Safety and immunogenicity of the ChAdOx1–MVA-vectored conserved mosaic HIVconsvX candidate T-cell vaccines in HIV-CORE 005.2, an open-label, dose-escalation, first-in-human, phase 1 trial in adults living without HIV-1 in the UK

Summary

Background

An HIV-1 vaccine is long overdue. Although vaccine research focuses on the induction of broadly neutralising antibodies, challenging infections such as HIV-1 could require parallel induction of protective T cells. It is important to recognise that not all T cells contribute to protection equally. Previously, we developed a T-cell immunogen-based bivalent mosaic vaccine, HIVconsvX, delivered by vaccine vectors ChAdOx1 and modified vaccinia Ankara. In this study, we tested the HIVconsvX vaccine regimen for the first time in humans. Other ongoing trials will assess the contribution of the vaccine-induced killer T cells to the control of HIV-1.

Methods

HIV-CORE 005.2 was an open-label, dose-escalation, first-in-human, phase 1 trial done at the Centre for Clinical Vaccinology and Tropical Medicine, University of Oxford, Oxford, UK. Eligible participants were healthy volunteers aged 18–65 years living without HIV-1 and at a low likelihood of acquiring it. Because it was the first administration of ChAdOx1.tHIVconsv1 (C1) to humans, participants were assigned stepwise to two groups. Volunteer group 1 received a low dose of C1 on enrolment. Following a satisfactory safety review 7 days after vaccination, volunteer group 2 received a full dose of C1 boosted by vaccines MVA.tHIVconsv3 (M3) and MVA.tHIVconsv4 (M4) 4 weeks later in regimen C1-M3M4 and were followed up until day 140. Focusing on the full vaccine doses in group 2, the primary outcome was the local and systemic safety of the vaccine. The secondary outcome was the frequency and breadth of epitope recognition by vaccine-induced T cells determined by IFN-γ ELISPOT assay using peripheral blood mononuclear cells (PBMC) at peak (1 and 2 weeks after the M3M4 boost) and at the end of the study, assessed against volunteer’s pre-vaccination levels. The HIV-CORE 005.2 trial is registered at ClinicalTrials.gov (NCT04586673) and is closed.

Findings

Between July 3, 2021, and Aug 3, 2022, 13 participants were recruited and assigned to group 1 (n=3) and group 2 (n=10). Low-dose C1 was safe and well tolerated in group 1, and all three vaccine components were well tolerated in volunteer group 2. There were no serious adverse events. Local and systemic reactogenicities were consistent with intramuscular needle administration of immunogenic substances. All volunteers responded, and their vaccine-elicited T-cell frequencies peaked at a median of 4433 (IQR 2750–5820) IFN-γ spot-forming units per 10⁶ PBMC and recognised a median of 9 (IQR 9-10) peptide pools out of 10, indicating that the responses were broadly specific and each vaccine recipient targeted at least nine epitopes on HIV-1. These frequencies were 7⋅4 times lower by day 140 (ie, 3 months later). T cells proliferated upon antigen re-exposure and displayed multiple effector functions, recognised variant epitopes, and inhibited HIV-1 from the four major global clades A, B, C, and D.

Interpretation

These results inform and support a programme of clinical evaluations of the HIVconsvX T-cell vaccines together with other cutting-edge tools for HIV-1 cure and prevention such as latency reactivating agents, passively infused combinations of broadly neutralising antibodies, and active Env-based vaccines or immunomodulators.

Funding

EU Horizon 2020 Research and Innovation programme, Medical Research Council and Foreign Common-wealth and Development Office Concordat agreement, European and Developing Countries Clinical Trials Partnership, National Institute for Health Research Oxford Biomedical Research Centre, and IAVI.

Introduction

Any package of interventions that could end the HIV-1 epidemic will need to include an effective vaccine. For easy-to-treat microbes, attaining protective levels of antibodies typically correlates with immunity, and is invariably the first tested strategy for vaccines against newly emerging pathogens. However, infections, for which the development of vaccines has been challenging for many years, require a deeper understanding of complex mechanisms leading to protection. The prevention or cure of such diseases might depend on effectively engaging the innate and both humoral and cell-mediated adaptive immune responses.¹

T cells impose selective pressure on HIV-1 and play an important role in controlling HIV-1 infection and clearance.^2–6 Therefore, their effector functions should be harnessed for vaccine development. However, not all T cells are equally protective. In natural infection, initial responses preferentially target highly variable and, therefore, easily escaping epitopes, thus building ineffective memory. Immunodominance limits the number of recognised epitopes and can hinder the development of immunity if the dominant epitopes are non-protective. The ineffectiveness of previous T-cell vaccines in preventing HIV-1 acquisition is generally regarded as being due to design failures rather than an intrinsic inability of fully functional T cells, which target vulnerable epitopes and are at the right place at the right time to slow or stop HIV-1 replication.

Our research aims to contribute to the development of an effective HIV-1 vaccine by finding a strategy for the induction of protective T cells.⁷ Over the past 25 years, we have used iterative improvements of vaccine immunogens and their vector delivery informed by small and rapid phase 1 and 2 clinical trials.^7–10 Our first T-cell immunogen tested in people was called HIVA (derived from African HIV-1 clade A). HIVA was a Gag p24-p17 fusion protein coupled to a string of CD8⁺ T-cell epitopes and was delivered by a DNA prime–poxvirus modified vaccinia virus Ankara (MVA) boost regimen.¹¹ The HIVA vaccines showed only a modest induction of T cells due to the weak DNA prime.

Once the enormous variability of HIV-1 sequences—currently nearly 3 million—fully emerged, we refined the vaccine strategy by focusing elicited T cells on the most vulnerable or functionally conserved sites of HIV-1, which the virus needs to keep unchanged to survive. These sites are sub-protein regions of the HIV-1 proteome with a high degree of similarity among all global variants accumulated in the Los Alamos National Laboratory (LANL) HIV sequence database. Furthermore, escape mutations in the conserved regions are much less common than in the immunodominant, highly variable and, therefore, less protective protein segments targeted initially in natural infection. The first-generation conserved immunogen, HIVconsv, was assembled from 14 conserved regions, whereby the regions used alternating clade consensus amino acid sequences, delivered by regimens combining plasmid DNA, simian (chimpanzee) adenovirus ChAdV63, and MVA.¹² HIVconsv vaccines elicited strong, broadly HIV-1 specific, and polyfunctional T cells capable of cross-clade inhibition of HIV-1 in vitro, which provided a statistically non-significant correlation with the post-antiretroviral treatment control of the virus rebound.¹³ Post-hoc analyses correlated virus remission with pre-existing methylation patterns,¹⁴ the microbiome composition¹⁵ and low levels of plasma CD33.¹⁶

Having lost access to the ChAdV63 vector, we moved to the ChAdOx1–MVA regimen, upgraded the insert to the second-generation bivalent conserved mosaic HIVconsvX immunogens of six regions, and improved epitope matching to the known global group M HIV-1 sequences.¹⁷ The high magnitude and broad specificity of the HIVconsvX-specific CD8⁺ T cells in people living with HIV-1 who have not yet received treatment were associated with good HIV-1 control (low plasma virus load) and preservation of the immune system (high CD4⁺ cell counts), and these correlations were statistically significant.^4,6,17,18 Similar vaccines delivering HIVACAT T-cell Immunogen¹⁹ by DNA, ChAdOx1, and MVA improved control of virus rebound by 32% compared with placebo in people living with HIV-1 without protective HLA alleles.²⁰ For HIVconsvX, the potent induction of T-cell responses in a small animal model has been comprehensively shown and characterised. Here, we describe for the first time the safety and immunogenicity of the HIVconsvX vaccines in a phase 1 trial in healthy adults living without HIV-1 in the UK.

Methods

Study and vaccine design

HIV-CORE 005.2 (NCT04586673) was a single-centre, open-label, dose-escalation, first-in-human phase 1 clinical trial testing the safety and immunogenicity of the candidate HIVconsvX HIV-1 vaccines. This trial was conducted at the Centre for Clinical Vaccinology and Tropical Medicine, University of Oxford, Oxford, UK. The trial was sponsored by the University of Oxford and was granted a favourable opinion by the East of England–Cambridge Central Research Ethics Committee (20/EE/0036). The study was approved by the UK Medicines and Healthcare Products Regulatory Agency (CTA 21584/0425/001-0001; EudraCT 2019-003973-25; IRAS 263882). The trial was conducted according to the principles of the 2008 Declaration of Helsinki and complied with the Good Clinical Practice guidelines developed by the International Council on Harmonisation. The clinical trial protocol (CTP) is provided in the appendix (p 30).

A family of chimeric proteins collectively called HIVconsvX (figure 1A) represent the second generation of conserved T-cell immunogens and overall the third generation of vaccines tested in humans developed by our team.^11,12 Two regions in Gag, including full-length p24, and four regions in Pol, were selected purely based on the sequence conservation without any consideration for the total number of known epitopes or the HLA coverage. Env was intentionally excluded because of the absence of sufficiently large regions associated with good virus control. The advantages of using protein regions instead of epitopes are that vaccine designers are not limited by the incomplete knowledge of epitopes and HLAs, and, at the same time, the regions are sufficiently large to contain multiple epitopes for every individual of any HLA haplotype to induce several T-cell specificities. The amino acid sequences were first computed for the whole Gag and Pol proteins as a bivalent mosaic, which differed by approximately 8% of amino acids²¹ and between them attained a perfect match to 80% of all potential T-cell epitope (PTE) variants found among group M HIV-1. The regional boundaries were adjusted to maximise the inclusion of protective epitopes identified in cohorts of individuals who had not yet received treatment in Spain, Peru, and South Africa associated with low plasma virus load.⁴ The regions were uniquely ordered to minimise the induction of T cells against irrelevant junctional neoepitopes. Genes coding for the HIVconsv1, HIVconsv3, HIVconsv4, and HIVconsv62 immunogens were inserted into replication-deficient vaccine vectors ChAdOx1 and MVA and delivered in a prime-boost regimen with demonstrated high in-human immunogenicity. The vaccine regimen consists of ChAdOx1.tHIVconsv1 (C1) plus ChA-dOx1.HIVconsv62 (C62) prime and MVA.tHIVconsv3 (M3) plus MVA.tHIVconsv4 (M4) boost, which is abbreviated to C1C62-M3M4.¹⁷ The C1-M3M4 regimen was tested in this study because of the delay in the C62 manufacture.

Figure 1. Vaccines and the regimen.

(A) Alignments of curated global HIV-1 protein sequences, present in the Los Alamos National Laboratory HIV Sequence Database as of around September, 2013, were used to compute bivalent mosaic sequences of the entire Gag and Pol HIV-1 proteins and define highly conserved regions. Induction of T cells against irrelevant junctional neoepitopes was minimised by reassorting the six regions into six unique orders that halved the dose of each junction per administration, whereas no junction in the priming dose would be boosted by the same junction in the boost. Arbitrarily, HIVconsvX genes were inserted into the simian adenovirus or poxvirus vectors to generate vaccines ChAdOx1.tHIVconsv1 (C1), ChAdOx1.HIVconsv62 (C62, not tested in this study), MVA.tHIVconsv3 (M3) and MVA.tHIVconsv4 (M4). t in the immunogen name indicates human tissue plasminogen activator leader sequence and Pk indicates C-terminal Pk (also known as SV5) tag recognised by monoclonal antibodies. (B) Trial design. In group 1 (n=3), volunteers received one-tenth of the standard adult dose of C1 and were carefully monitored for local and systemic reactogenicity before progressing to group 2 (n=10), who received the full adult doses of C1-M3M4. The timing of fresh IFN-γ ELISPOT assay is indicated by red drops below. Both groups were followed up for 112 days after vaccination. The schematic for vaccine derivation was modified from Beavis and colleagues.⁵⁰

Participants

Healthy adult male participants and female participants who were not pregnant were recruited if they were living without HIV-1 or HIV-2, at low likelihood of acquiring HIV-1, aged 18–65 years, fully comprehended the purpose and details of this study as provided in the participant information sheet, and able to give written informed consent. Eligibility depended on the results of laboratory tests, medical history review, physical exam results, and answers to questions about behaviours that could increase the chance of acquiring HIV-1. For COVID-19 vaccinations, participants must not have received a simian adenovirus-based COVID-19 or experimental vaccine before and during the trial or must have received their UK National Health Service COVID-19 vaccine offered at least 3 months after C1. There was no pre-selection of volunteers based on pre-existing neutralising antibodies to human adenovirus serotype 5 or MVA.

Assignment and masking

This was an open-label study. For the C1 vaccine component, HIV-CORE 005.2 was the first-in-human study. Volunteers were recruited sequentially and stepwise into two groups. The first volunteer in group 1 received a single low dose of 5 × 10⁹ virus particles of the vaccine ChAdOx1.tHIVconsv1 (C1) at enrolment. The other two volunteers were vaccinated at the same low dose at least 48 h after the vaccination, after the safety assessment of the first volunteer, and at least 1 h apart from each other. Following a satisfactory safety review of group 1 7 days after the C1 administration, ten new participants were enrolled in group 2, who received the standard adult dose of 5 × 10¹⁰ virus particles of C1 divided into two arms followed by a boost vaccination on day 28 with M3 and M4 at doses of 1× 10⁸ plaque-forming units (PFU) and 0⋅9× 10⁸ PFU, respectively (C1-M3M4). In group 2, the second volunteer only received the full dose of C1 at least 48 h after the first volunteer received the vaccine, and after a safety assessment of the first volunteer. The standard vaccine doses were determined and used in previous simian adenovirus and poxvirus-vectored vaccine trials.^7,8 The follow-up period was 112 days (figure 1B). The fresh ELISPOT and virus inhibition assays (VIA) were performed at two different laboratories using computer-generated volunteers’ identification numbers, which meant that the staff at the VIA laboratory were masked to the ELISPOT results.

Outcomes

The primary outcome was the vaccine safety assessment, measured by the proportion of volunteers with vaccine-related serious adverse events (SAEs) collected up to day 140 after enrolment, the proportion of volunteers with grade 3 or 4 unsolicited adverse events through 28 days after final vaccination, and local and systemic reactogenicity after up to 7 days after each vaccination. The severity of clinical and laboratory adverse events was assessed according to the scales in the Divisions of AIDS Table for Grading the Severity of Adult and Paediatric Adverse Events (corrected version 2.1 published in July, 2017). The secondary outcome was the vaccine immunogenicity in group 2, measured using the IFN-γ ELISPOT assay and VIA using eight infectious molecular clones derived from HIV-1’s representative of the four major global clades and engineered to express the Renilla reniformis luciferase reporter. For exploratory assays, the functionality of vaccine-elicited T cells was assessed using polychromatic flow cytometry. Gut microbiome composition and richness were assessed in group 2 and will be reported separately.

Procedures

In group 1, seven study visits for safety evaluations (the primary outcome) occurred over 112 days: screening (no longer than 42 days before the first vaccination), day 0 (D0; C1 low-dose vaccination), day 1 (D1), day 7 (D7), day 14 (D14; phone call only), day 28 (D28), and day 112 (D112). In group 2, twelve visits for safety evaluations and immunological analyses occurred over 140 days: screening (no longer than 42 days before the first vaccination), D0 (C1 standard-dose vaccination), D1, D7, D14 (phone call only), D28, (M3 and M4 vaccination), day 29 (D29), day 35 (D35), day 42 (D42), day 56 (D56), day 84 (D84), and day 140 (D140). Safety data included specified, solicited symptoms collected by diary cards during the 6 days after each vaccination; unsolicited adverse events collected up to 28 days after each vaccination; and SAEs collected until the end of the study (D140). All in-person visits occurred at the Centre for Clinical Vaccinology and Tropical Medicine, University of Oxford, Oxford, UK. At selected study visits, samples of venous blood were taken to evaluate biochemical or haematological parameters, or both (for volumes and parameters, see the CTP in the appendix [p 30]).

For the secondary and exploratory outcomes, group 2 was assessed for the C1-M3M4 regimen immunogenicity. Peripheral blood mononuclear cell (PBMC) samples were collected before vaccination (D0), at the peak of immune responses (D35 and D42), and at the end of the study (D140). Voluntary faecal samples (D0 and D35) were collected in group 2.

Statistical analysis

Analyses of variance were performed using Prism 10 (GraphPad). ELISPOT assay results were assumed to be non-Gaussian in distribution, so non-parametric tests were used throughout, and medians (IQR range) are shown. The Mann–Whitney U test was used for unpaired analyses. For paired analyses, either Wilcoxon matched-pairs signed ranktestor Friedmantest, multiple comparisons were used. Two-tailed p-values were used, and a p-value of less than 0⋅05 was considered statistically significant.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

The trial occurred between July 3, 2021, and August 3, 2022. 20 individuals were screened for eligibility and 13 (65%) were recruited stepwise into group 1 (n=3) and group 2 (n=10). Of the seven individuals who were excluded, four were excluded based on medicalhistory, one self-reported as being behaviourally vulnerable to HIV-1, and two had previously received a recombinant simian adenoviral vaccine. Volunteers’ median age at the first vaccination was 30 years (IQR 28–38). Eight (62%) participants were male, five (38%) participants were female, and all (100%) participants were White British (table 1). No participants withdrew before the end of the trial and all enrolled participants completed the follow-up (figure 2). The booster dose of participant 5201018 was delayed by 20 weeks due to COVID-19 and Christmas.

Table 1. Demographics.

	Group 1  (n=3)	Group 2  (n=10)	Total (n=13)
Sex
Male	2 (67%)	6 (60%)	8 (62%)
Female	1 (33%)	4 (40%)	5 (38%)
Ethnicity
White British	3 (100%)	10 (100%)	13 (100%)
Age, years
21‒30	1 (33%)	6 (60%)	7 (54%)
31‒40	2 (67%)	2 (20%)	4 (31%)
51‒60	0	1 (10%)	1 (8%)
61‒65	0	1 (10%)	1 (8%)

Data are n (%). Ethnicity data were collected by self-reported questionnaire.

Figure 2. CONSORT diagram.

C1=ChAdOx1.tHIVconsv1. C1-M3M4=full dose of C1 boosted by vaccines MVA.tHIVconsv3 and MVA.tHIVconsv4.

All three vaccines (C1, M3, and M4) were well tolerated. There were no SAEs. The most frequently reported adverse events were those typically seen in response to virus-vectored vaccines, including pain at the injection site, headache, fatigue, myalgia, nausea, and general malaise. A total of 134 solicited local and systemic adverse events were reported: 62 after the C1 prime (12 in group 1 and 50 in group 2) and 72 after the M3M4 boost (table 2). In group 2, 113 (93%) adverse events were mild or moderate and resolved within 72 h. All severe solicited adverse events occurred in group 2, and C1 was more reactogenic (causing seven SAEs in five participants) than the co-administered M3 and M4 (causing two SAEs in two participants). All nine SAEs in group 2 occurred after vaccinations and included feverishness, headache, nausea, malaise, and redness.

Table 2. Solicited adverse events over 6 days after vaccine administration.

	C1 (n=10) group 2				M3M4 (n=10) group 2
	Mild	Moderate	Severe		Mild	Moderate	Severe
Local reactogenicity
Warmth	0	0	3		5	0	0
Itch	6	0	0		4	6	0
Pain	1	0	0		1	0	0
Redness	1	0	0		5	0	0
Systemic solicited adverse events
Temperature	2	0	0		1	0	0
Arthralgia	3	1	0		6	1	0
Myalgia	6	2	0		3	4	0
Feverishness	3	3	0		6	1	1
Headache	6	1	1		7	0	1
Fatigue	5	4	0		6	2	0
Nausea	2	2	2		4	1	0
Malaise	4	3	1		4	3	0

C1=ChAdOx1.tHIVconsv1. M3=MVA.tHIVconsv3. M4=MVA.tHIVconsv4.

Unsolicited adverse events considered related to vaccinations included palpitations, gastrointestinal symptoms, and rhinitis and occurred in group 2. These unsolicited adverse events were moderate to severe but short-lived and all resolved within 48 h from onset (appendix p 24). A transient reduction in total lymphocyte count was observed within 24 h of administration of the prime and boost vaccines in all volunteers; six episodes (three after C1 regular dose and three after M3M4) were classified as severe (0⋅2–0⋅5× 10⁹ cells per L). The number of total lymphocytes dropped by a mean of 53% (SD 18) between D0 and D1, and by a mean of 61% (SD 8) between D28 and D29 (appendix p 4). These changes were not associated with any symptoms and the lymphocyte counts returned to normal by the following visit in all volunteers. The Data Management and Ethics Committee assessed that the changes in total lymphocyte count were not clinically significant. No other severe or systemic laboratory abnormality was detected.

The secondary trial objective evaluated vaccine immunogenicity in group 2. The frequencies of HIVconsvX-specific T cells among PBMC were determined before the vaccination on D0, at 1 and 2 weeks after the boost on D35 and D42, respectively, to assess the peak responses and at the end of the clinical phase on D140. A fresh ex-vivo IFN-γ ELISPOT assay using ten peptide pools (P1–P10) of 15-mer peptides overlapping by 11 amino acids (appendix p 25) enumerated vaccine-elicited T cells. The peptides corresponded to both mosaic 1 (HIVconsv1 of C1 and HIVconsv3 of M3) and mosaic 2 (HIVconsv4 of M4) and did not cross the regional junctions (appendix p 25). The C1-M3M4 regimen induced robust HIVconsvX-specific T cells in all ten (100%) volunteers who received the vaccine, the specific T-cell frequencies peaked at a median of 4433 (IQR 2750–5820) IFN-γ spot-forming units (SFU) per 10⁶ PBMC and recognised a median of 9 (9–10) peptide pools out of 10 (figure 3A–C; appendix p 5). The ex-vivo IFN-γ frequencies were, on average, 7⋅4 times lower by D140. Responses to a pool of 31 junctional peptides (appendix p 28) were a median of 77 (30–105) SFU per 10⁶ PBMC, corresponding to approximately 1⋅7% of the total HIVconsvX-specific response. Thus, HIVconsvX vaccines in the C1-M3M4 regimen induced strong and broad T-cell responses.

Figure 3. Strong and broad T-cell responses.

In HIV-CORE 005.2 group 2, ten healthy adults living without HIV-1 who were not behaviourally vulnerable to HIV-1 in Oxford, UK, received the C1 and M3M4 vaccines on D0 and D28, respectively. Vaccine-elicited T-cell responses were enumerated in an IFN-γ ELISPOT assay using partially overlapping HIVconsvX peptide pools P1–P10 (appendix p 25). (A) The total magnitudes are depicted at each timepoint with dotted lines connecting the timepoints of individual participants and a solid line showing the group median. The horizontal dotted line indicates pre-vaccination total HIVconsvX-specific frequencies of T cells plus three SDs. (B) Number of positive pools (net >50 SFU per 10⁶ PBMC) that participants responded to at each timepoint. (C) Responses to each HIVconsvX pool on D35 are shown as median, range, and IQR. (D) Responses to a single pool of 31 junctional peptides covering seven amino acids from each end of adjacent regions subtracting D0 from D35 for each volunteer shown on the y axis. In (A) and (B), the adjusted p values above are shown for the Friedman test with multiple comparisons. C1=ChAdOx1.tHIVconsv1. D0=day 0. D28=day 28. D35=day 35. D42=day 42. D140=day 140. M3=MVA.tHIVconsv3. M4=MVA.tHIVconsv4. PBMC=peripheral blood mononuclear cell. SFU=spot-forming units.

To estimatethe ability of T cells to proliferate and function, and to map stimulatory peptide pairs within each pool, frozen PBMC were thawed, cultured with individual pools for 10 days, and tested in IFN-γ ELISPOT assay against deconvoluted peptide pairs (for methods see appendix p 2). All volunteers showed responses greater than or equal to 750 SFU per 10⁶ short-term cell line cells to a median 17⋅5 (IQR 5⋅0–18⋅0) and mean 17⋅5 (SD 18⋅0) peptide pairs (appendix p 6). These numbers of recognised peptide pairs overestimate the actual number of targeted epitopes, because some epitopes can be shared between overlapping peptide pairs.

For 22 strong epitopes, we also assessed the degree of epitope variant recognition by predicting minimal epitopes in 15-mer peptide pairs based on the recipient’s HLA type using the Immune Epitope Database. For these predicted minimal epitopes, peptides corresponding to variants in the LANL HIV sequence database at frequencies greater than or equal to 1% were tested for recognition by unexpanded ex-vivo PBMC of D35. The depth varied for epitopes and volunteers, but most outcomes indicated reactivity to epitope variants beyond those present in the vaccine, especially if both mosaic 1 and 2 vaccine variant peptides were recognised, pointing to the benefits of the bivalent design (appendix p 7). Thus, recipients of HIVconsvX recognised multiple epitopes and their variants.

Next, frozen PBMC of D35 were thawed and stimulated with personalised pools based on results using short-term cell lines (appendix p 6) to separate reactive CD4⁺ and CD8⁺ T cells and assess their polyfunctionality in an intracellular cytokine staining assay. Specifically, vaccine-elicited T cells were assessed for production of IFN-γ, TNF-α, IL-2, CCL3 (also known as MIP-1α), and degranulation as judged by surface expression of CD107a known as LAMP-1. Overall, reactive CD8⁺ T-cell populations were polyfunctional with individual cells displaying one, two, three, and four functions at a median of 60% (IQR 41–67), 20% (15–26), 15% (13–22) and 4% (3–7) of total CD8⁺ cells, respectively, whereas CD4⁺ T cells showed the same respective functions in a median of 67% (60–86), 15% (14–34), 17% (9–23), and 0 of HIVconsvX-specific CD4⁺ T cells (figure 4A). Furthermore, for each volunteer, the strongest epitope variant of the mosaic pair for functional stimulation was identified for each volunteer’s ten strongest peptide pairs (appendix p 9).

Figure 4. Functionality and memory structure of vaccine-elicited CD8⁺ and CD4⁺ T cells.

Frozen PBMC samples were thawed, restimulated with personalised pools of stimulatory peptides overnight, and analysed in polychromatic flow cytometry. (A) Functionality. Group 2 median values of D35 (n=10) are shown for the proportions of multifunctional T-cell subpopulations (centre, scales of grey) with the total frequency as a percentage of all CD8⁺ and CD4⁺ cells given in the middle. Produced cytokines and chemokines tested are colour-coded (periphery; see appendix p 8 for gating). (B) For the same data, the sizes of functional subpopulations are shown for individual volunteers, cytokines, and chemokines, and for CD4⁺ and CD8⁺ T cells. (C) Memory subpopulations were defined as T_EFF (effector; CD45RA^PosCCR7^LoCD27^Pos), T_TD (terminally differentiated; CD45RA^PosCCR7^LoCD27^Neg), T_TEMRA (terminal effector memory; CD45RA^PosCCR7^HiCD27^Pos), T_CM (central memory; CD45RA^NegCCR7^HiCD27^Pos), T_TM (transitional memory; CD45RA^NegCCR7^LoCD27^Pos), and T_EM (effector memory; CD45RA^NegCCR7^LoCD27^Neg). Gating is shown in the appendix (p 20). D35=day 35. D140=day 140. PBMC=peripheral blood mononuclear cells.

For each volunteer in group 2, personalised peptide pools were also used to reveal the overall memory of structures of vaccine-elicited T cells (figure 4C; appendix p 20). There were clear distinctions between the CD8⁺ and CD4⁺ sub-populations producing IFN-γ and TNF-α, whereby CD4⁺ cells had proportionally expanded terminal effector memory (T_TEMRA), transitional memory (T_TM), and central memory (T_CM) T cells. By contrast, CD8⁺ cells developed larger effector (T_EFF) and terminally differentiated (T_TD) T-cell subpopulations (figure 4). The memory compositions changed over 6 months with noticeable relative expansions of CD8⁺ T_EFF and CD4⁺ T_TM cells.

Luciferase-based VIA measures the in-vitro ability of CD8⁺ T-cell effectors to inhibit HIV-1 replication in autologous CD4⁺ cells and was a secondary trial analysis. Replication of infectious molecular clones (IMCs) derived from HIV-1 isolates results in the production of luciferase enzyme while maintaining physiological levels and functions of Nef. In this study, we used eight previously characterised IMCs derived mostly from transmitted/founder HIV-1s isolated in different countries between 1983 and 2010 to represent the major global clades A, B, C, and D (appendix p 29). These IMCs were selected for not being overly sensitive to CD8⁺ T-cell inhibition ranging in log₁₀ relative light units (RLU) reduction values from 0⋅65 to 1⋅25.²² No outstanding intrinsic difference among the relative abilities of these IMCs to grow in the CD4⁺ cells of volunteers was detected (appendix p 21). Table 3 displays the log₁₀ reduction in RLU values in co-cultures of infected autologous CD4⁺ and CD8⁺ T cells compared with infected CD4⁺ T cells alone and the number of IMCs inhibited per volunteer at D35 and D140 after vaccination compared with D0 before vaccination. Overall, HIVconsvX vaccination induced CD8⁺ T cells, which inhibited a mean of 7⋅8 (SD 0⋅4) IMCs at peak responses and a mean of 6⋅4 (2⋅7) 3 months later (D140). Thus, in this in-vitro assay, vaccine-induced effectors displayed broad cross-clade HIV-1 inhib-ition. Alignment of the IMCs’ corresponding Gag and Pol sequences with the HIVconsvX regions showed a high-degree match endorsing the vaccine rationale (appendix p 22).

Table 3. Broad inhibition of representative viruses of the four major global HIV-1 clades A, B, C, and D in each participant distinguished by identification number on D0, D35, and D140.

	Log10 RLU reduction for CD4+ and CD8+ T-cell co-cultures compared with CD4+ T cells alone									Number of IMCs inhibited  per participant
	R.175019 (clade A)	R.175090 (clade A)	UG.191947 (clade A/D)	NL4-3 (clade B)	CH077  (clade B)	Z.235219  (clade C)	Z.235092  (clade C)	UG.191882 (clade D)		D35 (≥0.1 log10; ** and ***)	D35 (≥0.1 log10; ** and ***)
5201002
D0	0.76	0.96	0.85	0.77	0.86	0.81	0.86	0.61
D35	1.43**	1.43**	1.12**	1.15**	1.65**	1.02**	1.19**	1.13**		8
D140	1.08**	1.19**	1.00**	0.93**	1.12**	0.86*	1.01**	0.83**			7
5201005
D0	0.73	0.69	0.75	0.69	0.55	0.51	0.63	0.57
D35	1.17**	1.30**	1.13**	1.28**	1.02**	0.98**	1.07**	1.21**		8
D140	1.13**	1.19**	1.09**	1.23**	0.96**	1.02**	1.10**	1.15**			8
5201007
D0	0.85	0.79	0.76	0.68	0.78	0.85	0.53	0.79
D35	1.48**	1.31**	1.48**	1.85***	1.35**	1.15**	1.08**	1.58**		8
D140	1.17**	0.87**	1.06**	1.29**	0.97**	0.93*	0.76**	1.12**			7
5201008
D0	0.21	0.54	0.42	0.41	0.52	0.30	0.53	0.54
D35	0.43**	0.82**	0.77**	0.90**	0.51	0.63**	0.95**	0.87**		7
D140	0.32**	0.59*	0.47**	0.60**	0.49	0.32*	0.54*	0.58*			3
5201010
D0	0.72	0.82	0.73	0.73	0.77	0.53	0.61	0.61
D35	0.93**	1.05**	0.96**	1.63**	1.21**	0.90**	3.18***	1.27**		8
D140	0.82**	0.89*	0.81**	1.33**	0.96**	0.69**	1.93***	1.00**			8
5201013
D0	0.68	0.60	0.65	0.60	0.51	0.62	0.54	0.49
D35	0.86**	0.85**	0.80**	0.65*	0.71**	0.73**	0.81**	0.60**		7
D140	0.51	0.55	0.53	0.47	0.45	0.51	0.52	0.42			0
5201014
D0	0.49	0.63	0.69	0.54	0.65	0.45	0.72	0.49
D35	1.29**	1.03**	1.91***	1.29**	1.09**	0.88**	1.09**	1.36**		8
D140	0.95**	0.73**	1.68**	0.99**	0.77**	0.68**	0.93**	0.97**			8
5201017
D0	0.78	0.62	0.71	0.60	0.72	0.59	0.62	0.31
D35	1.04**	0.90**	1.00**	0.71**	0.90**	0.70**	0.78**	0.57**		8
D140	0.98**	0.72**	0.86**	0.53	0.93**	0.67*	0.73**	0.48**			7
5201018
D0	0.55	0.36	0.34	0.45	0.49	0.15	0.22	0.31
D35	0.72**	0.59**	0.70**	0.84**	0.67**	0.33**	0.47**	1.00**		8
D140	1.00**	0.76**	0.75**	0.85**	0.74**	0.56**	0.49**	0.97**			8
5201020
D0	0.45	0.38	0.38	0.38	0.39	0.27	0.28	0.24
D35	1.48***	1.45***	1.12**	1.38***	1.24**	0.54**	1.08**	1.09**		8
D140	0.93**	0.97**	0.72**	0.80**	0.76**	0.44**	0.66**	0.72**			8
Median (IQR)
D0	0.70 (0.49-0.76)	0.63 (0.54-0.79)	0.70 (0.42-0.75)	0.60 (0.45-0.69)	0.60 (0.51-0.77)	0.52 (0.30-0.62)	0.58 (0.53-0.63)	0.52 (0.31-0.61)
D35	1.11 (0.86-1.43)	1.04 (0.85-1.31)	1.06 (0.80-1.13)	1.22 (0.84-1.38)	1.06 (0.71-1.24)	0.81 (0.63-0.98)	1.08 (0.81-1.09)	1.11 (0.87-1.27)		8.0 (8.0-8.0)
D140	1.0 (0.82-1.08)	0.82 (0.71-0.97)	0.84 (0.72-1.06)	0.89 (0.60-1.23)	0.85 (0.74-0.96)	0.68 (0.51-0.86)	0.75 (0.54-1.01)	0.90 (0.58-1.00)			7.5 (7.0-8.0)

Increase in inhibition relative to pre-vaccination. *** indicates ≥1.0 log₁₀. ** indicates <1.0 log₁₀ to ≥0.1 log₁₀. * indicates <0.1 log₁₀, and no asterisk indicates no increase. D0=day 0. D35=day 35. D140=day 140. IMC=infectious molecular clone. RLU=relative light units.

Discussion

HIV-1 diversity is the biggest obstacle to developing an effective HIV-1 vaccine. Targeting the most conserved regions of the HIV-1 proteome by killer T cells could provide a substantial advantage for slowing or stopping viral replication, whereby a vaccine design based on regions rather than epitopes or amino acids alleviates the incomplete knowledge of defined epitopes and HLAs. In HIV-CORE 005.2, vaccines that used computed conserved bivalent mosaic of HIVconsvX delivered by ChAdOx1-MVA regimen C1-M3M4¹⁷ were tested in healthy adults living without HIV-1 in the UK. All vaccine components were well tolerated and together induced IFN-γ-producing T cells with a median of over 4000 SFU per 10⁶ PBMC, which were broadly HIV-1 specific, recognised epitope variants beyond those present in the immunogens, were polyfunctional, and were capable of proliferation followed by a display of effector functions. CD8⁺ T-cell effectors inhibited in-vitro HIV-1 IMCs representative of four major global clades (A, B, C, and D). These findings show that a highly rational vaccine design can induce robust responses to important albeit naturally subdominant epitopes.

To date, preventive efficacy trials have not succeeded in reducing HIV-1 acquisition except for RV144, of which the marginal protection of 31⋅2% (p=0⋅04) remains unconfirmed.²³ The first strength of our results is that all HIV-CORE 005.2 volunteers responded to the HIVconsvX vaccination with T-cell responses comparing favourably with those induced by previous vaccine candidates. It is possible that, had the previously tested vaccines induced T-cell responses akin to those elicited by the HIVconsvX vaccines along with anti-Env antibodies, the past trial outcomes might have been more encouraging. Although testing of the true protective efficacy of the HIVconsvX vaccine candidates is ongoing, based on our hypothesis for induction of protective killer T cells, the HIVconsvX vaccines use more rational immunogens and more potent heterologous vector delivery than the STEP²⁴ and Mosaico²⁵ study vaccines. HIVconsvX conserved regions do not waste responses on highly variable parts of full-size proteins and the use of bivalent mosaic is much more efficient if combined with regions of already low entropy; a bivalent mosaic alone cannot sufficiently compensate the variant PTE mismatch of full-size Gag-Pol-Env.²⁶ As for the vector delivery, repeated administrations of the same adenovirus vector decrease the vaccine intake compared with a heterologous prime–boost regimen, reflected mainly in the total vaccine-induced magnitude.

Targeting multiple HIV-1 epitopes at the same time is thought to be one of the prerequisites to successful HIV-1 control, which previously correlated with successful VIA.²⁷ Another strength of the present study is that recipients of the full C1-M3M4 vaccine regimen in group 2 responded to nine peptide pools, which upon in-vitro expansion deconvoluted to up to 18 peptide pairs per person across the HIVconsvX immunogen, of which likely half correspond to CD8⁺ T cells. In the STEP study, the small number of participants who seroconverted and developed a response against three or more Gag epitopes displayed lower plasma virus loads compared with those with a narrower response.²⁸ In the HVTN 054 trial,²⁹ volunteers responded to a median of three epitopes, whereas in the AELIX 002 trial,²⁰ people living with HIV-1 received a heterologous prime-boost regimen and recognised five epitopes. Immunodominance always narrows the vaccine-elicited T-cell responses, and vaccine-elicited epitopes inevitably establish a new immunodominance hierarchy in every individual. In this study, the strength is that HIVconsvX epitopes are conserved and, therefore, all likely contribute to protection.^4,6,30 Thus, HIV-consvX epitopes are not over-competed by responses to immunodominant hypervariable and less protective epitopes such as those typically recognised early during acute infection.^31,32

VIA quantifies collective protective functions of CD8⁺ T cells leadingto HIV-1 growth inhibition and can do so for a panel of viruses originating from diverse geographical regions, types (founder/transmitter vs lab-adapted), and tropisms (CCR5 vs CXCR4). The main contribution to inhibition is most likely the direct contact-dependent killing of virus-producing cells triggered by recognition of the peptide/HLA complex by T-cell receptor. This is combined with less specific mechanisms such as Fas and Fas ligand interaction,^33,34 and the release of soluble CXCR4-engaging and CCR5-engaging chemokines or non-lytic mechanisms, which worked through a transwell membrane.³⁵ A further strength of our results is that the HIVconsvX vaccination induced in most volunteers peak CD8⁺ T-cell effectors that suppressed all eight tested HIV-1 IMCs, albeitthisinhibition decreased by the end of the study 3 months later. For comparison, we previously assessed the ability of CD8⁺ T cells from 13 people living with HIV-1 who were naive to anti-retroviral treatment and had a range of plasma viral loads but were otherwise healthy at the time of sample donation to inhibit replication of 35 separate IMCs in VIA.³⁶ In that study, IMCs were inhibited by a range of log₁₀ RLU reduction values with an overall median of 1⋅07. A limitation of these results is that the translation of VIA into in-vivo HIV-1 control is not straightforward. However, the assay provides a useful indication of HIV-1 growth suppression in relative terms among different timepoints, isolates, vaccine recipients, and vaccine candidates, and supports cross-clade recognition by T cells induced by the HIVconsvX vaccination. Several publications argued for the correlation of VIA with in-vivo HIV-1 control.^33,34,37,38 Also notable was the ability of the STEP’s MRKAd5 vaccine to only suppress a matched virus.³⁹

Several T-cell features detected in this study were previously associated with protection, including proliferation, polyfunctionality, and the ability to recognise transmitted or founder viruses and target multiple conserved epitopes. In other studies, virus control was associated with high functional T-cell avidity,^40–44 although lower affinity clonotypes might provide a basis for cross-reactivity and escape control,⁴⁵ efficient killing of infected cells,^40,46,47 and production of multiple soluble antiviral factors.^40,47 Prophylaxis might differ from post-treatment control, requiring access to privileged sites of HIV-1 replication. A clear limitation of our work is the sole focus on PBMC, whereas the crucial surveillance and killing of HIV-1-infected cells occur in the lymphoid organs and tissues (eg, gut).⁴⁸ Furthermore, CD8⁺ T cells need optimal help from CD4⁺ T cells. Note that conserved HLA class II epitopes are equally well represented in the conserved regions.⁴⁹ Although the characteristics of T cells protecting against HIV-1 remain debated, we believe the HIVconsvX vaccines in HIV-CORE 005.2 might have induced at least some of them.

We are also aware of the HIV-CORE 005.2 limitations, which stem from the small sample size, which opens the possibility of a biased, well responding but atypical population, and the open-label design, which does not allow for masking the staff at each laboratory other than between laboratories.

Nevertheless, upgraded conserved mosaic T-cell candidate vaccines against HIV-1 were tested for the first time in humans, and their safety and immunogenicity encourage further development. The next crucial step inevitably requires validation of human protective efficacy. Initial analytical treatment interruption studies of the vaccination alone or together with other experimental strategies for HIV-1 cure could eventually guide the design of more expensive advanced prophylactic trials.

Research in context.

Evidence before this study

Vaccine-elicited T cells must act quickly and decisively to maximise the likelihood of early HIV-1 containment. To control HIV-1, T cells must recognise HIV-1-derived peptides, proliferate and reach protective numbers, target epitopes limited in their entropy and closely matching circulating viruses, exert protective functions, and be present in the right place at the right time. In the absence of simple, functional correlates of T-cell protection, vaccines should aim to imprint as many of the important features on T cells as possible, considering that any one being suboptimal could cause a failure of the whole vaccination strategy. Past failures of candidate vaccines in preventive phase 3 efficacy trials do not prove that an effective vaccine against HIV-1 is impossible or that T cells cannot protect against infection. None of the vaccine candidates tested to date for the prevention of HIV-1 acquisition used stabilised Env trimers, whether germline-targeting or ontogeny-guided, for the induction of antibodies. Similarly, better rationally designed immunogens for CD8⁺ T cells can specifically target vulnerable parts of HIV-1, and these can be delivered in more potent heterologous prime–boost regimens than those used to date. Reports on efficacy trials testing experimental interventions to prevent HIV-1 acquisition published on PubMed as of May 23, 2024, were considered.

Added value of this study

The central paradigm of the HIVconsvX T-cell vaccine strategy is focusing T cells on the functionally conserved regions of the HIV-1 Gag and Pol proteins, the potential T-cell epitopes of which are shared among 80% of global isolates and harbour fewer escape mutations. At the epitope level, the remaining variability within these regions is addressed computationally using the mosaic algorithm. The bivalent mosaic used in this study provides an additional advantage should a conserved vaccine be deployed across diverse geographical regions, but bivalency alone is not sufficient to tackle escape mutations in all highly variable and typically dominant regions of the full-length proteins, especially in those of Env. The results of HIV-CORE 005.2 support the claim that sub-dominant and, therefore, underused T-cell epitopes taken out of the context of the whole virus or full-length viral proteins can induce robust and broad T-cell responses when delivered by a strong, heterologous regimen such as ChAdOx1–MVA.

Implications of all the available evidence

Progress towards an effective HIV-1 vaccine has been slow and full of obstacles. However, the systematic development of vaccine components for neutralising antibodies and effective T cells, guided by human data, is starting to show encouraging results, supporting the field’s overall direction. In this study, we show that HIVconsvX-specific T cells can help slow and control diverse HIV-1’s in the laboratory. Cure studies with analytical treatment interruption in people living with HIV-1 are currently underway, assessing the HIVconsvX vaccine ability in combination with other anti-HIV-1 strategies to slow, delay, or control the virus rebound after stopping combination antiretroviral therapy. These studies are being conducted in collaboration with other partners and global networks.

Data sharing

De-identified participant data will be made available upon requests directed to the chief investigators PC and TH. Proposals will be reviewed and approved by the sponsor, chief investigators, and collaborators based on scientific merit. After approval of a proposal, data can be shared through a secure online platform after signing a data access agreement.