Genetically Encoded and Post-translationally Modified Forms of a Major Histocompatibility Complex Class I-restricted Antigen Bearing a Glycosylation Motif Are Independently Processed and Co-presented to Cytotoxic T Lymphocytes*

The mechanisms by which antigenic peptides bearing a glycosylation site may be processed from viral glycoproteins, post-translationally modified, and presented by major histocompatibility complex class I molecules remain poorly understood. With the aim of exploring these processes, we have dissected the structural and functional properties of the MHC-restricted peptide GP92–101 (CSANNSHHYI) generated from the lymphocytic choriomeningitis virus (LCMV) GP1 glycoprotein. LCMV GP92–101 bears a glycosylation motif -NXS- that is naturallyN-glycosylated in the mature viral glycoprotein, displays high affinity for H-2Db molecules, and elicits a CD8+ cytotoxic T lymphocyte response. By analyzing the functional properties of natural and synthetic peptides and by identifying the viral sequence(s) from the pool of naturally occurring peptides, we demonstrated that multiple forms of LCMV GP92–101 were generated from the viral glycoprotein and co-presented at the surface of LCMV-infected cells. They corresponded to non-glycosylated and post-translationally modified sequences (conversion of Asn-95 to Asp or alteration of Cys-92). The glycosylated form, despite its potential immunogenicity, was not detected. These data illustrate that distinct, non-mutually exclusive antigen presentation pathways may occur simultaneously within a cell to generate structurally and functionally different peptides from a single genetically encoded sequence, thus contributing to increasing the diversity of the T cell repertoire.

histocompatibility complex (MHC) class I molecules and short antigenic peptides generated from a multistep intracellular processing of viral proteins involving components present in the cytoplasm and the endoplasmic reticulum (ER). Processing of antigenic peptides from cytosolic or nuclear proteins has been extensively studied (for review, see Ref. 1). It begins in the cytosol with the breakdown of ubiquitinated proteins by the proteasome and possibly by nonproteasomal cytosolic proteases (2)(3)(4). Peptides generated by this degradation step are then translocated via peptide-specific transporters (TAP) into the ER, where association with newly synthesized MHC class I (5) and, eventually, further trimming of peptides (6,7) occur.
In contrast, the mechanisms by which MHC-restricted antigenic peptides are generated from transmembrane or secreted glycoproteins, which are co-translationally translocated into the ER during biosynthesis, are less understood; however, they frequently represent a source of antigenic peptides (8). The observation that, in most cases, their processing requires both proteasome activity and TAP-dependent translocation into the ER suggests that at least a certain amount of the glycoprotein may have access to the cytosol (9 -11). Both mistranslation in the cytoplasm (12) and retrograde transport from the ER to the cytosol have been proposed to explain the presence of glycoproteins in the cytosol (11,13).
It is thus of particular interest to study the processing from viral glycoproteins of peptides bearing a glycosylation motif (N-X-S/T, where X is any amino acid except P). In theory, such peptides can exist in diverse forms depending on their posttranslational state: non-glycosylated, N-glycosylated, or de-Nglycosylated. These forms are characterized, respectively, by the presence at the level of the N-X-S/T motif of an unmodified N, a glycan-branched N or a D (as the result of a peptide:Nglycanase activity that provokes deamidation of the side chain (14)). Further, the enzymatic activities required for N-glycosylation (N-glycosyltransferases) and de-N-glycosylation (peptide:N-glycanase) are thought to be located in the ER/Golgi (15) and cytosol/nucleus (14), respectively.
Infection of H-2 b mice with lymphocytic choriomeningitis virus (LCMV) generates a vigorous CD8 ϩ CTL response against three well defined H-2D b -restricted epitopes located in the glyco-(GP1 and GP2) and nucleo-(NP) proteins (16 -19). By screening these viral proteins for potential new epitopes, the LCMV GP92-101 (CSANNSHHYI) peptide was identified first as a strong H-2D b binder (20,21), and later as a non-immunodominant LCMV epitope (22). Interestingly, the central core motif GP95-97 of this antigen represents one of the glycosylation motifs, -NXS/T-, known to be N-glycosylated in the mature LCMV GP1 glycoprotein (23). These observations at both the structural and functional levels make this peptide a good model for studying the effect of post-translational modification on antigen processing and presentation, and the consequences it may have for the antiviral CTL response. This study aimed: (i) to analyze the effect of post-translational modification (N-glycosylation and de-N-glycosylation) on the MHC binding and CTL activation properties of LCMV GP92-101 and (ii) to identify the viral sequence(s) naturally processed and presented to CTLs at the surface of LCMV-infected cells. Binding studies performed with synthetic peptides corresponding to the three possible forms of GP92-101 indicated that neither N-glycosylation nor de-N-glycosylation affected the H-2D b -binding affinity of GP92-101. All three forms of the peptide were immunogenic and led to the generation of GP92-101-specific CTLs. Analysis of peptides eluted from the surface of LCMV-infected cells by a cytolytic assay coupled to HPLC led to the identification of both non-glycosylated (N95) and de-N-glycosylated sequences (D95) naturally presented at the surface of infected cells. The N-glycosylated (GlcNac-N95) form remained undetectable. Additionally, immunoreactive species of higher retention time derived from the N95 peptide were also observed, likely corresponding to oxidized forms of C92. The physiological relevance of this remains to be determined. Overall, the data presented here demonstrate for the first time that both genetically encoded and post-translationally modified forms of a viral glycopeptide may co-exist at the surface of virus-infected cells as a result of distinct, non-mutually exclusive processing pathways that can operate simultaneously.

MATERIALS AND METHODS
CTL, Cell Lines, and Virus-CTL were obtained from C57BL/6 mice immunized subcutaneously at the base of the tail with 50 -100 g of the indicated peptides mixed with 5 g of P30 T-helper epitope from tetanus toxoid (24) in incomplete Freund's adjuvant. After 1 week, draining lymph nodes were removed and CTL were restimulated weekly with irradiated (2,500 rads) C57BL/6 splenocytes and irradiated (10,000 rads) peptide-pulsed (1 M) RMA cells in the presence of 30 UI/ml IL2 (EL4 supernatant). Murine H-2 b mutant RMA-S cells and human T2 cells transfected with H-2D b (T2-D b ) were used in binding experiments. The murine H-2 b cell lines RMA and MC57 were used in in vitro cytotoxicity assays. Cells were grown in Dulbecco's modified Eagle's medium (RMA, RMA-S, MC57) or Iscove's modified Dulbecco's medium (T2-D b ) containing 5% bovine serum, L-glutamine (2 mM) and antibiotics (10 units/ml penicillin and 10 g/ml streptomycin). Geneticin (400 g/ml) was added to Iscove's modified Dulbecco's medium to maintain selection of T2-D b cells. Lymphocytic choriomeningitis virus Armstrong strain (LCMV Arm) was used to infect mice or cells.
Peptide Synthesis-Peptides were synthesized on an automated peptide synthesizer (Applied Biosystems 430A) by the solid-phase method using Fmoc chemistry. The glycopeptide analogue was obtained by using Fmoc-Asn(Ac 3 AcNH-␤Glc)-OH, a commercially available derivative of Asn bearing an N-acetyl-D-glucosamine moiety (Novabiochem). After standard protocols for solid phase synthesis, cleavage, and deprotection, the glycopeptide was de-O-acetylated with 0.1 M sodium hydroxide. Peptides were purified by HPLC on a RP300-C8 reversed-phase column (Brownlee Lab) and their identity confirmed by electrospray ionization mass spectrometry.
Binding Studies-Binding studies were performed as described previously (20). Briefly, T2-D b cells (1 ϫ 10 5 cells/well) were incubated in U-bottomed 96-well plates for 75 min at 37°C with 100 nM H-2D brestricted fluorescent probe FITC-KAIENAEAL (25) and increasing concentrations (10 Ϫ10 to 10 Ϫ4 M) of unlabeled competitors. Cells were then washed twice with bovine serum albumin-PBS and analyzed by flow cytometry. Total and nonspecific binding were measured in the absence or presence of 1 mM unlabeled LCMV NP396 -404. Specific binding to H-2D b was defined as the difference between the total and nonspecific binding. Percentage (%) of binding inhibition was calculated as 100 ϫ {1 Ϫ (fluorescence intensity (FI) in the presence of competitor Ϫ FI, nonspecific binding/FI, specific binding)}. IC 50 represents the peptide concentration inhibiting 50% of the specific binding of the fluorescent probe.
Extraction of Viral Peptides from Infected Cells-MC57 (H-2 b ) and control Balb/c (H-2 d ) cell lines (1-2 ϫ 10 9 cells) were infected or not with LCMV Arm for 48 h. Cells were washed three times in PBS, then resuspended in 0.1 M citrate/phosphate buffer at pH 3.0 for 2 min. The eluted material was desalted on a Sep-Pak column (Waters) according to the manufacturer's instructions, then transferred onto a Centricon 10 (Amicon) and centrifuged at 3500 rpm for 90 min at 4°C. Material less than 10 kDa was vacuum-concentrated and then resuspended in 20 l of 0.08% trifluoroacetic acid. Peptide separation was carried out on a reversed-phase C18 column (7 m, 2.1 ϫ 100 mm, Aquapore, Brownlee) using the Waters 600S controller system. Samples (10 l) were injected and separated using either system I (5-60% B in 60 min) or system II (6 -15% B in 30 min, then 15-57% B in 30 min). Solution A was 0.08% trifluoroacetic acid in H 2 O, and solution B was 0.08% trifluoroacetic acid in CH 3 CN (flow rate: 400 l/min). Fractions (200 l) were collected in U-bottomed 96-well plates, lyophilized, reconstituted in PBS, and stored at Ϫ80°C before analysis. In some cases, the eluted pool was treated with DTT (1 mM DTT in 0.1 M ammonium acetate, pH 8.5, under N 2 overnight at room temperature) or Endo H (50 milliunits/ml, pH 5.5, 24 h at 37°C) before separation.
Cytotoxicity Assays-RMA cells were incubated for 1 h at 37°C in medium containing 10-fold dilutions of peptides. MC57 cells were infected at a multiplicity of infection of 2 with LCMV Arm 48 h prior the assay. Peptide-pulsed and control RMA cells, and LCMV-infected and uninfected MC57 cells were 51 Cr labeled and used as targets (5 ϫ 10 3 /well) in chromium release assays. Effector CTL were added at 1.5 ϫ 10 4 /well or at the indicated E/T ratio. The 51 Cr content of supernatants was determined after 4 h of incubation at 37°C. The specific lysis was calculated as 100 ϫ [(experimental Ϫ spontaneous release)/(total Ϫ spontaneous release)].

N-Glycosylation and De-N-glycosylation Do Not Alter H-2D b
Binding Affinity of LCMV GP92-101-Three synthetic peptides corresponding to the non-glycosylated (N95), N-glycosylated ((GlcNac)-N95) and de-N-glycosylated (D95) forms of GP92-101 were synthesized (Table I). The N-acetyl-D-glucosamine moiety was chosen to mimic the glycosylated form because: (i) it is the first carbohydrate attached to N and is common to all eukaryotic glycosylation processes and (ii) it corresponds to the minimal sugar moiety left after Endo H digestion of glycosylated side chains, therefore allowing the detection of natural species. Peptide binding affinities for H-2D b were measured in a MHC binding competition assay on T2-D b cells (25). As shown in Fig. 1, N95 bound to H-2D b with a high affinity (IC 50 in the 10 nM range), in agreement with previous observations (20). The two modified analogues, (Glc-Nac)-N95 and D95, bound efficiently to T2-D b cells with doseresponse curves superimposable on that of the unmodified N95, thus indicating that all three forms of GP92-101 can potentially be presented by H-2D b at the surface of LCMV-infected cells expressing H-2D b .
N95, (GlcNac)-N95, and D95 Forms of LCMV GP92-101 Generate a CTL Response in H-2 b Mice-C57BL/6 mice were immunized subcutaneously at the base of the tail with N95, (GlcNac)-N95, and D95 in the presence of the P30 helper epitope and incomplete Freund's adjuvant. After 2 weeks of in vitro restimulation, CTL were obtained against each of the three peptides, indicating that they are all immunogenic. As shown in Fig. 2 (panel A), CTL generated against N95 killed target cells coated with the immunizing peptide with a halfmaximal lysis at concentrations in the 10 Ϫ9 to 10 Ϫ10 M range. These CTL also recognized (GlcNac)-N95 and D95, but at very different concentrations. (GlcNac)-N95 recognition required a concentration about 2 logs higher, and the maximal lysis plateau reached was lower than that for N95, whereas D95 was ϳ100-fold more potent than N95. CTL generated against (Gl-cNac)-N95 (panel B) and D95 (panel C) showed a 10-fold higher efficacy (half-maximal lysis in the range of 10 Ϫ11 to 10 Ϫ10 M range) and were fully specific for the immunizing peptide since no (or very weak) cross-reactivity with the other forms of GP92-101 was observed. The response generated against each form of the LCMV GP92-101 was peptide-specific since none of the CTLs recognized the unrelated GP33-41 sequence of the LCMV GP1 immunodominant epitope, even at the highest concentration tested (10 Ϫ6 M). Altogether, these results show that the three different forms of GP92-101 may represent distinct antigens. Obtention of these anti-GP92-101 CTLs then allowed us to search for the presence of GP92-101 sequence(s) naturally processed and presented at the surface of LCMV-infected cells.
LCMV-infected MC57 Cells Are Lysed by CTL Specific for Unmodified (N95) and Post-translationally Modified (D95) Analogs of LCMV GP92-101-The cytolytic activity of the anti-GP92-101 was tested against LCMV-infected or uninfected MC57 target cells. Fig. 3 shows that CTL generated against N95 (panel A) and D95 (panel C) but not (GlcNac)-N95 (panel B) were able to specifically lyse LCMV-infected cells, demon-strating that GP92-101 is naturally presented at the surface of LCMV-infected cells. This result is in agreement with the finding that anti-GP92-101-specific CTL are present in secondary antiviral CTL obtained after acute infection of C57BL/6 mice with LCMV (22,26). Given the pattern of reactivity of N95-and D95-specific CTL (see Fig. 1A), one can deduce that at least the post-translationally modified form (D95) of LCMV GP92-101 is naturally presented at the surface of LCMV-infected cells.
Both Unmodified (N95) and Post-translationally Modified (D95) Forms of LCMV GP92-101 Are Naturally Presented at the Surface of LCMV-infected Cells-Since LCMV-infected cells were recognized and lysed by anti-LCMV GP92-101 CTLs, peptides were extracted from the surface of LCMV-infected (or as a control, uninfected MC57 cells), separated by HPLC (Fig.  4, panel A) and analyzed in cytotoxicity assays (Fig. 4, panel B), in order to characterize further which form(s) of LCMV GP92-101 was(were) endogenously presented.
In a first step, CTL generated against N95 were used since they recognized all forms of GP92-101. As shown in Fig. 4 (panel B), these CTL killed H-2 b target cells coated with material eluted from LCMV-infected but not uninfected MC57 cells. Two peaks of activity corresponding to fractions 31-34 (retention time: 15-16.5 min) and to fractions 36 -37 (retention time: 17.5-18.0 min) were obtained, indicating that different forms of GP92-101 were processed in LCMV-infected cells and presented at their surface. N95-specific CTL did not recognize peptides eluted from LCMV-infected or uninfected Balb/c (H-2 d ) cells in control experiments (data not shown).
In a second step, fractions 20 -50 from an independent experiment were analyzed with either N95-or D95-specific CTL, both of which recognized LCMV-infected cells. As shown in Fig.  5 (panels A and B), a different profile of activity was observed with the two CTL lines. The D95-specific CTL activity profile (panel B) was very clear and restricted to fraction 32, which, as indicated by the arrow, corresponds to the retention time of the synthetic D95 peptide. This result confirms unambiguously that the D95 form of GP92-101 is present at the surface of LCMV-infected cells. The activity profile obtained with N95specific CTL (panel A) was more complex and showed that other forms of GP92-101 were present that were selectively recognized by N95-specific CTL but not D95-specific CTL.
Since the poor separation of the N95 and D95 forms obtained with the HPLC system I (see Table I) did not allow unambiguous identification of the peptide(s) present in the main peak recognized by anti-N95 CTL (fractions 31-34, Fig. 5A), we set up an optimized gradient (system II) allowing a better separation of N95 and D95 (see Table I). As shown in Fig. 6, endog-   (Table I). Fractions 32-34 rec-ognized by the two CTL lines contained the D95 form, as expected from results presented above and in Table I Surface of LCMV-infected Cells-Fractions 36 -37 recognized by N95-specific CTL do not co-migrate with any of the synthetic peptides corresponding to post-translationally modified analogs of GP92-101. We reasoned that this peak could correspond to a N-glycosylated form of GP92-101 different from the synthetic (GlcNac)-N95 modified peptide and thus not recognized by (GlcNac)-N95-specific CTL. To investigate this possibility, we treated the pool of eluted peptides with Endo H, which generates the (GlcNac)-N form from complex glycans attached to N, before performing the assay. Despite this treatment, none of the fractions was recognized by (GlcNac)-N95-specific CTL (shown in Fig. 7), indicating that no N-glycosylated form of GP92-101 was detectable among the eluted peptides. The synthetic control (GlcNac)-N95 peptide was recognized by specific CTL in fractions corresponding to the expected retention time.
Fractions 36 -37 Correspond to an Oxidized C92 Derivative of the N95 Form of LCMV GP92-101-The presence of a cysteinyl residue (C92) in the GP92-101 sequence may lead to oxidized species as a result of either antigen processing or storage of peptide eluates (27). As shown in Fig. 8, DTT treatment of peptide eluate abrogated the CTL activity against fractions 36 -40. To support this observation, we found that the dimerized N95 form of the GP92-101 synthetic peptide which eluted at an expected higher retention time (Table I) was still efficiently recognized by N95-specific CTL (data not shown). Thus, the DTT-reducible, N95-specific immunoreactivity found in fractions 36 -40 reflects the presence within the peptide eluate of chemically modified N95 forms of LCMV GP92-101 which likely correspond to oxidized C92 derivatives. Why oxidation of C92 only affected N95 and not D95 is unknown.
Altogether our results indicate that GP92-101 is present at the surface of infected cells as at least two different endogenously processed forms, a non-glycosylated one and a de-Nglycosylated one. A third one corresponding to a chemical modification of C92, whose natural occurrence remains to be established, was also detected. DISCUSSION In this study, we demonstrate, using the non-immunodominant H-2D b -restricted LCMV epitope GP92-101, that MHC class I-restricted antigens that bear a glycosylation motif -NXS-may be processed from a viral glycoprotein and presented at the surface of infected cells as at least two distinct sequences, genetically encoded (non-glycosylated) and posttranslationally modified (de-N-glycosylated).
To our knowledge, such a dual processing and presentation pathway has never been observed before. Very few (three) examples of naturally processed sequences bearing a glycosylation motif have been reported so far, consisting in each case of only one exclusive form, either non-glycosylated (28) or de- N-glycosylated (10,29). How may the two presentation pathways of LCMV GP92-101 be achieved? For the de-N-glycosylated form, processing might involve translation in the ER, export of full-length glycoprotein to the cytosol and transport of processed peptides by TAP for MHC binding, as shown for an HLA-A2 tyrosinase epitope (11). Proteasomal degradation of glycoproteins requires their retro-transport from the endoplasmic reticulum (where glycosylation occurs) to the cytosol (where both deglycosylation and degradation are achieved) (30,31). With regard to the non-glycosylated form, it is difficult to understand how a non-glycosylated peptide may be processed from the LCMV GP1 glycoprotein itself since all the five glycosylation motifs -NXS/T-of LCMV GP1 are glycosylated in the mature viral glycoprotein (23). Further, the peptide:N-gly-canases known so far, due to their ␤-aspartyl-glycosylamine hydrolase activity, always convert the glycosylated Asn to Asp and never restore the Asn initially present. One likely explanation is that the non-glycosylated form of LCMV GP92-101 does not originate from GP1 but from LCMV GP-C. The LCMV glycoprotein GP1 is generated, together with GP2, from a posttranslational cleavage of a common precursor GP-C. Unglycosylated GP-C is not cleaved (23); one may therefore hypothesize that the non-glycosylated form of GP92-101 originates from unglycosylated, uncleaved GP-C and not from GP1. The nonglycosylated and the de-N-glycosylated forms of GP92-101 would then be generated separately from two different proteins. Two other possibilities can also be proposed (28): (i) some of the glycosylation sites of LCMV GP1 may be randomly left unglycosylated within the viral glycoprotein, or (ii) parts of the glycoprotein may be aberrantly translated on free ribosomes in the cytosol and thus never access the ER. Determination of the relative proportion of the different forms of GP92-101 presented at the surface of infected cells would require highly sensitive detection methods not available in our laboratory. It is expected that the proportion of the different forms we observed would be determined by both (i) the efficacy of the processing pathways involved and (ii) the initial amount of substrate for these different pathways.
While the natural occurrence of glycosylated peptides presented by MHC class II molecules to ␣␤TCR has been well documented (32,33), the same remains to be demonstrated for classical MHC class I molecules. The presence of the de-Nglycosylated form of GP92-101 clearly indicates that a certain proportion of the peptide has previously been N-glycosylated, yet we were unable to detect any N-glycosylated GP92-101 sequence at the surface of infected cells. However, from our study and others using artificially glycosylated peptides, it is clear that glycopeptides can bind to MHC class I molecules and be immunogenic, indicating that a T cell repertoire exists for specific recognition of class I restricted glycosylated peptides (34 -37). It is thus quite possible that the absence of N-glycosylated peptides naturally presented by classical MHC class I molecules may be due to limitations at the level of the antigen processing pathway. For example, in the cytosol, only peptide: N-glycanase-treated glycoproteins but not untreated glycoproteins are substrates for proteasomal degradation and antigen processing in the ER is blocked as long as glycosylation is not pharmacologically or genetically inhibited (38). Whether or not O-glycosylated peptides produced in the cytosol can be presented by MHC class I molecules remains to be studied.
An oxidized form of the non-glycosylated peptide was also present in our extract. Even though the exact nature of this FIG. 7. Effect of Endo H treatment of peptide eluate on recognition by anti-(GlcNac)-N95-specific CTL. Material acid-eluted from the surface of LCMV-infected MC57 cells (1 ϫ 10 8 cell eq) treated (solid circles) or not (open circles) with Endo H as described under "Materials and Methods" was fractionated on RP-HPLC using gradient system I, and the fractions were tested for their ability to sensitize RMA cells to lysis by anti-(GlcNac)-N95 CTL at an E:T ratio of 10:1. Synthetic (GlcNac)-N95 peptide (dashed circles) whose retention time is indicated by the arrow was fractionated, treated, and analyzed under the same experimental conditions. or not (open circles) with DTT before HPLC fractionation as described under "Materials and Methods." Spontaneous chromium release from RMA cells in the presence or absence of DTT-treated blank sample was 27% and 2%, respectively. Anti-N95 CTL was then tested against fractions 20 -50 as described in the legend of Fig. 3. form remains to be fully characterized, we have shown that it was recognized by anti-N95 CTL but not by anti-95D CTL and that this peptide was absent from DTT-treated samples. It has to be noted that Cys residues can readily be oxidized in biological fluids or during storage of Cys-containing peptides. While in previous reports such a modification was produced during antigen processing (27,39), we lack evidence that such a mechanism occurs in our case.
What may be the physiopathological consequences of processing and presentation of multiple forms of a single antigenic sequence? In terms of MHC presentation, the observation that the high binding affinity of GP92-101 for H-2D b was left unaffected by either N-glycosylation or de-N-glycosylation is consistent with previous data showing that position 4 of H-2D brestricted peptides tolerates a wide variety of substitutions (20). The side chain of the residue at position 4 points out of the H-2D b binding groove and represents a TCR contact residue for most of the known H-2D b -restricted viral epitopes (40 -42). Generation of multiple forms of a peptide that display similar high MHC binding properties but have different TCR activation properties from a single sequence potentially increases the diversity of antigens presented to T cells. Thus, in terms of TCR recognition, N-glycosylation alone or followed by de-N-glycosylation can create distinct epitopes. Indeed, while CTL generated against N95 efficiently recognized both N95 and D95 (the latter even being recognized better than the immunogen), CTL generated against D95 selectively recognized this peptide. Immunization of C57BL/6 mice with N95 induces a protective CTL response against LCMV and splenocytes from LCMVinfected mice can be maintained in culture with N95 pulsed target cells and stained with N95-H-2D b tetramers (22,26). Presentation of multiple forms of the same viral antigen should be considered as an advantage for the host in its fight against viral infection. However, the observation that despite the presence of its multiple forms at the surface of LCMV-infected cells, an efficient GP92-101-specific primary CTL response is not observed in LCMV-infected C57BL6 mice against the unmodified N95 form (21,22) or post-translationally modified D95 and (GlcNac)-N95 forms (data not shown), demonstrates that the LCMV GP92-101 remains a non-immunodominant epitope. So, conversely, could processing and presentation of multiple forms of LCMV GP92-101 be a cause of its non-immunodominance? The observation that the protective capacity of CTL is influenced by the diversity of viral peptides generated within infected cells (43) would support this latter hypothesis. Another attractive and elegant hypothesis is that non-immunodominance may result from TCR antagonism naturally generated by multiple antigen processing from the same sequence. Both we and others have shown that altered peptide ligands resulting from mutation of an antigenic peptide can differentially affect TCR recognition, provoking partial agonism or antagonism (44 -46) and/or allowing viruses to evade the CTL response (41). Further, it has recently been shown that antigen processing could generate both stimulatory and antagonist peptides from a single class II model epitope, which may have important implications for T cell immunoregulation (47). These different hypotheses are currently under investigation in our laboratory.