SacII, respectively, and then subcloned back together into the pWE15 cosmid vector (BD Biosciences, Palo Alto, CA) that was first modified by adaptors in the BamHI site to create the sites SrfI, KpnI, SacII, and BamHI. The 33.3-kb HLA-DQ8 transgene fragment was excised free of vector sequences by digestion with NotI. The transgene DNA was purified and injected into fertilized eggs of MHC class II-deficient, I-Abβ−/− (hereafter, mII−/−) C57BL/6 (B6) mice (Ref. 22 ; Diabetes Endocrinology Research Center Transgenics Mouse Core Facility, Joslin Diabetes Center).
Three HLA-DQ8 founders (70, 275, and 73) were identified by PCR and Southern blot analysis. Cell surface expression of HLA-DQ8 was analyzed by staining PBLs with a pan anti-HLA-DQ mAb (Leu 10; BD PharMingen, San Diego, CA) and FACS analysis. Genotyping for HLA-DQ8 transmission was performed by PCR with a DQA1 exon 1 primer set: 5′-AACTCTTCAGCTAGTAACTGAG-3′ and 5′-GCTCTGGTTTCCTGAAGGAG-3′. The mII−/− knockout mutation was identified by PCR analysis as described (22). The three HLA-DQ8mII−/− lines (70, 73, and 275) were initially established on the B6 background and were then backcrossed onto the nonobese diabetic (NOD)4 background. Heterozygous carriers of the HLA-DQ8 and neo genes were used as breeders for each backcross generation. At the N2 generation, mice were analyzed with the microsatellite markers 5′-AGCCCCTTCCAAGTGTCTCT-3′ and 5′-GGTTTCGGAATGAGATGAGC-3′ and 5′-TGCTCCTCACCAGTCATGC-3′ and 5′-TAGTGTTTGCCATGGCTCTC-3′, which are linked to insulin-dependent diabetes (idd)3 and idd10, respectively. Segregants homozygous for allelic variants characteristic of NOD served as the progenitors for all mice in subsequent backcross generations. The animals used in the experiments described in this study are the intercrossed HLA-DQ8+/− (hereafter, HLA-DQ8+)mII−/− and HLA-DQ8−mII−/− littermates from the N5 to N8 backcross generations. The possibility that these transgenic mice expressed I-Ag7 was also excluded by negative staining of splenocytes with the I-Ag7-reactive mAb 10-3.6 (BD PharMingen). Because the severity of myocarditis did not increase in any of the three HLA-DQ8mII−/− lines from the N5 to N8 generations, the HLA-DQ8+mII−/− mice from each of the lines were pooled and are referred to as 70, 275, or 73. The nontransgenic (HLA-DQ8−) mII−/− littermates of these mice have been pooled and are referred to as mII−/−.
Recombination-activating gene (RAG)−/− NOD mice were generated by backcrossing the RAG-1−/− mutation, originally on a 129/SvJ background (Ref. 23 ; gift of S. Tonegawa (Massachusetts Institute of Technology, Cambridge, MA)), onto the NOD strain for nine generations after selecting for homozygosity at the idd1, idd3, and idd10 NOD loci. Line 73 HLA-DQ8+RAG−/−mII−/− NOD mice were generated by crossing line 73 HLA-DQ8+mII−/− N8F3 NOD mice with RAG−/− N10F7 NOD mice. Offspring heterozygous at both mII and RAG loci and transgenic for HLA-DQ8 were intercrossed to create HLA-DQ8+RAG−/−mII−/− NOD pups. All mice were maintained in a specific pathogen-free barrier facility in accordance with the guidelines established by the Institutional Animal Care and Use Committee of Harvard Medical School.
Histology and immunohistochemistry
Hearts from mice were fixed overnight in 10% formalin, embedded in paraffin, sectioned, and stained for H&E by conventional methods. Immunohistochemistry was performed on frozen sections as described (24) with biotinylated anti-CD4 (clone H129.19), anti-CD8 (clone 53-6.7), or IVD12 (anti-HLA DQ (7, 8, 9)) followed by peroxidase-conjugated streptavidin (all from BD PharMingen). Stains were developed with either 3-amino-9-ethylcarbazole (Sigma-Aldrich, St. Louis, MO) or 3,3′-diaminobenzidine substrates (BD PharMingen). Double staining was performed with biotinylated IVD12 and either B220 (clone RA3-6B2; BD PharMingen) or F4/80 (clone A3-1; Serotec, Oxford, U.K.), followed by alkaline phosphatase-conjugated streptavidin and peroxidase-conjugated goat anti-rat (both from Jackson ImmunoResearch, West Grove, PA). Stains were developed first with alkaline phosphatase substrate (Vector Laboratories, Burlingame, CA) and then with 3-amino-9-ethylcarbazole substrate (Sigma-Aldrich).
RAG−/− NOD hearts were fixed in 2% paraformaldehyde and embedded in OCT. Frozen sections (4-μm thick) were incubated with 1/50 dilutions of sera from HLA-DQ8+ (70, 275, or 73) or HLA-DQ8−mII−/− NOD mice, followed by detection with FITC-conjugated F(ab′)2 of goat anti-mouse IgG (Jackson ImmunoResearch). Slides were mounted using Vectashield and stained with 4′,6′-diamidino-2-phenylindole (Vector Laboratories) for identification of nuclei. Images were taken on a Zeiss LSM-410 confocal microscope (Zeiss, Oberkochen, Germany).
To avoid interference from endogenous Igs, all tissues were obtained from RAG−/− NOD mice. Myofibrillar extracts were prepared from hearts as previously described (25). Kidney and liver were homogenized in Laemmli sample buffer. Proteins, including porcine cardiac myosin (Sigma-Aldrich), were separated by SDS-PAGE (8%) and transferred onto polyvinylidene fluoride membranes (Millipore, Bedford, MA) overnight. Blots were incubated with sera from line 275, 73 (both at 1:10,000), or 70 (at 1:1,000) mice. This was followed by incubation with peroxidase-labeled F(ab′)2 of goat anti-mouse IgG and development with ECL (Amersham Biosciences, Arlington Heights, IL).
Plates were coated overnight with porcine cardiac myosin at 5 μg/ml. Plates were then washed and incubated with mouse sera starting at a 1/100 dilution with serial four-fold dilutions to 1/6400. Abs were detected with a peroxidase-conjugated F(ab′)2 of goat anti-mouse IgG, with ABTS (Sigma-Aldrich) as substrate. Absorbance was read at 405 nm.
For both cell and serum transfers, donors consisted of line 73 or 275 HLA-DQ8+mII−/− NOD mice that showed symptoms of advanced heart failure (labored breathing, lethargy, and cyanosis) and exhibited cardiomegaly at autopsy. For the cell transfers, splenocytes were depleted of RBC by ACK lysis and were injected i.v. into 6- to 8-wk-old line 73 HLA-DQ8+RAG−/−mII−/− NOD recipients (1.5–2 × 106 cells/recipient). For serum transfers, 300 μl of sera, pooled from the diseased mice that exhibited the highest titers of cardiac myosin Abs (>1:6400), was injected i.v. into 6- to 8-wk-old NOD or RAG−/− NOD recipients. All recipients were checked daily for manifestations of heart failure. At the onset of symptoms or at the termination of the experiments 8 wk after adoptive transfer, mice were euthanized and the hearts were removed for histology. Serum was collected from the recipients before and at weekly intervals after the cell/serum transfers, and at the time of sacrifice.
Comparisons between the Kaplan-Meier survival curves of the different transgenic lines were performed using the log-rank test. All other results are presented as means ± SEM using the unpaired Student’s t test.
Transgenic mice were generated by the microinjection of a 33-kb genomic fragment containing the genes encoding HLA-DQ8 (DQA1*0301/DQB1*0302), along with their endogenous promoters and flanking sequences (21), into embryos of MHC class II-deficient (I-Abβ−/−; hereafter, mII−/−) C57BL/6 (B6) mice (22). Three HLA-DQ8 transgenic lines (70, 275, and 73) were established and maintained on the mII−/− B6 background. In parallel, the HLA-DQ8 transgenes and the mII−/− knockout mutation were backcrossed onto the NOD mouse strain. The transgenic mice were then intercrossed to create mice expressing human HLA-DQ8 as the only class II molecule. The animals used in the experiments described in this study were intercrossed mice from the N5–N8 backcross generations. Because B6 and NOD mice carry the same promoter deletion in the I-Eα gene (26, 27), expression of HLA-DQ8 could be examined in the absence of both I-A and I-E. The possibility that the HLA-DQ8mII−/− transgenic mice expressed hybrid I-Aα-DQB1*0302 molecules was excluded by the lack of staining of splenocytes by FACS with an I-Aαb mAb (AF6-120.1); expression of endogenous I-Ab proteins was also biochemically excluded by the lack of signals after immunoprecipitation of [35S]methionine-labeled splenocytes with the mAb Y-3P (data not shown).
Immunohistochemical analysis of frozen thymic sections revealed that lines 73 and 275 displayed similar amounts of HLA-DQ8 staining in the thymic cortex, which was significantly greater than that seen in line 70 (Fig. 1,A). Consistent with these findings, line 275 and 73 mice showed near or complete restoration of the peripheral CD4+ T cell compartment (Fig. 1,B) with mean CD4:CD8 ratios of 2.5 ± 0.15 (n = 17) and 3.0 ± 0.13 (n = 14), respectively, similar to levels seen in control NOD mice (3.0 ± 0.08; n = 17). In contrast, CD4:CD8 ratios in line 70 (Fig. 1,B) were ∼4- to 5-fold lower at 0.61 ± 0.04 (n = 10; p < 0.0001, compared with line 73 or 275). As expected, mII−/− mice that lacked HLA-DQ8 (Fig. 1 B) had minimal CD4+ T cells with a mean CD4:CD8 ratio of 0.09 ± 0.2 (n = 10; p < 0.0001, compared with line 70). Cytofluorimetric analysis of splenic B220+ cell populations showed less striking differences between the lines with HLA-DQ8 expressed on similar percentages of B cells from lines 70 and 275, with mean levels of 40 ± 3.3% (n = 8) and 50 ± 4.2% (n = 9), respectively, compared with line 73, which showed mean levels of 76 ± 4.2% (n = 5). Immunization of HLA-DQ8+(70, 275, and 73)mII−/− mice with the known HLA-DQ8-restricted peptides HSV-2 VP16 (28) and insulin B9–23, revealed that line 70 and 275 mice made vigorous proliferative responses against both peptides with stimulation indices 7- to 20-fold above baseline, but weaker responses were noted in line 73, due to lower levels of HLA-DQ8 expression on dendritic cells (data not shown).
Unexpectedly, as lines 73 and 275 were bred onto the NOD background, the HLA-DQ8+mII−/− mice developed the acute onset of labored breathing, lethargy, edema, and cyanosis, symptoms consistent with congestive heart failure (Fig. 2,A). This was associated with a high incidence of premature sudden death, starting at ∼12 wk of age. By 40 wk of age, >50% of line 73 and 275 HLA-DQ8+mII−/− NOD mice died prematurely, compared with <5% of nontransgenic HLA-DQ8−mII−/− NOD control littermates (Fig. 2,B). Autopsy examination of the diseased mice revealed massively enlarged hearts with ischemic discoloration, varying degrees of pleural effusions, and enlargement and darkening of the liver. Surprisingly, marked cardiac enlargement was also seen in the majority of healthy appearing line 73 and 275 mice that survived to 40 wk (Fig. 2,C). Histological analysis of the hearts (Fig. 3,A) revealed extensive mononuclear infiltrates in the atria (RA), accompanied by cardiac myocyte degeneration and vacuolization, histopathological findings consistent with the diagnosis of active myocarditis in humans (29). Adjacent to the damaged tissues, massive mural thrombi were often observed (Fig. 3,A, t). The ventricles (Fig. 3,A, RV) were focally infiltrated, with line 73 exhibiting more severe pathology than line 275. Histological examination of the enlarged livers showed pathological changes characteristic of severe cardiac insufficiency with centrilobular hepatic necrosis and hemorrhage. Immunohistochemical analyses of the cellular composition of the ventricular lesions (Fig. 3,B) revealed large numbers of CD4+ and CD8+ T cells along with abundant F4/80+ macrophages, with only a few B cells present which often clustered in areas containing CD4+ T cells (B, arrow). A striking feature of the inflammatory heart lesions was the marked induction of MHC class II Ag expression (Fig. 3,B), potentially serving to amplify responses of primed CD4+ T cells in situ. Based on colocalization studies with F4/80, it also appeared that some of the recruited macrophages expressed HLA-DQ8 (Fig. 3 B). The de novo expression of MHC class II Ags on inflamed target cells is also a characteristic feature of human myocarditis lesions (4, 30).
In contrast to lines 73 and 275, line 70 HLA-DQ8+mII−/− NOD mice, even when followed up to the N8F1 generation, did not develop clinical symptoms of congestive heart failure or show significant premature mortality (Fig. 2,B). However, the majority of mice showed histological evidence of myocarditis with lymphocytic infiltrates and cardiac myocyte necrosis, especially in the atria (Fig. 3,A), indicating that the same basic disease process was occurring in all three lines. The development of myocarditis was HLA-DQ8- and NOD strain-specific, because it did not develop in nontransgenic mII−/− NOD mice, heterozygous HLA-DQ8+mII+/− NOD mice, or HLA-DQ8+mII−/− mice on the B6 background. We have recently generated fully congenic line 73 N10 HLA-DQ8+mII−/− NOD mice (n = 56) and line 275 N10 HLA-DQ8+mII−/− NOD mice (n = 37), and by Kaplan-Meier survival analysis and log-rank sum tests, there was no statistical difference in the time of onset or frequency of fatal myocarditis in either of these congenic HLA-DQ8 NODmII−/− lines compared with that of the N5–8 generation 275 and 73 mice shown in Fig. 2,B. Thus, the myocarditis phenotype we observed cannot be attributed to background genes other than those from the NOD strain. The immune-mediated damage was cardiac muscle specific, because infiltrates were not seen in skeletal (Fig. 4,A) or smooth muscle except in the pulmonary vein, which, in rodents, contains cardiac muscle. All three lines showed insulitis (represented in Fig. 4 B), but the extent of pancreatic β cell damage was not sufficient to result in clinical diabetes or the production of autoantibodies (data not shown). Infiltration of the salivary gland (sialitis) was also present but occurred with similar frequency and severity in nontransgenic HLA-DQ8−mII−/− NOD mice, consistent with its known control by non-MHC susceptibility genes (31).
Myocarditis was associated with the production of circulating IgG autoantibodies to cardiac tissues in all three lines of HLA-DQ8+mII−/− NOD mice. By indirect immunofluorescence and confocal microscopy on RAG−/− NOD heart sections, the staining pattern localized to the striations within myocytes, producing a distinctive myofibrillar pattern (Fig. 5,A). Although the prevalence and intensity of staining varied in mice from the different lines, the staining pattern was identical, suggesting that the autoantibodies were directed against similar antigenic determinants (Fig. 5,A). This was confirmed by immunoblot analysis (Fig. 5,B), which revealed that sera from HLA-DQ8+(70, 275, and 73)mII−/− mice, but not from NOD or control HLA-DQ8−mII−/− mice (data not shown), reacted predominantly against a ∼200-kDa protein in heart extracts that corresponded with the heavy chain of cardiac myosin, a major autoantigen in human myocarditis (25). Reactivity was not detectable in other tissues such as liver or kidney, but a ∼200-kDa protein was seen in skeletal muscle that corresponded to skeletal muscle myosin (data not shown). This cross-reactivity was not unexpected, because the α-cardiac and β-skeletal muscle myosin H chain isoforms in rodents are highly (>97%) homologous (32); similar cardiac and skeletal muscle Ab cross-reactivity is also seen in human myocarditis (33). These findings sharply contrasted with the cardiac specificity of this autoimmune disease process, because autoimmune infiltrates developed in the heart (Fig. 3), but not in skeletal muscle (Fig. 4,A). Comparison of the titers of cardiac myosin IgG autoantibodies by ELISA (Fig. 5 C), revealed that line 73 mice exhibited the highest prevalence and titers of autoantibodies (16 of 18 positive), followed by line 275 (12 of 18 positive), and finally, line 70 (5 of 24 positive), with no Abs detectable in control DQ8−mII− NOD mice (n = 16). In separate studies, we found that mice that exhibited macroscopic evidence of cardiomegaly (mean heart weight, 337 ± 16 mg (n = 9); compared with a mean heart weight, 154 ± 5.9 mg (n = 18) in HLA-DQ8−mII−/− littermates) exhibited titers >1:1600. Thus, the titers of the myosin autoantibodies appeared to serve as a good serological marker of disease severity in the different HLA-DQ8 transgenic lines and in individual mice. High-titer autoantibodies (>1:6400) were detectable in mice as young as 6 wk of age, and were highly predictive of early progression to fatal myocarditis.
Myocarditis did not develop in lymphocyte-deficient line 73 HLA-DQ8+mII−/−RAG−/− NOD mice (n = 6), confirming that the development of disease was dependent on the function of mature T and/or B lymphocyte subsets. Importantly, we found that myocarditis could be successfully transferred into healthy line 73 HLA-DQ8+mII−/−RAG−/− NOD mice by the i.v. injection of splenocytes (1.5–2.0 × 107) from affected donors. Starting at ∼3 wk after transfer, the recipients (n = 4) developed the acute onset of symptoms of severe heart failure accompanied by production of high-titer circulating autoantibodies against cardiac myosin. Histological analysis of the hearts revealed severe myocarditis (Fig. 6,A) with extensive infiltration of lymphocytes into the myocardium and cardiac myocyte destruction (Fig. 6,A, insets), reproducing the disease of the original donor mice. In contrast, the i.v. injection of serum (300 μl, i.v.) from severely diseased donors with the highest titers of anti-cardiac myosin Abs (>1:6400) into either immunocompetent NOD mice (n = 4) or RAG−/− NOD (n = 4) mice, did not cause myocarditis as assessed by the lack of clinical symptoms and histological changes in the heart (Fig. 6 B), and the steady decline in Ab titers to undetectable levels (data not shown). This cell transfer model contrasts with virus- and cardiac myosin-induced immunization models of myocarditis, in which the transfer of disease usually requires either in vitro activation of donor splenocytes or pretreatment of the recipients with immunoadjuvants (12, 34, 35). Our HLA-DQ8+RAG−/−mII−/− NOD transfer system should thus provide a useful in vivo assay for further identifying the relevant cell type(s) responsible for this spontaneous autoimmune disease process.
We describe a spontaneous animal model of autoimmune myocarditis that shares many clinical, histological, and immunological characteristics with human myocarditis (4, 12). Like humans, HLA-DQ8+mII−/− mice develop the sudden onset of heart failure, leading to their early demise. This was associated with the presence of inflammatory cellular infiltrates in the myocardium and adjacent myocyte damage, histological features reminiscent of active myocarditis in humans (29). The inflamed lesions also contained abundant macrophages and CD4+ and CD8+ T cells, and exhibited marked induction of MHC class II Ag expression, which are also characteristic of human myocarditis lesions (4, 12). In addition, the diseased mice developed high-titer IgG autoantibodies that were specifically targeted against cardiac myosin heavy chain, a major autoantigen in human myocarditis (12). Despite their differing levels of HLA-DQ8 expression, all three transgenic lines developed the same basic disease process, indicating that HLA-DQ8 plays a fundamental role in the observed spontaneous autoimmune disease. The autoimmune phenotype of this model sharply contrasts with that described in the dilated cardiomyopathy syndrome of mice lacking the negative immunoregulatory receptor PD-1, which was characterized by a noninflammatory, fibrotic reaction in the heart and the production of pathogenic autoantibodies against a 33-kDa protein that was recently identified to be cardiac troponin I (36, 37).
This study shows that a human MHC class II gene can mediate a spontaneous disease process in the absence of other transgenes or immunization. While this manuscript was under consideration, an article describing the development of autoimmune myocarditis in double-transgenic human CD4/HLA-DQ8mII−/− NOD mice appeared (38). However, only one transgenic NOD line was described, raising the possibility that this phenotype may have been caused by founder effects, e.g., due to the transgene insertion site (38). In addition, the expression of the HLA-DQ8 transgene was examined on only a single genetic background. We show that the expression of HLA-DQ8 alone in three independently derived transgenic lines results in the spontaneous development of autoimmune myocarditis, with the severity of disease correlating with levels of expression of the HLA-DQ8 transgene. In addition, we show that autoimmune myocarditis develops on the NOD—but not on the B6—genetic background, suggesting that both MHC and non-MHC background genes are necessary for disease expression (39, 40). These findings are consistent with studies on human type 1 diabetes that have shown that, although HLA-DQ8 is relatively common in the general population, only a small proportion of carriers, presumably those individuals with additional non-MHC susceptibility genes, develop disease (41, 42). Interestingly, despite the autoimmune propensity of the NOD strain and the close structural and functional similarities of HLA-DQ8 and I-Ag7 molecules (43, 44), NOD mice do not normally develop myocarditis. However, peptide binding studies have shown that, although the binding specificities of I-Ag7 and HLA-DQ8 are similar, they are not identical, because there are naturally processed peptides from NOD mice that do not bind to HLA-DQ8, and vice versa (43).
Similar to previous studies (45, 46), none of our three lines of HLA-DQ8mII−/− NOD mice developed spontaneous diabetes, and NOD mice heterozygous for I-Ag7 were also profoundly resistant to disease (47), likely a result of the inhibitory effects of the H-2b haplotype-linked genes that were introgressed intothe NOD background from the original mII−/− mutation made in129/Sv-derived ES cells (48). Interestingly, although line 70 HLA-DQ8+I-Ag7+/− NOD mice were also protected from diabetes, both line 275 and 73 HLA-DQ8+I-Ag7+/− NOD mice showed a markedly enhanced incidence of spontaneous diabetes (data not shown). Thus, line 73 and 275 NOD mice displayed enhanced susceptibility to two different organ-specific autoimmune disease processes. In this context, it is interesting to note that autoimmune myocarditis and diabetes in NOD mice share a number of features. Both conditions are MHC class II dependent but require additional non-MHC background genes for disease expression. Both diseases are associated with circulating organ-specific Abs that precede the onset of disease and serve as markers of the ongoing disease process, but are primarily T cell mediated. It will be of considerable interest to determine whether any of the key non-MHC idd loci that have been linked to susceptibility to type 1 diabetes (49) will also be associated with susceptibility to autoimmune myocarditis. Bottazzo and colleagues (9) have previously drawn parallels between the autoimmune features of human type 1 diabetes and those of familial dilated cardiomyopathy.
Autoimmunity has long been linked to myocarditis, but it has been postulated that the autoimmune responses are initiated by cardiac damage from infectious pathogens (12, 50). The results presented in this study suggest that some forms of myocarditis may be of primary autoimmune origin and mediated by HLA-DQ8 molecules. Interestingly, it has been shown that cardiac myosin-MHC class II complexes are normally constitutively expressed on residential cardiac APCs of healthy mice, suggesting that cell injury is not required for the exposure of cardiac Ags to the immune system (51). In addition, recent studies have suggested that cardiac myocytes undergo a remodeling process involving apoptosis and regeneration that plays an important role in normal cardiac development and tissue homeostasis (52). This raises the intriguing possibility that the physiological turnover of apoptotic cardiac myocytes may serve as an additional source of Ag to activate autoreactive CD4+ T cells as has been hypothesized to occur for pancreatic β cells in the initiation of autoimmune diabetes (53, 54). Our hypothesis is that HLA-DQ8 shapes the T cell repertoire in the thymus such that significant numbers of CD4+ T cells specific for cardiac Ags are selected that initiate this severe organ-specific inflammatory disease in the absence of infection with cardiotropic viruses. In keeping with this, the HLA-DQ8 transgenic lines with the highest levels of HLA-DQ8 expression in the thymic cortical epithelial cells and the largest numbers of peripheral CD4+ T cells develop the most aggressive form of the disease.
Despite extensive investigation, it has not yet been possible to identify how HLA-DQ8 molecules function in conferring susceptibility to human autoimmune diseases. However, given the close resemblance of spontaneous myocarditis in HLA-DQ8 transgenic NOD mice to human myocarditis, a detailed cellular and molecular analysis of this model should enhance our understanding of the role of HLA-DQ8 molecules in pathogenesis of this disease. This knowledge could lead to improved detection and treatment of this serious human condition.
We thank E. Boschetti, D. Neely, C. Cahill, and H. Thomas for their excellent technical help, and T. Lund and H. Wolpert for critically reading the manuscript.
This work was supported by National Institutes of Health Grant PO1 AI39619 (to K.W.W. and M.A.L.), the Iacocca Foundation (to M.F.M.), National Institutes of Health Grant P30 DK36836 (to Joslin Diabetes Center), and the Juvenile Diabetes Research Foundation (to M.A.L. and K.W.W.).
Abbreviations used in this paper: NOD, nonobese diabetic; RAG, recombination-activating gene; idd, insulin-dependent diabetes.