We have characterized an assay measuring CD8 T cell-mediated inhibition of human immunodeficiency virus (HIV) type 1 replication, demonstrating specificity and reproducibility and employing a panel of primary HIV-1 isolates. The assay uses relatively simple autologous cell culture and enzyme-linked immunosorbent assay, avoids generation of T cell clones, and can be performed with <2 million peripheral blood mononuclear cells. Efficient CD8 T cell-mediated cross-clade inhibition of HIV-1 replication in vitro was demonstrated in antiretroviral therapy-naive HIV-1-infected subjects with controlled viral replication in vivo but not in viremic subjects. An HIV-1 vaccine candidate, consisting of DNA and recombinant adenovirus 5 vectors tested in a phase I clinical trial, induced CD8 T cells that efficiently inhibited HIV-1 in a HLA-I-dependent manner. Assessment of direct antiviral T cell function by this assay provides additional information to guide vaccine design and the prioritizing of candidates for further clinical trials.
Because there is considerable evidence for T cell-mediated control of human immunodeficiency virus (HIV) type 1 replication in vivo, with maintenance of functional HIV-1-specific CD8 T cells associated with resolution of acute viremia and nonprogression to AIDS [1–11], efforts have focused on development of vaccine candidates designed to elicit HIV-1-specific T cells. However, until the efficacy of an HIV-1 vaccine candidate has been demonstrated in clinical trials, immune correlates of protection will remain speculative. Intracellular cytokine staining and interferon (IFN)-γ enzyme-linked immunospot (ELISPOT) assays are routinely employed to assess vaccine immunogenicity [12– 14]. However, ELISPOT responses do not correlate with in vivo virus control [10, 15–18] and did not predict the lack of efficacy in the Step Study trial of a recombinant adenovirus 5 (rAd5) candidate designed to induce HIV-1-specific T cells [12, 19]. Both ELISPOT and intracellular cytokine staining use high concentrations of exogenous peptides and may demonstrate T cells that recognize peptide-loaded but not HIV- or simian immunodeficiency virus (SIV)-infected cells expressing the same epitopes [15, 20, 21]. Therefore, T cells detected by these assays, although HIV-1 specific and theoretically antiviral, may have little effect on HIV-1 replication in vivo. An assay is needed that directly assesses the breadth of T cell-mediated antiviral activity against different HIV-1 isolates, correlates with in vivo virus control, and is akin to an antibody neutralizing assay, to enable better prediction of vaccine efficacy during clinical development.
An assay measuring CD8 T cell-mediated inhibition of HIV-1 replication has been described , and we have characterized it further. CD8 T cells from both HIV-1-infected subjects and those vaccinated in a phase I trial of a candidate DNA and rAd5 regimen, still untested in efficacy trials [23, 24], efficiently inhibited HIV-1 in vitro in an HLA-I-dependent manner. The assay may be used along with immunogenicity assays, such as ELISPOT assays, to determine the antiviral potential of vaccine elicited T cells, allowing selection of candidates under development for further testing.
Subjects. Chronically HIV-1-infected subjects (35 naive to antiretroviral therapy, 10 receiving highly active antiretroviral therapy [HAART]) were recruited from the Chelsea andWestminster Hospital and the Ragon Institute, with informed consent and local ethical approval (Table 1). HIV-1-uninfected subjects included blood transfusion volunteers and participants in a phase I trial of a DNA prime and rAd5 boost regimen, the International AIDS Vaccine Initiative V001 (http://clinicaltrials.gov/ct2/show/NCT00124007), conducted in Kenya and Rwanda. These vectors expressed HIV-1 Gag, Pol, and Nef from clade B and Env from clades A, B, and C, designed by the Vaccine Research Center, National Institutes of Allergy and Infectious Diseases (NIAID) [23, 24]. Seven vaccinees with HIV-1-specific IFN-γ ELISPOT peripheral blood mononuclear cell (PBMC) responses and a placebo recipient without a response were selected at 4 weeks after the third DNA vaccination and 6 and 12 weeks after rAd5 vaccination. PBMCs were isolated by density centrifugation, frozen in 10% dimethylsulphoxide/90% heat-inactivated fetal calf serum (Sigma), and stored in vaporphase liquid nitrogen.
Viruses. CXCR4-tropic clade B HIV-1IIIB was used as a common reference strain throughout all experiments. Other laboratory-adapted and primary isolates of various clades were used in a subset of experiments. HIV-1IIIB and HIV-1ELI (clade A/D) were cultured in H9 cells, and 50% tissue culture infectious doses (TCID50) were determined with C8166 cells. Chemokine (C-C motif) receptor 5 (CCR5)-tropic HIV-1Ba-L (clade B) was cultured in blood monocyte-derived macrophages. Primary isolates HIV-196ZM651 (clade C, CCR5-tropic), HIV-198IN022 (clade C, CCR5-tropic), HIV-197ZA012 (clade C, CCR5-tropic), HIV-198IN017 (clade C, CXCR4-tropic), and HIV-192BR021 (clade B, CCR5-tropic) were cultured in 3-day phytohemagglutinintreated PBMCs followed by interleukin (IL) 2 (Roche Diagnostics; 20 U/mL). The TCID50 for primary isolates was determined by using the reporter cell line TZM-bl .
HIV-1 isolates and cell lines used for culture and titration were obtained from the National Institutes of Health (NIH) AIDS Reference and Reagent Program. The TZM-bl cell line was obtained from John C. Kappes, Xiaoyun Wu, and Tranzyme; HIV-1ELI from Jean-Marie Bechet and Luc Montagnier; HIV-1IIIB and H9 from Robert Gallo; HIV-1Ba-L from Suzanne Gartner, Mikulas Popovic, and Robert Gallo; HIV-198IN022, HIV-197ZA012, and HIV-198IN017 from Robert Bollinger and the UNAIDS Network for HIV Isolation and Characterization; HIV-192BR021 from the UNAIDS Network for HIV Isolation and Characterization; and HIV-196ZM651 from Feng Gao and Beatrice Hahn. The C8166 cell line was provided by Robin Shattock, St George's Hospital Medical School, University of London.
Generation of CD4 target and CD8 effector T cells. PBMCs were resuspended at 1×106 cells/mL in Roswell Park Memorial Institute (RPMI) medium with 10% heat-inactivated fetal calf serum and 50 U of IL-2 and 0.5 µg/mL CD3/CD4 or CD3/CD8 bispecific antibodies (J.W., HarvardMedical School) for generation of CD8 or CD4 T cells, respectively [26, 27]. Culture volumes were doubled at days 3 and 6 by addition of fresh medium and IL-2. Cell numbers, purity, and HIV-1 replication were compared in 4- and 7-day cultures. CD4 T cells were infected with exogenous HIV-1 at a multiplicity of infection (MOI) of 0.01 for 3 h. A sample of each population was retained for flow cytometric determination of cell phenotypes.
Viral inhibition assay. Culture conditions were standardized across all inhibition experiments; 0.5×106 exogenously infected (MOI, 0.01) or uninfected 7-day antibody-expanded CD4 T cells were cultured with or without 0.5×106 autologous 7-day antibody-expanded CD8 T cells in 1 mL of medium with 50 U of IL-2 in 48-well plates. Half of the well supernatant was replaced with medium and IL-2 on days 3, 6, 8, and 10. Supernatant p24 content was measured on days 6 and 13 by enzyme-linked immunosorbent assay (ELISA) (PerkinElmer). CD8 T cell-mediated inhibition was expressed as the log10 reduction in p24 content of day 13 CD8 and CD4 T cell cocultures, compared with CD4 T cells alone. Cultures of uninfected CD4 T cells would reveal replication and inhibition of the subject's own endogenous virus. Such cultures were not required for HIV-1-uninfected subjects, including clinical trial subjects. For clinical trial subjects, antibody-expanded prevaccination CD4 T cells were used as common targets for HIV-1 infection in cocultures with pre- and postvaccination CD8 T cells. In some cases, CD4 and CD8 T cells were separated in 0.4-µm Transwell chambers (Corning), or major histocompatibility complex class I (MHC-I) blocking (W6/32) or isotype control antibodies were included (50 µg/mL; Serotec).
IFN-γ ELISPOT. V001 PBMC samples were tested with an IFN-γ ELISPOT assay, as described elsewhere , using overlapping 15-mer peptide pools provided by the Vaccine Research Center, matched to the HIV-1 clade B genes Gag, Pol, Env, and Nef inserted into the vaccine vectors. Bispecific antibody-expanded CD8 T cells were washed twice and rested for 24 h in RPMI with 20% heat-inactivated fetal calf serum before the ELISPOT assay.
Assay optimization and standardization. A 7-day culture using 0.5 µg/mL CD3/4 or CD3/8 bispecific antibody resulted in optimal expansion and purification of CD8 and CD4 T cells (Table 2). HIV-1IIIB replication was similar in cultures of phytohemagglutinin-stimulated PBMCs and 4- or 7-day cultures of antibody-expanded CD4 T cells. Increased HIV-1 MOIs (0.1 and 1) resulted in proportionately higher p24 release during the 3 days after infection, but similar plateau values were reached by day 10 (data not shown).
CD8 T cell-mediated HIV-1 inhibition assay specificity. To determine assay specificity and criteria for assay positivity, HIV-1IIIB inhibition mediated by CD8 T cells was assessed in 43 HIV-1-uninfected subjects. Typical data from 1 uninfected subject are displayed in Figure 1A. At day 13 of coculture, uninfected subjects had a median 0.70 log10 inhibition of p24 release from HIV-1IIIB-infected CD4 T cells, with a >99% confidence value of 1.13 log10. Inhibition detected above this value was subsequently defined as positive.
CD8 T cell-mediated inhibition of endogenous HIV-1 replication. Replication of the subject's own endogenous virus was detected by p24 release from CD4 T cells in 8 of 9 HAART-naive HIV-1-infected subjects with plasma viral loads (pVLs) of >10,000/mL (minimum day 13 p24 release, 92,000 pg/mL). CD8 T cell-mediated inhibition barely above the HIV-1IIIB positivity criterion was detected in only 3 of these subjects (1.2–1.5 log10). Endogenous HIV-1 replication was detected in only 7 of 26 HAART-naive subjects with pVLs of <10,000/mL, with 1 additional subject demonstrating minimal endogenous p24 release of 175 pg/mL. CD8 T cells inhibited endogenous virus replication in 5 of these 7 subjects, with inhibition above the 1.13 log10 positive value. With values of >1.13 log10 considered positive, there was no significant difference in the number of subjects scoring positive for endogenous HIV-1 inhibition between the low- and high-pVL groups (5 of 7 vs 3 of 8 subjects, respectively; P=.32, by Fisher's exact 2-tailed test). However, highly efficient inhibition of ⩾2 log10 was observed only in the low-pVL group (in 4 of 7 subjects vs 0 of 8 in the high-pVL group; P=.026).
CD8 T cell-mediated inhibition of exogenous HIV-1 replication. CD8 T cells from HAART-naive HIV-1-infected subjects with low pVLs (<10,000/mL) efficiently inhibited exogenous HIV-1IIIB replication (median, 3.17 log10). Similar inhibition was detected in subjects with pVLs of <2000 or 2000–10,000/mL (median, 3.17 and 3.20 log10, respectively). Minimal HIV-1IIIB inhibition (median, 1.12 log10) was detected in subjects with high pVLs (>10,000/mL). However, the ELISA would not distinguish between endogenous and exogenous p24; the true extent of HIV-1IIIB inhibition was therefore not determined in these cultures, because inhibition of this isolate may have been masked by endogenous virus replication. Compared with HIV-1-infected subjects with low pVL, significantly fewer (2 of 10) HAART-treated HIV-1-infected subjects (pVL, <50/mL) had HIV-1IIIB inhibition greater than the 1.13 log10 positive value (group median, 0.74 log10; P<.001, by Fisher's exact 2-tailed test). Typical data from 2 HAART-naive HIV-1-infected subjects, 1 with high and 1 with low pVL, are displayed in Figure 1B and 1C, with grouped data for HIV-1-uninfected and HAART-naive and HAART-treated HIV-1-infected groups in Figure 1D.
The ability of CD8 T cells to inhibit HIV-1 of different clades and coreceptor tropisms was assessed in 7 subjects who efficiently inhibited HIV-1IIIB (>2 log10). All 7 subjects also efficiently inhibited laboratory and primary isolates of clades A/D, B, and C with CXCR4 or CCR5 coreceptor tropisms (Table 3). Clade B CCR5-tropic HIV-1Ba-L was poorly inhibited by all subjects.
Interoperator variation was assessed by 2 operators, each thawing multiple identical PBMC vials on different days. HIV-1IIIB inhibition values did not differ significantly between operators, as indicated by overlapping 95% confidence limits (mean ± 1.96 standard errors). To assess the need for direct T cell contact, 3 HIV-1-infected subjects with >2 log10 HIV-1IIIB inhibition were tested in Transwell cultures. For HIV-1-infected subjects 1, 25, and 29 (Table 1), separation of CD4 and CD8 T cells greatly reduced inhibition, from 1.8, 4.2, and 2.4 log10 (median, 2.4 log10) to 0, 1.2, and 0.8 log10 (median, 0.8 log10), respectively. The same experiments in 6 uninfected subjects also reduced inhibition but to a lesser extent, from a median of 0.80 log10 (range, 0.4–1.1 log10) down to 0.3 log10 (range, 0–0.6 log10).
Application to prophylactic vaccine clinical trials. Samples from 7 vaccinees and 1 placebo recipient were selected from the phase 1 V001 trial. The placebo recipient and all 7 vaccinees before vaccination were classified as negative (<1.13 log10) for HIV-1IIIB inhibition (Figure 2). At 4 weeks after the third DNA vaccination, PBMCs were available from the placebo recipient and 5 of the 7 vaccinees and were negative for inhibition. At 6 and 12 weeks after rAd5 vaccination, CD8 T cells from all 7 vaccinees efficiently inhibited HIV-1IIIB (median, 2.22 and 1.48 log10, respectively) (Figure 2). At 6 weeks after rAd5 vaccination, all 7 vaccinees inhibited HIV-196ZM651 (median, 2.03 log10), and 4 of the 7 did so at 12 weeks (1.94–3.67 log10). At 6 and 12 weeks after rAd5 vaccination, only 3 vaccinees and 1 vaccinee, respectively, inhibited HIV-198IN017, at values between 1.21 and 1.89 log10 (median for the 6-week group, 0.93 log10). Sufficient antibody-expanded CD8 T cells for IFN-γ ELISPOT testing were available from 6 vaccinees and the placebo recipient 6 weeks after rAd5 vaccination. Responses were undetectable in the placebo recipient but detectable in all 6 vaccinees. There was no association between the overall magnitude of ELISPOT response and inhibition (Spearman correlation coefficient, r=0.29; P=.56) (Table 4). Different combinations of responses to HIV-1 genes were detected, with 1 subject targeting Gag, Pol, Env, and Nef, 2 targeting Gag, Pol and Env, 1 targeting Pol and Env, and 2 targeting either Gag or Env alone.
PBMC panel with blinding. The HIV Vaccine Trials Network, in collaboration with the Division of AIDS, NIAID, provided 9 PBMC samples and HIV-1NL4-3 as an evaluation of assays in use by laboratories to detect T cell-mediated inhibition of HIV-1. Participating laboratories were blinded to individual subject identities, but the panel was known to consist of 3 HIV-1-infected subjects, including a long-term nonprogressor (LTNP), and 6 uninfected subjects, including 2 vaccinees and 2 placebo recipients from a phase I trial (http://clinicaltrials.gov/ct2/show/NCT00270218) of the same vaccine candidate described above for trial V001. Efficient HIV-1NL4-3 inhibition was detected in samples 5, 6, and 9 (Figure 3) with endogenous HIV-1 in samples 1 and 2 and minimal inhibition typical of uninfected subjects in other samples. Although this was not required as part of the study, sample identities were predicted and communicated together with the data, using the criteria for assay positivity and characteristics of HIV controllers (no endogenous p24) and noncontrollers (endogenous p24). All sample predictions were shown to be correct after blinding was removed. The 3 samples demonstrating efficient inhibition were from both vaccinees and the LTNP. This efficient inhibition was reversed by MHC-I blockade, from a median of 2.73 to 0.37 log10. No such reversal in inhibition was detected in the placebo recipient and uninfected subjects. The isotype control had no effect (median, 3.04 log10). HIV-1IIIB inhibition was also reversed by MHC-I blockade, and blocking antibody had no effect on p24 release from HIV-1IIIB-infected CD4 T cells alone (data not shown).
IFN-γ ELISPOT and intracellular cytokine staining are reproducible assays of vaccine immunogenicity but are somewhat limited and did not predict failure of the Step Study vaccine candidate [12, 19]. We and others  believe that assays must be developed to assess cell-mediated inhibition of HIV-1. This study used an assay employing replication competent viruses, providing a physiologically relevant stimulus with good specificity and reproducibility. The procedure takes 3 weeks but is relatively simple, requiring cell culture and ELISA techniques and as few as 2 million PBMCs without knowledge of HLA type.
To define assay positivity criteria, CD8 T cell-mediated nonspecific HIV-1IIIB inhibition was assessed using 43 HIV-1-uninfected subjects, giving a >99% confidence interval of 1.13 log10, above which inhibition was considered HIV-1 specific (Figure 1D). Others have shown that CD8 T cells from uninfected subjects can mediate a 1-log10 inhibition of HIV-1 replication by noncytotoxic mechanisms [30–32].
The ability of CD8 T cells from HAART-naive HIV-1-infected subjects to inhibit HIV-1 replication was examined. Replication of the subject's own endogenous HIV-1 was undetectable in CD4 T cells from most subjects with pVLs<10,000/mL, but when detected it was efficiently inhibited by CD8 T cells. Endogenous HIV-1 replication was evident in 8 of 9 subjects with pVL>10,000/mL but was only minimally inhibited in 3 of these (Figure 1D). Although CD8 T cell-mediated inhibition was clearly lacking in viremic subjects, CD4 T cell function may also be deficient in such subjects, and it is unclear whether the lack of inhibition is due to CD4 or CD8 T cell functional deficiencies or both. Such a possibility would not apply to application of this assay to prophylactic vaccine clinical trials, which enroll only healthy subjects. The HIV-1 inhibitory ability of CD8 T cells from low-pVL subjects was clearly demonstrated by efficient HIV-1IIIB inhibition (1.9–4.7 log10). The minimal inhibition in HAART-treated HIV-1-infected subjects indicates that efficient inhibition in HAART-naive subjects with low pVL is not simply a consequence of low levels of virus and suggests that this inhibition contributes to control in vivo. Such findings are supported by a recent study demonstrating that proliferative and cytotoxic functions of HIV-specific CD8 T cells are not restored by antiretroviral therapy . CD8 T cells from high-pVL subjects mediated apparently poor HIV-1IIIB inhibition, but this inhibition may have been masked by endogenous p24 release. Therefore, the extent of HIV-1IIIB inhibition in these subjects was not determined in the present study.
HAART-naive subjects controlling HIV-1IIIB also controlled CCR5- and CXCR4-tropic clade B, C and A/D isolates (Table 3). The exception was B clade HIV-1BAL, with poor inhibition in all cases. The reasons for this are unclear, but culture in macrophages could have endowed this virus with a distinct biological phenotype. CD8 T cell-mediated cross-clade inhibition would highlight the potential benefit of inducing such cells by vaccination. Studies using T cell clones [34, 35] have demonstrated that minor epitope mutations may ablate CD8 T cell-mediated inhibition in vitro, but the present study used polyclonal cells more analogous to the in vivo situation and likely to be targeting multiple epitopes rather than peptide-selected clones. In addition, T cells from chronically infected subjects, as in the present study, target conserved HIV-1 epitopes [36, 37], offering an explanation for the efficient crossclade inhibition observed in the present study. Others suggest that candidate vaccines should target T cell responses toward conserved HIV-1 sequences [38, 39] which should induce inhibition of a broad spectrum of viruses. A key step in the development of this assay will be generation of a standard virus panel from populations where vaccines will be tested.
The assay was applied to samples from a phase I trial of a prophylactic DNA-rAd5 regimen untested in efficacy trials. HIV-1IIIB and 2 primary isolates of different coreceptor tropisms were chosen from the globally predominant C clade: HIV-1ZM96651 (CCR5) and HIV-198IN017 (CXCR4). Before vaccination and after DNA vaccination, all subjects demonstrated typical nonspecific HIV-1IIIB inhibition, as did the placebo recipient at all time points (Figure 2). At 6 and 12 weeks after rAd5, all vaccinees exhibited efficient HIV-1IIIB and HIV-196ZM651 inhibition. Only 3 vaccinees inhibited HIV-198IN017 to a lesser extent. This isolate was efficiently inhibited by HIV-1-infected subjects (Table 2), and the reasons for the differing inhibition are unclear. Published virus and vaccine protein sequences were compared using blast alignment software (http://www.ncbi.nlm.nih.gov/blast/bl2seq/wblast2.cgi). The vaccine clade B sequences were derived from Gag HXB2 (HIV-1IIIB), Pol HIV-1NL4-3, and Env clade A, B, and C strains HIV-192RW020, HXB2/HIV-1Ba-L, and HIV-197ZA012, respectively. HIV-1IIIB isolates Gag and Env matched the vaccine sequences and were 97% and 96% homologous to Pol and Nef. HIV-196ZM651 isolate matched the vaccine inserts by 83%, 90%, 81%, and 80% for Gag, Pol, Env, and Nef, respectively. HIV-198IN017 was the only clade C CXCR4-tropic isolate available in the laboratory, but published sequence data are unavailable. There was no correlation between viral inhibition and ELISPOT magnitude or breadth (HIV-1 Gag, Pol, Env and Nef genes targeted) (Table 4). Peptides and samples were not available for mapping ELISPOT responses. Although this study demonstrated that the regimen can induce CD8 T cells with HIV-1 inhibitory abilities, because vaccinee samples were selected for IFN-γ ELISPOT responses, the full inhibition-inducing profile of this and other regimens will require testing of broader sample cross-sections in future trials.
A blinded PBMC panel further demonstrated the potential of the assay, with efficient HIV-1NL4-3 inhibition detected in the LTNP and both vaccinees who received the same DNA-rAd5 regimen described above. The efficient inhibition was largely reversed by MHC-I blockade but not the nonspecific inhibition in HIV-1-uninfected subjects (Figure 3). Therefore, at least 2 inhibitory mechanisms are in operation: efficient antigen-specific inhibition by CD8 T cells recognizing MHC-I/peptides and comparatively less efficient nonspecific inhibition not involving MHC-I recognition. Separation of CD8 and CD4 T cells by Transwell chambers largely reversed but did not eliminate both specific and nonspecific inhibition, demonstrating that both mechanisms operate maximally when cells are in close proximity. The residual Transwell inhibition demonstrates that soluble factors contribute at least a minor component of the total. A recent study of LTNPs demonstrated that CD8 T cell-mediated elimination of infected CD4 T cells is associated with delivery of granzyme B and that, as in the present study, culture and cell activation are important, with maximal effects observed after 6 days of stimulation with HIV-1 peptides . We are working on the hypothesis that the HLA-dependent mechanism involves cytolysis and the nonspecific mechanism involves soluble factors.
Although it cannot be assumed that T cells inhibiting HIV-1 in vitro also inhibit it in vivo, low-pVL, HAART-naive chronically infected subjects with HIV-1 control in vivo have CD8 T cells that can efficiently inhibit endogenous and cross-clade exogenous HIV-1. In contrast, CD8 T cells from subjects with uncontrolled viremia did not inhibit endogenous virus in vitro. Such data provide additional support for the importance of CD8 T cells in controlling HIV-1 replication in vivo. This is the first report to describe vaccine-induced CD8 T cells with the ability to control HIV-1 in vitro. Other investigators have highlighted the requirement for additional functional assays of cellular immunity to be used in HIV-1 vaccine development [35, 41]. The relative simplicity of the technique, applicable to any donor and requiring minimal cell numbers, allows for application to clinical trials. We believe the viral inhibition assay will be a useful tool in the study of HIV-1 pathogenesis and vaccine development, complementing existing methods used to prioritize candidates for further trials.
We would like to thank the staff, volunteers, and patients at the St Stephens Trust HIV clinic. Assistance with the blinded PBMC panel was provided by Professor Julie McElrath, Holly Janes, Dr Natalie Zheng, Dr Patricia D'Souza, and Professor Otto Yang.