Self and viral peptides can initiate lysis by autologous natural killer cells

Abstract

Natural killer (NK) cells are inhibited by specific allotypes of class I major histocompatibility complex ligands recognized by polymorphic inhibitory receptors (e.g., NKIR1 and NKIR2). NK1- and NK2-specific clones recognize two groups of HLA-C allotypes that are distinguished by a dimorphism at residue 80 in the α1 helix (αLys-80 and αAsn-80, respectively). “Empty” HLA-Cw7 expressed in peptide transporter-deficient cells and HLA-Cw7 loaded with several peptides each functioned as inhibitory ligands for NK2 lines and clones. However, loading of HLA-Cw7 with two other peptides derived from glutamic acid decarboxylase or coxsackie virus (each of which has been associated with autoimmune diabetes mellitus) abrogated this inhibitory recognition. Both peptides contained Lys at P8 of the epitope. Substitution of P8 with Ala or two other basic amino acids, His and Arg, resulted in peptides that were inhibitory, as were peptides with P8 Val, Glu, or Asn. The manner in which a Lys at P8 might affect recognition is discussed, together with a hypothesis for a novel mechanism by which an autoimmune disease might be initiated.

Natural killer (NK) cells lyse cellular targets that include certain tumor cell lines and cells infected with intracellular pathogens (1). They are a distinct lymphocyte population that controls the early phase of infection by various pathogens and influence the subsequent T cell responses by secreting potent mediators of inflammation (2). Their ability to differentiate between virally infected and normal cells is controlled by class I major histocompatibility complex (MHC) glycoproteins expressed on target cells, the “missing self” hypothesis (3, 4). Unlike cytotoxic T lymphocytes, which are activated by specific class I MHC–peptide complexes, NK cells are inhibited from killing target cells that express certain class I MHC molecules (5, 6). The specific recognition of polymorphic determinants on class I MHC molecules by human NK cells is mediated by a family of receptors, NKIRs, belonging to the immunoglobulin (Ig) domain superfamily (7, 8). The receptors of NK groups 1 and 2 (NKIR1 and NKIR2, p58 proteins containing two Ig domains) recognize a dimorphism at position 80 of the α1 helix of HLA-C molecules (αLys-80 and αAsn-80, respectively; refs. 9–11). NK3-specific cells (which express NKIR3, a p70 protein) recognize HLA-B allotypes that contain the serologically determined Bw4 public epitope at α77–83, which includes αIle-80 (12). NK4-specific cells are also inhibited by cells with HLA-B molecules containing the Bw4 epitope but require αThr-80 for inhibition (13). Lastly, NK group 5 cells recognize HLA-A3 (14, 15). The NK receptor molecules can exist in one of two discrete isoforms with identical extracellular domains but distinct transmembrane regions and cytoplasmic tails (16, 17). The inhibitory receptors have an uncharged transmembrane span and a long cytoplasmic tail containing two ITIM motifs. Binding of inhibitory NKIR to class I proteins leads to the recruitment of protein tyrosine phosphatases and a signal that blocks NK cell killing (18–20). NK receptors with an activating function have a transmembrane domain that contains a lysine residue and a shortened cytoplasmic domain that lacks the ITIM motifs (8, 16). The lytic behavior of an individual NK clone is defined by the spectrum of NKIR that it expresses. All NK clones identified to date that express activating receptors also have inhibitory receptors, and the inhibitory function appears to be dominant (2).

Several viruses express genes that down-regulate expression of class I MHC proteins, thus providing them with a means of escape from cytotoxic T lymphocytes but rendering them susceptible to lysis by NK cells. The almost ubiquitous expression of class I MHC molecules provides one means by which NK cells discriminate among modified target cells that may have lost or down-regulated surface class I MHC molecules (3). The absence or dysregulation of class I MHC proteins may partially explain how NK cells survey tissue for the normal expression of class I MHC, the missing self hypothesis (3). Tumor cells down-regulate expression of class I MHC proteins, and those that express class I proteins are more resistant to killing by NK cells than derivatives that specifically lack class I MHC expression (5, 6). Class I negative mutant cell lines sensitive to NK lysis can be rescued by transfection of appropriate class I MHC molecules (6, 11). The mechanisms by which these allotype-specific recognition events translate into appropriate immune surveillance by NK cells remain to be elucidated. However, the simple loss or down-regulation of cell surface class I molecules does not always explain discrimination by NK cells. For example, killing of autologous human cells infected with human herpes virus 6 was observed despite no discernable changes in levels of surface class I MHC molecules (21).

Inhibition of human NK clones expressing p70 NKIR containing three Ig domains by HLA-B*2705 was reported to be peptide-specific to some extent. When HLA-B*2705 was expressed in peptide transporter (TAP)-deficient cells, many peptides rescued equivalent surface expression and inhibited lysis, but peptides with P8 Lys allowed lysis—i.e., inhibitory recognition was abrogated (22–24). The precise effect of peptide remains an unresolved question. In contrast, similar experiments using mouse NK cells revealed no effect of specific peptides (25, 26). However, mouse NKIR are type II transmembrane proteins that form a disulfide-linked dimer and are members of the C-type lectin family of proteins (27).

In the present study, an HLA-Cw7 homozygous prediabetic patient with consistent circulating NK cell percentages in the high normal range of 15–20% was initially identified. NK lines and clones from this individual and from two normal individuals have been used to investigate the role of peptide in recognition of HLA-C ligands by NK1 and NK2 cells, and particularly the role of P8 of the peptide bound to HLA-Cw7 on this recognition.

MATERIALS AND METHODS

Cell Lines and Antibodies.

The class I MHC negative human B cell line 721.221 was obtained from American Type Culture Collection. Priess was obtained from the XI International Histocompatibility Workshop Panel. RMA-S cells, which are TAP-deficient and therefore unable to load class I MHC molecules with peptide (28), were a kind gift of Kirsten Falk. Monoclonal antibody (mAb) HP3E4 (17) was a kind gift of M. Lopez-Botet, mAb L31 was a kind gift of P. Giacomini and A. Siccardi, mAb GAD6 was a kind gift of D. Monos, and mAb GL183 and EB6 (18) were purchased from Immunotech (Westbrook, ME).

NK cells were obtained from healthy adult donors MV (A2, A11; B18, B44; Cw5), NQ (A1, A29; B35, B51, Cw4), HTR (A1, A2; B7, B8; Cw7), and DP (A1, A3; B7, B8; Cw7), a prediabetic volunteer recruited from the Diabetes Prevention Trial. All studies were done in compliance with National Institutes of Health guidelines for human studies, Institutional Review Board Assurance of Compliance No. M1331. NK lines and clones were prepared as described (11).

Stable Priess and 721.221 and RMA-S transfectants expressing HLA-C molecules were generated as described (11). Surface expression of HLA-C molecules was analyzed by flow cytometry using the pan-class I mAb W6/32 or HLA-C-specific mAb L31.

Peptides.

Peptides were purchased from the Harvard Medical School Biopolymers Facility, and purity was assessed by HPLC and matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry.

Cytolytic Assays.

The cytolytic activity of NK lines and clones against the various HLA-C transfectants and mutants was assessed in 5-hr ³⁵S release assays (4 hr for the RMA-S targets) in which effector cells were admixed with 5 × 10³ [³⁵S]methionine-labeled targets at different effector:target (E:T) ratios in U-bottomed microtiter plates. Assays were terminated by centrifugation at 1,000 rpm for 10 min at 4°C, and 100 μl of the supernatant was collected for liquid scintillation counting. Specific lysis was calculated as follows: % lysis = (cpm experimental well − cpm spontaneous release)/(cpm maximal release − cpm spontaneous release) × 100. Spontaneous release was determined by incubation of the labeled target cells with medium. Maximal release was determined by solubilizing target cell in 0.1 M NaOH. In all presented experiments, the spontaneous release was <25% of maximal release. The range of triplicate values was 5% of their mean in each case. Each experiment was repeated 3–6 times. In experiments where NKIR-specific mAbs were used to block MHC/NKIR interaction, mAb was included in the medium to a final concentration of 5 μg/ml. For peptide-specific targets, RMA-S transfectants were incubated overnight with peptide at a final concentration of 10 μM and then used in cytotoxicity assays. Surface expression of HLA-C molecules complexed with exogenously added peptides was determined by flow cytometry using mAb W6/32 at the same time points used for cytotoxicity assays.

Computer Modeling.

The coordinates for the crystal structure of HLA-A2 complexed with HTLV-I Tax peptide LLFGYPVYV (29) were used to model the potential interaction of P8 Lys in the peptide epitope with αAsn-80 of the class I MHC heavy chain. Using the molecular modeling program o, the P8 Tyr was substituted by Lys and αThr-80 by Asn, respectively. Probable orientations of the side chain of P8 Lys based on phi/psi angles of the α-carbon backbone were tested. The final positioning, and that depicted in Fig. 8, was determined by a simple rotamer analysis.

Figure 8.

Model of the interaction between peptide epitope P8 Lys and αAsn-80. The model is based on the coordinates for the crystal structure of HLA-A2 complexed with HTLV-1 Tax peptide LLFGYPVYV (29) with P8 Tyr substituted by Lys and αThr-80 substituted by Asn.

RESULTS

To analyze the role of HLA-C-bound peptides in conferring resistance to NK cell lysis, RMA-S cells were transfected with the NK1-protective allele HLA-Cw6 or the NK2-protective allele HLA-Cw7 (11). RMA-S is a peptide transporter-deficient (tap-2) mouse cell line in which no surface class I MHC proteins are expressed at 37°C and is an excellent target for human NK cells. Expression of MHC complexes can be rescued in an apparently “empty” configuration by incubation at 26°C or by the addition of an appropriate exogenous peptide at 37°C (28). The NK2-specific lines from the prediabetic patient, DP, and from normal donors AH and HTR, as well as the NK1-specific line MV (11), were used to generate the clones described in this work. The cell line Priess (DR4, DQ8, Cw9 homozygous) used as a target carries the diabetes susceptibility gene, HLA-DQ8, and an allele protective for NK2-specific lysis, HLA-Cw9. This cell line had also been transfected with a recombinant glutamic acid decarboxylase (GAD) expression vector (Priess/pIiGAD); GAD expression was confirmed by Northern and Western blots and indirect immunofluorescence using the GAD6 mAb (S.B.W., unpublished work). Unexpectedly, Priess cells that expressed GAD after transfection were susceptible to lysis by all of the NK2-specific lines, whereas untransfected Priess cells, Priess cells to which soluble recombinant GAD was added exogenously to load class II MHC molecules, and an autologous Epstein–Barr virus (EBV)-immortalized B cell line from DP with or without exogenous GAD, were all resistant to lysis (Fig. 1A). Susceptibility of the Priess/pIiGAD to NK2 lysis was not due to alterations in surface level of total class I MHC proteins or HLA-C, since no differences in surface staining with W6/32, a pan-Class I mAb, and L31, an HLA-C-specific mAb, were seen (Fig. 1B). In addition, all lines were specifically lysed by the NK1-specific line MV and by clone MV 23, confirming that no nonspecific inhibition of NK cell cytotoxic effector function had occurred (data not shown).

Figure 1.

NK line lysis of B cell targets transfected or loaded exogenously with GAD. (A) The group 2-specific NK lines were prepared from donor DP and incubated with [³⁵S]methionine-labeled target cells for 5 hr at various E:T ratios. E:T ratios of 10:1 and 5:1 are shown. The figure shows one out of three representative experiments. (B and C) HLA-C expression on Priess and the transfectant Priess/pIiGAD. Priess cells were stained with the pan-class I mAb W6/32 antibody (boldface lines) or with the HLA-C specific mAb L31 (dotted line). Controls were the same cells stained with isotype matched IgG (filled histogram). All cells were then counterstained with goat anti-mouse IgG bound to fluorescein isothiocyanate (FITC). The figures show one of four representative experiments.

The NK2-specific lysis of the Priess line expressing recombinant GAD suggested that some GAD-related peptide might have been loaded onto HLA-Cw9 and abrogated the expected inhibition of lysis by NK2 cells. To investigate this question, potential HLA-Cw7 epitopes were identified by computer homology search of the GAD sequence using the published motif for HLA-Cw7 (30, 31). Because of the αAsn-80/Lys-80 dimorphism that determines NK1 and NK2 group specificity, epitopes that had Lys at P8 were particularly sought; P8 of the peptide is adjacent to αAsn-80 of the HLA-Cw9 protein and might be “mistaken” for αLys-80. Interestingly, a potential epitope with P8 Lys was identified that corresponded to the region of homology between GAD (amino acids 256–264) and a peptide from the coxsackie virus P2-C protein (32). Either of these peptides generated crossreactive T cells in the NOD mouse, a model of autoimmune diabetes (32). The GAD peptide, three previously identified HLA-Cw7-binding self peptides and two HLA-Cw6-binding self peptides were added to RMA-S target cells at 37°C (30). All of the HLA-Cw7-specific peptides (peptides 3–6) bound to this molecule, as demonstrated by nearly identical surface expression of class I molecules on RMA-S cells; none bound to HLA-Cw6 (Fig. 2). Peptide 1 bound only to HLA-Cw6, and peptide 2, which was isolated from HLA-Cw6, bound to both. Only those that bound to HLA-Cw7 and without Lys at P8 (peptides 2, 3, and 5: P8 Val, Glu, and Asn, respectively) protected the target from lysis by NK lines (Fig. 3). Loading RMA-S/Cw7 with the GAD peptide 6 (FKMFPEVKE) resulted in target cell lysis as robust as either the RMA-S control or, as expected, the HLA-Cw6-specific peptide 1 (YQFTGIKKL) that does not bind HLA-Cw7. Peptide 4 (NKADVIVLKY) had an intermediate profile. The differences in lysis seen with different peptides loaded on HLA-Cw7 are not a reflection of differences in the level of cell surface expression (Fig. 2).

Figure 2.

Binding of various peptides to HLA-Cw7 and HLA-Cw6. RMA-S cells transfected with HLA-Cw7 (A) or HLA-Cw6 (B) were cultured overnight at 37°C with (boldface line) or without (lightface line) added peptide at a final concentration of 10 μM. Cells were stained with mAb W6/32 and analyzed by fluorescence-activated cell sorting (FACS). HLA-C-specific fluorescence shift is indicated by increased intensity of staining compared with no added peptide control.

Figure 3.

Lysis of RMA-S/Cw7 complexed with various peptides by an NK2 line. [³⁵S]Methionine-labeled RMA-S/Cw7 cells were incubated with or without the indicated peptide and incubated with the DP line for 5 hr at various E:T ratios. The figure shows one of three experiments.

The observation that inhibition of lysis of RMA-S/Cw7 by NK2 lines depended on the peptide bound to HLA-Cw7 required confirmation using clones. A series of NK2 clones were generated from lines DP and HTR by limiting dilution, their NK2 specificities were verified, and their surface phenotype determined to be CD3−/CD56+/CD16+/NKIR2+ (GL183+) by fluorescence-activated cell sorting (FACS) analysis. None of the nine clones tested was inhibited by HLA-Cw7 containing peptides with Lys at P8 (Fig. 4). In particular, loading of RMA-S/Cw7 with the GAD peptide did not result in inhibition of lysis for any clone. All but one of the clones, DP 10.7, were inhibited by RMA-S/Cw7 loaded with peptides 2, 3, and 5 (P8 Val, Glu, and Asn, respectively). Clone 10.7 was inhibited by peptides 2 and 5 but not by peptide 3, which has P8 Glu.

Figure 4.

Lysis of RMA-S/Cw7 complexed with different peptides by various NK2 clones. [³⁵S]Methionine-labeled RMA-S/Cw7 cells were incubated with or without the indicated peptide and then with nine NK2 specific clones at an E:T of 1:1. The figure shows one of three experiments.

Coxsackie virus protein P2-C contains a peptide (VKILPEVKE) that is homologous to GAD 256–264, that also contains Lys at P8, and that rescues HLA-Cw7 surface expression in RMAS/Cw7. This Cw7/peptide complex also failed to protect from NK2 lysis (Fig. 5). If the presence of Lys at P8 prevents recognition of HLA-Cw7 by NK2 clones, then substitution of P8 Lys by Ala should convert the peptide to one that can promote a functional inhibitory complex. Conversely, replacement of residue P8 of an inhibitory peptide with Lys should convert that peptide to one incapable of generating an inhibitory signal in NK2 clones. The following substitutions were generated: (i) in the GAD peptide, FKMFPEVKE → FKMFPEVAE (GAD/K8A); (ii) in the coxsackie peptide, VKILPEVKE → VKILPEVAE (Cox/K8A); and (iii) in the inhibitory HLA-Cw7 self peptide 3, KYFDEHYEY → KYFDEHYKY (E8K). Substitution of Ala into the GAD and the coxsackie peptides converted them to peptides that formed inhibitory complexes with HLA-Cw7 for all clones tested, and conversely substitution of Lys into P8 of self peptide 3 converted it to one that could not form an inhibitory complex. Interestingly, clone DP10.7, which lysed complexes of HLA-Cw7 containing the peptide with P8 Glu, was also not inhibited by complexes with other peptides with P8 Lys, but peptide complexes with P8 Ala were inhibitory for this clone. Again, comparable cell surface expression of HLA-Cw7 was obtained for all the peptides, except that the coxsackie wild-type peptide gave slightly reduced expression of HLA-Cw7 (Fig. 6). Thus, NK2-specific cells were not inhibited by HLA-Cw7/peptide complexes with P8 Lys, and one clone was not inhibited by a peptide with P8 Glu.

Figure 5.

HLA-Cw7/peptide complexes with Lys at P8 do not inhibit lysis by NK2 clones. [³⁵S]Methionine-labeled RMA-S/Cw7 cells were incubated with or without the indicated peptide and reacted with nine NK2 specific clones at an E:T of 1:1. The figure shows one of three experiments.

Figure 6.

Binding of P8 Lys and Ala substituted peptides to HLA-Cw7. RMA-S cells transfected with HLA-Cw7 were cultured overnight with (boldface line) or without (lightface line) the added peptide at a final concentration of 10 μM. Cells were stained with mAb W6/32 and analyzed by fluorescence-activated cell sorting (FACS).

To show that the HLA-Cw7/peptide complexes were acting through the NKIR2 receptor and that the effect was specific for P8 Lys, an antibody blocking experiment was performed using P8-substituted peptide analogues of the GAD peptide. As shown above, HLA-Cw7 loaded with wild-type GAD did not inhibit lysis, and similarly the controls RMA-S and RMA-S/Cw7 without peptide (37°C) were all lysed in the absence of peptide (Fig. 7). The substituted GAD peptides P8 Ala, P8 Arg, and P8 His were all inhibitory. Thus, lysis of HLA-Cw7/peptide complexes was specific for P8 Lys within this range of basic amino acids, at least one of which (Arg) is considerably lager than Lys. The inhibition in each case was reversed by the addition of the NKIR2-specific mAb GL183, but not by the NKIR1-specific mAb HP3E4.

Figure 7.

Effect of NK-specific mAbs and homologous substitution at P8 of the peptide epitope. mAb GL183, but not HP3E4, reverses RMA-S/Cw7-mediated inhibition of an NK2-specific clone. The NK clone DP 10.5 was reacted at E:T ratios of 3:1 for the clones with various [³⁵S]methionine-labeled target cells for 5 hr in the presence or absence of the indicated antibodies (5 μg/ml) and peptides.

DISCUSSION

Lys at P8 of peptides derived from self and viral proteins (including both peptides from GAD and coxsackie virus P-2C protein) specifically abrogated the recognition of HLA-Cw7/peptide complexes by NKIR2 receptors on NK cells, and allowed lysis even though empty Cw7 molecules assembled at 26°C are able to inhibit (11). The observed interference with inhibition was NKIR2-specific, as demonstrated by extensive phenotyping of the NK2 clones used and by blocking experiments using the NKIR2-specific monoclonal antibody, GL183 (Fig. 6). The interference appeared to be Lys-specific, since peptides with Val, Glu, Asp, Ala, Arg, and His at P8 were fully inhibitory. One clone, DP10.7, was identified that, in addition to P8 Lys, also did not tolerate Glu at P8 (Figs. 4 and 5). NKIR2 from each of the clones has been sequenced by PCR; no obvious difference between NKIR2 from DP 10.7 and those from the other clones was found (O.M., unpublished work).

Residues α73, 76, 80, and 90 have all been shown to be important in recognition of HLA-C allotypes by NKIR1 and NKIR2 (39). All these residues are water-accessible and found on the top of the α1 helix (α73, 76, 80) or at the end of the strand (α90) that connects this helix to the beginning of the β-sheet formed by the α2 domain—i.e., the position in which NKIR1 or NKIR2 binds to HLA-Cw6 or HLA-Cw7 is similar to that of the superantigen TSST-1 complexed with HLA-DR1 (33), but slightly more C-terminal. The present data also suggest the binding region can be extended to P8 of the peptide that is adjacent to α80—i.e., the binding region may partially overlap the end of the epitope. Notably, the binding of TSST-1 is also affected by the nature of bound peptide (34). In the crystal structures of HLA-A2 complexed with five different individual peptides and in two HLA-B*3501 single peptide complexes, the P8 side chain (or its equivalent) points upward while the anchor side chain of the C-terminal amino acid, usually P9, is in the F pocket (29, 35–37). Val, Asp, His, Arg, Ala, and, with the exception of clone DP10.7, Glu all appear to be acceptable P8 side chains, permitting an HLA-Cw7/NKIR2 interaction that leads to inhibition, while P8 Lys permits lysis. A model for the effect of Lys at P8 can be proposed (Fig. 8). In this model, the mobile side chain of P8 Lys occludes recognition of the critical αAsn-80 residue and thereby interferes with specific recognition. An alternative model, that P8 Lys sterically interferes with recognition by NKIR, is less likely because the large basic side chain of Arg is compatible with inhibition. In addition to the dimorphism in HLA-C at α80, dimorphisms also occur in the sequences of the NKIR1 and NKIR2 receptors (Fig. 9). Seven dimorphic residues that correlate with NKIR specificity cluster in the first Ig domain at positions 54, 65, 67, 71, 88, 89, and 90 (and two more that do not occur at positions 56 and 66) and are likely to be important in the specificity of these receptors. Among these, the dimorphism at residue 67 is striking in that the reciprocal occurs at α80 in HLA-C—i.e., HLA-Cw6 has αLys-80, while NKIR1, with which it interacts, has Asn-67, and, reciprocally, HLA-Cw7 has αAsn-80, while NKIR2 has Lys-67, suggesting that these residues may interact in the ligand–receptor complex. If so, the role of peptide P8 Lys in preventing this interaction in NK2 cells would be clarified. In a related study, P8 Lys and Glu were shown to prevent the inhibition mediated by HLA-B2705 through the p70 NKIR (26). αThr-80 occurs in HLA-B2705, but immediately adjacent to it and also with the potential to interact with P8 Lys is αGlu-76, the HLA-B locus-specific residue (23). Interestingly, the region of homology of the p70 NKIR to p58 NKIR1 and NKIR2 occurs in the second of its three Ig domains.

Figure 9.

Alignment of a portion of the extra cellular Ig-domain sequences for NKIR1, NKIR2, and p70NKIR. The sequences for the first Ig domain of NKIR1 and NKIR2 and the second Ig domain of p70NKIR were obtained from the GenBank/EMBL database, and sequences corresponding to residues 50–91 of NKIR1 and NKIR2 were aligned. Boldface residues and asterisks highlight dimorphisms that correlate with NKIR1 and NKIR2 group specificities. The bullets (⋅) indicate those that do not correlate.

Do the present findings have any relevance to the problem of the initiation of autoimmunity? Two of the peptides studied were derived from a known diabetes autoantigen, GAD, and from a virus epidemiologically associated with diabetes, coxsackie virus (32). A major problem in considering the genesis of autoimmune diseases is the nature of the event that breaks self-tolerance. Molecular mimicry between a viral peptide and a peptide derived from a self protein has long been considered a possibility (31). In this view, the polyclonal T cell response that results from the viral stimulus would in some cases produce crossreactive T cells that would initiate disease resulting in breakdown of tissue that would release crossreactive autoantigen and sustain the response. Another possibility is suggested by the present findings. A viral infection in an individual who is homozygous for one group of HLA-C allotypes (e.g., Cw9-Cw9 or Cw1-Cw7; compare ref. 10) could result in processing of a viral peptide that contains P8 Lys, binds to the HLA-C molecules, and therefore abrogates its inhibitory recognition. The cells presenting this peptide would then be targets for lysis by autologous NK cells. This lysis could set up a cycle of tissue destruction, possibly augmented by a self peptide (e.g., a peptide derived from GAD). Moreover, the released proteins might provide a sufficient stimulus to break self-tolerance in T cells. Thus the process could be initiated by NK lysis and then sustained by a T cell response. Such a process may be useful to consider along with other possible mechanisms for the initiating event in autoimmunity.

ABBREVIATIONS

NK

natural killer

MHC

major histocompatibility complex

Ig

immunoglobulin

E:T

effector:target