Identification of a Kd-restricted antigenic peptide encoded by murine cytomegalovirus early gene M84

In spite of the enormous protein coding capacity of murine CMV (mCMV) and human CMV (hCMV) genomes (Rawlinson et al., 1996 ; Chee et al., 1990 ), antigenic peptides processed along the major histocompatibility complex class I (MHC-I) pathway and presented to CD8 T cells have so far been identified for only few proteins (reviewed by Reddehase, 2000 ). Specifically, the immediate-early (IE) proteins IE1-pp89 (ppm123) and IE1-pp72 (ppUL123) of mCMV (Del Val et al., 1991b ; Holtappels et al., 1998 ; Reddehase et al., 1989 ) and hCMV (Borysiewicz et al., 1988 ; Kern et al., 1999 ; Retiere et al., 2000 ), respectively, were found to be prominent antigens for the immune response mediated by CD8 T cells. The expression of IE proteins precedes the expression of early (E) proteins, which include immune evasion proteins interfering at various steps with the MHC-I pathway of antigen processing and presentation (reviewed by Hengel et al., 1998 ). Thus, the phenomenon of immune evasion operating in the E-phase of virus gene expression in infected fibroblasts could, in principle, explain the limited number of antigenic proteins as well as the immunological privilege of the IE1 proteins.

For mCMV, however, the specificity spectrum of cytolytic T lymphocytes (CTL) in pulmonary infiltrates predicted the existence of a multitude of subdominant antigenic peptides attributed to the E- or late (L)-phase (Holtappels et al., 1998 ). Specifically, infected target cells were preferentially lysed in the E-phase. A more recent report directly documented a leakiness of E-phase immune evasion functions in that the D^d-presented antigenic peptide ²⁴³YGPSLYRRF²⁵¹ was identified in the m04 gene product gp34 (Holtappels et al., 2000 ), an E-phase protein supposed to be involved in the surface transport of MHC-I molecules (Kleijnen et al., 1997 ).

For hCMV, the CD8 T-cell response to ppUL83 (pp65), a constituent of the virion tegument and a major component of dense bodies, appeared to be at least as prevalent as the response to IE1-pp72 (Boppana & Britt, 1996 ; Wills et al., 1996 ). In addition, an antigenic peptide has been described for the major virion envelope glycoprotein UL55-gB (Utz et al., 1992 ). While ppUL83 is an L-phase protein according to its expression kinetics, entry of dense bodies and virion uncoating after penetration can feed the alternative class I pathway of antigen processing and presentation before any virus gene expression (McLaughlin-Taylor et al., 1994 ). Therefore, processing of ppUL83 is not targeted by immune evasion mechanisms. It was thus an obvious and long overdue issue to determine an immunological role of mCMV homologues of hCMV ppUL83 and other hCMV virion proteins. It is the merit of D. H. Spectors group to have taken up that task. Two ORFs in the genome of mCMV were found to possess significant homology to hCMV UL83, namely ORFs M83 and M84. Based on percentage amino acid identity, pM84 (p65) is more related to ppUL83 than is the positional homologue ppM83 (pp105). However, only ppM83 proved to be analogous to ppUL83 by virtue of its late expression kinetics, phosphorylation and, most importantly, its virion association (Cranmer et al., 1996 ; Morello et al., 1999 ). In contrast, pM84 was identified as a nonstructural protein expressed in the E-phase (Morello et al., 1999 ) and should hence be targeted by immune evasion mechanisms operating in the E-phase. Notably, however, besides ORF m123-encoded IE1-pp89 (Gonzales Armas et al., 1996 ), only pM84 induced protective immunity in the H-2^d haplotype after genetic immunization of BALB/c mice with an M84 expression plasmid. Plasmids specifying the mCMV positional homologues of hCMV virion proteins UL32-pp150 (corresponding to mCMV gene M32), UL48-p212 (M48), UL56-p130 (M56), pUL69 (M69), UL82-pp71 (M82), UL83-pp65 (M83), UL85-mCP (M85), UL86-MCP (M86) and UL99-pp28 (M99) did not confer in vivo protection (Morello et al., 2000 ).

Based on these data, we started a survey for identifying an antigenic peptide processed from pM84 and presented by an MHC-I molecule of the H-2^d haplotype. A set of synthetic peptides deduced from the pM84 sequence was selected according to Rammensees antigenic motif forecast (Falk et al., 1991 ; Rammensee et al., 1997 ), which is essentially based on the identification of conserved anchor residues extending into hydrophobic pockets of MHC-I molecules. Specifically, the major nonameric motifs are x(Y or F)xxxxxx(I or L or V), xGPxxxxx(L or I or F) and x(P or S)xxxxxx(F or L or M) for the MHC-I molecules K^d, D^d and L^d, respectively. Based on empirical antigenic peptides, this forecast gets continuously refined and probability scores for antigenicity are assigned to the motifs. For example, antigenic peptides of mCMV defined by us previously, namely IE1 ¹⁶⁸YPHFMPTNL¹⁷⁶ (Reddehase et al., 1989 ) and m04 ²⁴³YGPSLYRRF²⁵¹ (Holtappels et al., 2000 ) got scores of 26 and 28, respectively. By using the Internet database SYFPEITHI version 1.0 (http://www.uni-tuebingen.de/uni/kxi/) for MHC ligands and peptide motifs, a refined forecast was made for pM84. For all motifs containing the major anchor residues (see above) and for all K^d and L^d motifs with a score of 20, peptides were synthesized and tested for antigenicity in a previously described microculture CTL assay system (Holtappels et al., 2000 ) involving three restimulations of memory CD8 T cells present in the spleens of BALB/c mice at 3 months after intraplantar infection (Fig. 1). In the first screening experiment, M84 nonapeptides starting at N-terminal positions 33 (score of 31) and 297 (score of 28) were found to be antigenic at restimulation concentrations of 10^-6 M and 10^-10 M, respectively, and there appeared to be further but minor candidates. A second screening experiment, performed with an independent group of memory T-cell donors, confirmed only the antigenic peptide at position 297. Potential candidates from either screening experiment were then tested for their ability to support the generation of a CTL line (CTLL) in a bulk culture system (Holtappels et al., 2000 ). The only peptide confirmed by the generation of a CTLL (see below) was the one starting at position 297 representing an MHC-I K^d motif. In conclusion, nonapeptide ²⁹⁷AYAGLFTPL³⁰⁵ represents an antigenic M84 peptide.

(99K): Fig. 1. Functional screening for antigenic peptides encoded by ORF M84. (a) Map location of ORF M84 in the genome of mCMV Smith strain (ATCC VR-194/1981) and distribution of nonapeptides with MHC-I binding motifs within the 587 amino acid 65 kDa protein pM84. Peptides with an antigenicity score of 21 are indicated by filled bars, peptides containing major motifs but with an antigenicity score of <21 are indicated by unfilled bars. Predicted MHC-I Kd, Dd and Ld motifs are highlighted by red, green and blue colouring, respectively. (b) All predicted antigenic peptides were synthesized and tested for their capability to generate CTLL in triplicate microcultures by three restimulations of memory CD8 T cells derived from the spleens of BALB/c mice at 3 months after intraplantar infection according to the method described by Holtappels et al. (2000) . Two experiments (Exp. 1 and Exp. 2) were performed with memory T cells derived from independent groups of donor mice. It should be noted that results were negative throughout when donor mice were unprimed (data not shown). For the sake of clarity, data are documented only for peptides corresponding to filled bars in the map (a). Peptides represented in the map by unfilled bars proved to be nonantigenic throughout. The specific lytic activity is shown for each individual microculture of the triplicates and data are given for the three indicated restimulation peptide molarities. Controls include nine microcultures with all components except peptide () as well as triplicate microcultures restimulated with the known antigenic peptides IE1 168YPHFMPTNL176 and m04 243YGPSLYRRF251. M84 nonapeptides are indicated by the protein map positions of the respective N-terminal residues.

Even though this new peptide should be presented by K^d according to the motif search, an MHC restriction analysis was needed for experimental verification. Long-term CTLL were established with the M84 peptide and, for controls, with the IE1 peptide and the m04 peptide. It should be noted that the generation of M84-specific CTLL proved to be very dependent upon the peptide concentration during restimulation in that we were successful only with 10^-10 M. Moreover, not every attempt led to a CTLL. These three CTLL were then tested for cytolytic activity against peptide-pulsed L-cell transfectants expressing selectively either L^d, D^d or K^d (Fig. 2). The known L^d and D^d restriction of the IE1 and the m04 peptides was confirmed and the predicted K^d restriction of the M84 peptide was verified. Thus, antigenic peptides of mCMV are now available for the whole set of MHC class-I molecules in the H-2^d haplotype.

(25K): Fig. 2. MHC-I restriction analysis for mCMV peptide-specific CTLL. Long-term CTLL were generated by restimulation of BALB/c (H-2d haplotype) memory CD8 T cells under bulk culture conditions according to the method described by Holtappels et al. (2000) . Optimized peptide concentrations used for the restimulations were 10-9 M, 10-8 M and 10-10 M, for the IE1, m04 and M84 peptides, respectively. Specific cytolytic activity was measured by a 4 h 51Cr-release assay at an effector-to-target cell ratio of 15:1, with peptide-pulsed L-cell transfectants L-Ld, L-Dd and L-Kd as target cells (for details, see Holtappels et al., 2000 ). The lytic activity (ordinate) is shown as a function of the peptide molarity used for target cell pulsing (abscissa).

We finally investigated the frequency of M84-specific memory CD8 T cells in the spleens of latently infected BALB/c mice (Fig. 3) by using an IFN-γ-based ex vivo ELISpot assay, a method previously described by Holtappels et al. (2000) . The frequency of IE1-specific memory cells ranged between ca. 300 and 500 per 10⁶ spleen cells. Since the proportion of CD8 T cells in the spleen is ca. 10%, according to cytofluorometric analysis (data not shown), the corrected frequency is ca. 30005000 per 10⁶ CD8 T cells. The frequency of m04-specific memory cells was found to range between ca. 250 and 550 per 10⁶ CD8 T cells. In both cases, the frequency in mice sensitized by prior mCMV infection was significantly higher than in naive spleen cell donors. Notably, the whole set of M84 peptides, including the already verified antigenic peptide ²⁹⁷AYAGLFTPL³⁰⁵, gave frequencies within the background of the assay, which was defined by the number of spleen cells secreting IFN-γ in the absence of a stimulating peptide. In addition, the frequencies were all in the range also measured for naive mice. This result gave us two messages. First, priming of naive T cells by the M84 peptide must be very inefficient, which may indicate that the amount of processed and presented peptide is too low during the transient and barely productive acute infection of immunocompetent mice. As a consequence, the M84 peptide induces a very low frequency of memory cells, even though it binds to the presenting K^d molecule at low molarity and effectively sensitizes target cells for lysis (Fig. 2). Second, and this is an important methodological conclusion, peptide screening based solely on an ELISpot assay can fail in identifying antigenic peptides. Nevertheless, as previously shown (Morello et al., 2000 ), M84 can induce protective immunity against mCMV infection in the context of an expression plasmid. With the knowledge presented herein, a predictable increase in the frequency of M84 peptide-specific memory CD8 T cells after genetic immunization can now be easily monitored. The low representation of the M84 specificity in memory T-cell populations induced by the infection of immunocompetent mice does not exclude the possibility that the M84 peptide becomes relevant under conditions of CMV disease when acute infection generates much greater amounts of E-phase peptides.

(26K): Fig. 3. Frequencies of mCMV peptide-specific memory T cells. The set of M84 peptides representing MHC-I motifs (see Fig. 1) was screened by an IFN-γ-based 16 h ELISpot assay (for the method, see Holtappels et al., 2000 ). Spleen cells were derived from BALB/c mice at 3 months after intraplantar infection or, as a negative control, from age-matched uninfected BALB/c mice. The assay was performed with 105 and 106 responder spleen cells in three independent assay cultures for each cell concentration. P815-B7 cells (Azuma et al., 1992 ) were pulsed with peptide at a concentration of 10-8 M, and 105 of these peptide-presenting cells were added to each assay culture. Controls included three cultures with responder spleen cells and P815-B7 cells that were not pulsed with peptide (). Also included were triplicate cultures in which the P815-B7 cells were pulsed with the known antigenic peptides IE1 168YPHFMPTNL176 and m04 243YGPSLYRRF251. M84 nonapeptides are indicated by the protein map positions of the respective N-terminal residues. The range of spot numbers counted in the triplicates is shown as a bar. Black bars represent the frequencies determined for memory spleen cells and shaded bars show the control data obtained with spleen cells from unprimed donor mice. The arrow points to the verified antigenic peptide M84 297AYAGLFTPL305.

With an antigenicity score of 28, the verified antigenic M84 peptide was predicted to bind very well to K^d. While peptides with a low score and thus with low MHC-I binding affinity are clearly inferior; a high score gives no guarantee of antigenicity. For example, the M84 peptide starting at the N-terminal position 33 has a score of 31, but this was not confirmed by the functional analyses. On the other hand, the immunodominant IE1 peptide has a score of 26. Besides binding to MHC-I molecules, further parameters are involved in determining antigenicity. Specifically, it is trivial that the respective protein must be expressed adequately for efficient priming of T cells. In addition, residues that flank the antigenic sequence in the protein can have a critical influence on the efficacy of proteasomal processing (Del Val et al., 1991a ; Eggers et al., 1995 ).

The fact that the antigenic M84 peptide is presented by the K^d molecule bears an interesting side aspect. Under conditions when the m152-encoded E-phase immune evasion gene product gp37/40 of mCMV mediates retention of peptide-loaded MHC-I molecules in a cis-Golgi compartment (Ziegler et al., 1997 ), the m04 E-phase gene product gp34 directs MHC-I molecules to the cell surface, wherefrom gp34MHC-I complexes can be immunoprecipitated (Kleijnen et al., 1997 ). It was an attractive speculation that gp34 may operate as a transporter for peptide-loaded MHC-I molecules to bypass m152-gp37/40-mediated retention and to aid presentation of E-phase peptides, including its own processing product, the m04 peptide ²⁴³YGPSLYRRF²⁵¹ (Holtappels et al., 2000 ). However, if gp34-mediated MHC-I surface transport were essential for peptide presentation in the E-phase, K^d-restricted peptides should not exist, because, unlike the situation seen with D^d and L^d, gp34K^d complexes were not detectable (Kleijnen et al., 1997 ).

Despite the high coding capacity of CMVs, only a few antigenic proteins and peptides derived therefrom have been defined for MHC-I presentation to CD8 T cells. For a more comprehensive understanding of immunity to CMVs, there is an urgent need to identify antigenic peptides assigned to different phases of the virus replication cycle. Previous work by Morello et al. (2000) has predicted an antigenic peptide in the nonstructural E-phase protein pM84, one of the two mCMV homologues of the hCMV virion tegument protein ppUL83. We have identified the K^d-restricted pM84-derived antigenic peptide ²⁹⁷AYAGLFTPL³⁰⁵. It represents the second E-phase peptide of mCMV that apparently escapes the immunosubversive mechanisms operative in the E-phase. Its existence underscores our previous notion that mCMV is held in check by a redundancy of CD8 T cells recognizing antigenic peptides in different phases of virus gene expression. The mCMV model predicts that further antigenic peptides derived from E- or L-phase proteins will exist also for hCMV.

We thank Stefan Stevanovic, University of Tübingen, Germany, for advice regarding the search for antigenic motifs. Support was provided by a grant to M.J.R. from the Deutsche Forschungsgemeinschaft, Sonderforschungsbereich 490, individual project B1 Immune control of latent CMV infection.