HIV-1 sequence diversity is considered a major hurdle for the design of an effective vaccine that could protect from heterologous viral infection . Centralized viral sequences, including consensus (corresponding to the most common character at each position in a nucleotide or protein alignment), ancestral (the most common recent ancestor, defined as the node at the base of an ingroup on a phylogenetic tree) and center-of-tree (the point of lowest average genetic distance to other sequences on a tree), have been proposed as potential immunogens owing to their overall reduced genetic distance from circulating viruses [2–5]. Among these, consensus sequences have been most widely used to assess T cell responses in HIV-infected individuals. However, neither the consensus nor the maximum likelihood-based sequences (such as ancestor and center-of-tree sequences) may have existed in nature, so their immunogenicity has been questioned in the past .
A number of recent studies have begun to address the potential implications of using different sequences both for in-vitro detection of responses and for in-vivo induction of T cell responses. In particular, consensus-, ancestral- and center-of-tree-based sequences representing different viral proteins and HIV clades have been tested for their respective immunogenicity in small animal models [7–11]. In addition, the functionality of the different proteins has been assessed (e.g., viral particle formation or human leucocyte antigen class I downregulation by centralized sequences encoding Nef), indicating that these centralized sequences encode protein sequences of intact functionality and generally good immunogenicity [7–12]. The observed interclade cross-reactivity of T cell responses elicited by these centralized sequences further supports their potential usefulness . Aside from animal models, centralized sequences have also been tested for their ability to detect HIV-1-specific T cell responses in infected humans. However, these analyses have been limited to relatively few individuals [14,15] and the use of single viral proteins .
It is not clear at present whether consensus, ancestral or center-of-tree sequences afford any advantage in terms of coverage of immunogenic epitopes in any population of HIV-1-infected individuals. To overcome some of these limitations and to improve definition of the relative in-vitro immunogenicity of centralized sequences, the present study took advantage of several well-established clade B-infected cohorts in the United States , Barbados  and Peru , as well as a clade C-infected cohort in South Africa , to compare T cell responses elicited by different clade B-, clade C- and group M-based centralized test sequences.
The study recruited 64 individuals infected with HIV-1 clade B and C from four cohorts, including 31 subjects from a largely non-Caucasian cohort in Boston , 11 from Barbados , 12 from Lima, Peru  and 10 from Durban, South Africa . Human subject protocols were approved by all participating hospitals and clinics, and all subjects provided written informed consent prior to enrollment.
Peptide sets were based on the described consensus clade B overlapping peptide set for the year 2001 (referred to as ConB′01), consisting of 410 adapted 18-mers overlapping by 10 residues each ( and unpublished data: http://www.HIV-1.lanl.gov/content/immunology/hlatem/study1/peptides.html). Peptides were synthesized for the consensus of clade B for 2002 (ConB′02), consensus C of 2001 (ConC), consensus for group M of 2004 (ConM), ancestral for clade B of 2002 (AncB) and group M of 2004 (AncM), center-of-tree for clade B of 2003 (COTB) and for group M of 2004 (COTM) that differed from consensus B′01 (ConB′01). Corresponding peptide regions had the same starting and ending position to ensure that the targeted epitopes were located at the same relative positions within the peptide throughout all test sets . Group M peptides were synthesized for Gag and Nef only. All peptide sets are shown in the supplementary Table 1 (available online). All peptides were synthesized at the MGH peptide synthesis facility using F-moc chemistry .
Assessment of T cell responses
Peripheral blood mononuclear cells (PBMC) were separated from whole blood and used in direct ex-vivo ELISpot assays to measure interferon-γ (IFN-γ)-producing cells as described , with an adjusted coating antibody concentration of 2 μg/ml. All assays were carried out on freshly isolated PBMC, including those from Peru, Barbados and South Africa. Thresholds for positive responses were set as at least 55 spot-forming cells per 106 input cells in the ELISpot assay, and responses exceeding the ‘mean of negative wells plus 3SD' or ‘four times the mean of negative wells’, whichever was highest. Assays with elevated background, defined as the ‘mean of negative wells > 30 spot-forming cells per 106’, were excluded from the analysis.
Phylogenetics and evolutionary analysis
Available gag sequences from the four geographic regions represented in this study were obtained from the Los Alamos HIV-1 database. GenBank accession numbers are as follows: 15 sequences from Boston are AF281682, AF281694, AF281706, AF281712, AF281733, AF281735, AF281760, AF281774, AF281788 and EF115501–EF115506; 16 sequences from Lima are EF680846–EF680861; 13 sequences from Barbados are EF102073–EF102085; and 16 clade C sequences from Durban, South Africa are AY463219, AY585265, AY772694, AY838565, AY878072, DQ011175, DQ056414, DQ093592, DQ093599, DQ164104, DQ275663, DQ351230, DQ369991, DQ369992, DQ369995 and DQ396397. The sequences were from different subjects to those sampled for T cell responses. All phylogenetic and evolutionary analyses were carried out on amino acids using the algorithm MUSCLE . To generate an estimate of the maximum likelihood phylogenetic tree, the algorithm was implemented in PHYML . All alignments were gap stripped so that ambiguously aligned positions were removed prior to tree estimation. Protein distances from each of the centralized sequences were obtained by pairwise methods using an HIV-1-specific amino acid model of evolution .
Statistical analyses for differences in ELISpots and protein distances were performed using GraphPad Prism 4 for Macintosh. All tests were two sided and nonparametric, including Friedman, Kruskal–Wallis, Wilcoxon matched pairs and Mann–Whitney U test. Bonferroni correction to account for multiple tests was applied to establish significance, uncorrected P values are shown in figures, and P values used as thresholds for significance given in the figure captions.
Response rates with consensus, ancestral and center-of-tree test sequences
Since consensus, ancestral and center-of-tree sequences are built on different principles, their ability to detect T cell responses in infected individuals may vary. To assess whether all of these sequences are equally suitable for vaccine design, 54 HIV clade B-infected individuals were tested for ELISpot responses against peptides representing the entire HIV proteome based on ConB′01, ConB′02, AncB or COTB (Cohort characteristics are summarized in Table 1).
Although the number of responses detected with any of the peptide sets varied considerably between individuals, none of the four peptide sets showed a significantly superior ability to detect more numerous (Fig. 1a) or stronger (Fig. 1b) responses. These data indicate that, despite different rationales supporting the design of these test sequences, all the sequences proved equally sensitive for the detection of HIV-1-specific IFN-γ-producing T cells in HIV-1-infected subjects, suggesting that they all provide equally powerful immunogens. Of note, no associations were found between breadth or magnitude of responses and markers of disease progression, age or gender of the studied subjects, in line with previous studies conducted in larger cohorts [17,24].
Influence of ethnicity or geographic provenance on frequency of recognition
The design of centralized sequences can be strongly influenced by the sequences that are available for their design and can in some instances change with the inclusion of relatively few additional sequences. Since most available clade B sequences in the Los Alamos HIV Sequence Database  are derived from subjects in the United States, Europe or Australia, centralized sequences may not reflect circulating sequences in other geographic regions of the world because of founder effects and locally independently evolving viral populations. Consequently, virus-specific T cell responses in subjects from these sites may not be detected as efficiently as responses in individuals from regions where viral populations are well represented in the centralized sequences. Therefore, the response rates to ConB′01, AncB and COTB in the 54 subjects were assessed stratified by their geographic origin. As shown in Fig. 2a, the number of peptides targeted did not differ significantly among the cohorts, even though no or very few sequences from Peru or Barbados had been included in the design of the peptide sets. As all sequences performed equally well in each of the studied cohorts, the data indicated that ancestral or center-of-tree sequences did not outperform consensus in populations that did not contribute substantial sequence data to the alignments used to build them.
Sequence divergence between Boston, Peru and Barbados and centralized gag sequences
The absence of significant differences in overall response rates to the various test sequences in individuals of different geographic provenance suggested that the phylogenetic distances between sequences from Boston, Peru and Barbados and the different centralized sequences may be comparable. An estimate of the maximum likelihood phylogenetic tree was, therefore, generated containing sequences from all three sites (Fig. 3). Since sequences were not available for subjects included in this study, GenBank gag sequences derived from Boston, Peru and Barbados were used for this analysis. The used sequences were obtained from the same cohorts from which patient samples were derived and temporally overlapped with time points from which immune data were derived, in an attempt to reduce potential confounding factors such as sequence evolution over the course of the epidemic. Protein distances from each of the centralized sequences to the independently derived sample sequences were determined in a pairwise fashion as described in the Methods. As suggested by the similar T cell response rates detected with the different peptide test sets, protein distances did not differ significantly between clade B sequences from Boston, Peru or Barbados and the tested centralized sequences (Fig. 2b).
Frequency of recognition of peptides based on group M sequences and clade B-specific sequences in clade B-infected subjects
A broadly applicable HIV-1 vaccine would ideally elicit responses that show cross-reactivity across different clades and may, therefore, be built based on group M sequences rather than clade-specific sequence information. To test the in-vitro immunogenicity of group M-based test sets, group M centralized sequences were generated and tested for their ability to elicit responses in HIV-1 clade B-infected individuals. As there were extensive differences between the centralized clade B and group M sequences, these analyses were restricted to Gag and Nef proteins, thereby including both a conserved and a more variable frequently targeted protein . To test for potential differences in response rates based on the geographic provenance of the study subjects, samples from the same three sites as above were included. Overall, group M-based consensus, ancestral and center-of-tree sequences did not differ in their ability to elicit in-vitro responses, similar to the finding using clade B-based centralized sequences; however, centralized group M Gag and Nef peptides were consistently less frequently recognized than their clade B counterparts (Fig. 4a–c). Group M-based sequences performed slightly better outside the Boston cohort, although this was not reflected by smaller protein distances of clade B sequences from regions where little sequence data had been collected (Fig. 4d–f).
Cross-clade and group M sequence recognition in clade C-infected individuals
To determine the degree of cross-reactivity of responses to clade B and C sequences, PBMC from 10 clade C-infected subjects from South Africa were tested against centralized clade B and clade C peptides covering all HIV-1 proteins. As shown in Fig. 5a, responses to consensus C peptides were detected significantly more frequently than to clade B-based consensus and center-of-tree peptides in the clade C-infected subjects. Although responses to clade C consensus peptides were also broader than responses to the clade B ancestral test set, this increase was not significant if a Bonferroni correction was applied (Fig. 5a; P = 0.049 uncorrected). The higher number of responses detected using AncB than other clade B sequences is likely explained by the more central position of the ancestral sequence on the phylogenetic tree (Fig. 3), placing it closer to clade C sequences. While all centralized clade B sequences were significantly more distant to clade C sequences than ConC (Fig. 5b; P = 0.0005 for all comparisons), AncB was significantly closer to the clade C sequences than ConB′01 or COTB (Fig. 5b; P = 0.0005 for both comparisons).
Since interclade peptides yielded significantly fewer responses than intraclade peptides, the question arose whether group M peptides would perform better than interclade peptides for the detection of virus-specific responses in individuals infected with viruses from different clades. The 10 clade C-infected subjects were, therefore, tested with ConC, ConB′01 and ConM Gag and Nef peptides. As shown in Fig. 5c,d, response rates were similar between ConC and ConM, whereas ConB′01 peptides performed significantly less well, suggesting that group M peptides represent a better alternative than interclade peptides. While this seems to confirm that response rates are dependent on the overall phylogenetic distance between infecting viral sequences and in-vitro test sets (Fig. 5e), the data also show that group M peptides performed relatively better in clade C-infected subjects than they did in clade B-infected individuals (Fig. 4a and Fig. 5c). This may be because the consensus C sequence used was an inferior match to actually circulating sequences in Durban since relatively few sequences from that area were available for the consensus design in 2001. Indeed, when comparing the distances of clade B sequences from all cohorts to ConB′01, and clade C sequences to ConC, clade B sequences were significantly closer to their respective consensus than clade C sequences (Fig. 5f; P = 0.0012). These data demonstrate the importance of considering local clustering of sequence polymorphisms, which ideally should be reflected in the design of in-vitro test reagents, and that group M-based test sequences may offer a reasonable alternative test set in the absence of sufficiently large sequence datasets. Yet, the data also show that even small increases in response rates were mirrored by significantly closer proximity of test reagents and locally circulating viral sequences, indicating the importance of developing adequate test sets for reliable detection of HIV-1-specific responses.
Increased response rates with combinations of sequences
Examination of whether combinations of sequences could significantly enhance the frequency of detection used only responses to Gag and Nef to allow comparisons with the group M sequences. Response rates to combined test sets were analyzed, including combinations of all clade B, all group M, or all B and M peptides.
In clade B-infected subjects, the observed increase in response rates detected by a combination of all clade B sequences compared with ConB′01 alone was consistently higher in all subjects tested (P < 0.0001), with some subjects doubling their number of positive responses using the combination of sequences (data not shown). Interestingly, a combination of all centralized group M sequences performed, on average, as well as ConB′01 alone, as equal or up to twice as broad responses were seen in 53% of subjects (P > 0.05). Combining all clade B and group M sequences in testing the clade B-infected subjects led to an increase of up to threefold in the breadth of response detected compared with using ConB′01 alone.
In clade C-infected subjects, combining all clade B sequences overcame some of the limitations of single interclade sequences, and the combination resulted in equal or better response rates than using the appropriate clade C consensus in half of the subjects. Combination of group M sequences improved the detection further, and combining all clade B and group M sequences led to a significant increase in response rates over ConC (P = 0.0020; data not shown). Together, the increased response rates suggest that existing single sequence peptide sets such as consensus-based peptides could be combined with broadly usable group M-based sequences to achieve significant improvement in the detection of responses in populations where limited sequence information on the circulating viral population would otherwise lead to the design of a suboptimal test reagent.
The sequence diversity of HIV-1 is considered a major block to building a broadly applicable HIV-1 vaccine , but despite the potentially crucial implications on vaccine efficacy, there is a profound lack of detailed comparisons of the in-vivo and in-vitro performance of possible immunogen sequences. Centralized, artificial sequences have been proposed to serve as immunogens since, on average, they halve the phylogenetic distance to naturally occurring viral strains [2,3]. However, their usefulness as immunogen sequences was initially controversial since they may not reflect naturally occurring, viable viral sequences. This issue seems largely resolved now, as consensus-, ancestral- and center-of-tree-based protein sequences have been shown to be functionally intact and immunogenic, at least in animal models [7–13]. However, their full ability to elicit recall T cell responses in humans or nonhuman primates had yet to be determined.
The comprehensive comparison of intraclade consensus, ancestral and center-of-tree sequences performed in the present study revealed that all of these sequences proved equally powerful for the detection of broadly directed and strong T cell responses in the ELISpot assays. While one or other sequence was superior in some of the 54 clade B-infected subjects tested, these differences evened out when evaluated in cross-section. In addition, individuals reacting strongly to one test sequence did also react with broad and strong responses to the other test sets, reflecting a wide overlap or cross-reactivity in the elicited responses. Although the detection of recall T cell responses is only a surrogate marker for the potential immunogenicity of these different sequences, recent data in the mouse model showed that immunization with different centralized sequences induced similar T cell response rates [9,11]. While the present ELISpot-based analyses do not allow distinction between CD4 and CD8 T cell-mediated responses, depletion studies and intracellular cytokine staining assays performed on samples from some of these chronically infected individuals have previously shown an absence of HIV-1-specific CD4 T cell activity, suggesting that the detected responses were largely CD8 T cell mediated . In addition, as the present analyses were based on IFN-γ ELISpot assays, it cannot be excluded that some responses escaped detection as they may not produce the full spectrum of effector functions . However, the phylogenetic associations explaining virtually all response rates, including subtle differences, suggest that missed IFN-γ-negative T cell responses were not a major determinant for the performance of the different test sequences.
Most of the available clade B sequences in the Los Alamos National Laboratory HIV-1 database  that informed the design of the centralized sequences used were derived from viruses from the United States, Europe or Australia. It was, therefore, questionable whether these peptide sets would perform equally well in other regions of the world, as founder effects and local host genetics may have led to viral populations less well covered by these test sequences. In addition, consensus sequences could, in theory, accumulate T cell epitope escape variants over time , potentially causing weaker responses than other test reagents. As shown by our data, geographic origin of the samples did not significantly influence the observed reactivities, suggesting that peptide sets based on large numbers of available sequences may be widely applicable throughout countries with predominantly clade B infections. Similarly, consensus sequences did not underperform compared with ancestral or center-of-tree sequences, indicating that the majority of circulating sequences maintain highly immunogenic epitope sequences. Interestingly, when testing clade C-infected subjects with clade B centralized sequences, the ancestral sequence detected slightly higher responses, suggesting that the position of the ancestor at the root of the phylogenetic tree relative to the clade B sequences, and thus with a diminished phylogenetic distance to the clade C sequences, provided a benefit in this setting. This observation can be generalized as even small, statistically insignificant, differences in the breadth of detected responses between sites and test sequences were generally reflected by differences in the phylogenetic proximity of the test sequence to the locally circulating viruses .
Group M-based Gag peptides have previously been shown to be effectively targeted in subjects infected with both clade B and clade C HIV-1 , suggesting that group M-based peptides could provide a valuable tool for comparison of response rates in subjects from regions where different clades are cocirculating. While, in our hands, responses to group M consensus did not score significantly lower in clade C-infected subjects than clade C peptides, group M peptides detected fewer responses in the clade B-infected subjects tested with clade B sequences. However, group M peptides could still prove equally applicable to compare responses across clades since the observed discrepancy between the relative performance in clade C- and B-infected individuals correlated with the higher distance of consensus C to clade C sequences compared with the distance of consensus B to clade B sequences.
Since the single intraclade centralized sequences performed equally well in detecting responses in several cohorts, the inability to detect more responses could have been the result of wide cross-reactivity between the sets. However, while cross-reactive T cell responses were frequently detected, some responses were lost or gained, depending on the test sequence used. This suggests that cross-reactivity was not complete, and that response rates could be optimized by combining sequences for increased sequence diversity coverage. Indeed, combinations of these centralized sequences significantly increased the response rates by up to threefold, suggesting that testing combined test sets could be a viable approach to improve the in-vitro detection of responses. However, as significant increases in response rates were obtained using these phylogenetically quite similar sequences, the data also suggest that additional sequence diversity coverage would further increase response rates. In this regard, a number of new approaches incorporating sequence diversity into new in-vitro test sets, as well as potential vaccine immunogens, are currently being evaluated and show promise to provide further improved tools in the future [30–33].
We would like to thank Hannah Hewitt, Rachel Allgaier, Laura Heath and Nancy Brown for excellent technical assistance, M. Tauheed Zaman and Eunice Pae for providing samples, and Ichih E. Wang for providing sequence data. We would also like to thank Philip Goulder, M. Anne St. John and Jorge Sanchez for their efforts in establishing the cohorts in South Africa, Barbados and Peru, respectively, from which samples were made available for these analyses; and Thomas Leitner for his assistance with the Peruvian sequence data. There are no conflicts of interest.
Sponsorship: This work has been funded in part with US federal funds from the National Institutes of Health under contracts N01-Al-15422, N01-Al-30024 on HLA typing and T cell epitope mapping relative to HIV-1 vaccine design, and R01-A1-067077. Additional funds came from the University of Washington Center for AIDS Research (P30-AI-27757 and P01-AI-57005) and an unrestricted gift from the Boeing Corporation to J.I.M.
Note: Nicole Frahm and David C. Nickle contributed equally to this work.
1. Garber DA, Silvestri G, Feinberg MB. Prospects for an AIDS vaccine: three big questions, no easy answers. Lancet Infect Dis 2004; 4:397–413.
2. Gaschen B, Taylor J, Yusim K, Foley B, Gao F, Lang D, et al
. Diversity considerations in HIV-1 vaccine selection. Science 2002; 296:2354–2360.
3. Nickle DC, Jensen MA, Gottlieb GS, Shriner D, Learn GH, Rodrigo AG, Mullins JI. Consensus and ancestral state HIV vaccines. Science 2003; 299:1515–1516.
4. Gao F, Korber BT, Weaver E, Liao HX, Hahn BH, Haynes BF. Centralized immunogens as a vaccine strategy to overcome HIV-1 diversity. Expert Rev Vaccines 2004; 3:S161–S168.
5. Mullins JI, Nickle DC, Heath L, Rodrigo AG, Learn GH. Immunogen sequence: the fourth tier of AIDS vaccine design. Expert Rev Vaccines 2004; 3:S151–S159.
6. Gao F, Bhattacharya T, Gaschen B, Taylor JH, Moore JP, Novitsky V, et al
. Author reply to: Consensus and ancestral state HIV vaccines by Nickle et al
. Science 2003; 299:1517–1518.
7. Gao F, Weaver EA, Lu Z, Li Y, Liao HX, Ma B, et al
. Antigenicity and immunogenicity of a synthetic human immunodeficiency virus type 1 group m consensus envelope glycoprotein. J Virol 2005; 79:1154–1163.
8. Doria-Rose NA, Learn GH, Rodrigo AG, Nickle DC, Li F, Mahalanabis M, et al
. Human immunodeficiency virus type 1 subtype B ancestral envelope protein is functional and elicits neutralizing antibodies in rabbits similar to those elicited by a circulating subtype B envelope. J Virol 2005; 79:11214–11224.
9. Kothe DL, Li Y, Decker JM, Bibollet-Ruche F, Zammit KP, Salazar MG, et al
. Ancestral and consensus envelope immunogens for HIV-1 subtype C. Virology 2006; 352:438–449.
10. Kothe DL, Decker JM, Li Y, Weng Z, Bibollet-Ruche F, Zammit KP, et al
. Antigenicity and immunogenicity of HIV-1 consensus subtype B envelope glycoproteins. Virology 2006; 360:218–234.
11. Rolland M, Jensen MA, Nickle DC, Yan J, Learn GH, Heath L, et al
. Reconstruction and function of ancestral center-of-tree (COT) HIV-1 proteins. J Virol 2007; 81:8507–8514.
12. Liao HX, Sutherland LL, Xia SM, Brock ME, Scearce RM, Vanleeuwen S, et al
. A group M consensus envelope glycoprotein induces antibodies that neutralize subsets of subtype B and C HIV-1 primary viruses. Virology 2006; 353:268–282.
13. Weaver EA, Lu Z, Camacho ZT, Moukdar F, Liao HX, Ma BJ, et al
. Cross-subtype T-cell immune responses induced by a human immunodeficiency virus type 1 group m consensus env immunogen. J Virol 2006; 80:6745–6756.
14. Rutebemberwa A, Currier JR, Jagodzinski L, McCutchan F, Birx D, Marovich M, Cox JH. HIV-1 MN Env 15-mer peptides better detect HIV-1 specific CD8 T cell responses compared with consensus subtypes B and M group 15-mer peptides. AIDS 2005; 19:1165–1172.
15. Sabado RL, Kilpatrick S, Ali A, Dagarag M, Ng HL, Cao H, Yang OO. Detection of HIV-1-specific CTL responses in clade B infection with clade C peptides and not clade B consensus peptides. J Immunol Meth 2005; 296:1–10.
16. Bansal A, Gough E, Ritter D, Wilson C, Mulenga J, Allen S, Goepfert PA. Group M-based HIV-1 Gag peptides are frequently targeted by T cells in chronically infected US and Zambian patients. AIDS 2006; 20:353–360.
17. Frahm N, Korber BT, Adams CM, Szinger JJ, Draenert R, Addo MM, et al
. Consistent cytotoxic-T-lymphocyte targeting of immunodominant regions in human immunodeficiency virus across multiple ethnicities. J Virol 2004; 78:2187–2200.
18. Zuniga R, Lucchetti A, Galvan P, Sanchez S, Sanchez C, Hernandez A, et al
. Relative dominance of Gag p24-specific cytotoxic T lymphocytes is associated with human immunodeficiency virus control. J Virol 2006; 80:3122–3125.
19. Kiepiela P, Leslie AJ, Honeyborne I, Ramduth D, Thobakgale C, Chetty S, et al
. Dominant influence of HLA-B in mediating the potential co-evolution of HIV and HLA. Nature 2004; 432:769–775.
20. Draenert R, Altfeld M, Brander C, Basgoz N, Corcoran C, Wurcel AG, et al
. Comparison of overlapping peptide sets for detection of antiviral CD8 and CD4 T cell responses. J Immunol Meth 2003; 275:19–29.
21. Edgar RC. MUSCLE: a multiple sequence alignment method with reduced time and space complexity. BMC Bioinformatics 2004; 5:113.
22. Guindon S, Gascuel O. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst Biol 2003; 52:696–704.
23. Nickle DC, Heath L, Jensen MA, Gilbert PB, Mullins JI, Kosakovsky Pond SL. HIV-specific probabilistic models of protein evolution. PLoS ONE 2007; 2:e503.
24. Addo MM, Yu XG, Rathod A, Cohen D, Eldridge RL, Strick D, et al
. Comprehensive epitope analysis of HIV-1-specific T cell responses directed against the entire expressed HIV-1 genome demonstrate broadly directed responses, but no correlation to viral load. J Virol 2003; 77:2081–2092.
25. Los Alamos National Laboratory. HIV Sequence Compendium 2005
. Los Alamos, NM: Los Alamos National Laboratory, Theoretical Biology and Biophysics Group; 2005. LA-UR 06–0680.
26. Brander C, Frahm N, Walker BD. The challenges of host and viral diversity in HIV vaccine design. Curr Opin Immunol 2006; 18:430–437.
27. Betts MR, Nason MC, West SM, De Rosa SC, Migueles SA, Abraham J, et al
. HIV nonprogressors preferentially maintain highly functional HIV-specific CD8+ T-cells. Blood 2006; 107:4781–4789.
28. Frahm N, Kiepiela P, Adams S, Linde CH, Hewitt HS, Sango K, et al
. Control of human immunodeficiency virus replication by cytotoxic T lymphocytes targeting subdominant epitopes. Nat Immunol 2006; 7:173–178.
29. Kesturu GS, Colleton BA, Liu Y, Heath L, Shaikh OS, Rinaldo CR Jr, Shankarappa R. Minimization of genetic distances by the consensus, ancestral, and center-of-tree (COT) sequences for HIV-1 variants within an infected individual and the design of reagents to test immune reactivity. Virology 2006; 348:437–448.
30. Frahm N, Kaufmann DE, Yusim K, Muldoon M, Kesmir C, Linde CH, et al
. Increased sequence diversity coverage improves detection of HIV-specific T cell responses. J Immunol 2007; 179:6638–6650.
31. Fischer W, Perkins S, Theiler J, Bhattacharya T, Yusim K, Funkhouser R, et al
. Polyvalent vaccines for optimal coverage of potential T-cell epitopes in global HIV-1 variants. Nat Med 2007; 13:100–106.
32. Nickle DC, Rolland M, Jensen MA, Pond SL, Deng W, Seligman M, et al
. Coping with viral diversity in HIV vaccine design. PLoS Comput Biol 2007; 3:e75.
33. Li F, Malhotra U, Gilbert PB, Hawkins NR, Duerr AC, McElrath JM, et al
. Peptide selection for human immunodeficiency virus type 1 CTL-based vaccine evaluation. Vaccine 2006; 24:6893–6904.