Of the estimated 50 million HIV infections that have occurred since the beginning of the pandemic, the vast majority have occurred in sub-Saharan Africa . All of the nine major subtypes and most of the 15 recognized circulating recombinant forms (CRF) of the HIV-1 M group have been isolated in sub-Saharan Africa . The identification of intersubtype genetic recombinants of HIV-1 occurring in regions where multiple subtypes are co-endemic means that the true extent of the genetic diversity of the virus is only now becoming clear . The magnitude and frequency of cross-subtype HIV-specific cytotoxic T lymphocyte (CTL) responses has been studied traditionally using chromium release assays based upon the ability of effector CTL from in-vitro stimulated peripheral blood mononuclear cell (PBMC) cultures to lyse peptide-pulsed or recombinant vaccinia-infected target cells. Using this approach, the presence of both broadly cross-subtype-reactive and subtype-specific CD8 T cells has been demonstrated in HIV-1-infected individuals [4–11]. An inherent drawback to this approach is that it measures only the cytolytic capacity of CD8 effector T cells, and is poorly quantitative because the in-vitro stimulation conditions used may skew the relative frequencies of different HIV-specific CD8 cells in the initial PBMC sample. Recently the ELISPOT assay and the flow cytometry-based intracellular cytokine (ICC) assay have facilitated the rapid identification of the function (upregulation of gene transcription) from both memory and effector CTL directly [12–17]. Within a heterogeneous population of cells, both assays permit a meaningful measure of the frequency of circulating HIV-specific CD8 T cells, as well as the relative degree of cross-recognition of different subtypes of HIV-1 by these cells.
In this report, data are presented that are relevant to vaccine trials that will be conducted in east Africa. The subtype-specificity of CD8 cellular immune responses in HIV-1-seropositive anonymous blood donors in Kenya have been assessed using the ICC assay for IFN-γ production. Employing a panel of recombinant vaccinia viruses (rVV) expressing gag, env and nef gene products from HIV-1 subtypes A–H, CD8 T cells specific for a particular HIV-1 gene product from any subtype were directly enumerated from PBMC samples. The overall frequency of responding CD8 T cells in peripheral blood, and the relative numbers of cells responding to different subtypes, was measured and the degree of cross-recognition was assessed. This unique cohort contains individuals infected with ‘pure’ (non-recombinant) subtype A, C and D HIV-1 viruses and with various unique inter-subtype recombinant viruses. Therefore, we have been able to investigate the comprehensive cross-subtype reactivity of CD8 T cells direct from the peripheral blood of individuals in a geographical region of multiple subtype endemicity.
HIV-1-positive blood units were collected between 1999 and 2000 from Kericho District Hospital (Kericho), Rift Valley Provincial Hospital (Nakuru), and Kenyatta National Hospital (Nairobi), all in southern Kenya, under a study approved by both Kenyan and US-based Institutional Review Boards. The names, personal information and medical conditions of all the subjects were not available and only alphanumeric characters identified blood units. HIV-1 positivity was assessed by Serostrip (Saliva Diagnostic Systems, Medford, NY, USA) and confirmed by enzyme-linked immunosorbent assay (Organon Teknika/BioMerieux, Inc., Marcy l'Etiole, France).
Recombinant vaccinia viruses
rVV expressing HIV-1 gene products gag, env and nef from subtypes A–H and CRF01_AE were obtained from the National Institutes of Health AIDS Research and Reference Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases, National Institutes of Health. The vaccinia recombinants used in the study outlining the clone of origin and the subtype and gene products each recombinant expresses have been listed in a previous study . The rVV vSC8 (NIH AIDS Research and Reference Reagent Program, cat. no. 357), which expresses β-galactosidase, was used as a negative control for background recombinant vaccinia-specific responses. To measure CD8 cell antigen-specific recall responses a rVV expressing the human cytomegalovirus (HCMV) matrix tegument phosphoprotein pp65 (a kind gift from S. Riddell, Fred Hutchinson Cancer Research Center, Seattle, WA, USA) was used. Quality control analysis for the efficiency of the Epstein–Barr virus (EBV)-transformed B-lymphoblastoid cell line (BLCL) infection was performed on six separate occasions using the entire panel of rVV, and revealed infection levels ranging from 36.9 to 94.6% of the BLCL (data not shown).
Recombinant vaccinia-based T cell stimulation assay
All assays were performed using RPMI-1640 medium containing 10% normal human serum, 100 U/ml penicillin and 100 μg/ml streptomycin (complete medium). The IFN-γ ICC assay was employed to measure the circulating frequency of HIV-specific CD8 T cells in the PBMC of infected individuals directly. Autologous BLCL were infected with 10 plaque-forming units per cell of each rVV and incubated for 14–16 h in complete medium at 37°C. Cryopreserved PBMC were recovered from liquid nitrogen, thawed rapidly, washed twice, and incubated at 5 × 106 cells/ml in complete medium for 14–16 h. After the overnight rest, the PBMC were re-counted before addition to the ICC assay plates at 0.5–1.0 × 106 cells per well. Ninety-six-well polypropylene tissue culture trays (no. 3790; Costar, Cambridge, MA, USA) were used for the assay co-incubation and subsequent staining procedure to facilitate efficient sample handling and a higher throughput. Vaccinia-infected BLCL were washed once after the overnight infection and distributed in duplicate wells at 1 × 105 cells per well (ratio of 5 : 1 to 10 : 1, PBMC : BLCL). As a positive control for the functional integrity of the cryopreserved cells staphylococcal enterotoxin-B was added to a single well at 5 μg/ml final concentration. Autologous BLCL infected with a rVV expressing the pp65 gene from HCMV also served as an additional positive control for functional integrity. The assays were incubated for 8 h at 37°C (5% carbon dioxide) in the presence of the protein transport inhibitor Brefeldin A (10 μg/ml; Sigma, St Louis, MO, USA), and were then stopped by incubation at 4°C overnight. Cells were stained for surface markers and intracellular IFN-γ expression the following day.
Intracellular IFN-γ detection and flow cytometry
Co-cultured PBMC (prepared above) were washed once with flow buffer (Dulbeccos's phosphate-buffered saline/0.1% bovine serum albumin/0.1% sodium azide) and incubated in the 96-well tissue culture tray wells for 10 min in 250 μl flow buffer at room temperature (same volume, temperature and base buffer used for all subsequent washings and incubations) containing 1 mM ethylenediamine tetraacetic acid. Cells were washed and fixed in 2% formaldehyde for 30 min and washed twice. Fixed cells were permeabilized with 0.5% saponin (Sigma) for 30 min, washed and resuspended in 0.5% saponin containing the following antibodies: fluorescein-isothiocyanate-conjugated anti-IFN-γ, peridin chlorophyll protein-conjugated anti-CD3 and activated protein C (APC)-conjugated anti-CD8 cells (BD Biosciences, San Jose, CA, USA). After 30 min incubation in the dark, cells were washed once with 0.5% saponin and twice with flow buffer and were finally re-suspended in 250 μl flow buffer. Cells were analysed within 24 h and data were acquired on a FACScalibur flow cytometer (Becton Dickinson, San Jose, CA, USA). Gating was performed using a total lymphocyte gate based on forward and side scatter characteristics and the acquisition of 50 000–200 000 cells within this gate. Color compensation was performed using similarly prepared cells from an HIV-1-seronegative donor and staining singly labeled cells with anti-CD3 labeled with fluorescein-isothiocyanate, peridin chlorophyll protein or APC fluorochromes (BD Biosciences). Data were analysed using FlowJo software version 3.6.1 (TreeStar, Cupertino, CA, USA).
Detection of high frequency broadly cross-subtype-reactive peripheral CD8 T cells
The frequency of cross-subtype-reactive CD8 T cells was determined for HIV-1 gag, env and nef genes in cryopreserved PBMC from subject NKU3006 (from the Rift Valley region). We and others have shown previously that rVV infection of autologous BLCL provides a consistent source of antigen-presenting cells for stimulating CD8 T cells in cryopreserved PBMC with high sensitivity and low background [18,19]. As the addition of rVV-infected BLCL to PBMC has proved to be an effective method for measuring IFN-γ ELISPOT responses from HIV-1-seropositive individuals, this co-culture method was used to measure IFN-γ production in an ICC assay format. Fig. 1 (a, b and c) shows the response of subject NKU3006 to the panel of rVV constructs, expressing HIV-1 gag, env and nef gene products from subtypes A, B, C, D, F, G and H, and CRF01_AE. PBMC were co-incubated for 8 h with autologous BLCL (ratio of 5 : 1 or 10 : 1 PBMC : BLCL), infected with the indicated rVV and subsequently analysed. Gating for the dot plots was based upon CD3+ FSChigh lymphocytes with each panel showing the percentage of cells that were CD8+/IFN-γ+. PBMC incubated with the β-galactosidase-expressing vSC8-infected BLCL showed a background of 0.38% IFN-γ+ CD8 T cells [bottom right panel of Fig. 1 (a, b and c)]. This background can most likely be attributed to a combination of EBV and vaccinia virus-specific CD8 T cells present in the circulation in this individual, because EBV and vaccinia virus antigens will also be expressed by the vSC8-infected autologous BLCL. High numbers of gag-specific CD8 T cells were detected in response to each of the eight gag constructs (Fig. 1a). HIV-1 gag-specific cells ranged from 2.41 to 11.35% of the CD8 T cell population (2.03–10.97% after subtraction of the vSC8 background). A high number of specific CD8 T cells were detected in response to each of the eight env constructs from subject NKU3006 (Fig. 1b). The range of HIV-1 env-specific cells was from 1.56 to 2.29% of the CD8 T cell population (1.18–1.91% after subtraction of the vSC8 background). Therefore subject NKU3006 has displayed a comprehensive cross-subtype CD8 cell response to both gag and env gene products. The response to the nef gene expressing constructs revealed the first example of subtype-specific CD8 cell responses for subject NKU3006. A positive CD8 cell response (defined in the section below) was detected against only four of the eight constructs (subtypes B, D, F and G all exceeding 0.43%). No significant responses were detected for nef subtypes A, C and H or CRF01_AE.
Population-based study of the frequency of HIV-1 cross-subtype-reactive and subtype-specific peripheral CD8 T cells
A further 20 subjects from the blood bank cohort were investigated for cross-subtype CD8 cell responses using the assay system described above. For all subjects, positive control responses were detected to SEB (range 1.58–21.68%, IFN-γ-positive CD8 T cells) confirming the viability and functional capacity of the cryopreserved cells. An additional positive control using a rVV expressing the pp65 gene of HCMV was included in the study. Fifteen out of 20 individuals responded to this construct (range 0.50–2.07%, IFN-γ-positive CD8 T cells). We found no relationship between the magnitude of response to the positive controls and the magnitude observed to the HIV constructs. The response of the 21 subjects to vSC8-infected BLCL was used to define a cut-off for positive antigen-specific responses as any value greater than the 99% confidence interval. All data were subsequently expressed as IFN-γ-positive CD8 T cells per 1000 total CD8 T cells acquired by flow cytometry. Therefore, for this data set any value greater than 4.3 (i.e. 4.3 IFN-γ positive CD8 T cells per 1000 total CD8 T cells) was considered positive (mean 2.9; range 1.8–3.8). The response of all subjects to the eight gag gene constructs is shown in Fig. 2a. High numbers of gag-specific CD8 T cells were detectable in response to all eight gag constructs. As outlined in Fig. 2a, at least eight (subtype H, 38.1%) and as many as 15 (subtype C, 71.4%) subjects responded to each of the constructs. The mean of the positive responders against each construct (only those that exceeded the 4.3 IFN-γ-positive CD8 T cells per 1000 total CD8 T cells cut-off value) was calculated for each subtype (range 14.7–22.7). Among the eight gag constructs, the mean positive CD8 T cell response was extremely consistent. Fig. 2 (b and c) shows a similar analysis for the env and nef response data. The env and nef constructs were both recognized by a similar number of responders and only marginally less often than gag. There was, however, a lower magnitude of the response to the env constructs, in particular when compared with gag (range 8.3–12.4), with all mean env responses being less than the lowest mean gag response. The mean nef construct response was also lower in general than the gag responses (range 7.5–14.5) with only the one nef construct (subtype H, mean 14.5) exceeding the lowest gag construct (subtype G, mean 14.4).
Table 1 shows a breakdown of the data in terms of the number of subtypes any given individual responded to (i.e. how many subtypes of a gene do CD8 T cells from a given subject recognize?). For the gag gene, eight subjects responded to at least seven subtypes, indicating that an effective cross-subtype CD8 cell response to gag is possible. However, six subjects responded to one or none of the gag constructs used here, and seven subjects showed a mixed response to three or four subtypes with a random pattern of recognition. Similarly, effective cross-subtype reactivity was also demonstrated for both env gene products (eight subjects recognized at least six subtypes) and for nef gene products (five subjects recognized all eight subtypes). Importantly, 12 of the 21 subjects responded to at least six of the subtypes from gag, env or nef. It should also be noted that two subjects recognized all 24 constructs, whereas three subjects recognized none of the 24 constructs. The three subjects who did not respond to any of the HIV-1 constructs all responded to the positive control (SEB). Totaling the data for the gag, env and nef responses to any given subtype (Table 1), it can be seen that there is little difference in the number of responders to any particular subtype. At least 13 subjects (subtypes D, G and H) and at most 16 subjects (subtype C) responded to at least one gene that is representative of a subtype. The presence of many unique inter-subtype recombinant viruses in this cohort suggests a complex history of HIV-1 infection in these individuals, complicating the interpretation.
Full-length HIV-1 genome sequence data had been generated in a previous study from 10 of the subjects examined here . Among these subjects, a pure A subtype sequence was isolated from five, a pure subtype C and a pure subtype D each from one, and a unique recombinant sequence was isolated from each of the remaining three subjects. One subject, KNH1209, demonstrated no response to any HIV-1 gene product in the ICC screening assay. For the remaining nine individuals, the total CD8 T cell response to the three HIV-1 gene products against the homologous (infecting) subtype was compared with the mean of the total response to the heterologous subtypes. As shown in Fig. 3, in seven out of nine subjects the response to the homologous subtype exceeds the mean response to the heterologous subtypes. The level of cross-reactivity between the subtypes was remarkable in that in all cases the mean response to the heterologous subtypes was at least half the magnitude of the response to the homologous subtype.
In the present study, high numbers of HIV-1-specific CD8 T cells with broad cross-subtype reactivity were detected in the peripheral blood of individuals infected with different subtypes and unique inter-subtype recombinant viruses. By using rVV expressing gag, env and nef gene products from HIV-1 subtypes A–H and CRF01_AE, CD8 T cells specific for a wide array of epitopes within a particular HIV-1 gene product from any subtype can be directly enumerated from a PBMC sample. To increase the sensitivity of the IFN-γ ICC assay, autologous BLCL were infected with rVV and then added as supplemental APC to the PBMC. The magnitude of the CD8 cell response between heterologous subtypes can be directly compared and the degree of cross-recognition assessed. As many as 10.97% of the circulating CD8 T cells were shown to be specific for a single HIV-1 gene product. Broadly cross-subtype-reactive CD8 cell responses were detected against all three gene products studied. As amino acid level epitope mapping was not performed in this study, it was not possible to determine how many, or which epitopes were being recognized within each of the constructs. The high frequency of peripheral HIV-1-specific CD8 T cells detected is in agreement with the findings of a recent study of HIV-infected progressors and non-progressors . The present study extends upon these findings by providing a comprehensive screening of the cross-subtype recognition and shows that in many cases the same gene product from as many as eight subtypes is recognized by a high frequency of CD8 T cells. Importantly, a high frequency of cross-subtype-reactive CD8 cells could be detected in individuals infected with a variety of non-subtype B viruses and unique inter-subtype recombinant viruses.
The number of HIV-1-specific T cells measured here are five to 100 times greater than reported previously in similar cohorts [12,15,16]. Using direct vaccinia addition to an ELISPOT assay, Cao et al.  and Larsson et al.  reported frequencies of 100–2000 IFN-γ-producing CD8 cells per million PBMC of individuals infected with subtype A or D and subtype B, respectively. Similar frequencies of IFN-γ−producing CD8 cells have been documented by screening subtype A, C, or D-infected individuals with overlapping and optimal length synthetic peptides [14–17]. The percentage of HIV-1-specific CD8 cells reported here corresponds to 2000–50 000 IFN-γ-producing CD8 cells per million PBMC. The higher frequency of cells detected in this study can be attributed to several factors. As we and others have shown [18,19], the use of vaccinia infected-BLCL increases the sensitivity of CD8 T cell response detection from cryopreserved PBMC. In addition, ELISPOT assays may underestimate the number of dead or dying cells in the PBMC fraction before assay set-up, which can be readily gated out in the analysis stage of flow cytometry-based assays. As the global response to a given HIV-1 gene product in the context of all MHC alleles from a patient has been measured, much higher frequencies of antigen-specific CD8 T cells may be detected than when screening with individual synthetic peptides.
The results of this study suggest that in HIV-1-seropositive blood donors in Kenya, CD8 cell responses are not only directed towards the homologous subtype, but are also directed towards many other heterologous subtypes. It is important to note that when the response of the population as a whole was assessed, no apparent subtype specificity was evident (Fig. 2). However, when the subtype specificity of the CD8 cell response was compared with the full-length sequence of infecting virus, subtype-specificity was apparent (Fig. 3). There was also no significant bias of the response towards any given subtype, despite the fact that in this cohort subtypes A, C and D represent 50, 20 and 30% of the total genetic makeup of HIV-1. These data may have significant implications for vaccine development, because the question of whether vaccine formulation should target the HIV-1 subtype circulating in a given geographical region is currently of considerable importance. In geographical regions where multiple subtypes circulate at a high frequency, the requirement for vaccine formulations with multiple subtype reactivity is paramount. Although considerable cross-subtype reactivity is detectable in this cohort, there is also an equal amount of subtype focussing on the CD8 cell responses. Many subtype-specific as well as non-responses were seen, despite using full-length HIV-1 gene products capable of being processed and presented in the context of any MHC allele. These results indicate that the use of a single subtype-based vaccine in a multi-subtype endemic region may not generate sufficient cross-subtype responsiveness, even when it matches one of the circulating subtypes. The importance of the fine specificity of the CD8 cell response to lentiviral infection has been highlighted in several recent reports. In an SIV model of HIV-1 infection, a single point mutation within a CD8 cell epitope was shown to be sufficient for viral escape from vaccine-induced protection . In humans, point mutation has also been associated with viral escape from CTL, control of infection, and the subsequent stable transmission of virus [24,25]. To contend with viral diversity, it has been postulated recently that use of consensus or putative ancestor sequences could be used in vaccine design to minimize genetic differences [26,27]. However, even this approach has the drawback of using a single sequence, and cannot represent all subtypes at any given CD8 cell epitope site. The use of vaccine modalities that maximize the breadth of the CD8 cell response will be critical to preventing point mutation from permitting viral escape from CD8 cell control. Considering the results presented here, a ‘cocktail’ approach to vaccine formulation involving a pool of constructs representing different subtypes may be an effective approach to generating sufficient cross-subtype CD8 cell responses in multi-subtype endemic locations.
Sponsorship: Financial support for this study was provided by Department of Defense Collaborative agreement DAMD17-98-2-8007. The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Departments of Army or Defense.
3. McCutchan FE. Understanding the genetic diversity of HIV-1. AIDS
4. Betts MR, Krowka J, Santamaria C, Balsamo K, Gao F, Mulundu G, et al. Cross-clade human immunodeficiency virus (HIV)-specific cytotoxic T-lymphocyte responses in HIV-infected Zambians. J Virol
5. Cao H, Kanki P, Sankale JL, Dieng-Sarr A, Mazzara GP, Kalams SA, et al. Cytotoxic T-lymphocyte cross-reactivity among different human immunodeficiency virus type 1 clades: implications for vaccine development. J Virol
6. Ferrari G, Humphrey W, McElrath MJ, Excler JL, Duliege AM, Clements ML, et al. Clade B-based HIV-1 vaccines elicit cross-clade cytotoxic T lymphocyte reactivities in uninfected volunteers. Proc Natl Acad Sci U S A
7. Lynch JA, deSouza M, Robb MD, Markowitz L, Nitayaphan S, Sapan CV, et al. Cross-clade cytotoxic T cell response to human immunodeficiency virus type 1 proteins among HLA disparate North Americans and Thais. J Infect Dis
8. McAdam S, Kaleebu P, Krausa P, Goulder P, French N, Collin B, et al. Cross-clade recognition of p55 by cytotoxic T lymphocytes in HIV-1 infection. AIDS
9. Sipsas NV, Kalams SA, Trocha A, He S, Blattner WA, Walker BD, et al. Identification of type-specific cytotoxic T lymphocyte responses to homologous viral proteins in laboratory workers accidentally infected with HIV-1. J Clin Invest
10. Wilson CC, Kalams SA, Wilkes BM, Ruhl DJ, Gao F, Hahn BH, et al. Overlapping epitopes in human immunodeficiency virus type 1 gp120 presented by HLA A, B, and C molecules: effects of viral variation on cytotoxic T-lymphocyte recognition. J Virol
11. Wilson SE, Pedersen SL, Kunich JC, Wilkins VL, Mann DL, Mazzara GP, et al. Cross-clade envelope glycoprotein 160-specific CD8+ cytotoxic T lymphocyte responses in early HIV type 1 clade B infection. AIDS Res Hum Retroviruses
12. Cao H, Mani I, Vincent R, Mugerwa R, Mugyenyi P, Kanki P, et al. Cellular immunity to human immunodeficiency virus type 1 (HIV-1) clades: relevance to HIV-1 vaccine trials in Uganda. J Infect Dis
13. Altfeld MA, Livingston B, Reshamwala N, Nguyen PT, Addo MM, Shea A, et al. Identification of novel HLA-A2-restricted human immunodeficiency virus type 1-specific cytotoxic T-lymphocyte epitopes predicted by the HLA-A2 supertype peptide-binding motif. J Virol
14. Altfeld MA, Trocha A, Eldridge RL, Rosenberg ES, Phillips MN, Addo MM, et al. Identification of dominant optimal HLA-B60- and HLA-B61-restricted cytotoxic T-lymphocyte (CTL) epitopes: rapid characterization of CTL responses by enzyme-linked immunospot assay. J Virol
15. Novitsky V, Rybak N, McLane MF, Gilbert P, Chigwedere P, Klein I, et al. Identification of human immunodeficiency virus type 1 subtype C Gag-, Tat-, Rev-, and Nef-specific elispot-based cytotoxic T-lymphocyte responses for AIDS vaccine design. J Virol
16. Goulder PJ, Brander C, Annamalai K, Mngqundaniso N, Govender U, Tang Y, et al. Differential narrow focusing of immunodominant human immunodeficiency virus gag -specific cytotoxic T-lymphocyte responses in infected African and caucasoid adults and children. J Virol
17. Goulder PJ, Addo MM, Altfeld MA, Rosenberg ES, Tang Y, Govender U, et al. Rapid definition of five novel HLA-A*3002-restricted human immunodeficiency virus-specific cytotoxic T-lymphocyte epitopes by elispot and intracellular cytokine staining assays. J Virol
18. Currier JR, deSouza M, Chanbancherd P, Bernstein W, Birx DL, Cox JH. Comprehensive screening for human immunodeficiency virus type 1 subtype-specific CD8 cytotoxic T lymphocytes and definition of degenerate epitopes restricted by HLA-A0207 and -C(W)0304 alleles. J Virol
19. Huang XL, Fan Z, Kalinyak C, Mellors JW, Rinaldo CR Jr. CD8(+) T-cell gamma interferon production specific for human immunodeficiency virus type 1 (HIV-1) in HIV-1-infected subjects. Clin Diagn Lab Immunol
20. Dowling WE, Kim B, Mason CJ, Wasunna KM, Alam U, Elson L, et al. Forty-one near full-length HIV-1 sequences from Kenya reveal an epidemic of subtype A and A-containing recombinants. AIDS
21. Gea-Banacloche JC, Migueles SA, Martino L, Shupert WL, McNeil AC, Sabbaghian MS, et al. Maintenance of large numbers of virus-specific CD8+ T cells in HIV-infected progressors and long-term nonprogressors. J Immunol
22. Larsson M, Jin X, Ramratnam B, Ogg GS, Engelmayer J, Demoitie MA, et al. A recombinant vaccinia virus based ELISPOT assay detects high frequencies of Pol-specific CD8 T cells in HIV-1-positive individuals. AIDS
23. Barouch DH, Kunstman J, Kuroda MJ, Schmitz JE, Santra S, Peyerl FW, et al. Eventual AIDS vaccine failure in a rhesus monkey by viral escape from cytotoxic T lymphocytes. Nature
24. Goulder PJ, Phillips RE, Colbert RA, McAdam S, Ogg G, Nowak MA, et al. Late escape from an immunodominant cytotoxic T-lymphocyte response associated with progression to AIDS. Nat Med
25. Goulder PJ, Brander C, Tang Y, Tremblay C, Colbert RA, Addo MM, et al. Evolution and transmission of stable CTL escape mutations in HIV infection. Nature
26. Gaschen B, Taylor J, Yusim K, Foley B, Gao F, Lang D, et al. Diversity considerations in HIV-1 vaccine selection. Science
27. Novitsky V, Smith UR, Gilbert P, McLane MF, Chigwedere P, Williamson C, et al. Human immunodeficiency virus type 1 subtype C molecular phylogeny: consensus sequence for an AIDS vaccine design? J Virol