Main

Many models have supported the mechanism of molecular mimicry by showing that foreign antigens can activate T cells that are crossreactive with self antigens, leading to the induction of autoimmune disease6,7,8,9,10,11,12. This has been further supported by structural studies that also suggest that the interactions between TCR, peptides and the major histocompatibility complexes (MHC) have a degree of flexibility13,14. Although several studies have assessed the importance of the affinity required for T-cell activation15,16,17,18,19,20, the role of affinity in the development of autoimmunity has not been extensively examined21,22. Therefore we wanted to evaluate the interactions between the TCR and peptide-MHC ligand required to initiate T-cell activation that will lead to substantial immunopathological damage. To monitor autoimmunity in vivo, we used a transgenic mouse model of diabetes (RIP-gp), which expresses the lymphocytic choriomeningitis virus glycoprotein (LCMV-gp) in pancreatic beta cells under the control of the rat insulin promoter (RIP)11. In a second transgenic model, P14, mice express a transgenic TCR specific for the gp33 epitope of the LCMV-gp (with the amino acid sequence KAVYNFATC) presented by H-2Db (ref. 23). Neither RIP-gp single-transgenic nor P14/RIP-gp double-transgenic mice become spontaneously diabetic. But immunization of RIP-gp or P14/RIP-gp mice with LCMV leads to the activation of T cells specific for LCMV-gp, resulting in beta-cell destruction and diabetes11. To evaluate whether the affinity between the TCR and peptide-MHC ligand is important in the development of autoimmunity, we defined the affinity of the P14 TCR for LCMV variant epitopes and the ability of these LCMV variants to induce an immune response and autoimmunity in vivo.

Here we focused on two LCMV variants derived from the wild-type LCMV (LCMV-L6F and LCMV-C4Y), which have an alteration in the major LCMV-gp epitope. The position 6 mutation of phenylalanine to leucine in the LCMV-L6F variant and position 4 mutation of tyrosine to cysteine in the LCMV-C4Y variant did not alter viral replication or CTL activation, as infection of C57Bl/6 mice with wild-type LCMV, LCMV-L6F or LCMV-C4Y led to a comparable expansion and elimination of virus from the spleen (see Supplementary Fig. 1 online). All viruses also induced comparable nucleoprotein peptide–specific CTLs in C57Bl/6 mice (Supplementary Fig. 1 online).

We characterized the properties of the variant peptide epitopes in several in vitro assays. The wild-type gp33 peptide and the variant L6F and C4Y peptides showed similar binding to H-2Db as determined by the level of MHC on RMA-S cells (Supplementary Fig. 2 online). Peptide binding stabilizes the expression of MHCI on the surface of RMA-S cells and thus the level of MHC on the surface is proportional to the affinity of the peptide for MHC. These results are consistent with the crystal structure of gp33 binding to H-2Db, in which the amino acids at positions 4 and 6 point out of the MHC groove toward the TCR19,24. Proliferation assays using P14 TCR transgenic splenocytes showed that the gp33 peptide induced the highest levels of proliferation, followed by L6F and then C4Y. We observed a similar pattern of IFN-γ production after stimulation of P14 T cells with the variant peptides (Supplementary Fig. 2 online). We also tested the ability of effector T cells taken from a P14 transgenic mouse infected with wild-type LCMV to recognize target cells pulsed with the wild-type or variant peptides (Supplementary Fig. 2 online). Activated P14 TCR transgenic T cells lysed both gp33- and L6F-pulsed EL4 target cells. Notably, C4Y peptide–pulsed target cells were lysed by P14 CTLs, although with slightly lower efficiency. The P14 CTLs could recognize L6F- and C4Y-pulsed target cells; this suggests that the orientation and conformation of the presented peptide is similar to the wild-type ligand. Collectively, these assays show that gp33, L6F and C4Y all bind to H-2Db. Although all peptides are recognized by activated P14 CTLs, gp33 is a stronger agonist than L6F. C4Y was an antagonist peptide that inhibited proliferation of P14 T cells to the gp33 agonist peptide (Supplementary Fig. 2 online).

To examine the strength of the interactions between the P14 TCR and MHC bound with gp33 or the variant ligands, we determined the affinities of these interactions with surface plasmon resonance (using Biacore apparatus) (Fig. 1a–c). The equilibrium dissociation constant (KD) of gp33 was 3.5 μM, indicating that it had the strongest affinity for the P14 TCR. L6F had an affinity that was approximately five times lower (KD = 19 μM), whereas the antagonist peptide C4Y had a KD of 69 μM. Therefore, these TCR-peptide–MHC measurements correlated with the results from the functional assays, in which gp33 was the strongest agonist ligand, followed by the L6F and C4Y peptides.

Figure 1: Characterizing P14 T-cell responses to variant viruses.
figure 1

(ac) Binding of soluble P14 TCR to (a) H-2Db gp33 (b) H-2Db L6F and (c) H-2Db C4Y. H-2Db-HY (to which the P14 TCR does not bind) was used as a control. (d) Expression of cell-surface markers Vα2, CD44, LFA-1 (CD11a), CD49d and CD62L on CD8+ T cells from P14 TCR transgenic mice that were immunized with wild-type LCMV, LCMV-L6F, LCMV-C4Y or vaccinia virus or were left uninfected. Profiles are shown from gated CD8+ populations. The values indicated are median fluorescence intensities. Representative data from six mice per treatment. (e) CTL activity from wild-type LCMV-, LCMV-L6F- or LCMV-C4Y-infected P14 transgenic mice was assessed on gp33-pulsed EL4 target cells. Lysis of negative control AV target peptide–pulsed cells was <5%. E/T ratio, effector/target ratio. (f) Percentage of CD8+Vα2+ T cells of the total CD8+ population, determined by flow cytometry.

Previous studies using an assay in which tetramer binding competed with MHC-peptide monomer have shown that 20-fold more L6F-Db monomer than gp33-Db monomer was required to compete with gp33-Db tetramer25. We carried out further competition experiments using a tetramer containing a strong agonist peptide, A3V25, to directly compare the relative ability of A3V to compete with gp33-Db, L6F-Db or C4Y-Db monomers. We found a 16- to 20-fold difference in the ability of L6F-Db and gp33-Db monomers to compete with the A3V-Db tetramer (Supplementary Table 1 and Supplementary Fig. 2 online). C4Y-Db was more than 35 times less efficient than gp33-Db, suggesting the same pattern (gp33 > L6F > C4Y) as the surface plasmon resonance assay. The fold differences in affinity between the monomer inhibition and surface plasmon resonance assays may reflect the cell-surface versus solution binding and the effect of CD8 on the stability of binding25,26.

To evaluate the ability of LCMV variants to activate transgenic P14 T cells, we immunized P14 TCR transgenic mice with wild-type LCMV, LCMV-L6F, LCMV-C4Y or vaccinia virus and compared the modulation of various markers on the cell surface of CD8+ T cells (Fig. 1d). Vaccinia virus acted as a negative control because it does not express known crossreactive ligands. Wild-type LCMV and LCMV-L6F infection led to similar changes in the activation markers CD44, LFA-1, CD49d and CD62L, as well as downregulation of the TCR in P14 T cells. No changes in activation markers were observed after infection with the low-affinity variant LCMV-C4Y or vaccinia virus.

To assess effector function, we assayed splenocytes from P14 mice immunized with wild-type LCMV, LCMV-L6F or LCMV-C4Y for cytotoxic function on EL4 target cells pulsed with gp33. Infection with wild-type LCMV and LCMV-L6F induced strong CTL activity (Fig. 1e). Notably, the low-affinity LCMV-C4Y variant induced efficient but not maximal CTL activity in P14 T cells. In vivo analysis showed that P14 T cells proliferated in the spleens of P14 LCMV-L6F immunized mice, although at lower levels than the mice immunized with wild-type virus (Fig. 1f). Therefore, the medium-affinity virus LCMV-L6F induced CTL function in vivo, despite a limited ability to promote proliferation. The low-affinity LCMV-C4Y was not as effective at inducing cytotoxic function in P14 T cells. Previous studies reported LCMV-L6F as a virus-escape variant from C57Bl/6 mice that were immunized with LCMV-GP peptides24,27. But our data show that the 'escape variant' virus that expressed an epitope with an affinity for the P14 TCR that was five times lower induced efficient T-cell activation and effector function.

To address the role of affinity in the induction of autoimmune disease, we immunized P14/RIP-gp mice with wild-type LCMV or LCMV variants. Immunization with wild-type LCMV resulted in diabetes in all the animals tested (Fig. 2a), but infection with the medium-affinity LCMV-L6F variant resulted in only 6 diabetic mice of the 14 infected (Fig. 2b). Notably, the kinetics of disease after LCMV-L6F were similar to the kinetics after wild-type LCMV infection. None of the mice immunized with the low-affinity LCMV-C4Y became diabetic (Fig. 2c). We carried out further experiments to ensure that diabetes in the P14/RIP-gp mice immunized with wild-type LCMV and LCMV-L6F resulted from activation of P14 transgenic T cells and not the nontransgenic lymphocyte population. Bone marrow from P14 Rag2−/− mice was used to reconstitute lethally irradiated RIP-gp mice. All chimeric animals infected with wild-type LCMV became diabetic, whereas only half of the animals (3 of 6) given LCMV-L6F developed diabetes (Supplementary Fig. 3 online). Despite the effective induction of CTLs by LCMV-L6F in P14 TCR transgenic animals, LCMV-L6F showed a reduced ability to induce proliferative responses in vivo, which corresponded with a lower incidence of diabetes. This suggests that the number of CTLs limited the destruction of the islet cells and is consistent with previous studies that have shown that the number of autoreactive CTLs can limit disease progression21,28.

Figure 2: Limited induction of diabetes by the LCMV-L6F variant virus despite effective ex vivo CTL function and islet infiltration.
figure 2

(ac) Diabetes (left panels) and insulitis (right panels) as assessed by immunohistochemistry for CD8+ infiltrates (stained red) in P14/RIP-gp mice that were immunized with (a) wild-type LCMV (b) LCMV-L6F (c) LCMV-C4Y. Each line depicts blood glucose of a single mouse. (d) Islets from P14 single-transgenic mice infected with wild-type LCMV. Representative profiles are shown. (e) Summary of the immunohistochemistry results for CD8+ infiltration of the islets. Results are from at least 6 mice and 20 islets per group.

Infection with either wild-type LCMV or LCMV-L6F resulted in substantial islet infiltration by CD8+ T cells (Fig. 2a,b,e). LCMV-C4Y infection resulted in less islet infiltration; however, even with this low-affinity virus, more than 50% of islets were highly infiltrated (Fig. 2c,e). The control single-transgenic P14 mice immunized with wild-type LCMV had virtually no insulitis (Fig. 2d,e). Therefore, the absence of diabetes induced by the lower-affinity LCMV-variant viruses did not result from the inability of P14 T cells to infiltrate the pancreatic islets.

We also carried out studies to examine the influence of LCMV-L6F on the induction of immunity and autoimmunity in mice with a normal T-cell repertoire. LCMV-L6F also induced efficient gp33-specific cytotoxic activity in C57Bl/6 mice, although the activity was lower than in mice infected with wild-type LCMV (Fig. 3a,b). Diabetes induction was also altered. All RIP-gp mice given wild-type LCMV became diabetic (Fig. 3c). Notably, infection with LCMV-L6F resulted in diabetes in fewer than half of the animals (4 of 10) (Fig. 3d). Islets from both wild-type LCMV- (Fig. 3c) and LCMV-L6F-infected (Fig. 3d) RIP-gp mice had a high degree of CD8+ T-cell infiltration, in contrast with uninfected RIP-gp (Fig. 3e) or wild-type LCMV-infected nontransgenic C57Bl/6 mice (data not shown). Therefore, in mice with a normal repertoire, infection with LCMV-L6F resulted in slightly reduced CTL activity specific for gp33 and a lower incidence of diabetes.

Figure 3: Effective CTL induction by the LCMV-L6F variant virus with limited diabetes.
figure 3

(a,b) Lysis of gp33 peptide–pulsed EL4 target cells by splenocytes from C57Bl/6 mice infected with (a) wild-type LCMV or (b) LCMV-L6F 8 d earlier. E/T ratio, effector/target ratio. Each line depicts CTL activity from an individual mouse (n = 6). (c,d) RIP-gp mice were infected with (c) wild-type LCMV or (d) LCMV-L6F and the onset of diabetes (left panels) and day 8 CD8+ T cell infiltration (right panels) were monitored. In c, n = 7; in d, n = 10. Symbols distinguish individual mice. (e) Pancreatic sections from uninfected RIP-gp mice are shown.

Notably, among both P14/RIP-gp and RIP-gp mice infected with LCMV-L6F, approximately 50% developed diabetes, despite the presence of functional LCMV-gp-specific CTLs and insulitis. One possible explanation is that the progression to diabetes may be limited by the natural negative regulation of the immune response. Cbl-b has an important role in negatively regulating T-cell responses in vivo, and its absence results in susceptibility to experimentally induced and spontaneous models of autoimmunity4,5. Further evidence showed that Cblb is an important susceptibility gene in the spontaneous type 1 diabetes animal model, the Komeda rat29. Therefore, to address the importance of negative regulation of T-cell function on autoimmunity, we generated RIP-gp/Cblb deficient mice. All RIP-gp/Cblb−/− or P14/RIP-gp/Cblb−/− mice infected with LCMV-L6F became diabetic (Fig. 4a and data not shown). Therefore the natural negative regulation of an immune response clearly limits autoimmunity in vivo.

Figure 4: Diabetes induction in RIP-gp mice is limited by Cbl-b.
figure 4

(a) Diabetes induction in RIP-gp Cblb−/− mice that were infected with LCMV-L6F. (b) In vivo proliferation, evaluated as percentage of CD8+Vα2+ cells of the total CD8+ population 7 d after infection as compared to uninfected control animals. (c,d) Diabetes in (c) RIP-gp Cblb−/− or (d) RIP-gp Cblb+/+ mice after wild-type LCMV infection. (e) CTL induction in Cblb+/+ and Cblb−/− mice 8 d after LCMV-L6F immunization. Each line represents one mouse. (f) Summary of CD8+ T-cell infiltration into the islets in RIP-gp Cblb+/+ and RIP-gp Cblb−/− on days 6, 8 and 15 after LCMV-L6F infection. Symbols represent individual mice from the same group and treatment.

To understand the reasons for the increased diabetes incidence in RIP-gp Cblb−/− mice, we analyzed the ability of Cbl-b-deficient T cells to respond to LCMV-L6F. Modulation of CD44, LFA-1, CD49d and CD62L upon LCMV-L6F infection was similar on both Cbl-b-deficient and wild-type T cells (Supplementary Fig. 4 online). In vivo, Cbl-b-deficient P14 T cells proliferated more in response to either wild-type LCMV or LCMV-L6F than P14 Cblb+/+ T cells (Fig. 4b). This correlated with faster kinetics of disease induction by wild-type LCMV in RIP-gp Cblb−/− animals than in RIP-gp Cblb+/+ (Fig. 4c,d and Supplementary Table 1 online). In 8 of 12 experiments, the Cblb−/− mice showed enhanced CTL activity after infection with LCMV-L6F (Fig. 4e). However, at this point it is not clear why enhanced CTL activity is not consistently seen directly ex vivo.

Infiltration of the islets by CD8+ T cells 6 or 8 d after infection was similar in RIP-gp/Cblb−/− compared to RIP-gp/Cblb+/+. But by day 15, more than 80% of islets in the RIP-gp/Cblb−/− mice, but only about 40% in RIP-gp/Cblb+/+ mice, were heavily infiltrated (Fig. 4f). Therefore, we have shown that Cbl-b has a crucial role in regulating CD8-mediated immune pathology in vivo by increasing the proliferative burst of CTLs and prolonged infiltration of activated CD8+ T cells.

In summary, we have used two virus variants, LCMV-L6F and LCMV-C4Y, to show that TCR affinity and negative regulatory molecules have key roles in preventing the induction of autoimmune disease. These studies lend further support to the notion that most autoimmune diseases are a combination of genetic and environmental factors. In general, the affinity between the TCR and activating peptide-MHC ligand is important and limits autoimmunity by molecular mimicry. The reason some individuals are predisposed to autoimmunity may relate to differences or defects in negative regulation, which normally limit T-cell responses and prevent disease.

Methods

Viruses.

Wild-type LCMV (WE strain) and LCMV-L6F, which is derived from LCMV-WE, were originally obtained from F. Lehmann-Grube27. The LCMV-C4Y variant was isolated from P14 TCR transgenic mice infected with 106 PFU of LCMV-WE, as described previously30. To make viral stocks, we grew viruses on L929 cells for 48 h and subsequently titrated as described previously31.

Mice and diabetes monitoring.

RIP-gp (C57Bl/6 background)11 and P14 TCR transgenic lines23 backcrossed 12 times to C57Bl/6 mice (R. Ahmed) were interbred to yield P14/RIP-gp mice. Cblb−/− mice were backcrossed 6 times to C57Bl/6 mice4. Diabetes progression in P14/RIP-gp or RIP-gp mice was monitored by blood glucose measurements 2–3 times per week after infection with 5 × 105 plaque-forming units (PFU) of wild-type LCMV, LCMV-L6F, or LCMV-C4Y. Single-transgenic RIP-gp mice were infected with 2,000 PFU. A mouse was considered diabetic when its blood glucose reached 14 mM. We measured blood glucose with Accu-Chek III gluco- meters (Boehringer Mannheim). All mice were maintained under specific pathogen-free conditions at the Ontario Cancer Institute Animal Resource Centre. All protocols used were approved by the Ontario Cancer Institute/University Health Network Animal Care Committee.

Peptide binding assays.

For assessment of peptide binding, we used the RMA-S class I stabilization assay32. NP118-127, which is an H-2Dd-restricted peptide, was used as a negative control. We calculated percent mean above background as (median fluorescence staining of sample – median fluorescence staining of negative control peptide)/(median fluorescence staining of negative control peptide) × 100%.

Cellular proliferation assays.

We cultured 105 P14 TCR transgenic splenocytes per well (96-well flat-bottomed plates) in triplicate in Iscove's Modified Dulbecco medium (IMDM) supplemented with 10% heat-inactivated fetal bovine serum (Sigma) together with 105 C57Bl/6 splenocytes ('feeder cells') and peptide. After 48 h, we pulsed the plates with 1 μCi [3H]thymidine (Perkin Elmer) per well overnight and harvested onto glass fiber filters. A scintillation counter (Topcount, Canberra Packard) measured [3H]thymidine uptake. The antagonist assays were done as previously described32.

51Cr-release assays.

EL4 target cells were pulsed with peptide for 90 min at 37 °C in the presence of 400 μCi/ml 51Cr (Perkin Elmer) in IMDM supplemented with 10% FCS. Cells were washed and 104 cells were transferred to a well of a round-bottom 96-well plate. Spleen cell suspensions from LCMV-infected mice were serially diluted and mixed with peptide-pulsed target cells. Plates were incubated for 5 h at 37 °C. We counted 70 μl of supernatant from each well using a Wallac Wizard counter (Perkin Elmer). Maximal release was induced by adding 1 M HCl to targets. We calculated percentage of specific lysis as (c.p.m. sample release – c.p.m. spontaneous release)/(c.p.m. maximal release – c.p.m. spontaneous release) × 100%.

Measurement of P14 soluble TCR binding.

P14-soluble TCR, linked by an engineered non-native disulfide bond, was produced as by Boulter et al.33 but using mouse P14 TCR. Biacore surface plasmon resonance assays were conducted as described previously34.

Tetramer binding measurements.

For the tetramer inhibition experiments, we prepared single-cell suspensions from spleens of P14 transgenic mice and then stained the splenocytes with 1nM strong agonist A3V-Db together with unlabeled pMHC monomer (10−5–10−9 M) and anti-TCRβ monoclonal antibody (H57-597) such that the monomer pMHC competes with the tetramer for TCR binding25. The difference between the inhibition curves of gp33-Db or L6F-Db was determined as a relative measure of affinity.

Flow cytometric analyses.

We immunized P14 TCR transgenic mice with 5 × 105 PFU LCMV. Three d later we harvested spleens and prepared single-cell suspensions. We stained lymphocytes with antibodies to Vα2, CD8, CD44, LFA-1, CD49d, CD62L and CD69 (BD Pharmingen), and performed detection of biotin-conjugated antibodies using streptavidin–Cy-Chrome (BD Pharmingen). Live events were collected based on forward- and side-scatter profiles on a FACScan flow cytometer (Becton Dickinson) and analyzed using CELLQuest software (Becton Dickinson).

Immunohistochemistry.

We immersed freshly removed pancreata in phosphate-buffered saline (PBS) and snap-froze them in liquid nitrogen. For the staining of cell differentiation markers, frozen tissue sections 5 μm thick were cut in a cryostat and stained as previously described with primary rat monoclonal antibodies to CD8 (YTS 169)35. Infiltration scale: 0, no infiltration; 1, periinsulitis; 2, minor (up to 30% of islet infiltrated); 3, partial (30–70%); 4, complete (>70% of islet infiltrated).

Generation and diabetes induction of the chimeras.

We removed and washed bone marrow from P14 Rag2−/− mice, resuspended it in HBSS and and injected 5 × 106 bone marrow cells intravenously into RIP-gp mice that were irradiated with 10 Gy. We analyzed blood from the chimeric mice 12 weeks later by flow cytometry for CD4, CD8 and Vα2 expression to ensure reconstitution (data not shown). We subsequently immunized mice with 2,000 PFU of wild-type LCMV or LCMV-L6F and monitored them for diabetes.

Note: Supplementary information is available on the Nature Medicine website.