A SNP in the Immunoregulatory Molecule CTLA-4 Controls mRNA Splicing In Vivo but Does Not Alter Diabetes Susceptibility in the NOD Mouse

The development of most autoimmune diseases, including type 1 diabetes, is influenced by both genetic and environmental factors. Such genetic susceptibility is complex. In addition to polymorphisms in MHC genes, there are many genetic variants that contribute to diabetes susceptibility, each of which contributes a relatively small additional risk. In man, genome-wide association studies have identified more than 40 genetic traits that affect type 1 diabetes susceptibility; a similar number of genetic intervals have been identified in the NOD mouse model (1). Variation that alters the mRNA splicing of CTLA-4 has been implicated in diabetes susceptibility in both man and the NOD mouse model (2). CTLA-4 polymorphisms are also associated with type 1 diabetes susceptibility in dogs (3).

CTLA-4 is an important immunoregulatory molecule with an essential role in the function of regulatory T cells (4,5) and a less well-defined role in effector T cells (6,7). In humans, two protein isoforms have been described: the canonical transmembrane homodimer (flCTLA4) and a soluble monomeric form (sCTLA4); both forms result from alternative mRNA splicing of the nascent CTLA-4 transcript, and both bind to the immunoglobulin superfamily ligands CD80 and CD86. In addition to these two forms, a third protein isoform, liCTLA-4, has been described in mice. This isoform results from alternative mRNA splicing in which exon 2 (encoding the ligand binding Ig domain) is spliced out to produce a predicted 72aa mature transmembrane protein capable of homo- or hetero-dimerization with li/flCTLA4 (8). Ueda et al. (2) found an association between lower expression levels of the soluble isoform of CTLA4 and susceptibility to autoimmune diseases, including type 1 diabetes, in humans and also showed that NOD mice but not NOD.B10Sn-Idd5 mice had lower levels of mRNA for the liCTLA4 isoform. Subsequent work by Wicker and colleagues has confirmed the close correlation between genetic variation in Ctla4, variation in liCTLA4 mRNA expression, and diabetes susceptibility in a number of NOD congenic mouse strains (9,10). Indeed, the protection from diabetes observed in NOD.B10Sn-Idd5.1 mice is now widely attributed to Ctla4, in particular, to the A/G single nucleotide polymorphism (SNP) at position 77 in Ctla4 exon 2 (11–13).

While the correlation between diabetes protection and increased levels of liCTLA is suggestive, it does not demonstrate a causal link. Furthermore, there has been no direct test of the hypothesis that the exon 2 SNP is responsible for the alteration in mRNA splicing. We recently developed germline competent embryonic stem (ES) cells from the NOD strain (14). We therefore decided to directly test whether this SNP is responsible for altering splicing efficiency and diabetes susceptibility. We have derived a novel “knock-in” transgenic strain on the NOD background that contains the candidate causative B10 SNP (Ctla4 e2_77A). Analysis of this mouse strain demonstrates that the SNP is fully responsible for the change in alternative mRNA splicing and that the increased expression of liCTLA4 does not alter the susceptibility to diabetes.

NOD/MrkTacDil, NOD.Cg-Tg(TcraBDC2.5,TcrbBDC2.5)1Doi/DoiJ, NOD.Cg-N(Foxp3-I-EGFP) and NOD 77A (official name NOD-Ctla4^tm1(77A)Holm) mice as well as their crosses were bred under specific pathogen-free conditions at the Biological Services Unit of the Department of Pathology, University of Cambridge. Mice had unrestricted access to standard laboratory food and water and were housed in open or air-filtered cages. All animal work was carried out in accordance with U.K. Home Office project and personal licenses and was approved by the Ethical Review Committee of the University of Cambridge.

Clone 16 NOD ES cells and their derivatives were maintained in N2B27 medium supplemented with 1 μmol/L mitogen-activated protein kinase inhibitor PD0325901, and 3 μmol/L glycogen synthase kinase-3 inhibitor CHIR99021 (2i; both synthesized in the Division of Signal Transduction Therapy, University of Dundee, Dundee, U.K.) plus Leukemia Inhibitory Factor (25 ng/mL, produced in house). We injected 10–15 trypsinized NOD ES cells into each C57BL/6 blastocyst and then transferred the blastocysts to the uterus of a pseudopregnant F1 female mouse at 2.5 days postcoitum. We assessed the resulting chimeras for germline transmission by mating them with NOD mice.

The NOD bacterial artificial chromosome (BAC) CH29-120A16 was obtained from the Children’s Hospital Research Center Oakland, Oakland, CA. A 10,047–base pair (bp) SpeI-SacI fragment that contained CTLA-4 was cut from the BAC sequence and inserted into a SacI-XbaI cut pNEB193 vector. This 10,047-bp fragment was in turn cut with SacI-KpnI and KpnI-PacI, producing two fragments that contained the 5′ and 3′ arms of the CTLA-4 gene, respectively, which were subcloned into pNEB193. The 3′ arm was modified by insertion of a selection cassette into the NsiI site in intron 2. The selection cassette consisted of a loxP-flanked neomycin resistance gene and a herpes simplex virus thymidine kinase gene, both driven by an enhancerless thymidine kinase promoter from plasmid, pNeoTKloxP (Peter McKinnon, St. Jude Children's Research Hospital, Memphis, TN). The 5′ arm was modified by substitution of a 236-bp DraIII-KpnI fragment containing a single G→A point mutation at position 77 of exon 2. The two arms were rejoined at the KpnI site and the resulting plasmid linearized with NotI and 10⁷ clone 16 male NOD ES cells (14) transfected by electroporation. Transfected cells were subjected to G418 selection in 2i media, and any surviving cells were cultured as individual clones in 96-well plates.

Homologous recombinants were identified by PCR screening with a three-primer assay that amplified distinct products from targeted and endogenous loci (primers P₁ + P₃ + P₄) (Table 1). Recombinants were confirmed by Southern blotting with both a 5′ flanking and internal probe, and the cointegration of the targeted 77A mutation was confirmed by sequencing. The loxP-flanked cassette was removed by transient transfection with a CAGGS-Cre-Ires-Blasticidin plasmid (Sandra Gomez-Lopez, Cambridge Stem Cell Institute). Twenty-four hours after transfection, cells were exposed to Blastocidin for 24 h, and candidate clones were picked from surviving cells. The structure of the recombined locus was confirmed by sequencing.

Urine glucose levels were measured using indicator strips, and readings of >28 mmol/L glucose in the urine were considered indicative of diabetes (rated according to the color glucose scale provided by the manufacturer). Glucose readings were taken between 1500 and 1800 h in the afternoon, except where stated otherwise. Only female mice were used for spontaneous diabetes incidence studies.

All mice were weighed and tested for hyperglycemia prior to injections. Only healthy male mice were injected with 20 mg/mL i.p. cyclophosphamide (CTX) solution at a dose of 250 mg/kg. Mice were held in a supine position, and injections were made in the lower left quadrant of the abdomen. After injection, mice were tested for hyperglycemia every 24 h for 4 weeks.

For template generation, cells were lysed using the QIAshredder kit and RNA was extracted using the RNeasy kit. DNA contaminants were removed using the TURBO DNA free kit, and the mRNA was reverse transcribed using the SuperScript III First Strand Synthesis System. RNA sequence bias was avoided by using Oligo (dT)20 primers for cDNA synthesis. RNA content in all reactions was optimized to maximize cDNA concentration.

Quantitative PCR (qPCR) assays were performed in triplicate or quadruplicate with water controls included on every plate and housekeeping gene controls included for every sample. Reactions contained 10 μL relevant primer mix (2×), 10 μL GoTaq qPCR Master Mix, and ∼1 ng cDNA template. Primer sequences are shown in Table 2.

Samples were analyzed using the Applied Biosystems 7500 Fast Real-Time PCR System calibrated for SYBR green detection. All plates were run using the same amplification program: 95°C for 15 min (94°C for 15 s, 60°C for 30 s, 72°C for 30 s) × 40 cycles followed by a dissociation stage. The threshold for detection was set within the linear phase of the logarithmic amplification plot.

For carboxyfluorescein succinimidyl ester (CFSE) labeling, CD4⁺ splenocytes were incubated with 1 μmol/L CFSE in PBS at a concentration of 10⁷ cells/mL. Cells were incubated for 15 min at 37°C and washed twice with cold Iscove's Modified Dulbecco's Medium (IMDM). Labeled cell pellets were resuspended in IMDM at a concentration of 5 × 10⁶ cells/mL and injected into the tail veins of recipient mice (10⁶ cells per mouse).

Cells were labeled with anti-CD4 (RM4–5) supplemented with 2.4G2 anti-FcγRII/III (produced in-house). Labeled cells were analyzed using the Cytek DxP8 flow cytometer.

Cells for use in qPCR were isolated using the MoFlo cell sorter, and cells for CFSE labeling were isolated using magnetic-activated cell sorting of anti-CD4(L3T4)–conjugated magnetic beads

For T-cell activation, culture dishes were coated with 10 μg/mL anti-CD3 antibody (145–2C11 produced in house) overnight. Purified T cells were cultured in IMDM medium supplemented with 10% FCS, 100 units/mL penicillin, 100 μg/mL streptomycin, 2 mmol/L glutamine, and 5 μg/mL anti-CD28 antibody (37.51 also produced in-house).

NOD strain ES cells were maintained in ground state by culture in 2i medium (14) and transfected with a Ctla4 targeting construct. The construct was prepared using the NOD BAC CH29-120A16 as an isogenic DNA source and contains a single G→A point mutation at position 77 in exon 2 (see 2research design and methods). The final genomic structure (R2) (Fig. 1A) retains a single loxP site within intron 2. Twenty-three male chimeras were produced by injecting clone 3E1/11 ES cells into C57BL/6 blastocysts. Eleven chimeras, chosen for strong NOD coat color contribution, were backcrossed to our NOD colony; 54% of offspring were ES cell derived, and 31.25% of these carried the targeted Ctla4^77A allele (difference in allelic transmission not significant). The breeding scheme for producing incidence cohorts for analysis is shown in Fig. 1B. Heterozygous 77A/G NOD offspring were intercrossed to produce both 77A/A and 77G/G homozygotes, from here on referred to as NOD 77A and NOD 77G. These first generation homozygotes were again intercrossed to produce sufficient female offspring for study (incidence cohort I). In this way, the experimental and control cohorts shared all ES-derived genetic material except Ctla4 and its linked genomic segment. A second cohort set was produced after a further backcross to our NOD colony, with the sex of the NOD parent reversed (Fig. 1B). First-generation 77A homozygotes were also crossed to our NOD.Cg-BDC2.5 and NOD.Cg-FoxP3^EGFP colonies. Sib mating of the offspring produced 77A homozygotes carrying the relevant transgene. FoxP3^EGFP was also made homozygous, whereas the BDC2.5 T-cell receptor (TCR) transgenes were maintained by screening.

We examined the effect of the 77A mutation on CTLA-4 mRNA splicing by measuring the mRNA levels of each CTLA-4 isoform in the T cells of NOD 77A Foxp3^EGFP and NOD 77G Foxp3^EGFP mice. Three different T-cell populations, conventional T cells (CD4⁺Foxp3⁻), regulatory T cells (CD4⁺Foxp3⁺), and in vitro–activated T cells (CD4⁺FoxP3⁻ stimulated with anti-CD3 plus anti-CD28) were investigated. CTLA-4 mRNA was most abundant in regulatory T cells, which, in comparison with in vitro–activated T cells and conventional T cells, contained 6-fold and 67-fold more CTLA-4 mRNA, respectively (Fig. 2A). Analysis of the different CTLA-4 isoforms revealed that the ligand-independent and soluble isoforms are rare compared with the full-length isoform, which represents between 86 and 99% of the total CTLA-4 mRNA (Fig. 2B–D).

In all analyzed T-cell populations, the introduction of the 77A mutation resulted in a significant increase of liCTLA-4 mRNA. A twofold increase in liCTLA-4 was observed in conventional and regulatory T cells, and a fourfold increase was observed in in vitro–activated cells (Fig. 2C). Considering that all isoforms are spliced from the same pre-mRNA, any increases in exon 2 splicing were expected to decrease the abundance of full-length and soluble isoforms (Fig. 2B and D). These results directly demonstrate, for the first time, that the 77A mutation in CTLA-4 exon 2 enhances the effectiveness of the exonic splicing silencer motif. Given the importance of CTLA-4 in immune regulation, the immunological relevance of this change in mRNA expression was explored.

Spontaneous diabetes incidence in female mice represents the most powerful immunological readout of the NOD model, and we therefore investigated spontaneous incidence in our gene-targeted NOD colonies. The first cohort study was conducted on animals that contained equal amounts of NOD ES cell–derived and “colony NOD”–derived genetic material (Fig. 1B). The main onset of diabetes in this cohort occurred around 15 weeks of age, and the total incidence leveled off between 60 and 70%. The 77A mutation had no significant effect on spontaneous diabetes development (Fig. 3A). Mice backcrossed a second time, now containing only 25% ES cell–derived genetic material, were investigated in a second incidence study (Fig. 3B). With diabetes onset at 12 weeks of age and a total incidence >80%, the backcrossed NOD 77A and 77G mice mirrored the diabetes incidence of the originator colony. Despite the increased cohort size, no difference between the NOD 77A and NOD 77G cohorts could be distinguished (Fig. 3B). The 77A mutation and altered levels of liCTLA-4 do not therefore affect spontaneous diabetes development. The difference in incidence between the initial and backcrossed cohorts may be due to variation in the environment, as all mice were moved from open cages into individually ventilated cages 20 weeks into the first study. Also, the possibility that both initial cohorts were affected by genetic or epigenetic changes occurring during ES cell culture, and lost after the second backcross, cannot be ruled out.

To explore the possibility of only certain functions of the immune system being linked to the 77A mutation, we compared the CTX-induced diabetes incidence in male NOD 77A and NOD 77G mice. CTX is thought to induce diabetes by depleting mature lymphocytes in the periphery, which leads to the preferential repopulation of lymphoid niches by diabetogenic effector T cells and, hence, the preclusion of regulatory T cells (15,16). Twelve-week-old mice received a single injection of 250 mg/kg i.p. CTX, and urinary glucose was monitored for 3 weeks. In both strains, the fastest rate of diabetes development was reached on day 10 postinjection and a total of 70–80% of mice developed diabetes by day 21 (Fig. 3C). No effect of the 77A mutation was observed, suggesting that the capacity of diabetogenic T cells to repopulate lymphoid niches was not affected by the 77A mutation or physiological changes in liCTLA-4 (Fig. 3C). These data appear at odds with the published result of Stumpf, Zhou, and Bluestone (17), who did see a reduction in CTX-induced diabetes in NOD background mice with similarly increased liCTLA4. There are a number of experimental differences, however. One of these was that younger mice were given two injections of 200 mg/kg CTX 7 days apart (17). We therefore used this protocol to determine whether the 77A mutation conferred any protection from diabetes onset. We could find no evidence for this (data not shown); indeed, the majority of mice given one injection of 200 mg/kg CTX at 8–10 weeks developed diabetes by 7 days later (see discussion for further consideration of these differences).

To directly assess the effect of liCTLA-4 mRNA variation on immune function, we used BDC2.5 transgenic NOD 77A and NOD 77G control mice. T cells in these mice express the TCR from BDC2.5, a diabetogenic T-cell clone (18). When transferred to adult NOD mice, BDC2.5 transgenic T cells undergo antigen-driven proliferation in the pancreatic lymph node (19). CD4⁺ T cells were isolated from BDC2.5 transgenic NOD 77A and NOD 77G mice, in vitro labeled with the proliferation dye CFSE and adoptively transferred into wild-type NOD recipients. After 72 h, cells were recovered from the pancreatic lymph nodes of recipient mice and analyzed by flow cytometry (Fig. 4). In all recipient mice, a proportion of transferred T cells were dividing, visualized by the typical pattern of division peaks (Fig. 4B–E). For direct comparison of the proliferative properties of 77A- and 77G-derived BDC2.5 T cells, proliferation profiles were averaged and overlaid (Fig. 4F). No significant effect of increased liCTLA-4 levels on T-cell proliferation could be observed. Alternative analysis of this data, comparing the number of cells that remained undivided, also showed no significant impact of liCTLA-4 mRNA (data not shown). These data strongly suggest that the TCR activation threshold, which represents the cell’s immunological state, as well as the dynamics of cell proliferation, is not affected by the 77A-induced changes to liCTLA-4 mRNA.

We also assessed the function of regulatory T cells (Tregs) using an in vivo adoptive transfer model. T cells from BDC2.5 mice induce rapid diabetes when transferred to NOD-SCID recipients; the cotransfer of CD25^high Tregs can prevent or delay diabetes onset in the NOD-SCID mice (20). BDC2.5 transgenic NOD 77A mice were crossed with the NOD.Cg-FoxP3^EGFP strain. This enabled the preparation of Treg-depleted effector populations (CD4⁺GFP⁻) and Tregs (CD4⁺CD25^highGFP⁺) from the same donors. Cotransfer of Tregs with effector T cells in a ratio of 1:3 delayed but did not prevent diabetes onset (Fig. 5). However, there was no discernible difference between the effectiveness of the Tregs from 77A and 77G genotypes (Fig. 5). This experimental model reinforces the conclusion that there is no difference in the ability of 77A effectors to cause diabetes and shows that effectors from 77A and 77G genotype NOD mice are equally responsive to regulation by cognate Tregs of the same CTLA4 genotype (Fig. 5).

A number of previous studies have provided evidence for the hypothesis that increased levels of mRNA splicing to the liCTLA4 isoform have functional consequences in T cells and, in congenic NOD strains, reduce the incidence of spontaneous and/or induced diabetes (8–10,17). We believe that our data (Fig. 3) clearly contradict the conclusion that a modest (two- to fourfold) increase in liCTLA4 levels produced by the G→A single nucleotide substitution at exon 2, position 77, alters the spontaneous or induced diabetes susceptibility in NOD mice. For the most part, significant experimental differences easily account for conclusions of our study differing from those previously published. In the work of Wicker and colleagues, NOD congenic strains were painstakingly derived with introgressed genomic DNA from the Idd5.1 region from the CAST and SWR strains in addition to the original B10Sn-Idd5.1 strain (10). NOD.CAST-Idd5.1 had a greater increase in liCTLA4 mRNA than the B10 congenic, while NOD.SWR-Idd5.1 mirrored the NOD level; the presence of a single copy of CAST-Idd5.1 was sufficient to reduce the spontaneous incidence of diabetes by nearly half, but two copies of SWR-Idd5.1 made no difference to diabetes susceptibility. However, in all the NOD congenics that shared the Ctla4 77A variant, it was accompanied by a very large number of other linked genetic variants differing from the NOD genome, so that the correlation was suggestive but by no means conclusive. In our mice, the only known genetic differences between our two cohorts are the 77A variant and the presence of a loxP site in Ctla4 intron 2.

Araki et al. (10) also found that NOD mice transgenic for a CD2-driven liCTLA4 cDNA developed diabetes at a reduced frequency. Here, the link of function to expression of liCTLA4 mRNA and protein is direct, and the mice are 100% NOD genetically except for the transgene. However, the transgenic construct produced an ∼2,000-fold increase in the liCTLA4 mRNA level and it is plausible that such an overexpression might have adventitious consequences, for example, by sequestering miRNAs. A more physiological increase in the level of liCTLA4 mRNA was produced by transgenesis with a C57BL/6 BAC clone encoding Ctla4. The BAC was modified so that exon 2 was flanked by loxP sites. Intercrossing with a Vav1-Cre deleter strain should produce mice whose hemopoetic cells contain a Cre-recombined transgene, capable of only producing liCTLA4 (17). These mice, termed NOD-Tg-Cre, were shown to have a twofold increase in liCTLA4 mRNA in CD4⁺CD25^highCD62L⁻ cells. Surprisingly, Stumpf, Zhou, and Bluestone (17) do not describe the effects of the transgene on the spontaneous incidence of diabetes. The most directly comparable experiments reported are reduced incidence of pancreatic infiltration (scored by histology) in 20-week-old prediabetic mice and a reduced incidence of diabetes in 8-week-old mice treated with a two-dose CTX regimen. Neither experiment reports the sex of the mice used, so we cannot be sure we are making correct comparisons; however, we saw no significant difference in immune cell infiltration in 12-week-old females (data not shown) and no difference in diabetes incidence in CTX-treated males (Fig. 3C). The BAC transgenic was produced by microinjection into NOD embryos; however, the Vav1-Cre transgene was made in (CBA/Ca×C57F7BL/10)F1 mice (21). Some non-NOD genomic DNA is present in the NOD-Tg-Cre mice even though the Vav1-Cre strain had been backcrossed to NOD. It also seems likely that the University of California, San Francisco, colony has a lower incidence of diabetes; their NOD mice took 16 days to develop diabetes after two injections of CTX, whereas we find that diabetes starts at about 1 week after a single injection when given at 8–10 weeks of age. In addition, we have to conduct our insulitis screen at 12 weeks, since 40–50% of our colony is diabetic by 20 weeks of age (Fig. 3B). These probable differences in diabetes incidence could be the result of genetic or environmental factors or a combination of both.

The data reported here are of the most direct (and simple) test of the functional effects of the modest increase in liCTLA4 mRNA produced by the exon 2 77A SNP, and we contend that the incidence data are conclusive in showing that this mRNA splicing change does not alter diabetes susceptibility. The clear implication is that one or more of the >700 other genetic differences between the NOD and NOD.B10/Sn-Idd5.1 strains (9), possibly affecting other genes, e.g., Icos, are required for the change in diabetes incidence previously reported (9,10). We do not exclude that the exon 2 77A SNP may be necessary for the diabetes phenotype, but neither it nor the splicing change is sufficient. This study demonstrates that NOD ES cells can provide a probative test of candidate mutations in diabetes susceptibility. Further studies using NOD ES cells carrying multiple or single genetic differences could resolve the ambiguities that surround causative mutations in genetic susceptibility to diabetes and provide models for probing mechanisms of protection, in many cases directly relevant to human disease (22).

Acknowledgments. The authors thank Peter McKinnon (St. Jude’s Children’s Research Hospital, Memphis, TN) for the plasmid pNeoTKloxP, Kim Hoenderdos (University of Cambridge) and Des Jones (University of Cambridge) for assistance with qPCR, Nigel Miller (University of Cambridge) for cell sorting, and Yvonne Sawyer (University of Cambridge) for assistance with animal husbandry.

Funding. F.J. was supported by a Medical Research Council Studentship, and the authors acknowledge the support of the Wellcome Trust and the Department of Pathology, University of Cambridge.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.J. and N.H. designed and performed experiments, analyzed data, and wrote the manuscript. K.J. and W.M. performed experiments and reviewed and edited the manuscript. J.N. and A.C. contributed to experimental design and interpretation and reviewed and edited the manuscript. N.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.