Quantitative and Qualitative Differences in the T Cell Response to HIV in Uninfected Ugandans Exposed or Unexposed to HIV-Infected Partners

Abstract

HIV-exposed and yet persistently uninfected individuals have been an intriguing, repeated observation in multiple studies, but uncertainty persists on the significance and implications of this in devising protective strategies against HIV. We carried out a cross-sectional analysis of exposed uninfected partners in a Ugandan cohort of heterosexual serodiscordant couples (37.5% antiretroviral therapy naive) comparing their T cell responses to HIV peptides with those of unexposed uninfected individuals. We used an objective definition of exposure and inclusion criteria, blinded ex vivo and cultured gamma interferon (IFN-γ) enzyme-linked immunospot assays, and multiparameter flow cytometry and intracellular cytokine staining to investigate the features of the HIV-specific response in exposed versus unexposed uninfected individuals. A response rate to HIV was detectable in unexposed uninfected (5.7%, 95% confidence interval [CI] = 3.3 to 8.1%) and, at a significantly higher level (12.5%, 95% CI = 9.7 to 15.4%, P = 0.0004), in exposed uninfected individuals. The response rate to Gag was significantly higher in exposed uninfected (10/50 [20.%]) compared to unexposed uninfected (1/35 [2.9%]) individuals (P = 0.0004). The magnitude of responses was also greater in exposed uninfected individuals but not statistically significant. The average number of peptide pools recognized was significantly higher in exposed uninfected subjects than in unexposed uninfected subjects (1.21 versus 0.47; P = 0.0106). The proportion of multifunctional responses was different in the two groups, with a higher proportion of single cytokine responses, mostly IFN-γ, in unexposed uninfected individuals compared to exposed uninfected individuals. Our findings demonstrate both quantitative and qualitative differences in T cell reactivity to HIV between HESN (HIV exposed seronegative) and HUSN (HIV unexposed seronegative) subject groups but do not discriminate as to whether they represent markers of exposure or of protection against HIV infection.

INTRODUCTION

Since the identification of HIV-1 as the cause of AIDS, there has been a remarkable accumulation of detailed knowledge on the interaction of this virus with the infected host. We know much less about abortive and failed infections, and yet such knowledge might be very useful for preventive approaches. It is relatively straightforward to develop vaccines against most infectious agents that induce natural immunity in at least a subset of infected individuals, but, unfortunately, natural HIV-specific immunity has been difficult to demonstrate, the main protective factors identified being genetic host factors such as CCR5 coreceptor deletion (1). There have been many reports of individuals that remain uninfected despite high-risk exposure. Since the sexual transmission of HIV is not very efficient (2–5) and varies according to the route of infection (6, 7), viral load (8), and subtype (9), the mere demonstration of exposure to HIV hardly implies that a persistently seronegative individual has protective immunity. Also, successful HIV-1 infection rapidly generates escape mutants circumventing and outpacing the relatively slower development of immune responses (10). This feature, conjointly with the transmission bottleneck (11, 12), results in a sharp demarcation between infection and noninfection, so that no instances of human HIV-1 infections controlled and cleared by natural immunity have been documented.

Evaluating transmission that did not occur is a major challenge: it is difficult to quantitate exposure from reported sexual behavior, and the low infectivity of HIV-1 requires large cohorts with long follow-up times. High-risk populations (e.g., sex workers) (13), discordant couples (14, 15), intravenous drug users (16), men who have sex with men (17), or populations with definite exposure, such as blood transfusion recipients that received contaminated blood (18) or babies born to HIV-1-infected mothers, have consequently been the focus of study (19). Although the exposed uninfected phenotype has been reproduced in macaques that remained persistently uninfected after multiple intrarectal inoculations with SIV, it did not correlate with the induction of adaptive responses (20). In uninfected humans, there have been reports of T cell responses to HIV-1, but there is still uncertainty on their magnitude and persistence and whether they have a protective role or merely indicate exposure that has not resulted in acquisition of HIV by other protective mechanism or by chance.

Studies of HIV-exposed seronegative individuals (HESN) have highlighted the importance of robust criteria for exposure to HIV and a very good definition of the study population. In the present study, we wanted to answer the question of whether the induction of T cell-mediated immune responses to HIV occur in exposed uninfected individuals with a well-defined heterosexual exposure to HIV compared to a companion study (21) involving predominantly homosexual exposure. To this end, we enrolled Ugandan HIV serodiscordant couples who declared ≥25 episodes of unprotected heterosexual sexual intercourse in the previous year. Appropriate controls were also essential, particularly in view of the high sensitivity of the main T cell response readout: the cultured gamma interferon (IFN-γ) enzyme-linked immunospot (ELISPOT) assay. We enrolled HIV-1-seronegative monogamous couples (either one or both partners) who were also negative for sexually transmitted diseases (STIs) as HIV-unexposed seronegative controls (HUSN). A detailed sexual behavior questionnaire with multiple cross checks, including interpartner corroboration and laboratory determinations of the viral load in the infected partners, were used to evaluate previous and ongoing exposure during the study (22). We used ex vivo and cultured IFN-γ ELISPOT assays to evaluate the response rates, magnitude of response, and breadth of recognition of HIV A and D consensus peptides, which represent the majority of circulating subtypes in Uganda. We also evaluated functional IFN-γ, macrophage inflammatory protein 1β (MIP-1β), tumor necrosis factor alpha (TNF-α), and interleukin-2 (IL-2) responses by multiparameter flow cytometry in a subset of HESN and HUSN. We detected reactivity to HIV epitopes in both HESN and HUSN, though reactivity was broader and occurred at a significantly higher rate in HESN. The proportion of multifunctional responses was also significantly different between the two groups.

MATERIALS AND METHODS

Human subjects.

We enrolled 72 serodiscordant couples and 55 unexposed uninfected participants (26 couples, 3 individuals). Not all participants enrolled could be tested due to limited peripheral blood mononuclear cells (PBMC) sample availability. Individuals included in this analysis were HESN sexual partners of HIV-infected individuals (n = 57) and HUSN members of couples enrolled together or as individuals (n = 46) at the AIDS Information Centre, Kampala, Uganda.

The inclusion criteria for HESN participants were as follows: (i) HIV-1-negative status as verified by antibody enzyme immunoassay (EIA) or nucleic acid testing (NAT)/p24 antigen studies; and (ii) the occurrence of ≥25 unprotected vaginal, anal, or oral sexual acts within the previous year, starting no later than 10 months before enrollment, with an HIV-infected partner diagnosed ≥1 year before screening and with a CD4 count of ≤450/mm³. The HUSN and their sexual partners were negative for HIV-1 antibody and virus (EIA, NAT, or p24 antigen), STI -free, and in a monogamous relationship for ≥1 year. The demographic statistics of the participants are summarized in Table 1. Participants were followed up for 9 months, with visits every 3 months. T cell responses were evaluated cross sectionally at the second visit. HIV status was evaluated at each visit. Written informed consent was provided. The study had ethical and scientific clearance from the Uganda Virus Research Institute Science and Ethics Committee and Uganda National Council for Science and Technology and was regularly monitored by the Division of Acquired Immunodeficiency Syndrome (through Pharmaceutical Product Development Limited Liability Company and Clinical Laboratory Services audits) for adherence to good clinical laboratory practice.

Table 1.

Summary of participant characteristics at enrollment

Characteristic	Median (IQR)a
	HESN (n = 57)	HUSN (n = 46)	HIV+ (n = 57)
Age (yr)	30 (27–40)	32 (26–37)	34 (28–40)
No. of males (%)	26 (45.6)	25 (54.3)	31 (54.4)
No. of black African subjects (%)	57 (100)	46 (100)	57 (100)
Relationship duration (mo)b	63.7 (37.7–147.3)	74.5 (43.7–121.2)	65.3 (37.6–151.1)
pVL (RNA copies/ml)	NA	NA	1,485 (400–14,000)c
CD4 counts (cells/μl)	NA	NA	263 (191–418)d

^aValues are expressed as the “median (interquartile percentiles [IQR])” at enrollment except as noted otherwise in column 1. NA, not applicable.

^bThe duration of the relationship as reported by either partner was different for some couples, resulting in different values for HESN and HIV⁺ partners.

^cMeasurements were available for 46 participants. The detection cutoff was 400 copies/ml.

^dMeasurements were available for 53 participants.

Cells.

PBMC obtained 3 months after enrollment were used in the present study. PBMC were isolated from acid citrate dextrose (ACD) anticoagulated blood (Vacutainer; BD) by centrifugation on a Histopaque (Sigma) gradient, cryopreserved within 8 h of collection in 10% dimethyl sulfoxide (DMSO; Sigma) and 90% fetal bovine serum (FBS; Gemini Bio-Products, West Sacramento, CA), and stored long term in the vapor phase of a liquid nitrogen tank.

Peptides and control stimuli.

The consensus proteome of HIV-1 subtypes A1 and D (obtained from the Los Alamos National Laboratory HIV database [http://www.hiv.lanl.gov]) was covered using synthetic peptides of 18-amino-acid length overlapping by 10 amino acids (Sigma). Peptides were dissolved in DMSO and pooled by the major HIV antigens envelope (Env), polymerase (Pol), group-specific antigen (Gag), and Nef, plus all accessory proteins (NA), and diluted to a final individual concentration of 0.7 to 2 μg/ml in RPMI 1640 with 10% human serum AB, 2 mM l-glutamine, 1 mM sodium pyruvate, 10 mM HEPES, 100 IU of penicillin/ml, and 100 μg of streptomycin/ml (RAB10; all components were from obtained Sigma) in order to stimulate PBMC or short-term T cell lines (STCL). An influenza virus, Epstein-Barr virus, and cytomegalovirus (FEC) peptide pool consisting of known CD8 T cell epitopes (23) was used to comparatively evaluate T cell responses to common, non-HIV epitopes. All responses to peptides were evaluated against negative controls of 0.45% DMSO (maximum concentration present in pools [Mock]) and positive controls (phytohemagglutinin [PHA] 10 μg/ml). All stimuli in ELISPOT assays were distributed from prepared 96-well plates at 2× final concentrations and stored at −80°C (<2 years).

Ex vivo ELISPOT assays.

Cryopreserved PBMC were thawed, rested 24 h in R10 (containing the same components as RAB10, but with 10% FBS instead of human serum AB), distributed at 2 × 10⁵ cells/well, and stimulated with HIV peptides, FEC, PHA, or Mock stimulus in IFN-γ ELISPOT assays for 18 to 20 h.

Cultured ELISPOT assay.

The cultured ELISPOT assay includes the preexpansion of antigen-specific T cells and is more sensitive than the ex vivo assay (24).

Briefly, cryopreserved PBMC were thawed, rested 24 h in R10, and then cultured in the presence of 25 ng of rhIL-7 (R&D Systems)/ml and peptide pools consisting of HIV subtypes A and D for Env, Gag, Pol, NA, or FEC (0.7 to 2 μg/ml) in RAB10 for 3 days, after which cultures received fresh RAB10 and 1,800 IU of rh-IL-2 (Novartis)/ml on days 3 and 7. STCL were harvested on day 10, washed extensively, and incubated for 29 to 34 h in RAB10 before distributing 4 × 10⁴ STCL/well and restimulating with peptide pools in IFN-γ ELISPOT assays.

ELISPOT assays (procedures common to ex vivo and cultured assays).

Mouse anti-human IFN-γ antibodies (MabTech, clone 1-D1K, 1 μg/well) were bound to MAIPS4510 96-well plates (Millipore) and R10 blocked for 1 to 3 h before adding cells in RAB10. The cells were distributed into wells and rested for 30 min in a humidified incubator at 37°C with 5% CO₂ before stimuli were added.

Plates were replaced in the incubator for 20 h. The cells were discarded, the wells were washed six times with 200 μl of phosphate-buffered saline (PBS; Sigma) and 100 μl of biotinylated mouse anti-human IFN-γ antibody (MabTech clone 7-B6-1, 0.1 μg/well)/well in PBS–0.5% bovine serum albumin (Sigma), followed by incubation at room temperature for 2 to 4 h. The wells were then washed six times with 200 μl of PBS–0.05% Tween (Sigma), followed by incubation for 1 h at room temperature with 100 μl of peroxidase complex (Vector Laboratories)/well. The wells were next washed three times with 200 μl of PBS–0.05% Tween, followed by three times 200 μl of PBS, before the addition of 100 μl of 3-amino-9-ethylcarbazole substrate (Sigma)/well for a 5-min incubation at room temperature, followed by washing with distilled water and drying. Spot counts were acquired on an ELISPOT reader ELRIFL04 (AID, Strassburg, Germany) using AID ELISPOT software version 4.0, after calibration with a reference plate. The counts were initially acquired with automated settings and subsequently quality controlled by an administrative user blinded to the exposure status of the participant.

Rejection and response criteria in ELISPOT assays.

Quality control filters were applied to spot counts for each STCL. The results were excluded if the spot counts were ≤30/well for PHA, >10/well for Mock, <3 replicate wells for any HIV antigen or Mock, or if the replicate variance/(median + 1) was >50 spot-forming units (SFU). Responses were scored as positive if the mean SFU were three times Mock and ≥12 SFU/well above Mock (i.e., “biologically significant responses”). For HIV responses in cultured ELISPOT analyses, an additional criterion required that biologically significant responses be consistently repeated with different peptide pools versus seen only in one of two pools containing the same peptide. The subpool rule can have two forms—“or” and “and”—as shown in Table 2. The “and” version is useful in evaluating the subtype specificity of responses.

Table 2.

Example of scoring responses to Env using “or” and “and” versions of the subpool rule^a

Response on individual Env pool and subpools			Response scoring
A/D	A	D	“Or”	“And”
1	1	1	Env A, Env D, and Env A/D	Env A, Env D, and Env A/D
1	1	0	Env A and generic “Env”	Env A
1	0	1	Env D and generic “Env”	Env D
1	0	0	No response	No response
0	1	1	No response	No response
0	1	0	No response	No response
0	0	1	No response	No response
0	0	0	No response	No response

^aA biologically significant responses to each single pool or subpool is indicated by “1”, and a lack of response is indicated by “0”. We used the “or” version for scoring response generically to Env, Gag, Pol, and NA. The “and” version was used for determining subtype specificity.

Flow cytometry on STCL.

Because of the limited availability of cells, flow cytometry was performed using leftover STCL remaining after setting up the cultured ELISPOT assays. After being washed and after 29 to 34 h of incubation in RAB10, the STCL of each participant were pooled and stimulated for 6 h in U-bottom, 96-well plates with ENV_AD, GAG_AD, POL_AD, NA_AD, and FEC peptide pools, SEB (0.2 μg/ml; Sigma), and phorbol myristate acetate (PMA)-ionomycin (0.1 and 1 μg/ml, respectively; Sigma) as positive controls and an unstimulated control (RAB10) in the presence of 1 μg of brefeldin A (Sigma)/ml, 1 μg of anti-CD28 (BD)/ml, and 1 μg of anti-CD49d (BD)/ml in a total volume of 100 μl. The STCL were subsequently treated with EDTA (Fastimmune; BD), and the dead cells were stained with fixable viability dye (ViViD; Invitrogen), fixed in 0.35% buffered formalin (Sigma) at room temperature for 10 min, and either frozen at −80°C in sealed aluminum foil-wrapped plates or processed directly. Frozen STCL were thawed at 37°C and immediately washed, permeabilized with Perm Solution 2 (BD) for 10 min at room temperature, washed again, and stained for 30 min at room temperature with an antibody panel consisting of anti-CD3 Alexa 488 (BD Biosciences [, San Jose, CA], catalog no. 557694), anti-CD4 Quantum Dot 605 (Invitrogen, catalog no. Q10008), anti-CD8 PE Cy7 (BD Biosciences, catalog no. 557746), anti-IFN-γ PE (BD Biosciences, catalog no. 340452), anti-IL-2 APC (BD Biosciences, catalog no. 341116), anti-TNF-α Alexa 700 (BD Biosciences, catalog no. 557996), anti-MIP-1-β PerCP Cy5.5 (BD Biosciences, catalog no. 560688), anti-CD14 Pacific Blue (Invitrogen, catalog no. MHCD1428), and anti-CD19 Pacific Blue (Invitrogen, catalog no. MHCD1928). After washing, the samples were acquired on a customized LSR II (BD Biosciences) using a high-throughput unit and FACSDiVa software v6.1.3.

Analysis of flow cytometry data.

Events were gated on singlets (FSC-A versus FSC-H), lymphocytes (FSC-A versus SSC-A), live CD3⁺ cells (CD3 versus a Pacific Blue/ViViD dump channel for monocytes, macrophages, B lymphocytes, and dead cells), and CD4⁺ or CD8⁺ cells. The average numbers of live CD3⁺ cells analyzed were as follows: Env, 28,000; Gag, 30,000; NA, 27000; Pol, 25,000; and unstimulated, 29,000. Cytokine-producing cells within the CD4⁺ and CD8⁺ subsets were identified using single cytokine gates to generate Boolean sets of gates for all possible “with or without” combinations. Analyses were performed using FlowJo (v9.2; TreeStar) and the programs Pestle (v1.6.2) and SPICE (v5.2), which were developed by Mario Roederer, Vaccine Research Center, National Institutes of Health, Bethesda, MD.

Statistical analysis.

The ex vivo ELISPOT readouts were analyzed as described previously (25). Cultured ELISPOT response rates were analyzed using classical two-group proportion tests (z-statistics) and the Fisher exact test. Magnitudes (expressed as the frequencies of SFU) and breadth (expressed as the proportions of the peptide pools recognized) of responses were evaluated using two-group t tests with equal variances as determined by STATA 10.0 SE (StataCorp LP, release 10.0). Multiparameter flow cytometry data were analyzed using the built-in permutation analysis, Wilcoxon test, and Student t test in SPICE. Significance levels were set at P < 0.05.

RESULTS

IFN-γ response rates.

Because of PBMC number limitations, cultured ELISPOT assays were prioritized over ex vivo ELISPOT assays, leading to different numbers of participants tested in the two assays. The responses of 41 HESN and 28 HUSN to pools of HIV-1 subtype A and D peptides were evaluated using ex vivo IFN-γ ELISPOT assays (Fig. 1). Neither HESN nor HUSN responded to HIV-1 peptide pools, whereas 19/29 (65.5%) HESN and 11/19 (57.9%) HUSN responded to the FEC pool, demonstrating easily detectable CD8 T cell responses to common pathogens (influenza virus, Epstein-Barr virus, and human cytomegalovirus). In control experiments using the ex vivo ELISPOT assay, PBMC from the HIV-infected partners responded to at least one HIV pool in 30/31 (96.8%, 95% confidence interval [CI] = 83.3 to 99.9%) assays (data not shown), suggesting that a positive response in the ex vivo ELISPOT assay might require a relatively high frequency of specific T cells, such as induced in an infection.

Fig 1.

Ex vivo IFN-γ ELISPOT responses in HESN and HUSN. PBMC were stimulated with pools of subtype A or D peptides, grouped by main antigens. “A&D shared” included shared peptides that have the same sequence in A and D subtypes. Bars indicate response rates (HESN, dark gray; HUSN, light gray); black vertical lines indicate the 95% CI. The number of individuals with positive responses/number of individuals tested is shown above each bar.

In order to detect HIV-specific memory T cells that might be present at low frequency in the peripheral blood of uninfected individuals, responses were investigated in 57 HESN and 46 HUSN using the more sensitive cultured IFN-γ ELISPOT assay. In preliminary experiments carried out to compare response rates between the Entebbe and Oxford laboratories, we had found that 9/30 (30%, 95% CI = 16.7 to 47.9%) of Ugandan HIV-1-seronegative individuals and 0/6 (0%; 95% CI = 0 to 39.0%) of U.S. leukapheresis donors responded to any HIV A or D subtype pool in the cultured IFN-γ assay (data not shown), confirming that the cultured ELISPOT assay could detect responses to HIV peptides in HIV-seronegative Ugandan individuals.

PBMC were stimulated with subtype A+D peptide pools for Env, Gag, NA, or Pol (each antigen separately) or FEC and then washed, rested, and tested in IFN-γ ELISPOT assays with the same antigen and HIV/FEC cross-tests. In order to minimize false-positive responses, strict quality control filters were applied. Of 57 HESN and 46 HUSN for whom cultured ELISPOT analyses were completed, 53 HESN and 45 HUSN had HIV results that passed the filters, and 54 HESN and 46 HUSN had valid FEC results (Fig. 2). Flow cytometric evaluation was performed on STCL from 15 HESN and 10 HUSN for whom sufficient cells were available. All biologically significant responses of HIV HESN and HUSN controls in cultured ELISPOT assays are included in Fig. 3, which shows the overlap of sets of biologically significant responses to the individual antigen pools of A, D, and A+D subtypes and the effect of the subpool rule on response rates. These Venn diagrams show that T cell repertoires in HESN and HUSN individuals contain memory cells capable of responding to both A and D subtypes of HIV-1. Env antigens appeared predominant in inducing responses, both in HESN and HUSN, whereas Pol-specific responses were the least frequent. A marked difference between HESN and HUSN can be seen in Gag-specific responses, with 1/35 (2.9%) HUSN responder compared to 10/50 (20.0%) HESN (P = 0.023, Fisher exact test).

Fig 2.

Overview of cultured ELISPOT assays.

Fig 3.

Venn diagrams of responses of HESN and HUSN in cultured IFN-γ ELISPOT assays. Short-term T cell lines stimulated with subtype A and D peptides pooled by antigen (Env, Gag, NA, and Pol) were tested in cultured IFN-γ ELISPOT assays for recognition of the same pool (A+D) and its separate A or D subpools. Responses were classified as positive based on biological significance criteria as described in Materials and Methods. HESN are represented by full symbols, and HUSN are represented by open symbols. The gray circular areas encompass the total number of individuals tested for each antigen (ENV [n = 80], 46 HESN and 34 HUSN, 1 HESN and 1 HUSN were not tested on A; GAG [n = 85], 50 HESN and 35 HUSN, 2 HUSN were not tested on D; NA [n = 67], 38 HESN and 29 HUSN, 1 HESN not tested on A/D; POL [n = 63], 37 HESN and 26 HUSN, 1 HESN and 1 HUSN not tested on A/D), and the overlapping sets labeled A, D, and AD include all individuals with positive responses to the corresponding pools. Responses validated by the “or” subpool rule are in the sectors highlighted by a white background, while the central white sector with triple overlap corresponds to the more selective “and” subpool rule.

After applying the subpool rule (the “and” form), the overall response rates to “any HIV antigen” were higher in HESN than in HUSN, with a mean response rate of 12.55% (95% CI = 9.67 to 15.42%, n = 510) versus 5.71% (95% CI = 3.33 to 8.08%, n = 368), two-sample test of proportions P = 0.0004. Stratification by antigen (Fig. 4) showed that the highest response rate in HESN was elicited by Env, whereas the highest response in HUSN was against the NA D pool. Significant differences between HESN and HUSN were seen for Env A/D (15.22% [n = 46] versus 2.94% [n = 34]; P = 0.0352), Gag A (14.00% [n = 50] versus 2.86% [n = 35]; P = 0.0417) and Gag D (12.00% [n = 50] versus 0.00% [n = 33]; P = 0.0194). Responses to the FEC pool were similar in HESN and HUSN (82.22% [n = 45] versus 78.38% [n = 37]; P = 0.3310), suggesting comparable exposure to common viruses influenza virus, cytomegalovirus, and Epstein-Barr virus and no inherent differences in T cell responsiveness in the two participant groups.

Fig 4.

Response rates of HESN and HUSN to HIV antigen pools. PBMC were stimulated in cultures with Env A/D Gag A/D, NA A/D, and Pol A/D peptide pools, before testing in ELISPOT assays with homologous superpool and subpools, using the “and” version of the subpool rule. FEC responses are shown for comparison. Dark gray and light gray bars represent mean response rates of HESN and HUSN, respectively; black lines represent the 95% CI. The P values of statistically significant differences are underlined.

Responses to the FEC pool were commonly observed in cultures stimulated with HIV-1 pools, but only in individuals that had responses to FEC detectable by ex vivo ELISPOT assay, suggesting that FEC-reactive cells might be retained in the cultures in the absence of specific stimulation. The response rates, shown in Fig. 5, were approximately half those seen after FEC stimulation (HESN [82.2%] versus HUSN [78.4%], P = 0.331, Fig. 4), and there was no significant difference between HESN and HUSN for any of the culture stimuli, with P values of >0.05. The reverse (HIV-specific responses in FEC-stimulated cultures) was observed in one HESN (response to Pol) and one HUSN (response to NA),who also had positive responses with STCL stimulated with Pol and NA, respectively. These cultures were not analyzed further.

Fig 5.

Response rates to FEC in STCL stimulated with HIV antigen pools. Replicate wells were set up with FEC test antigen, and biologically significant responses were scored. Bars represent the mean response rates of HESN (dark gray) and HUSN (light gray). Black lines show the 95% CI, and P values are indicated for each pairwise comparison.

Magnitude of IFN-γ responses.

The mean magnitude of responses to HIV pools was generally higher in HESN compared to HUSN but did not reach statistical significance (Fig. 6). The A-versus-D plots (Fig. 7) show that STCL stimulated by the HIV antigens included a balanced response to both A and D subtypes, indicating the presence of cells capable of recognizing A and D epitopes in most responder individuals.

Fig 6.

Magnitude of responses to HIV antigens. PBMC were stimulated in cultures with Env A/D, Gag A/D, NA A/D, and Pol A/D peptide pools, before testing in ELISPOT assays with homologous superpools and subpools. FEC-specific responses are shown for comparison. Bars represent mean responses (HESN, dark gray; HUSN, light gray); black lines represent the 95% CI. All P values for HESN-HUSN comparisons were >0.05.

Fig 7.

Magnitude of IFN-γ responses to A and D subtypes. The magnitude of subtype A- and D-antigen-specific responses in SFU/million PBMC was plotted on orthogonal logarithmic scales for each participant. HESN are represented by diamonds, and HUSN are represented by squares. The black symbols with interval bars indicate mean magnitudes and the 95% CI. The means for HESN and HUSN were not significantly different.

Breadth of IFN-γ responses.

In order to compare the breadth of HIV antigen recognition in HESN and HUSN, the average number of different peptide pools responded to by each participant, of the 12 HIV peptide pools tested, was calculated by the progressively stricter criteria of biological activity and the “or” and “and” forms of the subpool rule. By all three criteria, there was a significantly broader recognition of HIV peptide pools in HESN compared to HUSN (Fig. 8).

Fig 8.

Breadth of response to HIV antigen pools. Responses of HESN (n = 53) and HUSN (n = 45) to the Env, Gag, NA, and Pol peptide pools containing A, D, and A+D subtype peptides (12 pools) were scored using the three progressively stricter response criteria of simple biological activity, the subpool rule (“or” version), and the subpool rule (“and” version). Mean numbers of pools giving a response were compared by using the t test. Bars represent mean numbers of pools producing a positive response in HESN (dark gray) or HUSN (light gray) overlapped with 95% CI. P values are shown for each comparison.

Lack of correlation between partner viral load and HESN responses to HIV.

Since plasma viral load in the HIV⁺ partners might be taken as a rough proxy for the HESN′s exposure to HIV, we evaluated whether plasma viral load in the HIV⁺ partner could be correlated with IFN-γ responses to HIV epitopes in the HESN partner (Table 3). A total of 23/50 (46.0%) HIV⁺ partners for whom viral load measurements were available (from plasma samples obtained between 129 days before and 238 days after the PBMC sample) had viral loads above the detection limit of 400 RNA copies/ml of plasma (i.e., median, 11,800; minimum, 594; and maximum, 134,000). A total of 18/48 (37.5%) HIV⁺ partners were not on antiretroviral therapy (ART). The background subtracted SFU of the HESN partner were cumulated by single antigens and also by any HIV reactivity. A pairwise correlation analysis was performed for the whole HESN group, and a separate one was performed for just the HESN subset whose partners were not on ART. No significant correlation was observed in either case.

Table 3.

Lack of correlation between partner viral load and magnitude of IFN-γ ELISPOT response in HESN^a

Specificity of response	Correlation coefficient (rho)	Significance (P)	No. of observations
All HESN
Env	–0.0120	1.0000	48
Gag	–0.0777	1.0000	50
NA	0.0918	1.0000	42
Pol	–0.0530	1.0000	36
Cumulative (all HIV)	–0.0373	1.0000	50
HESN whose partner was not on ART
Env	0.1525	1.0000	16
Gag	–0.0180	1.0000	18
NA	0.6085	0.3140	14
Pol	0.0261	1.0000	11
Cumulative (all HIV)	0.2410	1.0000	18
HESN whose partner's viral load was >1,000
Env	0.1877	1.0000	17
Gag	–0.1598	1.0000	19
NA	0.3974	1.0000	15
Pol	–0.1207	1.0000	11
Cumulative (all HIV)	0.1107	1.0000	19

^aThe IFN-γ ELISPOT SFU for HIV antigens in HESN were cumulated and compared by pairwise correlation with viral load measurements in the HIV⁺ partner. The Bonferroni correction for multiple comparisons was applied. The interval between samples for viral load and for PBMC was between days −129 (before the PBMC sample) and 238.

Cytokine polyfunctionality.

Since we had observed higher response rates to HIV peptides in HESN compared to HUSN based on IFN-γ cultured ELISPOT assays, we wanted to verify whether the “quality” of response in terms of coexpression of a group of cytokines (IFN-γ, TNF-α, MIP-1β, and IL-2) by individual T cells was also different and whether different T cell subsets (CD4⁺ and CD8⁺) could be involved. To this end, all STCL left over after setting up the cultured ELISPOT assays for 15 HESN and 10 HUSN were pooled for each individual. Equal aliquots of the pool (0.2 × 10⁶ to 1.4 × 10⁶ cells) were separately stimulated for 6 h with ENV_AD, GAG_AD, POL_AD, and NA_AD, with FEC peptide pools, SEB, and PMA-ionomycin as positive controls, and with an unstimulated control (RAB10) in the presence of a secretion blocker and costimulatory antibodies. The STCL were then stained for viability, fixed, permeabilized, and stained for CD3, CD4, CD8, CD14, CD19, IFN-γ, IL-2, TNF-α, and MIP-1β. The percentages of CD4⁺ and CD8⁺ T cell subsets expressing the four cytokines were defined using binary Boolean gates. Among the possible 15 combinations of positive responses, single cytokine responses predominated within both CD4 and CD8 T cell subsets. T cells producing only IFN-γ were significantly more frequent in HUSN than in HESN (data not shown). HESN had a larger proportion of cells expressing >1 cytokine than HUSN. The overall profiles of cytokine responses in HESN and HUSN were compared by permutation analysis (26), obtaining P values that represent the proportion of permutations producing distributions that are less alike than the experimentally observed one. These findings are summarized in SPICE visualizations (Fig. 9). With the exception of CD8 responses to Env stimulation, significant differences with P values between 0.02 and 0.0346 were found across multiple HIV stimulation conditions. Responses to SEB were not significantly different between HESN and HUSN, whereas CD8 responses to FEC were significantly different (P = 0.0430), with higher IFN-γ responses in HUSN compared to HESN. The distribution of spontaneous responses (background) was not significantly different in the two groups (data not shown).

Fig 9.

Multifunctional responses. Boolean gates were created for IFN-γ, IL-2, TNF-α, and MIP-1β responses in CD4 and CD8 subsets of STCL cultures stimulated with Env, Gag, Pol, and NA. The responses were threshholded. Pie charts represent the proportions of cells expressing one, two, three, and four cytokines simultaneously. P values indicating significantly different (<0.05) pie chart comparisons are shown.

DISCUSSION

This study found evidence of specific T cell responses to HIV epitopes in uninfected individuals exposed to HIV through heterosexual intercourse within serodiscordant couples in a cohort composed exclusively of Ugandan participants. T cell responses were not detectable by conventional ex vivo IFN-γ ELISPOT assays and required short-term culture of PBMC in the presence of HIV peptides, IL-7 and IL-2, indicating a very low frequency of circulating HIV-specific memory effectors in the peripheral blood. The high sensitivity of the cultured ELISPOT assay highlighted the presence of low-level reactivity to HIV epitopes both in HESN and in HUSN used as a control, but the response rate and the magnitude and breadth of the responses were significantly higher in HESN compared to HUSN. Gag-specific responses were only detected in 1/35 (2.9%) HUSN compared to 10/50 (20.0%) HESN, showing a qualitative difference between the two groups in antigen specificity. The qualitative “flavor” of cytokine response was also significantly different between HESN and HUSN in a subset of participants that was tested by flow cytometry and intracellular cytokine staining for IFN-γ, MIP-1β, TNF-α and IL-2, with single cytokine responses being a larger proportion of the total response in HUSN compared to HESN.

These observations suggest that differences in exposure to HIV might be involved in the different reactivities of HESN and HUSN. The response rates and frequencies of T cells specific for influenza virus, cytomegalovirus, and Epstein-Barr virus were higher than those specific for HIV and not significantly different between HESN and HUSN, suggesting that HUSN were not particularly sheltered from environmental exposure to common viruses, nor did they have had intrinsic T cell responsiveness defects. Within the limits of the cultured ELISPOT assay, the response rate to HIV subtypes A and D in HUSN parallels findings in studies with B subtype in the United Kingdom and C subtype in Malawi (21). In that study and in the present one, strict criteria for the definition of the HESN and HUSN groups and for the quality control of ELISPOT responses were used; thus, the baseline responses observed in HUSN are likely to represent cross-reactivity with other undefined environmental exposures. Indeed, we observed a trend, albeit not attaining statistical significance, for higher response rates to HIV peptides in PBMC of Ugandan seronegatives compared to the PBMC of seronegative leukapheresis donors from the United States, suggesting it might be worthwhile to further investigate geographical and/or environmental correlations of HIV-specific cross-reactive responses in HUSN. We have not formally excluded the expansion of naive T cells present at very low frequency in the repertoire; however, the culture conditions used are not expected to support the in vitro priming of naive T cells, and a similar expansion of naive cells would be expected in HESN and HUSN. Our cultured ELISPOT assays primarily evaluated memory responses: more sensitive methods would be needed to measure frequencies of HIV-specific precursors in the naive repertoire. Other studies have detected HIV reactivity among HESN using ex vivo ELISPOT assays (19, 27, 28), and a possible explanation for our failure to do so might be a very low frequency of responder T cells in PBMC from seronegative partners in HIV-discordant couples, compared to high-risk populations of sex workers or breast-fed children of HIV-positive mothers. Our findings with STCL differ from those in Zambian exposed uninfected seronegatives (29), where after 10 days of expansion in vitro in the presence of HIV peptides, STCL from 8 HESN did not show reactivity to HIV in flow cytometric assays. The discrepancy might be due to inclusion of rhIL-7 in the initial phase of culture, which might have prevented the early apoptosis of specific T cells in our study. In fact, we observed efficient detection of FEC-specific T cells even from cultures stimulated with HIV antigens (and sporadically HIV-specific T cells in cultures stimulated with FEC), showing that culture conditions were such as to preserve nonstimulated T cells over at least 11 days. It is unclear whether the higher responses observed in HESN can afford relative protection against HIV infection compared to HUSN, and the design of the present study is not appropriate for demonstrating such protective effects. Since the efficiency of HIV transmission is low, with a distinctive transmission bottleneck (11, 30), it is likely that a marginal increase in T cell responses to noninfectious viral antigen or nonproductive infections would become detectable in HESN before sufficient exposure for transmission has occurred. In the present study, exposure was well defined, and an HIV viral load of >400 RNA copies/ml of plasma could be measured in 60% of the infected partners. However, the magnitude of cumulative IFN-γ responses detected in HESN did not correlate with the plasma VL load measured in the infected partners. Under these conditions no case of HIV transmission within the serodiscordant couples occurred during a 9-month follow-up period, which sets an upper bound on the incidence rate at <1.9/100 person years (py). Since bleeds for the present study were obtained 3 months into the follow-up, we can rule out the possibility that responses greater than the baseline detected in HESN might constitute initial HIV infection, which would have been detected at the 6- or 9-month visit. In another study of serodiscordant couples conducted by us in a parallel cohort at the same AIDS Information Centre in Kampala, a transmission rate of 2.3/100 py was observed when the seropositive partners were on ART, and a transmission rate of 2.9 was observed when they were not on ART (31). Although ours was the largest study of HESN to date using the cultured ELISPOT assay in a serodiscordant-couple cohort, evaluating the correlation between T cell responses and viral loads was not a primary objective of the study design, and thus this specific analysis lacked statistical power. Other limitations are the use of ART in over one-third of the infected partners and the use of plasma rather than mucosal viral load determinations. In comparison to high-risk cohorts, the HESN in the present study could represent the low exposure end of a spectrum, with lower incidence rates and plasma viral loads. The fact that HIV-specific responses above HUSN levels can still be detected reinforces the view that such reactivity is the result of exposure to viral antigen. The most common responses in HESN observed here were directed against envelope epitopes, and a significant difference between HESN and HUSN was the detection of Gag-specific responses almost exclusively in HESN, but it is problematic to interpret these responses as having a protective effect against HIV acquisition. A pattern of strong Env-specific and low Gag-specific responses was associated with poor control of progression in HIV-infected individuals (32), whereas good control correlates with Gag-specific responses (33). Viremic control in animal models (34, 35) correlated with Gag-specific T cell responses, and protection from acquisition was correlated with Env-specific responses. However, the possibility that these potential immune correlates are simply surrogate markers cannot be ruled out. If the association between real, mechanistic correlates and surrogates were steady, it would still be useful for vaccine development; however, this is not the case. In the STEP and Phambili Vaccine Trial, T cell responses to Gag, Pol, and Nef detectable by ex vivo ELISPOT assay—and therefore more than 1 order of magnitude higher than those described here—were present in vaccinated individuals that subsequently became infected (36), showing that the mere presence of IFN-γ reactivity to HIV antigens does not necessarily afford protection against infection. Indeed, a recent analysis of the infection rates in the STEP trial (37) shows that vaccinated uncircumcised individuals with preexisting immunity to the adenovirus 5 vector were initially more susceptible to HIV infection, suggesting a detrimental, transient biologic effect contemporaneous with the induction of T cell responses by the vaccine. In the RV144 trial, 31.2% protection against acquisition of HIV was correlated with plasma IgG against the V1V2 region of Env, antibody-dependent cellular cytotoxicity, and CD4⁺ responses, whereas plasma IgA against Env correlated with increased risk of HIV acquisition, and neither CD8⁺ responses nor broadly neutralizing antibodies were induced (38).

In view of these cautionary findings, our conservative conclusion is that the higher HIV-specific IFN-γ responses that we detected in HESN compared to HUSN might be a marker of exposure rather than protection against HIV infection and claims of protective effects will have to be supported by direct demonstration of antiviral activity of the responding T cells or by the correlation of lower seroconversion rates with preexisting T cell responses in high-risk individuals.