Adenovirus Vector-Induced Immune Responses in Nonhuman Primates: Responses to Prime Boost Regimens¹

The quest for an efficacious vaccine to the HIV-1 virus continues despite a large number of theoretical and practical obstacles. Given the present inability to generate broadly reactive HIV envelope-specific neutralizing Abs, most current HIV-1 vaccine candidates focus on eliciting protective CD8⁺ T cell responses (1). In a recent phase IIb clinical trial, termed STEP trial, the most promising of such vaccines, an E1-deleted adenovirus (Ad)⁴ vector of the human serotype 5 (AdHu5), not only failed to protect against infection, but instead showed a trend to render male volunteers with preexisting neutralizing Abs (NAs) to the vaccine carrier more susceptible to infection (2). The negative result of the STEP trial has raised considerable doubts about the validity of the concept of CD8⁺ T cell-mediated protection against HIV-1 infection (3, 4). Additionally, the STEP trial has triggered intense studies aimed at identifying the mechanisms underlying the vaccine’s “facilitating” effect on HIV-1 transmission linked to the presence of preexisting anti-AdHu5 Abs (5).

To circumvent the effects of NAs on the vaccine carrier in individuals that are infected during childhood with human Ad viruses such as AdHu5 (6), we developed E1-deleted Ad vectors from chimpanzee serotypes (AdC) (7, 8). We derived these vectors from several AdC serotypes to allow for booster immunization with heterologous AdC vectors. The molecular organization and basic biology of AdC viruses are similar to those of human Ad viruses (9, 10). In mice and nonhuman primates (NHPs), AdC vectors were shown to induce robust transgene product-specific T and B cell responses (7, 8, 11). Most importantly, NAs to AdC viruses are rarely detected in humans (12), and thus these vectors may outperform AdHu5 vectors in clinical trials.

Here, we compared two AdC-HIV-1 gag vectors (AdC6 and AdC7 serotypes) in an alternating boost protocol to a dual immunization with an AdHu5-HIV-1 gag vector in rhesus macaques that had or had not been preexposed to AdHu5. The results show that heterologous booster immunizations with the AdC vectors induce markedly higher gag-specific T and B cell responses compared with repeated immunization with the AdHu5 vector, and that responses to the AdC vectors, unlike those to the AdHu5 vector, are not impaired by preexisting NAs to AdHu5.

The vaccine vectors express a codon-optimized gag of HIV-1 clade B. Ad vectors were derived from AdHu5, AdC6, or AdC7. Vectors were E1-deleted and generated from viral molecular clones by viral rescue on HEK 293 cells, which were grown, purified, titrated, and quality controlled as described (8).

Two- to 3-year-old Chinese origin Macaca mulatta were purchased and housed at Bioqual. All procedures involving handling and sacrifice of animals were performed according to approved protocols.

PBMCs and lymphocytes from tissues were isolated as described. They were tested immediately after isolation ELISPOT assays. Remaining cells were frozen in 90% FBS and 10% DMSO (Sigma-Aldrich) at −80°C.

NA titers were determined as described (11) on HEK 293 cells infected with Ad vectors expressing GFP.

The ELISA assays were conducted on plates coated with HIV gag protein as described (13).

HIV clade B consensus sequence Gag peptides, 15-mers overlapping by 11 aa, were obtained from the National Institutes of Health Research and Reference Reagents Program.

The ELISPOT assays for IFN-γ and IL-2 were conducted as described (13). Spots were counted using the CTL Series 3A Analyzer and ImmunoSpot 3.2 (Cellular Technology). The minimum spot size was set to 0.0016 mm², and the maximum spot size was set to 0.0878 mm². The criteria for determining positive samples included that for every 10⁶ cells stimulated with peptides, at least 55 spots after subtraction of background spots (spots without antigenic stimulation) had to be detected. The number of spots upon peptide stimulation had to be at least 3-fold higher than that seen in control wells. Data shown on graphs represent values of peptide-stimulated wells from which background values have been subtracted.

Frozen cells were thawed and immediately washed with HBSS supplemented with 2 U/ml DNase I, resuspended with RPMI 1640, and stimulated for 6 h with anti-CD28, anti-CD49d, and brefeldin A (10 μg/ml each), with or without 1 μg/ml peptide of the gag HIV-1 peptide pools at 37°C, 5% CO₂. After incubation, cells were stained with violet-fluorescent reactive dye with Pacific Blue (Invitrogen), anti-human (h) CD14-Pacific Blue, anti-hCD16-Pacific Blue, anti-hCD8-PerCP-Cy5.5, anti-hCD95-PE-Cy5, and anti-hCD28-Texas Red (Beckman Coulter) for 30 min at 4°C. After fixation and permeabilization with Cytofix/Cytoperm (BD Biosciences) for 20 min at 4°C, cells were stained with anti-hIFN-γ-allophycocyanin, anti-hIL-2-PE, anti-hTNF-α-PE-Cy7, and anti-hCD3-allophycocyanin-Cy7 for 30 min at 4°C. Cells were washed twice, fixed with BD Stabilizing Fixative (BD Biosciences), and then analyzed by FACS using LSRII (BD Biosciences) and Diva software. Postacquisition analyses were performed with FlowJo (Tree Star). Single-color controls used CompBeads anti-mouse Igκ (BD Biosciences). Unless otherwise noted, Abs were purchased from BD Biosciences.

Cells from immunized NHPs were thawed and washed with HBSS supplemented with 2 U/ml DNase I and then stained as described for ICS. Cells were sorted by a FACSVantage SE using DiVa software (both from BD Biosciences) in a biosafety level 3 laboratory (University of Pennsylvania).

Genomic DNA was extracted as described (14). GAPDH was quantified by real-time PCR. Samples were adjusted to equal amounts of GAPDH (per 10⁶ copy) and then hexon sequences from AdHu5, AdC6, and AdC7 were amplified by PCR (30 cycles of 95°C for 40 s, 50°C for 40 s, and 72°C for 40 s) using primers that distinguished between the viruses. The amplicon (1 μl) from the first PCR product was used as template for a nested PCR (30 cycles of 95°C for 30 s, 52°C for 30 s, and 72°C for 30 s). The following primers were used for first PCR (forward and reverse, respectively): AdHu5, 5′-ATCATGCAGCTGGGAGAGTC, 5′-ACACCTCCCAGTGGAAAGCA; AdC6, 5′-ATCGGTCTTATGTACTAC, 5′-GTCCATGGGGTCCAGCGACC; AdC7, 5′-AGGTACAGATGACAGTAGCTC. The following primers were used for the nested PCR: AdHu5, 5′-GACTCCTAAAGTGGTATTGT-3′, 5′-GTCTTGCAAATCTACAACAG-3′; AdC6, 5′-TCCCAGCTGAATGCTGTG-3′, 5′-GCCGTCCAAGGGGAAGCAAT-3′; AdC7, 5′-ACAGACCCAACTACATTGGC-3′, 5′-GATTCCACATACTGAAATACC-3′. Amplicons were checked on 1% agarose gels; in most samples no specific band could be detected after the first PCR. The amplicons (1 μl) from the first PCR product were used as templates for a second real-time PCR (40 cycles at 95°C for 5 s, 52°C for 10 s, 72°C for 12 s, and 83°C for 4 s). The amplicon was run on a 1% agarose gel, and samples that showed a band of the expected size were viewed as being positive. Some of the bands were sequenced to confirm the specificity of the reaction. For some experiments the copy numbers of hexon in each sample were quantified by normalization in comparison to GAPDH sequences as described (14).

Significance in ICS was determined by two-tailed Student’s t tests and ANOVA analysis. ELISPOT statistical analyses were conducted using two-tailed Student’s t tests or Wilcoxon rank-sum tests, as appropriate. All analyses were conducted using SAS 9.0.

Two experiments involving a total of 28 juvenile Chinese rhesus macaques (M. mulatta) divided in four groups of animals were conducted as part of this study. The four experimental groups each contained seven animals, of which six received an HIVgag immunization protocol and one control animal that was vaccinated with the same vaccine vector containing an unrelated transgene (Table I). All of the animals were prescreened for lack of NAs to the Ad vector vaccines. Before the HIVgag immunization phase, half of the macaques were immunized with 10¹¹ virus particles (vps) of an AdHu5 vector expressing α1 anti-trypsin (A1AT) given intratracheally. NA titers to AdHu5 were measured 2 wk later from serum. Animals with titers <1:10 were reimmunized with 10¹¹ vps of AdHu5-A1AT and thereafter seroconverted. In the first experiment, six AdHu5 preexposed and six non-preexposed animals were immunized with 10¹¹ vps of AdC6gag and then boosted at 3-mo intervals with 10¹¹ vps of AdC7gag, followed by 10¹¹ vps of AdC6gag, followed by 10¹¹ vps of AdC7gag. All vaccines were given i.m. The control animals received the same vaccine carriers expressing the rabies virus glycoprotein using the same regimen. In the second experiment, six AdHu5 preexposed and six non-preexposed animals were immunized i.m. with 10¹¹ vps of AdHu5gag vector. They were boosted 3 mo later i.m. with the same dose of AdHu5gag vector. Control animals received an AdHu5 vector expressing the rabies virus glycoprotein.

For both experiments, after each dose of vaccine, plasma and PBMC samples were collected in 2- to 4-wk intervals. Between 2 and 3 mo after the last immunization, animals were euthanized and lymphocytes were isolated from tissues.

T cell responses to gag peptide pools were tested by IFN-γ and IL-2 ELISPOT assays (Fig. 1, A and B) as described (13). By wk 10 after immunization with the AdC6gag vector, IFN-γ responses were positive (>55 spots/10⁶ PBMCs) in most animals and had reached peak frequencies. After booster immunization with AdC7gag, all animals developed detectable IFN-γ responses within 4 wk and in most animals, responses peaked by wk 6. After the third and fourth immunizations peak frequencies (Fig. 1 C) remained below those obtained after the first boost. The response to the AdHu5gag vector came up more rapidly in non-preexposed animals, most of which developed peak frequencies of IFN-γ-producing T cells within 2 wk after immunization. Responses were higher in the non-preexposed than in the AdHu5 preexposed group. The second immunization with the AdHu5gag vector increased responses in some of the animals; this increase was largely transient. Statistical analyses revealed no significant differences in the IFN-γ response of nonexposed and AdHu5 preexposed AdC-immunized groups (p = 0.904), whereas responses in the AdHu5gag-vaccinated groups were significantly affected by preexposure to AdHu5 (p < 0.001).

Most of the animals developed IL-2 responses after the initial AdC6gag immunization, and all of them became positive after booster immunization with AdC7gag. Further booster immunizations were relatively ineffective (Fig. 1 D). After the first immunization with the AdHu5gag vector, only a fraction of the animals developed low frequencies of IL-2-producing T cells; these responses declined in all animals to <55 spots/10⁶ PBMCs by wk 14 after immunization. After the boost, some animals had positive IL-2 responses, which in most disappeared by wk 6. Overall there was no significant difference in the IL-2 response of preexposed and non-preexposed animals in the AdC-vaccinated (p = 0.668) and the AdHu5-vaccinated (p = 0.92) animals, although the response was higher and more sustained in animals vaccinated with the AdC vectors (p = 0.010). PBMCs from control animals were tested in parallel in both experiments, and responses were <55 spots for IFN-γ or IL-2 per 10⁶ PBMCs at all time points tested (data not shown).

Frequencies of IFN-γ- and IL-2-producing CD8⁺ and CD4⁺ T cells were analyzed by ICS (Fig. 2) and were in general comparable to those obtained by ELISPOT. Gag-specific CD8⁺ T cells (Fig. 2, A and B) were dominated by cells producing IFN-γ (Fig. 2,A), and frequencies were, at most time points tested, higher in animals immunized with the AdC vectors than with the AdHu5 vector. The booster effects after repeated immunization with the AdC vectors were more pronounced than seen by ELISPOT. A detectable CD4⁺ T cell response was elicited only in AdC-immunized animals (Fig. 2, C and D). After the first and second immunization, CD4⁺ T cells produced mainly IL-2 (Fig. 2 D), while the third and fourth immunizations resulted in induction of CD4⁺ T cells producing IFN-γ in the AdC-immunized groups.

To test if the effectiveness of booster immunizations correlated with NA titers to the vaccine carrier, plasma from animals was tested before each boost for neutralizing Abs to the vaccine carrier (Fig. 3). As expected after each immunization, NA titers to the vaccine carrier increased. Unexpectedly, the primary Ab response to AdC7 was less pronounced than that to AdC6. There was no clear correlation between the increase of T cell responses following booster immunization and titers of NAs to the vaccine carrier at the time of vaccination.

Ab titers to gag were measured after each vaccine dose. The AdC6gag vector induced a gag-specific Ab response in all of the NHPs that increased after the AdC7gag boost (Fig. 4). Additional booster immunizations with AdC6gag and AdC7gag did not cause any further increase. Responses in AdHu5 preexposed and non-preexposed NHPs were comparable in both AdC-immunized groups (p > 0.05, two-tailed Student’s t test) at all time points tested. AdHu5gag vectors only induced a moderate Ab response to gag that did not increase after the second dose. In AdHu5 preexposed NHPs, the Ab response to gag presented by AdHu5 vectors was significantly lower than in non-preexposed NHPs (p < 0.02 at all time points). After the second immunization, titers in both groups (preexposed and non-preexposed) were significantly higher in AdC-vaccinated animals (p = 0.005 for preexposed NHPs, p = 0.036 for non-preexposed NHPs).

After euthanasia, lymphocytes were isolated from various tissues and tested for responses to gag peptides by ELISPOT and ICS. Responses to gag were detectable in all AdC-immunized animals in blood, spleen, and liver. For most AdC-immunized animals, responses were comparable in blood and spleen, low in lymph nodes (data not shown), and, as reported previously, high in liver (13). Responses in tissues were low in the AdHu5-immunized animals. Lymphocytes isolated from the intestine or the genital tract had high background activity, and these results could not be interpreted (data not shown).

IL-2 responses were overall lower than IFN-γ responses (Fig. 5). In most of the AdC vector-immunized animals, IL-2 responses were correlated between blood, spleen, and liver. A few animals had high IL-2 responses in iliac lymph nodes (data not shown). AdHu5-immunized animals had low IL-2 responses at the time of euthanasia.

ICS analyses showed that most of the IFN-γ was derived from CD8⁺ T cells (Fig. 6), although some animals also had IFN-γ⁺CD4⁺ T cells in various compartments. IL-2-producing CD4⁺ and CD8⁺ T cells were detected at low frequencies in blood of non-preexposed AdC-immunized animals, but not in PBMC samples from the other groups. A fraction of the AdC-immunized animals had IL-2-producing CD8⁺ or CD4⁺ T cells in their spleen, most had detectable frequencies of IL-2-producing CD8⁺ T cells in liver, and some had hepatic IL-2-producing CD4⁺ T cells. Conversely, IL-2-producing CD4⁺ or CD8⁺ T cells could not be detected in most AdHu5gag-immunized animals.

Neither the IFN-γ nor IL-2 responses in the individual tissues tested reached statistical significance when comparing the preexposed and non-preexposed groups of animals receiving either vector (p > 0.05 for each tissue). Overall, the IFN-γ responses induced in the liver of animals vaccinated with the chimpanzee vector regimen were slightly higher than in the liver of animals vaccinated with the AdHu5 vaccine regimen (p = 0.044). Also, at necropsy the IL-2 responses induced in the blood, liver, and spleen of animals vaccinated with the chimpanzee vector regimen were higher than in these tissues of animals vaccinated with the AdHu5 vaccine regimen (p = 0.009, p = 0.002, and p = 0.003, respectively).

In summary, the response in tissues was dominated by CD8⁺ T cells producing IFN-γ in AdC-immunized animals regardless of preexposure to AdHu5, and these responses were more robust than in AdHu5-immunized animals. CD4⁺ T cell responses to the transgene product were only detected in AdC-vaccinated NHPs.

Gag-specific T cells isolated from blood at two time points (wk 6 and 10) after each immunization were characterized in more depth from three of the NHPs (nos. 3, 9, and 11) that mounted high responses to AdC immunization (Fig. 7,A). One of the animals (no. 3) had been preexposed to AdHu5, while the other two had not. PBMCs were stained with Abs to CD3, CD8, CD95, and CD28 to determine frequencies of effector (CD95⁺CD28⁻) and memory (CD95⁺CD28⁺) T cells (Fig. 7,A, line graphs) to gag. Effector CD8⁺ T cells were more common than memory CD8⁺ T cells throughout the course of the immunizations in NHP nos. 9 and 11, while NHP no. 3 had slightly higher levels of memory CD8⁺ T cells. Memory CD8⁺ T cell cytokine production profiles in blood varied between the animals but were remarkably stable through the course of vaccination in each of the animals (Fig. 7 A, pie charts). NHP nos. 3 and 11 had mainly memory CD8⁺ T cells that produced IFN-γ or IFN-γ and TNF-α, although CD8⁺ T cells that produced IL-2 or all three cytokines could also be detected; NHP no. 9 had higher frequencies of CD8⁺ T cells producing TNF-α or IL-2.

Spleen- (Fig. 7, B and C) and liver- (Fig. 7, D and E) derived CD8⁺ T cells from the three NHPs described above, as well as from NHP no. 7, which had been preexposed to AdHu5, were also analyzed to determine frequencies of CD95⁺CD28⁻ (effector) and CD95⁺CD28⁺ (memory) T cells (Fig. 7, C and E) and their cytokine production (Fig. 7, B and D) in response to gag. In spleens and liver, frequencies of effector and memory CD8⁺ T cells were roughly comparable; NHP nos. 3 and 11 had in both compartments slightly higher frequencies of memory cells, while NHP no. 9 had slightly higher frequencies of effector cells. Effector and memory CD8⁺ T cells to gag from spleens and liver of all four NHPs produced mainly IFN-γ, TNF-α, or IFN-γ and TNF-α. CD8⁺ T cells that produced IL-2 alone or together with IFN-γ or that produced all three cytokines were mainly found in the memory subset. There was no clear difference in subset distribution or the cytokine production profile between preexposed and non-preexposed animals.

A similar analysis was conducted for CD3⁺CD8⁻ cells (CD4⁺ T cells) (Fig. 8). Throughout the course of the immunizations, animal no. 11 had mainly circulating gag-specific CD95⁺CD28⁺ memory CD4⁺ T cells, and comparatively lower levels of the more activated CD95⁺CD28⁻ effector CD4⁺ T cells. In contrast, NHP no. 3 had comparable levels of both subsets. In NHP no. 9, the ratio of effector to memory CD4⁺ T cells fluctuated. Additionally, CD4⁺ T cells were stained for intracellular IFN-γ, IL-2, and TNF-α (Fig. 8 A, pie charts). Gag-specific CD4⁺ T cells from these three animals differed in their cytokine production. NHP no. 3 had a very mixed profile, with CD4⁺ T cells that produced two or all three cytokines being frequent. NHP no. 9 had at most time points a predominance of CD4⁺ T cells producing either TNF-α or IL-2. Most CD4⁺ T cells from NHP no. 11 produced either IL-2 or TNF-α. In the latter two animals, CD4⁺ T cells that produced only one cytokine were more common than those positive for two or three cytokines.

Spleen- (Fig. 8, B and C) and liver- (Fig. 8, D and E) derived T cells from the three NHPs described above, as well as from NHP no. 7, which had been preexposed to AdHu5, were also analyzed to determine frequencies of CD95⁺CD28⁻ (effector) and CD95⁺CD28⁺ (memory) T cells (Fig. 7, C and E) and their cytokine production (Fig. 8, B and D) in response to gag. In spleens NHP nos. 3, 7, and 9 had predominantly gag-specific memory CD4⁺ T cells, while NHP no. 11 had approximately equal numbers of memory and effector CD8 T cells. Effector CD4⁺ T cells produced mainly IFN-γ or IFN-γ and TNF-α. Frequencies of CD4⁺ T cells producing more than one cytokine were higher in memory CD4⁺ T subset. Frequencies of IL-2-producing memory CD4⁺ T cells were higher in preexposed than in the non-preexposed animals.

In the liver of all of the NHPs, gag-specific memory CD4⁺ T cells were more frequent than gag-specific effector CD4⁺ T cells. Effector CD4⁺ T cells producing only IFN-γ predominated, and those producing only TNF-α or IFN-γ and TNF-α were also common. As in spleens, memory CD4⁺ T cells showed a mixed profile, with substantial frequencies of CD4⁺ T cells producing two or three cytokines. Subset distribution and cytokine profiles in blood were poorly predictive for those in tissues.

Levels of cytokine production were analyzed from blood-derived T cells at various times after each immunization and from spleens and liver (Fig. 9). In general, CD8⁺ (Fig. 9,A) or CD4⁺ (Fig. 9 B) T cells positive for two or three cytokines produced markedly higher levels of IFN-γ and/or IL-2 than did those that were only positive for one cytokine.

Lymphocytes and tissue fragments of spleens, liver, and intestine were analyzed for vector sequences using primers specific for the hexon coding sequence of the different Ad vectors. As shown in Table II, most samples were positive for the Ad vectors used for immunization. Additional studies on sorted lymphocytes showed that vector sequences were mainly present in CD8⁺ T cells (Table II). Previous mouse studies had shown that Ad vectors preferentially persist in activated CD8⁺ T cells responding to the transgene product of the Ad vector (14). During ICS analyses of lymphocytes from AdC-immunized animals, we detected a population of IFN-γ-producing CD8⁺ T cells that was CD95^low or CD95^int (Fig. 10,A), which is unexpected for Ag-induced T cells. Upon sorting of CD8⁺ T cells isolated from the blood of NHP no. 11 harvested at the time of euthanasia into CD95^low, CD95^int, or CD95^high populations, most of the vector sequences were detected by real-time PCR in the CD95^low population (Fig. 10 B).

The HIV-1 pandemic continues virtually unabated and by now is thought to have caused the death of more than 22 million people. It is estimated that each day an additional 14,000 humans become infected, mainly in developing countries, which cannot afford to provide antiretroviral therapy. For these reasons, an efficacious vaccine to HIV-1 is direly needed to stem the HIV-1 epidemic. Preventative vaccines to most viruses are designed to elicit protective NA responses. Unfortunately, the high diversity and structural plasticity of the HIV-1 envelope protein, its potential for extensive glycosylation, and the structural changes that occur upon receptor binding have defied the development of a vaccine that elicits protective titers of broadly cross-reactive NAs to HIV-1 (15, 16). Vaccine efforts have thus largely concentrated on induction of cellular immune responses to relatively conserved Ags of HIV-1 (17, 18, 19). However, the first large-scale clinical trial of a CD8⁺ T cell-inducing vaccine to HIV-1, the STEP trial, showed that an AdHu5-based HIV vaccine that induced a robust CD8⁺ T cell response to conserved Ags of HIV-1 was not only unable to prevent disease, but may have increased the risk of transmission in individuals with preexisting moderate- to high-titer NAs to AdHu5.

Our work has focused on candidate HIV-1 vaccines based on AdC vectors to which humans in the United States and Asia lack NAs, while low titers of such Abs are detectable in small percentages of humans residing in Central Africa (12). The rationale behind the development of these AdC-based vaccines was to circumvent the reduction of vaccine Ag-specific immune responses in individuals with preexisting NAs to human serotypes of Ad virus such as AdHu5. In the United States, ∼40% of human adults carry NAs to AdHu5, while in some countries of Asia and Africa 80–90% of adults have such Abs (12).

In this study, we tested two AdC vectors, AdC6 and AdC7 expressing gag of HIV-1, in a heterologous prime boost regimen, in comparison to an AdHu5 vector given twice in a homologous prime boost regimen. In both immunization protocols, the animals were divided into naive and AdHu5 preexposed cohorts. The results of this study clearly show that the AdC heterologous prime boost regimen induces higher frequencies of gag-specific T cells in blood and tissues and higher titers of gag-binding Abs in blood compared with a repeated immunization with the AdHu5gag vector. Additionally, preexisting immunity to AdHu5 reduced the Ab and CD8⁺ T cell response to gag presented by the AdHu5 vector without affecting responses to the AdC vectors. Memory CD4⁺ T cell responses have been implicated in previous NHP studies to play a major role in protection against challenge with SIV (20). In our study animals developed higher and more sustained CD4⁺ T cell responses to AdC immunization than upon immunization with AdHu5. We assume that this reflects in part the distinct biology of different Ad vectors that, as we and others reported previously, also affects activation of innate responses (21, 22), which in turn influences activation of the adaptive immune system.

Studies from HIV-1-infected humans have shown that protection against disease progression is not linked to T cell frequencies but to T cell functionality, with T cells producing multiple cytokines being more protective compared with those secreting only one cytokine (23, 24). The role of such subsets in vaccine-induced protection remains unknown. The AdC immunization regimen resulted in a multifaceted T cell response. In two of the three animals tested, most circulating CD8⁺ T cells belonged to effector T cell subsets, while in one animal memory CD8⁺ T cells predominated in blood. Circulating CD8⁺ T cells produced mainly IFN-γ or TNF-α or both. It was remarkable that the profile of cytokine production showed limited variability during the lengthy course of the experiment in individual animals, although there was pronounced variability between animals. In spleens and tissues, most of the NHPs tested had both memory and effector CD8⁺ T cells that secreted mainly IFN-γ or TNF-α or both. IL-2-producing CD8⁺ T cells and CD8⁺ T cells that produced two or three cytokines were more common in memory CD8⁺ T cell subsets than in the more activated effector subset. Circulating CD4⁺ T cells, which belonged to the memory subset, showed a diverse cytokine profile of IFN-γ, IL-2, TNF-α, or combinations thereof in only one NHP. In the other two NHPs the response was more restricted to T cells producing only one of the cytokines at most time points tested. In spleens and livers CD4⁺ T cells belonged mainly to the memory cell subset except for those from spleens of NHP no. 11, which also had high frequencies of effector-like CD4⁺ T cells. As was seen for CD8⁺ T cells, the effector CD4⁺ T cells produced mainly one cytokine (i.e., IL-2), IFN-γ, or TNF-α, while memory CD4⁺ T cells showed a more diverse profile with subsets producing one, two, or three cytokines. CD4⁺ and CD8⁺ T cells from liver and spleen that were positive for two or three cytokines produced significantly higher levels of these cytokines compared with those that were positive for only one cytokine, suggesting that induction of these subsets by a vaccine may be highly desirable. Neither CD8⁺ nor CD4⁺ T cells in blood showed a clear shift in subset distribution or cytokine profiles during the course of immunization, as had been shown previously in mice (25, 26), nor were these parameters affected by preimmunization to AdHu5 virus.

We previously reported that Ad vectors persist in mice and NHPs at low levels, mainly in activated CD8⁺ T cells directed to the Ad-encoded transgene product (14). Again in this study, we could detect Ad vector sequences in all of the animals in liver, spleen, and intestine as well as in lymphocytes derived from these tissues. Upon analysis of T cell subsets we observed in blood and spleens three populations of CD8⁺ T cells producing IFN-γ that could be distinguished by levels of CD95 expression. Activated T cells express high levels of CD95, which facilitates their apoptosis (27); the presence of Ag-induced CD8⁺ T cells expressing only low or intermediate levels of CD95 upon Ad vector vaccination was therefore unexpected. Further analyses showed that vector sequences could mainly be detected in the CD95^low population. The E3 encoded receptor internalization and degradation (RID) complex of Ad viruses can internalize CD95 and facilitate vector persistence (28, 29), and we suggest that the same mechanism may in part explain persistence of E1-deleted Ad vectors. The long-term persistence of the Ad genome that, as we showed previously in mice (14), remains transcriptionally active presumably allows for the exceptionally sustained transgene product-specific T cell responses upon Ad vector immunization. Many of these T cells remain activated at the effector or effector memory stage rather than transitioning into central memory. Whether vaccine-induced effector-like CD8⁺ and CD4⁺ T cells are advantageous to prevent an infection with HIV-1 by providing an immediate layer of defense or are detrimental by providing targets to HIV-1 remains to be investigated.

Vectors based on replication-defective AdHu5 were viewed as the most promising candidates for prevention of HIV-1 infections, as they were found to elicit potent and sustained T cells responses to their transgene product. Although their immunogenicity was attenuated in individuals with preexisting NAs to AdHu5, early clinical trials yielded sufficiently promising results to advance an AdHu5-based HIV-1 vaccine into a phase II trial (STEP trial), designed to assess vaccine efficacy in volunteers with or without NAs to AdHu5 at high risk to contract HIV-1. The vaccine was found to lack efficacy in reducing HIV-1 acquisition or postinfection viral load, which raises the question if the concept of vaccine-induced T cell-mediated protection against HIV-1 infection is flawed. Available evidence argues that this is unlikely. Surrogate challenge models in NHPs show that the solid protection induced by attenuated SIV against challenge with a virulent heterologous SIV is linked to T cell responses (30, 31, 32, 33). Even more importantly, in humans control of HIV-1 infection can be mediated by CD8⁺ T cells (34, 35, 36). One could thus argue that the AdHu5 vaccine used in a homologous prime boost regimen failed by not inducing a potent enough immune response, especially in individuals with preexisting NAs to the vaccine carrier, which may be overcome by heterologous prime boost regimens such as the one described herein.

Although the first highly immunogenic HIV-1 vaccine based on AdHu5 vectors failed to induce protective immunity (37, 38), more innovative vaccine regimens based on different vaccine platforms that allow for heterologous prime boost regimens are expected to induce even more potent and potentially functionally distinct immune responses to HIV-1 that may yet provide T cell-mediated protection against HIV-1 infections of humans.

We thank staff of the flow facilities of the Wistar Institute and University of Pennsylvania for assistance with flow cytometry and cell sorting, Mark Lewis at Bioqual for his assistance with the nonhuman primates, and Christina Cole and Colin Barth (Wistar Institute) for preparation of the manuscript.

The authors have no financial conflicts of interest.