Rhesus Macaque MHC Class I Molecules Present HLA-B-Like Peptides1
http://www.100md.com
免疫学杂志 2005年第13期
Abstract
SIV-infected Indian rhesus macaques (Macaca mulatta) are an important animal model for humans infected with HIV. Understanding macaque (M. mulatta class I (Mamu)) MHC class I-peptide binding facilitates the comparison of SIV- and HIV-specific cellular immune responses. In this study, we characterized the endogenous peptide-binding properties of three Mamu-A (A*02, A*08, A*11) and three Mamu-B (B*01, B*03, B*12) class I molecules. Motif comparisons revealed that five of the six macaque class I molecules (A*02, A*08, A*11, B*01, and B*03) have peptide-binding motifs similar to those of human class I molecules. Of the 65 macaque endogenous peptide ligands that we sequenced by tandem mass spectroscopy, 5 were previously eluted from HLA class I molecules. Nonamers predominated among the individual ligands, and both the motifs and the individual ligands indicated P2, P9, and various ancillary anchors. Interestingly, peptide binding of the Mamu-A and Mamu-B molecules exhibited cross-species peptide-presentation overlap primarily with HLA-B molecules. Indeed, all of the macaque class I molecules appeared HLA-B-like in peptide presentation. Remarkably, the overlap in macaque- and HLA-peptide presentation occurred despite divergent class I peptide-binding grooves. Macaque and human class I differing by up to 42 aa (13–23%) within the -1 and -2 domains, including substantial divergence within specificity pockets A-F, bound the same endogenous peptide. Therefore, endogenous peptide characterization indicates that macaque class I molecules may be the functional equivalents of HLA-B molecules.
Introduction
Class I MHC molecules mediate antiviral immune responses by first binding short viral peptides (7–14 aa) within the infected host cell and then transiting to the cell surface where they present their peptide cargo to CTL (1, 2, 3, 4, 5, 6). Viruses have developed a myriad of means to avoid the presentation of viral peptides by class I MHC molecules (7, 8, 9, 10, 11), and as an immune countermeasure, individuals within each mammalian species tend to differ in their class I MHC genotype and phenotype (12, 13, 14, 15). The polymorphism that distinguishes one class I molecule from the next alters the peptide ligands that can be accommodated within a class I molecule’s peptide-binding groove (16, 17). Therefore, MHC class I molecules tend to differ functionally from individual to individual within a species.
Animal models represent a valuable tool for studying host-pathogen relationships, and the Indian rhesus macaque represents an important animal model for a number of viruses that infect humans (18, 19, 20, 21, 22, 23). Notably, infection of macaques with SIV results in an AIDS-like illness similar to that of humans infected with HIV (24). Use of the Indian rhesus macaque as a model for HIV infection confirms the important role of class I MHC in mediating antiviral immune responses and has led to the development of molecular class I typing methods and the application of synthetic peptide binding assays for macaque MHC molecules (24, 25, 26, 27, 28). With the recent discovery of as many as 16 active rhesus MHC class I gene loci, assays for determining macaque class I MHC gene types and peptide-binding specificities will play an expanded role in characterizing antiviral immune responses in this primate (29).
In humans, class I MHC peptide-binding specificities have been empirically elucidated through a combination of 1) the Edman sequencing of pooled endogenous ligands, 2) the mass spectrometric sequencing of individual endogenous ligands, and 3) competitive synthetic peptide-binding assays (30, 31, 32). By comparison, the peptide specificity of MHC class I molecules in the Indian rhesus macaque has been derived almost exclusively from in vitro competitive-binding studies with synthetic peptides (10, 26, 27, 33, 34). Peptide-binding assays delineate the major peptide anchors preferred by a given class I molecule and have led to predictions of peptides that are presented in vivo by a particular molecule. Sequencing of endogenous ligands complements synthetic peptide-binding data by demonstrating that ligands of a particular sequence endogenously arise, traffic, survive, and are loaded into MHC molecules within a cell.
To characterize endogenous peptide-binding specificities for Indian rhesus macaque class I molecules, and to better relate the peptide-binding properties of rhesus monkey and human class I molecules, we have established endogenous-peptide motifs for six macaque class I molecules. Additionally, we sequenced >60 endogenously loaded peptide ligands, representing the first comprehensive set of naturally loaded MHC class I ligands identified in nonhuman primates. Like the human, the macaque class I molecules bound primarily nonamers with distinct P2 and P9 anchor preferences. Surprisingly, the macaque and human class I MHC molecules that bound similar (and in some cases identical) peptides did so despite divergent peptide-binding grooves. Even a substantial number of nonconservative amino acid changes within specificity pockets did not abrogate the binding of similar peptides by particular pairs of macaque and human class I molecules. Finally, the three Macaca mulatta class I-A (Mamu-A)4 and three Mamu-B class I molecules studied here are all HLA-B functional equivalents. The implications of this similarity in peptide binding is discussed.
Materials and Methods
Mamu protein production
Cloned DNA for Mamu-A*02, -A*08, -B*01, and -B*03 was amplified using the 5MamuKE (5'-GGGCGAATTCCCGCCGCCACCATGGCGCCCCGAACCC-3') and 3VLDLRHIN (5'-CCGCAAGCTTTCAGGCAAGGTCGTCGTCGGTGCTGACGACGCTCCATCTCAGGGTGAGGGG-3') or 5Mamu1KE (5'-GGGCGAATTCCCGCCGCCACCATGCGGGTCATGGCTCC-3') and 3VLDLRHIN (5'-CCGCAAGCTTTCAGGCAAGGTCGTCGTCGGTGCTGACGACGCTCCATCTCAGGGTGAGGGG-3') for Mamu-A*11 and -B*12. Both primer pairs resulted in an amplicon of 900 bp corresponding to a truncated protein lacking the cytoplasmic and transmembrane domains; the latter primer pair also encoded the epitope tag SVVSTDDDLA which was used for protein purification. Each amplicon was cloned into the mammalian expression vector pCDNA 3.1- (Invitrogen Life Technologies) and sequenced to confirm insert integrity. Constructs were electroporated into the class I-negative cell line 721.221 as previously described (35, 36). Stable transfectants were created for each macaque molecule and cultured in CP2500 hollow-fiber bioreactors (Biovest International). Soluble macaque class I molecules and their bound peptides were collected in the Cell Pharm supernatant for purification.
Peptide isolation
Once 25 mg of soluble macaque molecules (sMamu) were collected, the supernatant was clarified by centrifugation and purified as previously described over W6/32 affinity columns (37, 38). Because of weak recognition of sMamu-A*11 and sMamu-B*12 by W6/32, these two molecules were purified using 6A6 (American Type Culture Collection) affinity columns against the epitope tag (encoded by PCR) coupled to a Sepharose 4B matrix (Amersham Pharmacia Biotech) according to the manufacturer’s protocol. Molecules were eluted from the column in the same manner as the W6/32 column (with 0.2 N acetic acid), and were processed identically after elution. Eluted molecules were brought to a final concentration of 10% acetic acid and heated to boiling for 10 min. Peptides were then purified by passage through a stirred cell ultrafiltration device equipped with a 3-kDa cut-off membrane (Millipore) before loading onto a Jupiter Proteo 4 micrometer C18 reversed-phase HPLC column (Phenomenex).
Peptide characterization
Eluted peptides were first subjected to nine cycles of Edman degradation analysis to produce a peptide motif from endogenous ligands. Motifs were generated according to the description of Barber et al. (39). Briefly, residues were classified as dominant if they exhibited a 3.5-fold or greater increase in picomoles over the previous round of sequencing, strong if they exhibited a 2.5- to 3.5-fold increase, and weak if they possessed a 2.0- to 2.5-fold increase. Mass spectral analysis was performed on a Q-ToF Q-Star (PerSceptive Sciex) mass spectrometer as previously described (38).
Pocket analysis
Macaque molecules were analyzed for their peptide binding in accordance with the B- and F-pocket descriptions of Saper et al. (40). Numbering of residues of the class I H chain began with position 1 being the first residue of the mature H chain (excluding the leader sequence). Anchor residues involved in B-pocket binding were therefore 7, 9, 24, 25, 34, 45, 63, 66, 67, 70, and 99; residues involved in F-pocket binding were 77, 80, 81, 116, 123, 143, 146, and 147. Additionally, the A-pocket residues were 5, 7, 59, 63, 66, 99, 159, 163, 167, and 171. The C-pocket residues were 9, 70, 73, 74, and 97. The D-pocket residues were 99, 113, 114, 155, 156, 159, and 160. The E-pocket residues were 97, 114, 147, 152, and 156. Separate protein alignments were performed with each macaque class I molecule and the human class I molecule(s) possessing matching motifs and/or binding identical ligands. Dashes in the alignments indicate identity. Nonidentities are indicated by the single letter amino acid code of the differing residue.
Results
Endogenous peptide motifs of six macaque molecules exhibit properties similar to HLA molecules
To characterize the peptides endogenously bound by Indian rhesus macaque class I molecules and to compare their peptide repertoires to those in human MHC molecules, we produced transfectants of sMamu-A*02, -A*08, -A*11, -B*01, -B*03, and -B*12 in the human, class I-negative cell line 721.221 (35). Through hollow-fiber bioreactor culture of stable transfectants, we produced 25 mg of each molecule, purified molecules by affinity chromatography, and isolated and purified 0.5 mg of naturally bound peptides. We then recorded peptide motifs for each of the molecules in accordance with previously published procedures for human class I molecules (39) (Fig. 1).
Perhaps not surprisingly, Mamu-B molecules also possess HLA-B-like motif specificities. Mamu-B*01 closely resembles HLA-B*3701. Although B*3701 exhibits a slightly different preference for peptides possessing a D or M at P2 (no dominant preference for E is seen), other HLA motifs indicate that the negatively charged residues D and E are largely interchangeable at P2. Therefore, it is likely that the extended characterization of HLA-B*3701 would find an E at P2 (30, 41). Mamu-B*03 possesses a P2 R and P9 L as do members of the HLA-B*27 family; the P9 L is accommodated by most members of the B*27 family and is a dominant anchor in B*2706, 2707, and 2709. Additionally, HLA-B*1402 also possesses the same combination of a P2 R and P9 L as Mamu-B*03.
The single motif outlier in these studies is Mamu B*12. This macaque molecule strongly prefers peptides with an N at P2. No MHC class I motif, including both human and nonhuman primate, has been described with this anchor preference. Interestingly, multiple HLA class I alleles present individual peptide ligands with a P2 N, including HLA-B8, -B51, -B62, -B*1503, and B*1801 (41). The chimpanzee Patr-B*16 molecule also presents a peptide with a P2 N (42). Because a motif is not yet described for this molecule, it may reflect this P2 preference or it may simply be a subdominant peptide class capable of binding as seen in the HLA-B molecules presenting an N-P2 peptide. The single known ligand previously reported for Mamu-B*12 also possesses a P2 N (43). At this point, it is not possible to correlate the peptide motif of Mamu-B*12 with a given HLA-class I locus.
Macaque class I molecules expressed in human cells load and present endogenous peptides
To understand the nature of individual peptides endogenously bound by Mamu MHC class I, we next sequenced a number of peptides (65) eluted from the six macaque class I molecules (Table II). Peptides bound by all molecules varied in length from 7 to 11 residues with an average length of 9 (Fig. 2), in accordance with previously published human peptides (Refs. 38 and 44 and www.hlaligand.ouhsc.edu). Mamu A*08 bound a particularly short heptamer, TVLTSKY, derived from hemoglobin , but this molecule also accommodated the decamer VPWSVQQDSP from the KIAA0960 protein. In general, Mamu A*08 favored shorter peptides (octamers) than the other macaque molecules studied here (Fig. 2). The longest peptides presented by any macaque molecule were two undecamers bound by Mamu-B*03 (Table II). Thus, the macaque class I molecules were closely allied with human class I molecules in their peptide-length preferences.
The range of peptides bound by each macaque molecule also resembled those bound by HLA molecules with regard to peptides not indicated by the motif. For instance, although Mamu B*03 presented mostly peptides with the strong anchor of R predicted by motif analysis, peptides with a K at P2 were also identified (Table II). Likewise, peptides bound by Mamu B*12 contained the dominant anchor of N predicted by motif analysis but peptides with S at P2 were also abundant (Table II). As seen with studies of human class I peptides, motifs provide an incomplete picture of diversity within the macaque class I-peptide population.
Macaque and human class I molecules present identical peptides
To identify individual macaque ligands that could also be naturally presented by HLA molecules, we searched the MHC class I databases, HLA Ligand/Motif Online Database and SYFPEITHI(www.syfpeithi.de and Refs. 45 and 46). Surprisingly, we identified five peptides that were endogenously bound by both macaque and HLA molecules. Mamu-A*11 peptide GEHGLIIRV is also presented by HLA-B*4006, while Mamu-A*11 peptide SEFGLKISFV is likely presented by HLA-B*1801; we found a longer version of the nonamer SEFGLKISF that bound by B*1801 and, given unlimited peptide for analysis, would likely find the nonamer version as well. Thus, Mamu-A*11 presents documented HLA-B peptides.
Three Mamu-B endogenous peptides are also presented by HLA-B molecules. The peptide KRFSGTVRL is presented by Mamu-B*03 and HLA-B*2703 (Table II). Peptide KRTTVVAQL is presented by Mamu-B*03 and HLA-B*2709 (47, 48, 49). The Mamu-B*12 peptide YAFNMKATV (not represented by the B*12 motif) is presented by HLA B*5101. Therefore, Mamu-B molecules present peptides identical to several HLA-B molecules.
The mass spectrometric identification of multiple peptides bound by both human and macaque molecules demonstrates that the macaque molecules are functionally equivalent to HLA molecules in terms of peptide binding and presentation. All five peptides with dual macaque and human presentation were bound by HLA-B molecules. Therefore, the macaque molecules in this study are predominantly HLA-B functional equivalents at the individual ligand level.
Macaque molecules possess divergent peptide-binding pockets when compared with human molecules
To better understand the structural basis for the accommodation of the same peptides by macaque and human class I molecules, we analyzed the pocket structure of both sets of molecules. Using the previously established A-F specificity pocket residues set forth by Saper (40), we produced and aligned protein sequences containing only pocket-specific residues. Each macaque molecule and its human peptide-binding equivalent was compared in an individual alignment. We hypothesized that these alignments would indicate that the macaque and human molecules that bound identical peptides or that shared motifs would predominantly differ outside of the peptide-specificity pockets.
We first analyzed the five pairs of macaque and human class I molecules that bound identical peptides. We initially counted residue differences between these pairs that lied within specificity pockets A-F (Fig. 3). The highest divergence of specificity pockets was seen in Mamu-B*12 and HLA-B*5101 (which present the same peptide) with 54.5% of the specificity pocket residues differing. Although not as disparate as 54.5%, all macaque/human class I pairs presenting identical ligands differed significantly in their specificity pocket residues: Mamu B*03 and HLA-B*2703 differ in 40.9% of residues involved in peptide binding. Mamu B*03 and HLA-B*2709 differ by 38.6%, and Mamu-A*11 and B*1801 diverge by 31.8%. Mamu-A*11 and B*4006 were the closest macaque/human pair in terms of shared specificity-pocket residues yet still differed in 18.2% of residues (Fig. 3). Therefore, substantial divergence existed in the peptide-binding pockets of macaque and human molecules despite the ability of the pairs to bind identical peptides.
We next analyzed all amino acid positions that were not conserved between the macaque and human class I molecules possessing similar peptide-binding specificities (44 positions) and tried to identify regions of amino acid conservation between the groups (Fig. 4). In the Mamu-A*02/HLA-A*1101, B*1516, B*1517 peptide-specificity group, we found several small, conserved groups of residues such as IAQA at positions 23, 24, 32, and 42 of the class I H chain and RNMA at 65, 66, 67, and 69 that possibly contribute to the binding of similar peptides. However, sequence conservation at these two sets of residues is not seen in the other macaque/human groups. Likewise, the conserved NTTYE at amino acids 70, 71, 73, 74, and 76 in the Mamu-A*11 group is not conserved in other macaque/human peptide-specificity groups. Therefore, among the macaque/human groups, we were unable to identify conserved positions in the class I H chain that consistently contributed to overlapping motifs or endogenous-ligand loading.
Although Fig. 4 provides a linear comparison of the primary sequence of the macaque/human peptide-specificity groups, it remained possible that discontinuous regions of conservation might become localized in the three-dimensional structure of the class I H chains. Therefore, we compared protein structures of the macaque/human pairs binding identical peptides based upon existing crystal structures of the molecules (Fig. 5). Interestingly, polymorphisms between different macaque/human peptide-binding partners are distributed throughout the three-dimensional structures of the molecules. Additionally and as predicted by specificity pocket analysis, many of the differing residues directly contact the bound peptide. Through structural analyses, we saw no clear patterns of amino acid conservation among pairs of macaque and human class I molecules that overlap in peptide presentation.
Discussion
We have expanded our current knowledge of the MHC class I peptide-binding properties of the Indian rhesus macaques through the characterization of endogenous ligands and motifs. In combination with synthetic peptide-binding assays, these data will provide a more complete foundation for understanding the dynamics of peptide presentation and viral evasion in SIV-infected animals (10, 50, 51). These data also suggest that the conventions governing macaque class I peptide overlaps and divergences have yet to be elucidated.
To describe the peptide-binding properties of six macaque molecules (three Mamu-A and three Mamu-B molecules), we eluted naturally processed and loaded peptides from soluble Mamu molecules produced in tissue culture. Peptides eluted from each molecule were then characterized for their pooled motif and for their individual ligand sequences. The peptides presented by macaque class I MHC molecules resembled peptides bound by human molecules in several important ways. Primarily, the macaque peptides possessed the same length preferences of 7–14 aa. Anchor residues were also a prerequisite for endogenous peptide binding by the macaque molecules, most of which exhibited a dominant P2 anchor, a C-terminal anchor, and dispersed ancillary anchors. Like most human molecules, the peptides were derived from a variety of endogenous cellular proteins and not limited to a small subset of cellular proteins (Table I and Refs. 38 , 52 , and 53). Additionally, each macaque molecule in this study possessed a unique pooled Edman motif as well as distinct individual peptide ligands. There is considerable peptide-binding overlap among divergent class I HLA molecules (54). Because there are peptide-binding similarities between rhesus and human MHC class I molecules, we posit that further study of the macaque class I and their peptide ligands will reveal further peptide-binding overlaps among divergent MHC class I molecules in this species.
The initial finding of physical similarity between macaque and human peptides, along with observations from other studies, prompted us to search for human counterparts to the macaque class I ligands and motifs described here. Remarkably, all of the macaque molecules possessed a motif or individual ligands identical to one or more human class I molecule. Even more surprisingly, both Mamu-A and Mamu-B molecules possessed HLA-B-like peptide specificities: motif analysis of the Mamu-A locus molecules revealed peptide specificities similar or identical to those of HLA-B locus molecules B*1516/17, B*7801 and B*1801 or B*4402. Likewise, Mamu-B locus molecules possessed specificities akin to B*1402, B*2703/09, and B*3701. The identification of multiple ligands bound by macaque molecules that were also identified in human molecules strengthened this human B locus equivalency. Indeed, all of the ligands identified as endogenously presented by both human and macaque molecules were bound by human B locus molecules, even those presented by the macaque A locus molecule Mamu A*11. In terms of endogenous ligand binding, these data indicate that the Mamu-A and -B molecules both resemble HLA-B molecules in their peptide-binding specificities.
The peptide elution experiments performed in this study were accomplished with macaque class I produced in human cells known to generate peptide ligands for a large number of different HLA class I molecules (38, 55, 56, 57, 58, 59, 60). We used human cells to facilitate a direct comparison of the peptide-binding properties of the macaque class I molecules to those of human class I molecules, although the cell line used could impact the peptide presentation of macaque class I molecules through the intracellular protein/peptide-processing machinery and/or the association of the macaque H chain with endogenous 2-microglobulin. Therefore, the peptides characterized in this study may not be identical to those presented by macaque class I molecules expressed in a macaque cell line; however, the underlying motif and ligand sequence data (based on the ability of the macaque molecule to physically bind peptides) will be the same in a cell line of either origin. Additionally, the differences between human 2-microglobulin and macaque 2-microglobulin lie in regions not involved in association with the H chain (61). Thus, binding of human 2-microglobulin should not alter the three-dimensional structure of the macaque H chain and peptide-binding groove such that different peptides are presented. Further studies with macaque cell lines will determine the impact, if any, of using a human cell line to characterize peptides for comparative purposes.
Because particular pairs of macaque and human molecules possessed similar peptide-binding specificities, we hypothesized that amino acid identity/similarity within the peptide-binding groove, especially at positions in specificity pockets A-F, would facilitate similar peptide accommodation. Sequence comparison of residues within the binding groove showed no such similarity. Some macaque/human class I pairs that bound the same peptide differed by 55% in the peptide-binding groove. Although we were able to identify small stretches of conserved sequence in the human/macaque MHC class I pairs, these were not conserved in all pairs presenting identical peptides. Likewise, comparison of the three-dimensional structures of identical peptide-presenting pairs produced no obvious mechanism for divergent grooves binding the same peptide. Finally, residue differences between the human/macaque MHC class I pairs directly contacted the bound peptide. It is difficult to understand why MHC class I molecules with such striking similarities in the physical nature of the peptides that they bind would not use a similar peptide-binding groove.
It is interesting to speculate about the biological impact of multiple HLA-B-like macaque-peptide specificities. Several recent studies have pointed to a significant role for HLA-B in the resolution of viral infection. For instance, HLA-B molecules are proposed to place significantly more pressure on HIV than HLA-A, and HLA-B molecules are also thought to have a dominant involvement in influencing HIV disease outcome (62). Additionally, HLA-B molecules (and not HLA-A) present immunodominant CMV- and EBV-derived peptide epitopes (63, 64). Perhaps the presence of a large number of HLA-B-like molecules in the macaque contributes to their long-term survival in a natural setting despite high SIV viral loads (65), and perhaps other nonhuman primates will exhibit a propensity for HLA-B-like peptide ligands as well.
In summary, macaque MHC class I molecules bind and present peptides with similar physical properties to the peptides bound by human class I molecules. In this study, we provide the first demonstration of rhesus and human MHC class I molecules binding identical endogenous peptides. This suggests that there is functional equivalency between particular MHC class I molecules. We also show that the macaque class I molecules present peptides that are distinctly HLA-B-like. The mechanisms behind this overlap in peptide binding in these two species are not entirely understood. Additional studies should reveal how cross-species peptide binding occurs despite divergent binding grooves.
Acknowledgments
We thank Kenneth Hatter of the Molecular Biology Resource Facility at the William K. Warren Medical Research Foundation for mass spectrometer technical assistance, Melva Gonzalez for critical reading of the manuscript, and Sreedevi Chandrasekaran for help in preparation of the manuscript.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported by a National Institutes of Health Grant 5R24RR015371 (to D.I.W.). H.D.H.-M. was supported by National Institutes of Health Training Grant T32AI07633.
2 Current address: Laboratory of Viral Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD.
3 Address correspondence and reprint requests to Dr. William H. Hildebrand, Department of Microbiology and Immunology, University of Oklahoma Health Sciences Center, 975 N.E. 10th Street, Oklahoma City, OK, 73104. E-mail address: William-hildebrand@ouhsc.edu
4 Abbreviations used in this paper: Mamu, Macaca mulatta class I; sMamu, soluble Mamu.
Received for publication February 11, 2005. Accepted for publication April 13, 2005.
References
Townsend, A. R. M., F. M. Gotch, J. Davey. 1985. Cytotoxic T cells recognize fragments of the influenza nucleoprotein. Cell 42: 457-467.
Townsend, A. R. M., J. Bastin, K. Gould, G. G. Brownlee. 1986. Cytotoxic T lymphocytes recognize influenza haemagglutinin that lacks a signal sequence. Nature 324: 575-577.
Nixon, D. F., A. R. M. Townsend, J. G. Elvin, C. R. Rizza, J. Gallwey, A. J. McMichael. 1988. HIV-1 gag specific cytotoxic T lymphocytes defined with recombinant vaccinia virus and synthetic peptides. Nature 336: 484-487.
McMichael, A. J., B. D. Walker. 1994. Cytotoxic T lymphocyte epitopes: implications for HIV vaccines. AIDS 8: S155-S163.
Brodsky, F. M., L. E. Guagliardi. 1991. The cell biology of antigen processing and presentation. Annu. Rev. Immunol. 9: 707-744.
Brodsky, F. M., L. Lem, A. Solache, E. M. Bennett. 1999. Human pathogen subversion of antigen presentation. Immunol. Rev. 168: 199-215.
Nowak, M. A., R. M. May, R. E. Phillips, S. Rowland-Jones, D. G. Lalloo, S. McAdam, P. Klenerman, B. Koppe, K. Sigmund, C. R. Bangham, et al 1995. Antigenic oscillations and shifting immunodominance in HIV-1 infections. Nature 375: 606-611.
Collins, K. L., D. Baltimore. 1999. HIV’s evasion of the cellular immune response. Immunol. Rev. 168: 65-74.(Heather D. Hickman-Miller)
SIV-infected Indian rhesus macaques (Macaca mulatta) are an important animal model for humans infected with HIV. Understanding macaque (M. mulatta class I (Mamu)) MHC class I-peptide binding facilitates the comparison of SIV- and HIV-specific cellular immune responses. In this study, we characterized the endogenous peptide-binding properties of three Mamu-A (A*02, A*08, A*11) and three Mamu-B (B*01, B*03, B*12) class I molecules. Motif comparisons revealed that five of the six macaque class I molecules (A*02, A*08, A*11, B*01, and B*03) have peptide-binding motifs similar to those of human class I molecules. Of the 65 macaque endogenous peptide ligands that we sequenced by tandem mass spectroscopy, 5 were previously eluted from HLA class I molecules. Nonamers predominated among the individual ligands, and both the motifs and the individual ligands indicated P2, P9, and various ancillary anchors. Interestingly, peptide binding of the Mamu-A and Mamu-B molecules exhibited cross-species peptide-presentation overlap primarily with HLA-B molecules. Indeed, all of the macaque class I molecules appeared HLA-B-like in peptide presentation. Remarkably, the overlap in macaque- and HLA-peptide presentation occurred despite divergent class I peptide-binding grooves. Macaque and human class I differing by up to 42 aa (13–23%) within the -1 and -2 domains, including substantial divergence within specificity pockets A-F, bound the same endogenous peptide. Therefore, endogenous peptide characterization indicates that macaque class I molecules may be the functional equivalents of HLA-B molecules.
Introduction
Class I MHC molecules mediate antiviral immune responses by first binding short viral peptides (7–14 aa) within the infected host cell and then transiting to the cell surface where they present their peptide cargo to CTL (1, 2, 3, 4, 5, 6). Viruses have developed a myriad of means to avoid the presentation of viral peptides by class I MHC molecules (7, 8, 9, 10, 11), and as an immune countermeasure, individuals within each mammalian species tend to differ in their class I MHC genotype and phenotype (12, 13, 14, 15). The polymorphism that distinguishes one class I molecule from the next alters the peptide ligands that can be accommodated within a class I molecule’s peptide-binding groove (16, 17). Therefore, MHC class I molecules tend to differ functionally from individual to individual within a species.
Animal models represent a valuable tool for studying host-pathogen relationships, and the Indian rhesus macaque represents an important animal model for a number of viruses that infect humans (18, 19, 20, 21, 22, 23). Notably, infection of macaques with SIV results in an AIDS-like illness similar to that of humans infected with HIV (24). Use of the Indian rhesus macaque as a model for HIV infection confirms the important role of class I MHC in mediating antiviral immune responses and has led to the development of molecular class I typing methods and the application of synthetic peptide binding assays for macaque MHC molecules (24, 25, 26, 27, 28). With the recent discovery of as many as 16 active rhesus MHC class I gene loci, assays for determining macaque class I MHC gene types and peptide-binding specificities will play an expanded role in characterizing antiviral immune responses in this primate (29).
In humans, class I MHC peptide-binding specificities have been empirically elucidated through a combination of 1) the Edman sequencing of pooled endogenous ligands, 2) the mass spectrometric sequencing of individual endogenous ligands, and 3) competitive synthetic peptide-binding assays (30, 31, 32). By comparison, the peptide specificity of MHC class I molecules in the Indian rhesus macaque has been derived almost exclusively from in vitro competitive-binding studies with synthetic peptides (10, 26, 27, 33, 34). Peptide-binding assays delineate the major peptide anchors preferred by a given class I molecule and have led to predictions of peptides that are presented in vivo by a particular molecule. Sequencing of endogenous ligands complements synthetic peptide-binding data by demonstrating that ligands of a particular sequence endogenously arise, traffic, survive, and are loaded into MHC molecules within a cell.
To characterize endogenous peptide-binding specificities for Indian rhesus macaque class I molecules, and to better relate the peptide-binding properties of rhesus monkey and human class I molecules, we have established endogenous-peptide motifs for six macaque class I molecules. Additionally, we sequenced >60 endogenously loaded peptide ligands, representing the first comprehensive set of naturally loaded MHC class I ligands identified in nonhuman primates. Like the human, the macaque class I molecules bound primarily nonamers with distinct P2 and P9 anchor preferences. Surprisingly, the macaque and human class I MHC molecules that bound similar (and in some cases identical) peptides did so despite divergent peptide-binding grooves. Even a substantial number of nonconservative amino acid changes within specificity pockets did not abrogate the binding of similar peptides by particular pairs of macaque and human class I molecules. Finally, the three Macaca mulatta class I-A (Mamu-A)4 and three Mamu-B class I molecules studied here are all HLA-B functional equivalents. The implications of this similarity in peptide binding is discussed.
Materials and Methods
Mamu protein production
Cloned DNA for Mamu-A*02, -A*08, -B*01, and -B*03 was amplified using the 5MamuKE (5'-GGGCGAATTCCCGCCGCCACCATGGCGCCCCGAACCC-3') and 3VLDLRHIN (5'-CCGCAAGCTTTCAGGCAAGGTCGTCGTCGGTGCTGACGACGCTCCATCTCAGGGTGAGGGG-3') or 5Mamu1KE (5'-GGGCGAATTCCCGCCGCCACCATGCGGGTCATGGCTCC-3') and 3VLDLRHIN (5'-CCGCAAGCTTTCAGGCAAGGTCGTCGTCGGTGCTGACGACGCTCCATCTCAGGGTGAGGGG-3') for Mamu-A*11 and -B*12. Both primer pairs resulted in an amplicon of 900 bp corresponding to a truncated protein lacking the cytoplasmic and transmembrane domains; the latter primer pair also encoded the epitope tag SVVSTDDDLA which was used for protein purification. Each amplicon was cloned into the mammalian expression vector pCDNA 3.1- (Invitrogen Life Technologies) and sequenced to confirm insert integrity. Constructs were electroporated into the class I-negative cell line 721.221 as previously described (35, 36). Stable transfectants were created for each macaque molecule and cultured in CP2500 hollow-fiber bioreactors (Biovest International). Soluble macaque class I molecules and their bound peptides were collected in the Cell Pharm supernatant for purification.
Peptide isolation
Once 25 mg of soluble macaque molecules (sMamu) were collected, the supernatant was clarified by centrifugation and purified as previously described over W6/32 affinity columns (37, 38). Because of weak recognition of sMamu-A*11 and sMamu-B*12 by W6/32, these two molecules were purified using 6A6 (American Type Culture Collection) affinity columns against the epitope tag (encoded by PCR) coupled to a Sepharose 4B matrix (Amersham Pharmacia Biotech) according to the manufacturer’s protocol. Molecules were eluted from the column in the same manner as the W6/32 column (with 0.2 N acetic acid), and were processed identically after elution. Eluted molecules were brought to a final concentration of 10% acetic acid and heated to boiling for 10 min. Peptides were then purified by passage through a stirred cell ultrafiltration device equipped with a 3-kDa cut-off membrane (Millipore) before loading onto a Jupiter Proteo 4 micrometer C18 reversed-phase HPLC column (Phenomenex).
Peptide characterization
Eluted peptides were first subjected to nine cycles of Edman degradation analysis to produce a peptide motif from endogenous ligands. Motifs were generated according to the description of Barber et al. (39). Briefly, residues were classified as dominant if they exhibited a 3.5-fold or greater increase in picomoles over the previous round of sequencing, strong if they exhibited a 2.5- to 3.5-fold increase, and weak if they possessed a 2.0- to 2.5-fold increase. Mass spectral analysis was performed on a Q-ToF Q-Star (PerSceptive Sciex) mass spectrometer as previously described (38).
Pocket analysis
Macaque molecules were analyzed for their peptide binding in accordance with the B- and F-pocket descriptions of Saper et al. (40). Numbering of residues of the class I H chain began with position 1 being the first residue of the mature H chain (excluding the leader sequence). Anchor residues involved in B-pocket binding were therefore 7, 9, 24, 25, 34, 45, 63, 66, 67, 70, and 99; residues involved in F-pocket binding were 77, 80, 81, 116, 123, 143, 146, and 147. Additionally, the A-pocket residues were 5, 7, 59, 63, 66, 99, 159, 163, 167, and 171. The C-pocket residues were 9, 70, 73, 74, and 97. The D-pocket residues were 99, 113, 114, 155, 156, 159, and 160. The E-pocket residues were 97, 114, 147, 152, and 156. Separate protein alignments were performed with each macaque class I molecule and the human class I molecule(s) possessing matching motifs and/or binding identical ligands. Dashes in the alignments indicate identity. Nonidentities are indicated by the single letter amino acid code of the differing residue.
Results
Endogenous peptide motifs of six macaque molecules exhibit properties similar to HLA molecules
To characterize the peptides endogenously bound by Indian rhesus macaque class I molecules and to compare their peptide repertoires to those in human MHC molecules, we produced transfectants of sMamu-A*02, -A*08, -A*11, -B*01, -B*03, and -B*12 in the human, class I-negative cell line 721.221 (35). Through hollow-fiber bioreactor culture of stable transfectants, we produced 25 mg of each molecule, purified molecules by affinity chromatography, and isolated and purified 0.5 mg of naturally bound peptides. We then recorded peptide motifs for each of the molecules in accordance with previously published procedures for human class I molecules (39) (Fig. 1).
Perhaps not surprisingly, Mamu-B molecules also possess HLA-B-like motif specificities. Mamu-B*01 closely resembles HLA-B*3701. Although B*3701 exhibits a slightly different preference for peptides possessing a D or M at P2 (no dominant preference for E is seen), other HLA motifs indicate that the negatively charged residues D and E are largely interchangeable at P2. Therefore, it is likely that the extended characterization of HLA-B*3701 would find an E at P2 (30, 41). Mamu-B*03 possesses a P2 R and P9 L as do members of the HLA-B*27 family; the P9 L is accommodated by most members of the B*27 family and is a dominant anchor in B*2706, 2707, and 2709. Additionally, HLA-B*1402 also possesses the same combination of a P2 R and P9 L as Mamu-B*03.
The single motif outlier in these studies is Mamu B*12. This macaque molecule strongly prefers peptides with an N at P2. No MHC class I motif, including both human and nonhuman primate, has been described with this anchor preference. Interestingly, multiple HLA class I alleles present individual peptide ligands with a P2 N, including HLA-B8, -B51, -B62, -B*1503, and B*1801 (41). The chimpanzee Patr-B*16 molecule also presents a peptide with a P2 N (42). Because a motif is not yet described for this molecule, it may reflect this P2 preference or it may simply be a subdominant peptide class capable of binding as seen in the HLA-B molecules presenting an N-P2 peptide. The single known ligand previously reported for Mamu-B*12 also possesses a P2 N (43). At this point, it is not possible to correlate the peptide motif of Mamu-B*12 with a given HLA-class I locus.
Macaque class I molecules expressed in human cells load and present endogenous peptides
To understand the nature of individual peptides endogenously bound by Mamu MHC class I, we next sequenced a number of peptides (65) eluted from the six macaque class I molecules (Table II). Peptides bound by all molecules varied in length from 7 to 11 residues with an average length of 9 (Fig. 2), in accordance with previously published human peptides (Refs. 38 and 44 and www.hlaligand.ouhsc.edu). Mamu A*08 bound a particularly short heptamer, TVLTSKY, derived from hemoglobin , but this molecule also accommodated the decamer VPWSVQQDSP from the KIAA0960 protein. In general, Mamu A*08 favored shorter peptides (octamers) than the other macaque molecules studied here (Fig. 2). The longest peptides presented by any macaque molecule were two undecamers bound by Mamu-B*03 (Table II). Thus, the macaque class I molecules were closely allied with human class I molecules in their peptide-length preferences.
The range of peptides bound by each macaque molecule also resembled those bound by HLA molecules with regard to peptides not indicated by the motif. For instance, although Mamu B*03 presented mostly peptides with the strong anchor of R predicted by motif analysis, peptides with a K at P2 were also identified (Table II). Likewise, peptides bound by Mamu B*12 contained the dominant anchor of N predicted by motif analysis but peptides with S at P2 were also abundant (Table II). As seen with studies of human class I peptides, motifs provide an incomplete picture of diversity within the macaque class I-peptide population.
Macaque and human class I molecules present identical peptides
To identify individual macaque ligands that could also be naturally presented by HLA molecules, we searched the MHC class I databases, HLA Ligand/Motif Online Database and SYFPEITHI(www.syfpeithi.de and Refs. 45 and 46). Surprisingly, we identified five peptides that were endogenously bound by both macaque and HLA molecules. Mamu-A*11 peptide GEHGLIIRV is also presented by HLA-B*4006, while Mamu-A*11 peptide SEFGLKISFV is likely presented by HLA-B*1801; we found a longer version of the nonamer SEFGLKISF that bound by B*1801 and, given unlimited peptide for analysis, would likely find the nonamer version as well. Thus, Mamu-A*11 presents documented HLA-B peptides.
Three Mamu-B endogenous peptides are also presented by HLA-B molecules. The peptide KRFSGTVRL is presented by Mamu-B*03 and HLA-B*2703 (Table II). Peptide KRTTVVAQL is presented by Mamu-B*03 and HLA-B*2709 (47, 48, 49). The Mamu-B*12 peptide YAFNMKATV (not represented by the B*12 motif) is presented by HLA B*5101. Therefore, Mamu-B molecules present peptides identical to several HLA-B molecules.
The mass spectrometric identification of multiple peptides bound by both human and macaque molecules demonstrates that the macaque molecules are functionally equivalent to HLA molecules in terms of peptide binding and presentation. All five peptides with dual macaque and human presentation were bound by HLA-B molecules. Therefore, the macaque molecules in this study are predominantly HLA-B functional equivalents at the individual ligand level.
Macaque molecules possess divergent peptide-binding pockets when compared with human molecules
To better understand the structural basis for the accommodation of the same peptides by macaque and human class I molecules, we analyzed the pocket structure of both sets of molecules. Using the previously established A-F specificity pocket residues set forth by Saper (40), we produced and aligned protein sequences containing only pocket-specific residues. Each macaque molecule and its human peptide-binding equivalent was compared in an individual alignment. We hypothesized that these alignments would indicate that the macaque and human molecules that bound identical peptides or that shared motifs would predominantly differ outside of the peptide-specificity pockets.
We first analyzed the five pairs of macaque and human class I molecules that bound identical peptides. We initially counted residue differences between these pairs that lied within specificity pockets A-F (Fig. 3). The highest divergence of specificity pockets was seen in Mamu-B*12 and HLA-B*5101 (which present the same peptide) with 54.5% of the specificity pocket residues differing. Although not as disparate as 54.5%, all macaque/human class I pairs presenting identical ligands differed significantly in their specificity pocket residues: Mamu B*03 and HLA-B*2703 differ in 40.9% of residues involved in peptide binding. Mamu B*03 and HLA-B*2709 differ by 38.6%, and Mamu-A*11 and B*1801 diverge by 31.8%. Mamu-A*11 and B*4006 were the closest macaque/human pair in terms of shared specificity-pocket residues yet still differed in 18.2% of residues (Fig. 3). Therefore, substantial divergence existed in the peptide-binding pockets of macaque and human molecules despite the ability of the pairs to bind identical peptides.
We next analyzed all amino acid positions that were not conserved between the macaque and human class I molecules possessing similar peptide-binding specificities (44 positions) and tried to identify regions of amino acid conservation between the groups (Fig. 4). In the Mamu-A*02/HLA-A*1101, B*1516, B*1517 peptide-specificity group, we found several small, conserved groups of residues such as IAQA at positions 23, 24, 32, and 42 of the class I H chain and RNMA at 65, 66, 67, and 69 that possibly contribute to the binding of similar peptides. However, sequence conservation at these two sets of residues is not seen in the other macaque/human groups. Likewise, the conserved NTTYE at amino acids 70, 71, 73, 74, and 76 in the Mamu-A*11 group is not conserved in other macaque/human peptide-specificity groups. Therefore, among the macaque/human groups, we were unable to identify conserved positions in the class I H chain that consistently contributed to overlapping motifs or endogenous-ligand loading.
Although Fig. 4 provides a linear comparison of the primary sequence of the macaque/human peptide-specificity groups, it remained possible that discontinuous regions of conservation might become localized in the three-dimensional structure of the class I H chains. Therefore, we compared protein structures of the macaque/human pairs binding identical peptides based upon existing crystal structures of the molecules (Fig. 5). Interestingly, polymorphisms between different macaque/human peptide-binding partners are distributed throughout the three-dimensional structures of the molecules. Additionally and as predicted by specificity pocket analysis, many of the differing residues directly contact the bound peptide. Through structural analyses, we saw no clear patterns of amino acid conservation among pairs of macaque and human class I molecules that overlap in peptide presentation.
Discussion
We have expanded our current knowledge of the MHC class I peptide-binding properties of the Indian rhesus macaques through the characterization of endogenous ligands and motifs. In combination with synthetic peptide-binding assays, these data will provide a more complete foundation for understanding the dynamics of peptide presentation and viral evasion in SIV-infected animals (10, 50, 51). These data also suggest that the conventions governing macaque class I peptide overlaps and divergences have yet to be elucidated.
To describe the peptide-binding properties of six macaque molecules (three Mamu-A and three Mamu-B molecules), we eluted naturally processed and loaded peptides from soluble Mamu molecules produced in tissue culture. Peptides eluted from each molecule were then characterized for their pooled motif and for their individual ligand sequences. The peptides presented by macaque class I MHC molecules resembled peptides bound by human molecules in several important ways. Primarily, the macaque peptides possessed the same length preferences of 7–14 aa. Anchor residues were also a prerequisite for endogenous peptide binding by the macaque molecules, most of which exhibited a dominant P2 anchor, a C-terminal anchor, and dispersed ancillary anchors. Like most human molecules, the peptides were derived from a variety of endogenous cellular proteins and not limited to a small subset of cellular proteins (Table I and Refs. 38 , 52 , and 53). Additionally, each macaque molecule in this study possessed a unique pooled Edman motif as well as distinct individual peptide ligands. There is considerable peptide-binding overlap among divergent class I HLA molecules (54). Because there are peptide-binding similarities between rhesus and human MHC class I molecules, we posit that further study of the macaque class I and their peptide ligands will reveal further peptide-binding overlaps among divergent MHC class I molecules in this species.
The initial finding of physical similarity between macaque and human peptides, along with observations from other studies, prompted us to search for human counterparts to the macaque class I ligands and motifs described here. Remarkably, all of the macaque molecules possessed a motif or individual ligands identical to one or more human class I molecule. Even more surprisingly, both Mamu-A and Mamu-B molecules possessed HLA-B-like peptide specificities: motif analysis of the Mamu-A locus molecules revealed peptide specificities similar or identical to those of HLA-B locus molecules B*1516/17, B*7801 and B*1801 or B*4402. Likewise, Mamu-B locus molecules possessed specificities akin to B*1402, B*2703/09, and B*3701. The identification of multiple ligands bound by macaque molecules that were also identified in human molecules strengthened this human B locus equivalency. Indeed, all of the ligands identified as endogenously presented by both human and macaque molecules were bound by human B locus molecules, even those presented by the macaque A locus molecule Mamu A*11. In terms of endogenous ligand binding, these data indicate that the Mamu-A and -B molecules both resemble HLA-B molecules in their peptide-binding specificities.
The peptide elution experiments performed in this study were accomplished with macaque class I produced in human cells known to generate peptide ligands for a large number of different HLA class I molecules (38, 55, 56, 57, 58, 59, 60). We used human cells to facilitate a direct comparison of the peptide-binding properties of the macaque class I molecules to those of human class I molecules, although the cell line used could impact the peptide presentation of macaque class I molecules through the intracellular protein/peptide-processing machinery and/or the association of the macaque H chain with endogenous 2-microglobulin. Therefore, the peptides characterized in this study may not be identical to those presented by macaque class I molecules expressed in a macaque cell line; however, the underlying motif and ligand sequence data (based on the ability of the macaque molecule to physically bind peptides) will be the same in a cell line of either origin. Additionally, the differences between human 2-microglobulin and macaque 2-microglobulin lie in regions not involved in association with the H chain (61). Thus, binding of human 2-microglobulin should not alter the three-dimensional structure of the macaque H chain and peptide-binding groove such that different peptides are presented. Further studies with macaque cell lines will determine the impact, if any, of using a human cell line to characterize peptides for comparative purposes.
Because particular pairs of macaque and human molecules possessed similar peptide-binding specificities, we hypothesized that amino acid identity/similarity within the peptide-binding groove, especially at positions in specificity pockets A-F, would facilitate similar peptide accommodation. Sequence comparison of residues within the binding groove showed no such similarity. Some macaque/human class I pairs that bound the same peptide differed by 55% in the peptide-binding groove. Although we were able to identify small stretches of conserved sequence in the human/macaque MHC class I pairs, these were not conserved in all pairs presenting identical peptides. Likewise, comparison of the three-dimensional structures of identical peptide-presenting pairs produced no obvious mechanism for divergent grooves binding the same peptide. Finally, residue differences between the human/macaque MHC class I pairs directly contacted the bound peptide. It is difficult to understand why MHC class I molecules with such striking similarities in the physical nature of the peptides that they bind would not use a similar peptide-binding groove.
It is interesting to speculate about the biological impact of multiple HLA-B-like macaque-peptide specificities. Several recent studies have pointed to a significant role for HLA-B in the resolution of viral infection. For instance, HLA-B molecules are proposed to place significantly more pressure on HIV than HLA-A, and HLA-B molecules are also thought to have a dominant involvement in influencing HIV disease outcome (62). Additionally, HLA-B molecules (and not HLA-A) present immunodominant CMV- and EBV-derived peptide epitopes (63, 64). Perhaps the presence of a large number of HLA-B-like molecules in the macaque contributes to their long-term survival in a natural setting despite high SIV viral loads (65), and perhaps other nonhuman primates will exhibit a propensity for HLA-B-like peptide ligands as well.
In summary, macaque MHC class I molecules bind and present peptides with similar physical properties to the peptides bound by human class I molecules. In this study, we provide the first demonstration of rhesus and human MHC class I molecules binding identical endogenous peptides. This suggests that there is functional equivalency between particular MHC class I molecules. We also show that the macaque class I molecules present peptides that are distinctly HLA-B-like. The mechanisms behind this overlap in peptide binding in these two species are not entirely understood. Additional studies should reveal how cross-species peptide binding occurs despite divergent binding grooves.
Acknowledgments
We thank Kenneth Hatter of the Molecular Biology Resource Facility at the William K. Warren Medical Research Foundation for mass spectrometer technical assistance, Melva Gonzalez for critical reading of the manuscript, and Sreedevi Chandrasekaran for help in preparation of the manuscript.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported by a National Institutes of Health Grant 5R24RR015371 (to D.I.W.). H.D.H.-M. was supported by National Institutes of Health Training Grant T32AI07633.
2 Current address: Laboratory of Viral Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD.
3 Address correspondence and reprint requests to Dr. William H. Hildebrand, Department of Microbiology and Immunology, University of Oklahoma Health Sciences Center, 975 N.E. 10th Street, Oklahoma City, OK, 73104. E-mail address: William-hildebrand@ouhsc.edu
4 Abbreviations used in this paper: Mamu, Macaca mulatta class I; sMamu, soluble Mamu.
Received for publication February 11, 2005. Accepted for publication April 13, 2005.
References
Townsend, A. R. M., F. M. Gotch, J. Davey. 1985. Cytotoxic T cells recognize fragments of the influenza nucleoprotein. Cell 42: 457-467.
Townsend, A. R. M., J. Bastin, K. Gould, G. G. Brownlee. 1986. Cytotoxic T lymphocytes recognize influenza haemagglutinin that lacks a signal sequence. Nature 324: 575-577.
Nixon, D. F., A. R. M. Townsend, J. G. Elvin, C. R. Rizza, J. Gallwey, A. J. McMichael. 1988. HIV-1 gag specific cytotoxic T lymphocytes defined with recombinant vaccinia virus and synthetic peptides. Nature 336: 484-487.
McMichael, A. J., B. D. Walker. 1994. Cytotoxic T lymphocyte epitopes: implications for HIV vaccines. AIDS 8: S155-S163.
Brodsky, F. M., L. E. Guagliardi. 1991. The cell biology of antigen processing and presentation. Annu. Rev. Immunol. 9: 707-744.
Brodsky, F. M., L. Lem, A. Solache, E. M. Bennett. 1999. Human pathogen subversion of antigen presentation. Immunol. Rev. 168: 199-215.
Nowak, M. A., R. M. May, R. E. Phillips, S. Rowland-Jones, D. G. Lalloo, S. McAdam, P. Klenerman, B. Koppe, K. Sigmund, C. R. Bangham, et al 1995. Antigenic oscillations and shifting immunodominance in HIV-1 infections. Nature 375: 606-611.
Collins, K. L., D. Baltimore. 1999. HIV’s evasion of the cellular immune response. Immunol. Rev. 168: 65-74.(Heather D. Hickman-Miller)