Type 1 diabetes (T1D) is characterized by T cell–mediated destruction of the insulin-producing β cells of the pancreatic islets. Among the loci associated with T1D risk, those most predisposing are found in the MHC region. HLA-B*39:06 is the most predisposing class I MHC allele and is associated with an early age of onset. To establish an NOD mouse model for the study of HLA-B*39:06, we expressed it in the absence of murine class I MHC. HLA-B*39:06 was able to mediate the development of CD8 T cells, support lymphocytic infiltration of the islets, and confer T1D susceptibility. Because reduced thymic insulin expression is associated with impaired immunological tolerance to insulin and increased T1D risk in patients, we incorporated this in our model as well, finding that HLA-B*39:06–transgenic NOD mice with reduced thymic insulin expression have an earlier age of disease onset and a higher overall prevalence as compared with littermates with typical thymic insulin expression. This was despite virtually indistinguishable blood insulin levels, T cell subset percentages, and TCR Vβ family usage, confirming that reduced thymic insulin expression does not impact T cell development on a global scale. Rather, it will facilitate the thymic escape of insulin-reactive HLA-B*39:06–restricted T cells, which participate in β cell destruction. We also found that in mice expressing either HLA-B*39:06 or HLA-A*02:01 in the absence of murine class I MHC, HLA transgene identity alters TCR Vβ usage by CD8 T cells, demonstrating that some TCR Vβ families have a preference for particular class I MHC alleles.
Type 1 diabetes (T1D) is characterized by T cell–mediated destruction of insulin-producing β cells (1). Both CD4 and CD8 T cells are important for T1D pathogenesis, with CD8 T cells requiring the presentation of β cell epitopes by class I MHC molecules to interact with and eliminate the β cells (2, 3). It is thus unsurprising that although multiple genetic loci have been found to contribute to T1D development, those most predisposing to T1D can be found in the MHC region (4). Several class I MHC alleles have been found to be predisposing to T1D, including HLA-A*02:01 and HLA-B*39:06 (5–9). Although the presentation of β cell epitopes by HLA-A*02:01 has long been known and extensively studied (10), HLA-B*39:06 has only more recently gained attention as a T1D-associated allele, and much remains to be understood about its ability to confer T1D risk.
Although T1D associations have been observed at all HLA class I loci (9), HLA-B*39:06 is the most predisposing HLA class I allele (7, 9) and, importantly, is associated with an early age of onset (11). Furthermore, HLA-B*39:06 is most common among the Latin American population (12), in which T1D incidence has been rising (13–15). Development of an HLA-B*39:06–transgenic mouse model is thus of the utmost importance to understand the relationship between HLA-B*39:06, genetic risk background, and T1D pathogenesis. A transgenic model is also essential for the preclinical testing of HLA-B*39:06–targeted treatments.
Given the multiple risk factors associated with T1D predisposition, it is important to study HLA-B*39:06 in a translationally relevant manner. The NOD mouse is considered by many to be a good model for human T1D (16, 17). For example, the NOD class II MHC H2-Ag7 shares striking similarities with several T1D-associated human class II MHC alleles, such as HLA-DQ8 (18). Among other similarities, both NOD mice and human T1D patients display reduced regulatory T cell function and reduced IL-2 signaling (17, 19, 20). Most importantly, T cells from HLA-transgenic NOD mice may target similar or even identical β cell epitopes to those found in T1D patients (21–23). However, to most accurately model HLA-B*39:06 in the context of human T1D, it is preferable to incorporate additional human non-MHC risk alleles. In humans, the non-MHC locus that confers the most susceptibility to T1D is the variable number of tandem repeats (VNTR) region of the insulin gene (24–26). Shorter VNTR sequences are known as class I, whereas longer VNTR sequences are known as class III. Class I VNTR sequences are associated with T1D risk and with a decrease in thymic insulin mRNA levels compared with the longer class III VNTR alleles, which are protective (24). The decrease in thymic insulin expression correlates with impaired negative selection of high-avidity insulin-specific T cells in humans (27).
The reduced thymic insulin expression associated with class I VNTR sequences in humans has been modeled in mice through introduction of two Insulin 2 (Ins2) knockout (KO) alleles (28–32). Mice possess two insulin genes, Insulin 1 (Ins1) and Ins2. Although expressed in the pancreas, little (28, 33) to no (34, 35) Ins1 expression occurs in the thymus. In contrast, Ins2 is expressed in both the thymus and the pancreas (28). Upon Ins2 ablation, both NOD mice and mouse strains not prone to T1D exhibit diminished T cell tolerance to insulin, as evidenced by enhanced T cell reactivity to insulin and insulin-derived peptides (28–30, 32). We have shown that NOD mice even just heterozygous (Het) for the Ins2KO allele exhibit decreased thymic insulin expression, as seen in human T1D patients (36). In the context of HLA-A*02:01, we have previously found that NOD mice with reduced thymic insulin expression display increased T1D incidence, islet infiltration, and CD8 T cell responses to insulin-derived peptides (30, 36), speaking to the importance of examining multiple risk alleles simultaneously.
In this study, we have developed HLA-B*39:06–transgenic NOD mouse models and have demonstrated that HLA-B*39:06 is able to independently mediate the development of CD8 T cells required for T1D onset. We found that in the context of reduced immunological tolerance to insulin (i.e., Ins2 ablation), HLA-B*39:06–transgenic NOD mice develop T1D at an accelerated rate compared with mice with wild-type (WT) thymic insulin expression. We excluded secondary causes for the enhanced disease by verifying that blood insulin levels were normal even though Ins2 was ablated, as has been observed elsewhere (37), and that no gross alterations in lymphocyte composition or TCR Vβ family usage were exhibited. Rather, with a decrease in thymic insulin expression, HLA-B*39:06 will be less able to negatively select insulin-specific CD8 T cells. Thus, by generating HLA-B*39:06–transgenic NOD mice in the presence of reduced thymic insulin expression, we show the development of models that will provide excellent tools for the examination of HLA-B*39:06’s impact on T1D and for the preclinical testing of HLA-B*39:06–targeted therapies.
Materials and Methods
To develop HLA-B*39:06–transgenic NOD mice, we prepared a monochain chimeric HLA-B*39:06 construct, comprising the α1- and α2 peptide–binding domains of HLA-B*39:06 linked to the α3 CD8–binding and transmembrane domains of H2-Db, with human β2-microglobulin (β2m) linked covalently to the α1 domain. Chimeric constructs of this design are designated as human β2m/HLA/H2-Db (HHD) (38). This HLA-B*39:06 HHD construct was injected into NOD zygotes, and founder mice were identified by PCR of tail-tip DNA using these HLA-B*39:06 primers: 5′-CTTCATCTCAGTGGGCTAC-3′ and 5′-CGGTCAGTCTGTGTGTTGG-3′. Positive progeny were further assessed for HLA-B*39:06 expression on their peripheral blood leukocytes by flow cytometry using anti–HLA-A, -B, and -C (W6/32; BioLegend). Founder 45, with the highest expression of HLA-B*39:06, was selected for further investigation and was crossed with an NOD mouse. Progeny of this cross were assessed for the presence of the transgene by PCR of tail-tip DNA; mice hemizygous (Hemi) for the transgene were designated NOD.HLA-B*39:06Hemi. To maintain this strain, NOD.HLA-B*39:06Hemi mice were crossed with NOD littermates. NOD.HLA-B*39:06Hemi females were also crossed with male mice from NOD.β2mKO (39) or NOD.β2mKO.Ins2KO strains (36). To fix the HLA-B*39:06 transgene to homozygosity, the resulting progeny were interbred as appropriate to generate HLA-B*39:06 homozygous (Hom) mice (HLA-B*39:06Hom) with the following genotypes: NOD.HLA-B*39:06Hom.β2mKO, NOD.HLA-B*39:06Hom.β2mKO.Ins2Het, and NOD.HLA-B*39:06Hom.β2mKO.Ins2KO. As we and others have found that female NOD.β2mKO mice breed poorly (40), we crossed male β2mKO mice with female β2mHet mice whenever possible. The WT and KO β2m and WT and KO Ins2 alleles were identified by PCR using the following primer pairs: 5′-GAAACCCCTCAAATTCAAGTATACTCA-3′ and 5′-GACGGTCTTGGGCTCGGCCATACT-3′ (β2mWT), 5′-GAAACCCCTCAAATTCAAGTATACTCA-3′ and 5′-TCGAATTCGCCAATGACAAGACGCT-3′ (β2mKO), 5′-GGCAGAGAGGAGGTGCTTTG-3′ and 5′-AGAAAACCACCAGGGTAGT_TAGC-3′ (Ins2WT), and 5′-GGCAGAGAGGAGGTGCTTTG-3′ and 5′-ATTGACCGTAATGGGATAGG-3′ (Ins2KO). NOD.HLA-A*02:01 HHD.β2mKO mice have been previously described (21).
Assessment of HLA-B*39:06 homozygosity by real-time PCR
Mouse tails were numbed with ethyl chloride (Gebauer) and the tail tips were removed. Tails were digested in 200 μl proteinase K (Roche) solution overnight at 56°C. The reaction was stopped by placing tails at 95°C for 10 min. The resultant DNA (1 μl) was mixed with PrimeTime Gene Expression Master Mix (Integrated DNA Technologies) and each of the following primers and TaqMan probes: HLA-B*39:06 primers (5′-TTCATCTCAGTGGGCTACG-3′ and 5′-TGTGTTCCGGTCCCAATATTC-3′) and probe [5′-[(6-FAM)-TCGCTGTCGAACCTCACGAACTG-(Zen probe with Iowa Black)-3′] and internal positive control primers (5′-CACGTGGGCTCCAGCATT-3′ and 5′-TCACCAGTCATTTCTGCCTTTG-3′) and probe [5′-(Cy5)-CCAATGGTCGGGCACTGCTCAA-(Black Hole Quencher 2)-3′]. Real-time quantitative PCR was performed in triplicate using an iQ5 real-time PCR detection system (Bio-Rad). Amplification was carried out as follows: initial denaturing at 94°C for 2 min, followed by 38 cycles of 20 s at 94°C, 15 s at 60°C, and 10 s at 72°C. Copy numbers were calculated using the 2ΔΔCt method.
Assessment of T1D
Mice were monitored weekly from 4 to 30 wk for glucosuria using Diastix reagent strips (Bayer). Mice were considered diabetic following two consecutive positive tests. The first positive test was recorded as the date of diabetes onset.
Pancreata were fixed in Bouin’s solution, sectioned at three nonoverlapping levels, and stained with aldehyde fuchsin and H&E. Islets were scored for insulitis by a blinded observer as previously described (41): 0, no visible lesions; 1, peri-insular or noninvasive leukocytic aggregates; 2, <25% islet destruction; 3, 25–75% islet destruction; 4, >75% islet destruction. A mean insulitis score was determined for each mouse by dividing the total score for each pancreas by the total number of islets examined. Diabetic mice were assigned a score of 4.
Blood collection and staining of peripheral blood leukocytes
Blood (10 μl) was collected from the mouse tail vein and added to 50 μl PBS (pH 7.2; Life Technologies) with 1 mM EDTA (Sigma-Aldrich). Samples were mixed well and erythrocytes were lysed for 2–3 min with 200 μl ammonium chloride–potassium (ACK) lysis buffer (Lonza). Plates were centrifuged at 700 × g for 3 min and ACK lysis was repeated. Following centrifugation, samples were washed twice with PBS containing 1% FBS (HyClone) and 0.1% (w/v) sodium azide. All subsequent washes and dilutions were performed using this buffer. Cells were stained with Fc Block (BD Biosciences), followed by anti-CD8α (53-6.7; BD Biosciences) and anti–HLA-A, -B, and -C (B9.12.1; Beckman Coulter) and incubated on ice for 15–20 min. Samples were washed twice, suspended in 1 μg/ml DAPI, and incubated on ice for 15–30 min. Samples were filtered through a 35-μm cell strainer prior to data collection on a BD LSR II flow cytometer with five lasers (355, 405, 488, 561, and 640 nm). Data were analyzed using FlowJo software (version 8.8.6).
Serum collection and insulin ELISA
Blood (20–40 μl) was collected from the mouse tail vein and allowed to clot at room temperature for 1 h. Samples were centrifuged for 15 min at 960 × g at 4°C. Serum was stored in aliquots at −20°C. Blood insulin levels were measured using the Mouse Ultrasensitive Insulin ELISA (ALPCO). Absorbance of each well at 405 nm was detected using an Emax precision microplate reader (Molecular Devices), and the results were analyzed using GraphPad Prism 7 software.
Preparation of splenocytes and pancreatic lymph node cells and flow cytometry
Mice were euthanized using CO2 asphyxiation, followed by cervical dislocation. Spleens (and for some experiments, pancreatic lymph nodes) were harvested and placed in ice-cold RPMI 1640 (Life Technologies) supplemented with 10% FBS, 1% sodium pyruvate (Life Technologies), 1% nonessential amino acids (Life Technologies), 50 U/ml penicillin, and 50 μg/ml streptomycin (Life Technologies). Organs were crushed, passed through a 40-μm cell strainer, and washed with RPMI. Samples were centrifuged at 486 × g for 5 min. Erythrocytes were lysed in ACK lysis buffer (Lonza) for 4 min at room temperature and washed with RPMI. The resultant cells were centrifuged and washed twice with PBS. Prior to the final wash, samples were passed through a 40-μm cell strainer. Cells were counted and suspended in PBS. Samples prepared in the above manner were added to a V-bottom plate and centrifuged at 486 × g for 5 min. Samples were washed once in PBS containing 2% FBS (HyClone). This buffer was used for all subsequent washing and dilution steps. Cells were stained with Fc Block (BD Biosciences) on ice for 10 min and washed once. For monitoring of class I MHC expression, cells were incubated on ice for 20 min with labeled anti–HLA-A, -B, and -C (B9.12.1; Beckman Coulter), anti-pan murine class I MHC (M1/42; The Jackson Laboratory), or an appropriate isotype control Ab (mouse IgG2a for B9.12.1 and rat IgG2a/κ for M1/42). For analysis of splenic immune cell populations, cells were stained with labeled anti-CD19 (6D5; BioLegend), anti-TCRβ (H57-597; BD Biosciences), anti-CD8α (53-6.7; BD Biosciences), anti-CD4 (GK1.5; BD Biosciences), and anti-CD25 (PC61.5; eBioscience). For study of TCR Vβ usage, an anti-mouse TCR Vβ screening panel was used (BD Biosciences) in conjunction with labeled anti-CD19 (6D5; BioLegend), anti-CD3ε (145-2C11; BD Biosciences), anti-CD8α (53-6.7; BD Biosciences), anti-CD4 (GK1.5; BD Biosciences), and anti-CD25 (PC61.5; eBioscience). Samples were washed twice, incubated in 1 μg/ml DAPI for 15 min on ice, and filtered through a 35-μm cell strainer prior to data collection. Data were collected on a BD LSR II flow cytometer with five lasers (355, 405, 488, 561, and 640 nm) and analyzed using FlowJo (version 8.8.6) and GraphPad Prism 7 software.
NOD mice transgenic for HLA-B*39:06 are susceptible to T1D
To begin to study the association of HLA-B*39:06 with T1D, we first developed NOD.HLA-B*39:06 mice using a monochain HLA-B*39:06 construct. We tracked these mice for susceptibility to disease to ensure that the integration of HLA-B*39:06 did not interfere with T1D development. We found no decrease in disease susceptibility compared with non-transgenic littermates in either females (Fig. 1A) or males (Fig. 1B). The earliest age of onset among female mice was 14 wk, with 82% diabetic by 30 wk. In males, the earliest age of onset was at 13 wk, although as expected, incidence was reduced compared with females, with only 57% converting to disease by 30 wk of age. Because females were more susceptible to disease than males, we used female mice for our subsequent experiments.
HLA-B*39:06 allows for the selection of CD8 T cells in NOD mice
To examine the influence of HLA-B*39:06 on T1D without the complicating factor of the concomitant expression of murine class I MHC molecules, we developed a model in which the transgenic HLA-B*39:06 was expressed in the absence of murine β2m by breeding with the NOD.β2mKO strain (39). Because the transgenic HLA-B*39:06 HHD molecules contain covalently bound human β2m, HLA-B*39:06 can fold without reliance on murine β2m, whereas the endogenous H2-Kd and H2-Db cannot. To maximize the expression of HLA-B*39:06 and the thymic selection of CD8 T cells, we sought to fix the HLA-B*39:06 transgene to homozygosity (HLA-B*39:06Hom). To do so, we first examined the level of human class I MHC on peripheral blood leukocytes from female NOD.HLA-B*39:06.β2mKO mice (Fig. 2A). Although all mice tested expressed human class I MHC, there appeared to be two groups of mice: one with high levels of class I MHC, with an average geometric mean fluorescence intensity (MFI) of 1534, and one with low class I MHC levels, with an average MFI of 607, suggesting that the mice with increased human class I MHC levels were HLA-B*39:06Hom. We used real-time PCR for the HLA-B*39:06 transgene to ensure that the high expressers were indeed Hom for HLA-B*39:06 (Fig. 2B). We found that the average copy number of the low expressers was 1.6. This value was consistent with previous experiments with NOD.HLA-B*39:06.β2mKO mice that were known to be Hemi (data not shown), confirming that the mice with low levels of human class I MHC were HLA-B*39:06Hemi. Mice with high levels of human class I MHC had a copy number of 3.1, nearly double what was seen in the HLA-B*39:06Hemi mice and indicating that these mice were, in fact, Hom for HLA-B*39:06. We hypothesized that HLA-B*39:06Hom mice would be capable of developing increased amounts of CD8 T cells relative to HLA-B*39:06Hemi mice. We therefore examined the percentage of blood CD8 T cells in female NOD.HLA-B*39:06.β2mKO mice (Fig. 2C). HLA-B*39:06Hemi mice had 1.5% CD8 T cells among their peripheral blood leukocytes, whereas HLA-B*39:06Hom mice had nearly double that amount with 2.5% CD8 T cells, indicating that increased HLA-B*39:06 expression can mediate the development of a higher percentage of CD8 T cells. Mice Hom for HLA-B*39:06 were used for all subsequent experiments.
Having observed CD8 T cells in the peripheral blood of NOD.HLA-B*39:06.β2mKO mice, we next sought to confirm the lack of cell-surface expression of murine class I MHC on splenocytes using the pan murine class I MHC Ab M1/42. Spleens from NOD and NOD.β2mKO mice and the previously characterized NOD.HLA-A*02:01.β2mKO strain (21) were also examined. As expected, only NOD splenocytes showed expression of murine class I MHC (Fig. 3A). The absence of murine class I MHC in NOD.β2mKO mice results in a lack of CD8 T cells (Fig. 3B, 3C), as reported previously (39). However, we observed a partial restoration of CD8 T cell development in the NOD.HLA-B*39:06.β2mKO strain (Fig. 3B, 3C), demonstrating that HLA-B*39:06 is indeed able to mediate CD8 T cell development.
HLA-B*39:06 mediates T1D in NOD mice lacking murine β2m
We next examined the ability of HLA-B*39:06 to mediate the development of T1D. NOD.β2mKO mice are protected from T1D because they lack CD8 T cells (39, 42–44). However, Hom expression of HLA-B*39:06 in NOD.β2mKO mice partially restored a disease phenotype (Fig. 4A), with the earliest age of onset at 20 wk and with 17% of NOD.HLA-B*39:06.β2mKO mice diabetic at 40 wk. We therefore show in this study for the first time, to our knowledge, that HLA-B*39:06 is able to independently lead to the development of T1D in mice. As previously reported (21), Hom expression of HLA-A*02:01 (HHD) also allowed for partial restoration of disease susceptibility (Fig. 4A). The incidence curves for the two strains were statistically indistinguishable (p = 0.17) (Fig. 4A), and examination of insulitis in nondiabetic mice of each strain revealed similar amounts of islet infiltration (p = 0.15) (Fig. 4B). Representative islets from HLA-A*02:01 and HLA-B*39:06 mice are shown in Fig. 4C and 4D, respectively. Consistent with previous results (36), the majority of NOD.HLA-A*02:01.β2mKO mice displayed insulitis (Fig. 4B). Similarly, histological examination of islets from 40-wk-old NOD.HLA-B*39:06.β2mKO mice revealed that despite not all progressing to overt T1D, all mice displayed some degree of insulitis, with 81% of mice fully infiltrated (Fig. 4B).
Decreased thymic insulin expression results in earlier T1D onset in HLA-B*39:06–transgenic mice
NOD.Ins2KO mice are characterized by greatly diminished thymic insulin expression (32). Compared with WT littermates, they display accelerated T1D onset, increased insulitis, and impaired immunological tolerance to insulin manifested as increased T cell reactivity to insulin and insulin-derived peptides (30–32). The impact of Ins2 deficiency on disease is dependent on the genetic context, as NOD.HLA-A*02:01.β2mKO.Ins2KO mice have a faster disease onset than NOD.Ins2KO mice (45), indicating that the effects of multiple risk alleles can combine to increase risk. We find that in conjunction with HLA-B*39:06, Ins2 deficiency leads to a rapid onset of disease (Fig. 5), demonstrating the importance of examining T1D in the context of multiple risk factors. Female NOD.HLA-B*39:06.β2mKO mice had an earliest age of onset of 25 wk, with 58% diabetic at 30 wk. In contrast, NOD.HLA-B*39:06.β2mKO.Ins2KO mice had an earliest age of onset of 12 wk, with 100% of this strain being diabetic by 16 wk. We have previously noted that Ins2Het NOD mice exhibit a modest decrease in thymic insulin expression compared with WT NOD mice (36). Comparison of incidence curves (Fig. 5A) and age at onset (Fig. 5B) showed a trend for NOD.HLA-B*39:06.β2mKO.Ins2Het mice to exhibit a disease phenotype intermediate between that of their KO and WT counterparts, although the differences between the Het and WT mice did not reach statistical significance with the sample sizes available.
Differing amounts of thymic insulin expression do not grossly alter lymphocyte populations
It was likely that the increased disease susceptibility seen in the NOD.HLA-B*39:06.β2mKO.Ins2KO mice was due solely to the impaired T cell tolerance to insulin that is well known to be associated with Ins2 ablation (28–30, 32). However, to exclude a possible secondary cause, which had not been explored in prior work (i.e., gross changes in lymphocyte populations), we examined the impact of differing amounts of thymic insulin expression on splenic B cell, CD8 T cell, and CD4 T cell populations (Fig. 6A). As previously reported (21), NOD.HLA-A*02:01.β2mKO mice have a reduced percentage of splenic CD8 T cells and an increased percentage of B cells and CD4 T cells relative to NOD mice, and this was also observed in the NOD.HLA-B*39:06.β2mKO strain (Fig. 6B). However, when HLA-B*39:06 mice with differing Ins2 genotypes were compared, no significant changes in splenocyte subset percentages were observed (Fig. 6B). Furthermore, the percentage of CD4+CD25+ T cells was consistent across all groups, suggesting that the change seen in disease susceptibility was not due to a differing proportion of largely regulatory T cells. Together, these data suggest that the increase in disease incidence seen in the NOD.HLA-B*39:06.β2mKO.Ins2KO mice compared with their Ins2Het and Ins2WT counterparts was not due to gross changes in lymphocyte composition.
Thymic insulin expression does not alter TCR Vβ usage, but HLA transgene identity does
We next wished to exclude the unlikely possibility that changes in TCR Vβ usage accompanied the enhanced disease observed in the NOD.HLA-B*39:06.β2mKO.Ins2KO mice. For this purpose, splenocytes from these mice and their Ins2WT counterparts were stained with a panel of anti-mouse TCR Vβ Abs. Separate examination of CD8 and CD4 T cells revealed no significant differences in TCR Vβ usage between these two strains of mice (Fig. 7A, 7B). When the CD4+CD25+ (largely regulatory) T cell population was examined individually, the Ins2KO mice showed a small but significant increase in the use of TCR Vβ8.1/2 when compared with Ins2WT mice, but no other changes were noted (Fig. 7C).
The availability of both NOD.HLA-B*39:06.β2mKO and NOD.HLA-A*02:01.β2mKO mice presented a unique opportunity to examine the influence of HLA transgene identity on TCR Vβ usage. Examination of CD8 T cells revealed significant differences in usage of Vβ2 and Vβ11 (higher in the presence of HLA-B*39:06) and Vβ6 and Vβ8.1/2 (higher in the presence of HLA-A*02:01) (Fig. 7A). In contrast, there were no differences in the usage of these TCR Vβ families when CD4 T cell populations were studied (Fig. 7B, 7C). These findings demonstrate that some TCR Vβ families have a preference for particular class I MHC alleles.
To determine whether certain Vβ families were enriched among β cell–specific CD8 T cells in our HLA-transgenic models, we attempted to perform TCR Vβ analysis on islet infiltrates directly ex vivo. However, these experiments proved unfeasible because of limited cell numbers. As an alternate approach, we investigated whether certain Vβ families were enriched among CD8 T cells in the pancreatic lymph nodes of the mice, where β cell–specific CD8 T cells can be primed and induced to expand (46). For the NOD.HLA-B*39:06.β2mKO strain, Ins2Het mice were used, as they trend toward an intermediate disease phenotype compared with Ins2KO and Ins2WT mice (Fig. 5A). CD8 T cells in the pancreatic lymph nodes used a broad repertoire of TCR Vβ families, as observed for splenic CD8 T cells, and no Vβ families were enriched in the pancreatic lymph nodes relative to the spleens of the same animals (Fig. 8A). Similarly, the TCR Vβ family usage among CD8 T cells from the pancreatic lymph nodes of NOD.HLA-A*02:01.β2mKO mice was indistinguishable from that in the spleens (Fig. 8B). These results suggest that β cell-specific CD8 T cells in both HLA-transgenic models may use a broad TCR repertoire.
Despite altered thymic insulin expression, HLA-B*39:06–transgenic NOD mice retain typical blood insulin levels
Because of the compensatory changes observed in pancreatic Ins gene expression when the total number of Ins genes is reduced from four to two (28), we considered it unlikely that insufficient pancreatic insulin expression contributed to the earlier age of disease onset seen in the NOD.HLA-B*39:06Hom.β2mKO.Ins2KO mice. To confirm this, we measured blood insulin levels in young mice (5–6.5 wk old), well prior to disease onset (Fig. 9). We found that the level of insulin expression was statistically indistinguishable between NOD.HLA-B*39:06Hom.β2mKO mice and their Ins2Het and Ins2KO counterparts, with an average concentration of 0.8 ng/ml, consistent with previous reports for other mouse strains (37, 47). This supports the notion that the changes in disease onset are solely due to diminished T cell tolerance to insulin associated with Ins2 ablation (28–30, 32), and they are not also influenced by an inherently decreased ability to produce insulin.
Multiple loci are associated with T1D risk, including a number of class I and class II MHC alleles (4). Among these, HLA-B*39:06 is not only the most predisposing class I HLA allele in T1D patients (7, 9) but also leads to an earlier age of disease onset (11). However, because of the rarity of this allele among the populations studied thus far, investigation of the direct impact of HLA-B*39:06 on T1D pathogenesis has not been possible (6, 7). It is important to note that HLA-B*39:06 is more common among Latin American populations, with allele frequencies of 0.03 among Mexican Americans, 0.02–0.06 among Hispanic Americans, and 0.01–0.09 among Mexican individuals (12). The Venezuela Perja Mountain Bari population has an allele frequency of 0.24. Although T1D is relatively rare within Latin American countries, incidence is rising worldwide and new patients from these populations can be expected (14, 15). Similarly, patients carrying this genetic variant can increasingly be found in countries where T1D incidence is highest (48). Although genetic background is important, environment is as well; when individuals from areas with low T1D incidence move to areas with high incidence, they assume some of the risk of their new environment (49, 50). Therefore, inclusion of HLA-B*39:06–positive patients in treatment studies is essential. As such, the development of a mouse model for the study of HLA-B*39:06 is important, as this resource can provide a useful preclinical tool for the testing of HLA-B*39:06–directed treatments in the absence of sufficiently powered patient studies.
We have previously used an NOD.β2mKO-based model to study the contribution of HLA-A*02:01 to T1D development (21). In the current study, we found that NOD.HLA-B*39:06Hom.β2mKO mice develop similar amounts of CD8 T cells as their HLA-A*02:01–transgenic counterparts (Fig. 3B, 3C), suggesting that HLA-B*39:06 is as efficient at leading to CD8 T cell development as the more well-studied HLA-A*02:01. To our knowledge, we show in this study for the first time that HLA-B*39:06 can directly mediate T1D in an NOD.β2mKO model (Figs. 4A, 5A). However, unlike HLA-A*02:01 (21, 51), when expressed in the presence of the NOD class I MHC alleles H2-Db and H2-Kd, HLA-B*39:06 did not accelerate disease (Fig. 1). Although this may be due to strain-specific differences (e.g., transgene integration site), it also may speak to the importance of other aspects of the genetic environment that are known to be important for HLA-B*39:06–related susceptibility in humans. HLA-B*39:06 has been found to exert its effect on T1D risk in patients with specific class II MHC haplotypes, namely HLA-DR8/DQ4 (6, 52). Depending on the population studied, these class II MHCs may be independently predisposing, in which case HLA-B*39:06 accelerates disease progression, or may have a neutral impact on T1D, in which case HLA-B*39:06 lends risk to such patients. It has been well established that H2-Ag7, the NOD class II MHC, bears great similarity to the human T1D–associated HLA-DQ8 (18). HLA-DR8, part of a haplotype associated with HLA-B*39:06, has similar peptide binding characteristics to both H2-Ag7 and HLA-DQ8 (53). Although HLA-DR8 has been hypothesized to be the T1D-causative allele in the HLA-DR8/DQ4 haplotype, other evidence suggests that HLA-DQ4 is associated with risk of greater disease progression (54–56). Given that genetic context is important for the association between HLA-B*39:06 and T1D risk, the lack of a class II MHC molecule similar to HLA-DQ4 in the NOD mouse model may explain why similar diabetes incidence curves were observed for NOD and NOD.HLA-B*39:06 mice (Fig. 1). However, despite potentially not having an ideal genetic environment, HLA-B*39:06 is still able to mediate disease and islet infiltration, as confirmed by our findings in the NOD.HLA-B*39:06.β2mKO strain (Figs. 4A, 4B, 4D, 5A). It is important to appreciate that although HLA-B*39:06 allows the development of a diabetogenic CD8 T cell repertoire when expressed in NOD mice (Figs. 1, 5A), not all HLA class I molecules can do so. For example, when expressed along with H2-Db and H2-Kd in NOD mice, HLA-B*27 inhibits the development of T1D (51).
To more accurately model the genetic background of patients with T1D, we incorporated reduced thymic insulin expression into the NOD.HLA-B*39:06.β2mKO model. We found that NOD.HLA-B*39:06.β2mKO.Ins2KO mice are susceptible to disease at a younger age compared with their Ins2WT counterparts (Fig. 5). Based on our findings that the gross lymphocyte populations (Fig. 6B), TCR Vβ usage (Fig. 7), and blood insulin levels (Fig. 9) do not differ dramatically between these strains, the earlier age of onset in the context of reduced insulin expression is likely to be solely explained by the decrease in insulin tolerance that is associated with Ins2 ablation (28–30, 32). CD8 T cells are necessary for the development of T1D (2, 39, 42–44). As the expression of HLA-B*39:06 restores T1D susceptibility to NOD.β2mKO mice (Figs. 4A, 5A) and is enhanced in the Ins2KO mice (Fig. 5A), it is likely that reduced thymic insulin expression results in an increase in CD8 T cell reactivity toward insulin, as it has been shown to do in NOD.Ins2KO and NOD.HLA-A*02:01.β2mKO.Ins2KO mice. However, increased CD4 T cell reactivity to insulin could also be a contributing factor. We propose that an increased HLA-B*39:06–restricted reactivity to insulin may also contribute to the earlier age of onset seen in HLA-B*39:06–positive patients. These points will be clarified by future investigations.
The NOD.HLA-B*39:06.β2mKO model can be used in a variety of ways to probe the influence of HLA-B*39:06 on T1D susceptibility. As we have successfully done for HLA-A*02:01 (21, 22, 30), the model will allow us to identify the β cell peptides recognized by HLA-B*39:06–restricted T cells without the potentially confounding presence of murine class I MHC molecules. That the peptide-binding motif for HLA-B*39:06 has recently been identified may simplify the identification of HLA-B*39:06–restricted epitopes (57, 58). We have previously identified HLA-A*02:01–restricted epitopes in an NOD.β2mKO–based model (21, 22, 30); these epitopes were the same or similar to those found in human T1D patients expressing this class I variant (21, 23). Thus, the use of the NOD.HLA-B*39:06.β2mKO models could provide a direct translational impact. Identification of HLA-B*39:06–restricted epitopes can enable their use in epitope-directed therapies; these are an attractive option, as they can allow for treatments targeted at specific epitopes without the risk of off-site effects. Furthermore, knowledge of targeted epitopes permits the tracking of response to therapy (e.g., through the use of peptide–MHC tetramers). Such therapies require preclinical testing, representing another future use of the NOD.HLA-B*39:06.β2mKO models.
We did not find differences in TCR Vβ gene usage among splenic CD8 T cells when comparing NOD.HLA-B*39:06.β2mKO mice that were KO, Het, or WT for Ins2 (Fig. 7A). However, when we compared the TCR Vβ usage among splenic CD8 T cells in NOD.HLA-B*39:06.β2mKO and NOD.HLA-A*02:01.β2mKO mice, we found that four TCR Vβ families were differentially expressed (Fig. 7A). Specifically, usage of Vβ2 and Vβ11 was increased in the presence of HLA-B*39:06, whereas usage of Vβ6 and Vβ8.1/2 was decreased (Fig. 7A). This was initially a surprising finding, as until recently it was not generally thought that a given TCR Vβ family had any preference for a particular MHC allele (59). Recently, however, usage of TCR Vβ (and Vα) genes has been found to be associated with the MHC genotype in humans, leading to the proposal that different TCR V gene products may indeed have a bias toward particular MHC alleles (60). Our results using the NOD.HLA-B*39:06.β2mKO and NOD.HLA-A*02:01.β2mKO strains (Fig. 7A) support this view and represent a valuable system to explore this phenomenon further. As it would be most relevant to examine this question in the context of human TCRs, previously described mice expressing a polyclonal human TCR repertoire (61) would enhance such studies. In the case of HLA-A*02:01–transgenic mice, human TCRs have been shown to better select CD8 T cells than murine TCRs (61), and the same could hold true for HLA-B*39:06 mice as well.
In sum, we have established that HLA-B*39:06 can directly mediate T1D in the NOD mouse model, confirming the results seen in multiple genome-wide association studies (5, 7). We have furthermore developed models for HLA-B*39:06 in a genetic context more relevant to human disease by incorporating reduced thymic insulin expression. These models will allow a detailed investigation of the influence of HLA-B*39:06 on T1D development.
We thank Denisa Ferastraoaru for assistance with the preparation of the monochain chimeric HLA-B*39:06 construct. We thank The Jackson Laboratory Transgenic Genotyping Core for developing the real-time PCR assay for monitoring HLA-B*39:06 homozygosity.
This work was supported by National Institutes of Health Grants R01 DK064315, R01 DK094327, and R03 AI119225 (to T.P.D.); R01 DK046266, R01 DK095735, and U54 OD020351-5022 (to D.V.S.); F30 DK103368 (to J.S.); F32 DK111078 (to J.J.R.); and T32 GM007288 (which supported J.S.). Additional support was provided by National Institutes of Health Grants P30 CA013330, which supports the Cancer Center of the Albert Einstein College of Medicine; P30 CA034196, which supports the Cancer Center of The Jackson Laboratory; and P60 DK020541, which supports the Diabetes Research Center of the Albert Einstein College of Medicine; and by American Diabetes Association Grant 1-16-IBS-069 (to T.P.D.), JDRF Fellowship 3-PDF-2017-372-538 A-N (to J.J.R.), and Diabetes Research Connection Grant DRC 006887 JR (to J.J.R.). T.P.D. is the Diane Belfer, Cypres & Endelson Families Faculty Scholar in Diabetes Research.
Abbreviations used in this article:
mean fluorescence intensity
type 1 diabetes
variable number of tandem repeats
The authors have no financial conflicts of interest.