Major histocompatibility complex class I-restricted T cells are required for all but the end stages of diabetes development in nonobese diabetic mice and use a prevalent T cell receptor α chain gene rearrangement

Abstract

Nonobese diabetic (NOD) mice develop insulin-dependent diabetes mellitus due to autoimmune T lymphocyte-mediated destruction of pancreatic β cells. Although both major histocompatibility complex class I-restricted CD8⁺ and class II-restricted CD4⁺ T cell subsets are required, the specific role each subset plays in the pathogenic process is still unclear. Here we show that class I-dependent T cells are required for all but the terminal stages of autoimmune diabetes development. To characterize the diabetogenic CD8⁺ T cells responsible, we isolated and propagated in vitro CD8⁺ T cells from the earliest insulitic lesions of NOD mice. They were cytotoxic to NOD islet cells, restricted to H-2K^d, and showed a diverse T cell receptor β chain repertoire. In contrast, their α chain repertoire was more restricted, with a recurrent amino acid sequence motif in the complementarity-determining region 3 loop and a prevalence of Vα17 family members frequently joined to the Jα42 gene segment. These results suggest that a number of the CD8⁺ T cells participating in the initial phase of autoimmune β cell destruction recognize a common structural component of K^d/peptide complexes on pancreatic β cells, possibly a single peptide.

Insulin-dependent diabetes mellitus (IDDM) is an autoimmune disease characterized by T cell-mediated destruction of pancreatic islet β cells (1). The nonobese diabetic (NOD) mouse (2) constitutes a widely studied model system for IDDM, as it shares many of the characteristics of the human disease. For example, human patients and NOD mice both develop lymphocytic infiltration of islets (insulitis) and subsequent β cell destruction mediated by T lymphocytes. Cell-surface αβ T cell receptors (TCRs) enable such T lymphocytes to recognize specific antigens on the surfaces of target cells in the form of peptides complexed with major histocompatibility complex (MHC) molecules, with CD8⁺ cytotoxic T lymphocytes being restricted to class I MHC molecules and CD4⁺ T cells to class II.

Certain unusual MHC class II alleles provide the strongest genetic component of IDDM susceptibility in both humans and NOD mice (3). Thus, it is not surprising that autoreactive CD4⁺ T cells are essential for IDDM development in NOD mice (4). However, several lines of evidence indicate that MHC class I-restricted CD8⁺ T cells also are required for this process. First, when splenic T lymphocytes are isolated from young prediabetic NOD donors, both CD8⁺ and CD4⁺ T cells are required to adoptively transfer diabetes to normally resistant NOD-scid recipients (4). Second, NOD mice depleted of CD8⁺ T cells by treatment with an anti-CD8 antibody do not develop diabetes (5), nor do β2-microglobulin-deficient NOD mice, which lack MHC class I expression and therefore do not develop CD8⁺ T cells (6–9). Third, β cell-cytotoxic CD8⁺ T cells can be isolated from the insulitic lesions of NOD mice (10–13). Finally, T cells from young prediabetic NOD donors can transfer IDDM to recipients having MHC class I-positive, but not class I-negative, pancreatic β cells (14). Thus, diabetogenesis in NOD mice depends on CD8⁺ T cell recognition of antigens presented by MHC class I molecules expressed on β cells. The development of such diabetogenic CD8⁺ T cell responses is promoted most readily by expression of the particular MHC class I alleles (K^d and D^b) of the NOD H2^g7 haplotype (15). However, although the requirement for CD8⁺ T cells in IDDM development is clear, whether they are only needed to initiate the earliest events of β cell destruction or are critical to all stages of diabetogenesis remained unknown and is one of the questions addressed in this study.

It is also necessary to define the characteristics and specificities of MHC class I-restricted effectors participating in the earliest initiative phases of IDDM. In view of the importance of understanding the nature of these initiating T cells, we have developed a technique for the isolation of monoclonal and oligoclonal populations of NOD β cell-cytotoxic CD8⁺ T cells that uses islets from a newly developed stock of NOD-scid.RIPB7 mice as a potent source of stimulating antigen. This approach has enabled us to isolate cytotoxic CD8⁺ T cells from prediabetic mice in the earliest stages of insulitis and to establish their MHC class I allelic restriction, the diversity of their TCR α and β chain repertoires, and their degree of clonality within and among individual NOD mice. We show that class I-dependent T cells are critical to all but the final stages of diabetes development in NOD mice and that such β cell-cytotoxic effectors in early insulitic lesions use a prevalent TCR α chain gene rearrangement.

MATERIALS AND METHODS

Mice.

NOD/Lt mice (H2^g7=K^d, I-A^g7, I-E^null, D^b) are maintained at The Jackson Laboratory by brother–sister mating. Currently, IDDM develops in 90% of female and 63% of male NOD/Lt mice by 1 year of age. A diabetes-resistant stock of NOD mice congenic for the H2^nb1 MHC haplotype (K^b, I-A^nb1, I-E^k, D^b) derived from the NON/Lt strain (designated NOD.H2^nb1) is maintained at the N21 backcross generation (16). T and B lymphocyte-deficient NOD-scid mice (official designation NOD-Prkdc^scid) are maintained at the N11 backcross generation (4, 17). NOD-scid mice deficient in MHC class I expression because of a functionally inactivated β2-microglobulin allele (14, 18) also are maintained at the N11 backcross generation (official designation NOD-Prkdc^scid.B2m^tm1Unc and here denoted as NODscid.B2m^null). NOD mice whose pancreatic β cells express B7–1 T cell costimulatory molecules under the control of the rat insulin promoter (NOD-RIPB7; ref. 19) kindly were provided by R. Flavell, C. Janeway, Jr., and F. S. Wong (Yale University). These served as progenitors for an N8 backcross stock of NOD-scid.RIPB7 mice. While generating the NOD-scid.RIPB7 stock, segregants homozygous for the scid mutation were identified by fluorescence-activated cell sorting (FACS) (FACScan, Becton Dickinson) for an absence of T cells among peripheral blood leukocytes by using the fluorescein isothiocyanate-conjugated CD3-specific mAb 145–2C11. The RIPB7 transgene was detected as an 880-bp PCR product by using the primer set 5′-TGAAGCCATGGGCCACAC-3′ and 5′-GACACTGTTATACAGGGC-3′. All mice were maintained under specific pathogen-free conditions and were allowed free access to food (National Institutes of Health diet 31A, Purina) and acidified drinking water. All stocks of scid mice were treated for 3 days per week with trimethoprim-sulfamethoxazole (Sulfatrim, Barre-National, Baltimore) in the drinking water.

Adoptive Transfer of Autoimmune IDDM to NOD-scid Stocks.

Aliquots of 2 × 10⁷ pooled splenic leukocytes, prepared as described (4) from NOD female donors of the indicated ages, were injected i.v. into 4-week-old NOD-scid and NODscid.B2m^null female recipients. The recipients were monitored weekly for the development of diabetes as defined by the onset of glycosuria by using Ames Diastix (kindly supplied by Miles Diagnostics, Elkhart, IN). At diabetes onset or at 17 weeks post-transfer, CD4⁺ and CD8⁺ T cell reconstitution was verified by FACS typing of recipient splenocytes as described (14).

Isolation of MHC Class I-Restricted CD8⁺ T Cells Contributing to the Initiation of Autoimmune β Cell Destruction in NOD Mice.

Pancreatic islets were isolated as described (20) from NOD-scid.RIPB7 donors. Aliquots of 15 irradiated (2000 R) NOD-scid.RIPB7 islets then were seeded in 100 μl of the previously described culture medium (21) into round-bottom 96-well plates and were incubated for 48 hr at 37°C in a 95% air/5% CO₂ humidified atmosphere. Aliquots of 10 pancreatic islets isolated from individual 5- to 6-week-old standard NOD female mice then were seeded in an additional 100 μl of medium on top of the now-adherent lawn of NOD-scid.RIPB7 islets. Human recombinant interleukin (IL) 2 and IL-7 were added at final concentrations of 50 units/ml and 5 ng/ml, respectively. (IL-7 kindly was provided by R. Sun, Corixa.) A saturating concentration of the anti-CD4 antibody GK1.5 also was added to each well to block the outgrowth of CD4⁺ T cells. Sporadic wells were characterized by outgrowths of T cells from the standard NOD islets, which were expanded further by weekly passage on NOD-scid.RIPB7 islets as antigen source in the presence of IL-2, IL-7, and anti-CD4. When expanded to sufficient numbers, the T cells were typed by FACS for CD4 and CD8 expression by using the mAbs YTS191 and 53–6.72, respectively. Only CD8⁺ T cells were detected. In one case, the three clonotypes found in a single T cell line were separated by flow cytometric sorting (FACStar) by using appropriate TCR Vβ-specific antibodies. Cytotoxic activity of the various T cell isolates against the specified target cells was assessed in 4- or 20-hr ⁵¹Cr-release assays as described (22). Before being labeled with ⁵¹Cr and used as targets, pancreatic islets were dispersed into single cell suspensions by treatment with enzyme-free cell dissociation buffer (GIBCO/BRL).

Determination of TCR Vα and Vβ Gene Usage and TCR α and β Chain Sequences.

Messenger RNA was purified from 2 × 10⁵−1 × 10⁶ T cells by using the QuickPrep Micro mRNA Purification Kit (Pharmacia) and was reverse-transcribed into single-strand cDNA by using MMuLV reverse transcriptase and oligo(dT)₁₅ as primer. To determine the TCR Vα gene families used by each T cell line, cDNAs were amplified by PCR by using a TCR Cα primer (5′-TCGGAGTCCCATAACTGACAG-3′) paired with 1 of 20 Vα primers, each specific for a particular Vα gene family and designed to recognize all known members of that family (23). The Vα primer sequences, given 5′ to 3′, follow: Vα1, GCACTGATGTCCATCTTCTC; Vα2, AAAGGGAGAAAAAGCTCTCC; Vα3, AAGTACTATTCCGGAGACCC; Vα4, CAGTATCCCGGAGAAGGTC; Vα5, CAAGAAAGACAAACGACTCTC; Vα6, ATGGCTTTCCTGGCTATTGCC; Vα7, TCTGTAGTCTTCCAGAAATC; Vα8, AGACAACAAGAGGACCGAGC; Vα9, CAAATACATCACAGGAGACACC; Vα10, AACGTCGCAGCTCTTTGCAC; Vα11, CCCTGCACATCAGGGATGCC; Vα12, TCTGTTTATCTCTGCTGACC; Vα13, ACCTGGAGAGAATCCTAAGC; Vα14, GCACGCTGCACATCACAGCC; Vα15, AGAAGCTGGAAAGGGTCTCC; Vα16, CATTCGCTCAAATGTGAACAG; Vα17, GGAAAATGCAACAGTGGGTC; Vα18, CAAATGAGAGAGAGAAGCGC; Vα19, TGGTTTGAAGGACAGTGGGC; and Vα20, GACATGACTGGCTTCCTGAAGGCCTTGC. To determine the Vβ gene families used by each T cell line, cDNAs were PCR-amplified by using a Cβ primer (5′-CACTGATGTTCTGTGTGACAG-3′) paired with 1 of 20 Vβ primers. The Vβ primer sequences were as follows: Vβ1, CCCAGTCGTTTTATACCTGAATGC; Vβ2, TCACTGATACGGAGCTGAGGC; Vβ3, CCTTGCAGCCTAGAAATTCAGTCC; Vβ4, GCCTCAAGTCGCTTCCAACCTC; Vβ5, TGTATTCCCATCTCTGGACAT; Vβ6, CTCTCACTGTGACATCTGCC; Vβ7, TACAGGGTCTCACGGAAGAAGC; Vβ8, CATGGGCTGAGGCTGATCCAT; Vβ9, TCTCTCTACATTGGCTCTGCAGGC; Vβ10, ATCAAGTCTGTAGAGCCGGAGGAC; Vβ11, GCACTCAACTCTGAAGATCCAGAGC; Vβ12, GAAGATGGTGGGGCTTTCAAGGATC; Vβ13, AGGCCTAAAGGAACTAACTCCAC; Vβ14, ACGACCAATTCATCCTAAGCAC; Vβ15, CCCATCAGTCATCCCAACTTATCC; Vβ16, CACTCTGAAAATCCAACCCAC; Vβ17, CCTATCCTCTGAAGAAGACG; Vβ18, CAGCCGGCCAAACCTAACATTCTC; Vβ19, AGAGATCCAGTCCAGCAAGC; and Vβ20, CCTGGGAATCAGAACGTGCG. PCR products (0.5–0.8 kbp) were detected by ethidium bromide staining of 1.5% agarose gels. The Vα and Vβ gene families are designated throughout according to Arden et al. (23). The use of certain of the Vα and Vβ primers has been described (24).

To confirm the Vα and Vβ gene family assignments and discriminate between productive and nonproductive rearrangements, PCR products purified by using the QIAquick PCR Purification Kit (Qiagen, Chatsworth, CA) were sequenced from both directions at the DNA Sequencing Facility of the Albert Einstein College of Medicine (Bronx, NY). The sequencing primers used were the appropriate Vα or Vβ PCR primer and a Cα (5′-TGCTGTCCTGAGACCGAGGAT-3′) or Cβ (5′-TGCAATCTCTGCTTTTGATGG-3′) primer. In some cases, PCR products first were cloned into pCR2.1 (Invitrogen), and the sequences in multiple resulting plasmids were determined.

RESULTS

MHC Class I-Restricted T Cells Are Required for All but the End Stages of Diabetogenesis in NOD Mice.

In previous studies, splenocytes from young prediabetic NOD donors were found to adoptively transfer IDDM to MHC class I-positive, but not MHC class I-negative, NOD-scid recipients (14). In contrast, splenocytes from overtly diabetic NOD donors transferred disease to both class I-positive and class I-negative recipients. These results demonstrated that the initiation of autoimmune β cell destruction in NOD mice requires contributions from CD8⁺ T cells but that a population of CD4⁺ T cells eventually develops that is able to elicit β cell destruction in an MHC class I-independent fashion. Whether the requirement for CD8⁺ T cells is a transient early event or critical to all stages of diabetogenesis remained unknown. To address this, we compared the ability of splenocytes from NOD female donors of various ages to transfer IDDM to MHC class I-positive NOD-scid or MHC class I-negative NOD-scid.B2m^nullrecipients (Fig. 1). As reported (14), we found that splenocytes from 4- to 5-week-old NOD donors transferred IDDM to a majority (73%) of NOD-scid recipients but not to MHC class I-negative NOD-scid.B2m^nullrecipients. On average, it required 14.5 weeks for splenocytes transferred from 4- to 5-week-old NOD donors to induce IDDM in standard NOD-scid recipients. Splenocytes from 6- to 8-week, 10- to 11-week, and 12- to 14-week-old NOD donors transferred IDDM at increasingly faster rates to the majority (80–100%) of NOD-scid recipients but not to any NOD-scid.B2m^nullrecipients. Only splenocytes from overtly diabetic NOD female donors (>14 weeks of age) could transfer IDDM to both MHC class I-positive and MHC class I-negative recipients. Thus, between 5 weeks of age and the onset of overt IDDM, there is clearly an expansion in the number and/or types of β cell-autoreactive T cells, as evidenced by the decreasing mean time to disease onset in recipients of T cells from progressively older NOD donors (Fig. 1B). However, autoreactive CD4⁺ T cells capable of mediating pancreatic β cell destruction in an MHC class I-independent manner are only detectable within this expanding set of diabetogenic effectors during the very late stages of the pathogenic process. Thus, in this adoptive transfer model, MHC class I-restricted T cells are essential for all but the final stages of diabetes development.

Figure 1.

MHC class I-restricted T cells are required for all but the terminal stages of diabetogenesis in NOD mice. Splenocytes from NOD donors of various ages were transferred to 4-week-old MHC class I-positive NOD-scid and MHC class I-negative NOD-scid.B2m^null female recipients. The total number of recipients and those that became diabetic by 17 weeks post-transfer are plotted in A, and the mean time to disease onset post-transfer is plotted in B.

CD8⁺ T Cells from Early Insulitic Lesions of NOD Mice Are β Cell-Cytotoxic and H-2K^d-restricted.

In view of the requirement for MHC class I-restricted T cells in initiating IDDM, we sought to isolate and characterize CD8⁺ T cells from the islets of prediabetic NOD mice at an early stage of insulitis. In NOD females housed at The Jackson Laboratory, the initial histological signs of insulitis are observed at ≈5 weeks of age, with <10% of the total pancreatic islets affected. Thus, 5- to 6-week-old NOD females were deemed to be appropriate donors to isolate MHC class I-restricted T cells contributing to the initiating events of autoimmune β cell destruction. Our culture system was based on a report that, although a diverse array of T cell clonotypes can be detected in pooled islets from a single NOD pancreas, the T cells within any individual islet are less diverse (25). Given that <10% of the islets exhibit insulitic infiltrates at 5 weeks, we reasoned that, when cultured in groups of 10 in the presence of a CD4-specific antibody, islets isolated from individual 5- to 6-week-old females might give rise to outgrowths of monoclonal or oligoclonal MHC class I-restricted β cell-autoreactive T cells. Because Wong et al. (13) had reported that MHC class I-positive islets expressing transgene-encoded B7–1 costimulatory molecules provided a potent antigenic source for propagating diabetogenic CD8⁺ T cells from NOD mice, we developed a NOD-scid.RIPB7 stock of mice as a source of B7-expressing islets. We then used these islets in our T cell cultures to provide stimulating antigen without the possibility of introducing contaminating T cells.

FACS analysis demonstrated that all of the T cell populations generated in the culture system described above were entirely of the CD8 phenotype (data not shown). To assess their β cell cytotoxicity and MHC class I allelic restriction, the T cell lines in Table 1 were evaluated for cytotoxic activity against dispersed pancreatic islet cells from NOD-scid (expressing K^d and D^b MHC class I molecules) and NOD.H2^nb1congenic mice (expressing K^b and D^b). Each T cell line displayed activity against NOD-scid islet cells, but not against NOD.H2^nb1islet cells, at an effector to target cell ratio of 25. The lack of activity against NOD.H2^nb1islets indicates that these cytotoxic T cell lines recognized β cell autoantigens presented by K^d rather than D^b MHC class I molecules. NOD.H2^nb1islet cells do not exhibit a generalized resistance to cytotoxic T cell-mediated killing, as both NOD.H2^nb1and NOD-scid islet cells are killed by the previously described D^b-restricted T cell clone B/L (22), which recognizes the H3a^b minor histocompatibility antigen expressed by NOD mice (data not shown). Our β cell-autoreactive T cell lines only exhibited detectable cytotoxicity against NOD-scid islets in 20-hr, but not in 4-hr, ⁵¹Cr release assays. Three separate groups have made similar observations with other CD8⁺ β cell-autoreactive T cell lines from NOD mice (10, 12, 13). This may be the result of the finding that at least some diabetogenic CD8⁺ T cells in NOD mice destroy β cells primarily through a Fas/Fas ligand interaction rather than through a more efficient perforin-mediated mechanism (26). We generated 10 T cell lines from the islets of young prediabetic NOD mice and all visually destroyed islets after several days in culture. However, two of the lines failed to exhibit detectable cytotoxicity on NOD-scid islets in our ⁵¹Cr release assay system and were not studied further.

Table 1.

CD8⁺ T cells isolated from the islets of 5- to 6-week-old NOD mice exert K^d-restricted cytotoxic activity against pancreatic β cells

Effector T cells	% specific cytotoxicity
	NOD-scid islet cells (Kd, Db)	NOD.H2nb1 islet cells (Kb, Db)
AI4	19.8, 36.8, 21.4	0
AI11.A5	13.2	0.6
AI11.A7	11.0	0
AI11.B2	12.3	1.1
AI12.B1	16.8, 23.9, 18.0	0.4
AI15.A10	18.7	0

Data represent the percentage of specific cytotoxicity exerted against the indicated target cells by the listed CD8⁺ T cells at an E/T ratio of 25:1 in 20-hr chromium release assays. 

In an Early Stage of Insulitis, β Cell-Autoreactive CD8⁺ T Cells Use a Prevalent TCR α Chain Gene Rearrangement.

For any individual prediabetic NOD mouse in the earliest stages of insulitis, it was not known whether the islet-infiltrating CD8⁺ cytotoxic T cells were monoclonal or polyclonal. The degree of diversity present across many individuals was also not known. To address these questions, we used PCR-based oligonucleotide typing of TCR cDNAs from the T cell populations that we had derived and determined which and how many Vα and Vβ gene families were used.

Most of our β cell-autoreactive CD8⁺ T cell lines used only one to a few TCR Vα and Vβ gene families (Table 2), suggesting that our isolation protocol had enabled us to derive populations that were monoclonal or oligoclonal. However, when considering all isolates from each individual, β cell-autoreactive CD8⁺ T cells isolated from three of five mice used multiple TCR Vα and Vβ gene families. These results indicate that, within any individual and across multiple NOD mice, the CD8⁺ T cells contributing to the earliest initiation of autoimmune β cell destruction are not monoclonal. Nonetheless, all but two of the CD8⁺ T cell lines contained a TCR α chain rearrangement that used a Vα17 gene segment. This is significant because, in general, the Vα17 gene family is not used frequently by CD8⁺ T cells in NOD mice (ref. 12; see Discussion). Table 2 also shows that five of eight of our lines used a Vβ8 gene segment. However, the significance of this finding is less clear because FACS analysis has shown that a high percentage (≈30%) of all splenic T cells from 5-week-old NOD mice are also Vβ8⁺ (27). This indicates that Vβ8 is a commonly used gene family in this strain.

Table 2.

TCR Vα and Vβ families used by β cell-cytotoxic CD8⁺ T cell lines

Mouse	T Cells	Vα	Vβ
1	AI4	8	2
2	AI11.A5	17	16
	AI11.A7	17	n.d.
	AI11.B2	2, 17, 18	8, 13
3	AI12.B1∗	4, 10, 17	8, 11, 16
4	AI15.A3	1, 3, 4, 5, 13, 17, 20	6, 8, 10, 11, 13, 14
	AI15.A10	5	8
5	AI15.F5	17	8

CD8⁺ T cells were isolated from the islets of prediabetic 5- to 6-week-old NOD mice, and TCR Vα and Vβ family usage was determined by reverse transcription–PCR. PCR products found to represent nonproductive gene rearrangements are not included. Vα and Vβ gene families were assigned according to Arden et al.(23). n.d., not determined. 

^*This line was later sorted into three clonotypes: Vα10/Vβ16, Vα4/Vβ11, and Vα17/Vβ8. 

In the case of the β cell-cytotoxic lines that were clearly not monoclonal (AI11.B2, AI12.B1, and AI15.A3), we cannot be sure that all of the cells in each line were cytotoxic because the cytotoxicity reported in Table 1 might reflect the activity of only a subset of the T cell clonotypes making up each line. However, all of the T cells are β cell-autoreactive, as stimulation with islet antigen promoted their survival and growth in vitro.

To further characterize the TCR repertoire of autoreactive CD8⁺ T cells isolated from early insulitic lesions, we partially sequenced the amplified TCR α and β chain cDNAs from our T cell populations. The results of these analyses are shown in Tables 3 (α chains) and 4 (β chains). The lengths of the predicted complementarity-determining region 3 (CDR3) loops, expected to contact the antigenic peptide (32–34), were quite variable for both the α and β chains. The TCR β chains did not bear conserved amino acid sequence motifs in their CDR3 regions, although AI15.A3 and AI15.F5, isolated from two different mice, had Vβ8⁺ TCRs with identical CDR3β sequences. However, the junctional sequences of the TCR α chains were more homogeneous, in that four of the six Vα17 gene rearrangements used the Jα42 gene segment. This same Jα was found joined to a Vα10 gene segment in another T cell line (AI12.B1), and the highly homologous Jα53 was used in yet another (AI15.A3). The recurrent use of these Jα gene segments led to a subset of TCRs having CDR3α loops of identical lengths and all carrying the Jα-encoded amino acid sequence motif GGSN(A/Y)KLT. These results reveal preferential usage of a CDR3α motif among β cell-reactive CD8⁺ T cells in the earliest stage of autoimmune β cell destruction in NOD mice.

Table 3.

TCR α chain CDR3 sequences used by β cell-autoreactive CD8⁺ T cells

Mouse	T Cells	Vα	FW	CDR3	FW	Jα
1	AI4	8	CAL	RTAGANTGKLT	FG	52
2	AI11.A5	17	CAM	RVSGGSNAKLT*	FG	42
	AI11.A7	17	CAM	REGGNYAQGLT	FG	26
	AI11.B2	2	CAA	SAWNNNAPR	FG	43
		17	CAM	RVSGGSNAKLT*	FG	42
		18	CAT	ETTASLGKLQ	FG	24
3	AI12.B1.1	10	CAM	ERWGGSNAKLT	FG	42
	AI12.B1.2	4	CAL	SGTGGYKVV	FG	12
	AI12.B1.3	17	CAM	RDSGGSNAKLT	FG	42
4	AI15.A3	1	CAV	SASGGSNYKLT	FG	53
		3	CAV	SAAGNTRKLI	FG	37
		4	CAL	TTGGNNKLT	FG	56
		17	CAM	REGGGSNAKLT	FG	42
		20	CAA	MATGGNNKLT	FG	56
	AI15.A10	5	CAV	IIYQGGRALI	FG	15
5	AI15.F5	17	CAM	REAGTQVVGQLT	FG	5

TCR cDNAs were amplified by PCR and were sequenced. Vα gene families were assigned according to Arden et al. (23). Jα genes were numbered according to Koop et al. (28), except for Jα26 because their compilation of Jα genes shows a frameshift deletion for Jα26, but the sequence present in AI11.A7 matched that described in the original publication (29) and did not have this deletion. The frequently observed Jα-encoded CDR3 motif GGSN(A/Y)KLT is underlined. FW, framework residues. 

^*AI11.A5 and AI11.B2 had TCR α chain CDR3s encoded by identical nucleotide sequences, but they were not the same clonotype, as their Vβ usage was different (Table 2). 

DISCUSSION

It was demonstrated earlier that, when splenic T lymphocytes are isolated from young prediabetic NOD donors, both CD8⁺ and CD4⁺ T cells are required to adoptively transfer diabetes to NOD-scid recipients whereas CD4⁺ T cells from overtly diabetic NOD donors can transfer disease independently of CD8⁺ T cells (4). Consistent with these results and also suggesting a role for CD8⁺ T cells, it subsequently was shown that splenic T cells from young prediabetic NOD donors could transfer IDDM to MHC class I-positive, but not MHC class I-negative, NOD-scid recipients unless the class I-negative recipients previously had been engrafted with class I-positive islets (14). Thus, recognition of pancreatic β cells by MHC class I-restricted T cells is essential to the initiation of autoimmune IDDM, but, as the pathogenic process progresses, populations of CD4⁺ T cells eventually develop that independently can cause disease. However, it remained unknown at what point the diabetogenic process no longer required contributions from MHC class I-restricted T cells. We addressed this question by demonstrating that T cells from 4- to 14-week-old prediabetic NOD donors can transfer disease to MHC class I-positive, but not MHC class I-negative, NOD-scid recipients (Fig. 1A). Only T cells from NOD donors that already have developed overt diabetes can transfer disease in an MHC class I-independent manner. Thus, if this adoptive transfer model accurately reflects the spontaneous disease process in standard NOD mice, MHC class I-restricted CD8⁺ T cells are required for all but the final stages of IDDM development. However, even though not strictly required, CD8⁺ T cells also participate in the end stages of diabetes development in NOD mice, as T cells from diabetic NOD donors transfer disease into MHC class I-negative recipients less efficiently than in MHC class I-positive recipients (Fig. 1A; ref. 14).

To understand the basis for the MHC class I-dependent T cell requirement in IDDM, we sought to define the characteristics of the β cell-autoreactive CD8⁺ T cells participating in the initial stages of diabetogenesis. Although NOD islets express both K^d and D^b MHC class I molecules, the β cell-cytotoxic CD8⁺ T cell lines for which we determined MHC restriction were all K^d-restricted (Table 1). One or both of the specific MHC class I alleles expressed by NOD mice previously was shown to be important for disease development because a stock of NOD mice congenic for the H2^ct haplotype sharing class II but not class I alleles with H2^g7 is relatively diabetes-resistant (15). Studies in humans also have begun to show that certain MHC class I alleles are associated with an increased risk for IDDM development (35–37). Our work demonstrates the importance of H-2K^d MHC class I molecules in presenting peptides to CD8⁺ T cells during the initiation phase. It is of interest that, in studies on overtly diabetic mice, several β cell-cytotoxic CD8⁺ T cell lines also were K^d-restricted (12). Of course, these findings do not preclude a role for H-2D^b in IDDM development, and it should be noted that a recent preliminary report indicated that D^b-restricted β cell-cytotoxic CD8⁺ T cells could be isolated from the spleens of NOD mice (38).

To refine our understanding of the pathogenic process, we characterized the early diabetogenic CD8⁺ T cells with regard to their degree of clonality and their TCR Vα and Vβ gene usage and junctional sequences. Previous studies had used reverse transcription–PCR to investigate the TCR repertoire of islet-infiltrating T cells in 4- to 5-week-old prediabetic NOD mice and had showed that multiple TCR Vβ gene families were used (27, 39). However, these studies did not analyze MHC class I-restricted CD8⁺ and class II-restricted CD4⁺ T cells separately, nor could they discriminate true β cell-autoreactive T cells from those that may have been merely trafficking through the pancreas. Although an earlier study did examine specifically the TCR repertoire of β cell-autoreactive CD8⁺ T cells, it surveyed clonotypes isolated from islets of both overtly diabetic and prediabetic NOD mice of unspecified age (12). Given that age-dependent “clonotype spreading” occurs at least among diabetogenic CD4⁺ T cells (40, 41), we thought it important to perform a TCR gene usage survey that focused solely on MHC class I-restricted T cells present at the early initiative phases of autoimmune β cell destruction.

We found that β cell-cytotoxic CD8⁺ T cells isolated from the earliest insulitic lesions of NOD mice do use a diverse set of TCRs. Thus, 8 Vβ gene families and 11 Vα gene families were represented (Table 2). However, these β cell-autoreactive CD8⁺ T cells were characterized by a TCR α chain repertoire that was restricted in several important respects (Table 3). Of the sequenced α chains, 6 of 16 had the amino acid sequence motif GGSN(A/Y)KLT in their CDR3 loops because of recurrent usage of the highly homologous Jα42 and Jα53 gene segments. The TCR α chain repertoire also exhibited a prevalence of Vα17 family members and, in four of six Vα17⁺ TCRs, the Jα42 gene segment. These restricted features of the TCR α chain repertoire are significant because the Vα17, Jα42, and Jα53 gene segments are not used commonly by CD8⁺ T cells in the NOD mouse, as demonstrated by studies of Santamaria et al. (12), who sequenced 31 functional TCR α chain gene rearrangements from CD8⁺ NOD spleen cells and found that none contained Vα17 gene segments or the CDR3α motif GGSN(A/Y)KLT. (In the report by Santamaria et al., the Vα17 gene segments are referred to as Vαn1.1 and Vαn1.2.)

One possible explanation for the restricted TCR α chain repertoire is that these particular α chains in some way facilitate interaction between the TCR and the K^d molecule and therefore are used preferentially by K^d-restricted T cells. However, as mentioned, α chains containing Vα17, Jα42, or Jα53 gene segments were not found among 31 functional TCR α chain gene rearrangements from CD8⁺ NOD spleen cells, which should have contained at least some K^d-restricted T cells (12). Further, a published analysis of the TCR Vα and Jα gene segments used by 28 K^d-restricted T cell clones derived from BALB/c mice showed that their TCR α chains were quite heterogeneous and that none contained Vα17 gene family members or the Jα-encoded GGSN(A/Y)KLT motif (24). Therefore, although it is possible that the selection for particular TCR α chains among β cell-autoreactive T cells reflects their K^d-restriction in the setting of the NOD autoimmune-prone background, we have no evidence in favor of this argument.

Alternate hypotheses are suggested by the recently solved crystal structures of TCR/peptide/class I MHC complexes (32–34). All have shown that the TCR α chain contacts the N-terminal half of a class I MHC-bound peptide by using its CDR1 and CDR3 loops, which are respectively encoded within the germline V gene segment and at the V-J junction. Amino acid residues in the α1 and α2 helices of the class I MHC molecule near the N-terminal half of the bound peptide also are contacted by all three CDR loops of the TCR α chain. Thus, the restricted TCR α chain repertoire among β cell-autoreactive T cells might imply that a subset of these cells is recognizing a common structural feature of certain K^d/peptide complexes on β cells. Alternatively, T cells with related TCR α chains may be recognizing a single peptide in the context of the K^d molecule. Our finding of a much more diverse TCR β chain repertoire does not preclude this possibility, as it is known that, for certain MHC/peptide complexes, the responding T cells express limited TCR α chains but far less restricted β chains (42–45). Further, crystallographic studies show that, when an identical MHC/peptide complex is being recognized by two different TCRs, the TCR β chain residues that interact with peptide/MHC residues can vary in each case (32, 34).

Independent evidence supporting the biological relevance of our TCR gene usage results comes from Santamaria et al. (12). Using a very different cell isolation protocol, they found that β-cell autoreactive CD8⁺ T cells isolated from two overtly diabetic NOD mice and one prediabetic mouse of unspecified age used Vα17-Jα42 gene rearrangements (referred to as Vαn1.1-Jα34 in ref. 12). Further, our AI12.B1.3 clone and the β cell-autoreactive T cells (NY7.2, NY8.3, and NY9) isolated by Santamaria et al. not only all expressed TCR α chains encoded by Vα17-Jα42 gene rearrangements, but they also all had identical CDR3α sequences (11, 12). Even more compelling evidence for the biological importance of T cells bearing Vα17-Jα42 TCR α chains comes from studies of NOD mice transgenic for the TCR β chain of NY8.3. These mice show an earlier onset of IDDM and are characterized by islet-derived, but not peripheral, CD8⁺ T cells that show a prevalent usage of an endogenously derived Vα17-Jα42 TCR α chain identical to that of the original NY8.3 T cell clone (46). Thus, when considered along with these independent observations, our finding that CD8⁺ T cells from early insulitic lesions commonly use a Vα17-Jα42 gene rearrangement strongly implicates T cells bearing such TCR α chains in the pathogenesis of IDDM in NOD mice.

Our studies have shown that class I MHC-dependent T cell responses are required for all but the terminal stages of diabetogenesis in NOD mice. Although a number of the antigens recognized by β cell-autoreactive CD4⁺ T cells have been characterized (47), to date no antigens recognized by β cell-cytotoxic CD8⁺ T cells have been identified. It is not known whether a limited or diverse set of antigenic peptides will be recognized, nor is it known whether these peptides will be derived from a few or many β cell proteins. However, considering the importance of class I MHC-restricted T cell responses in the pathogenesis of IDDM, the eventual identification of the antigens recognized by these T cells in the earliest phase of β cell destruction should lead to a better understanding of how disease is initiated. Our isolation and characterization of β cell-cytotoxic CD8⁺ T cells present at the inception of IDDM development should provide valuable reagents for this necessary next step.

Table 4.

TCR β chain CDR3 sequences used by β cell-autoreactive CD8⁺ T cells

Mouse	Cells	Vβ	FW	CDR3	FW	Jβ
1	AI4	2	CTC	SAHRGLGNTLY	FG	1.3
2	AI11.A5	16	CAS	SSDWGYEQY	FG	2.6
	AI11.B2	8S1	CAS	SSDTYEQY	FG	2.6
		13	CAS	SRTGGRGEQY	FG	2.6
3	AI12.B1.1	16	CAS	SLARPSAETLY	FG	2.3
	AI12.B1.2	11	CAS	SLGDWGNEQY	FG	2.6
	AI12.B1.3	8S3	CAS	SGDRYEQY	FG	2.6
4	AI15.A3	6	CAS	SRSGTGTEVF	FG	1.1
		8S1	CAS	SGTGGQNTLY	FG	2.4
		10	CAS	SLGQYEQY	FG	2.6
		11	CAS	SWTGGGTLY	FG	2.4
		13	CAS	IKLGGQNTLY	FG	2.4
		14	CAW	SHSYYNSPLY	FA	1.6
	AI15.A10	8S2	CAS	GDAGGDQAPL	FG	1.5
5	AI15.F5	8S1	CAS	SGTGGQNTLY	FG	2.4

TCR cDNAs were amplified by PCR and were sequenced. Vβ gene families were assigned, and specific Vβ8 gene segments were numbered, according to Arden et al. (23). Jβ genes were assigned according to Malissen et al. (30) and Gascoigne et al. (31). FW, framework residues. 

ABBREVIATIONS

NOD

nonobese diabetic

IDDM

insulin-dependent diabetes mellitus

TCR

T cell receptor

MHC

major histocompatibility complex

CDR

complementarity-determining region

FACS

fluorescence-activated cell sorting

IL

interleukin