This study was approved by the Institutional Review Board (IRB) of Johns Hopkins University. Written informed consent was obtained from all participants prior to their enrollment. The study cohort included 13 people with HIV who were started on suppressive antiretroviral therapy during the chronic phase of infection (chronic progressors or CPs) and 6 HIV negative donors (HNDs). Peripheral blood mononuclear cells (PBMCs) were collected from whole blood after Ficoll-Paque PLUS gradient centrifugation. The median CD4 count of CPs was 830, with a range from 225 to 1550. The median age of participants was 53 years, with a range from 28 to 66 years. The median time on ART was 16 years, with a range from 2 to 23 years. Detailed characteristics of the participants are provided in Supplementary Table 1 .

A total of 14 Gag ARFP, 7 Pol ARFP, and 3 Env ARFP peptides were included in this study. ARFP sequences were obtained from two sources. ARFP peptides were either selected from previously reported polypeptides (2, 3) or were derived from long uninterrupted polypeptide sequences within the Gag +1 and Gag +2 reading frames ( Table 1 ). The peptides were synthesized by Genscript Biotech Corporation (Piscataway NJ) with a purity of > 85%.

CD8-depleted PBMCs were cultured for 10 days in RPMI media with 10% Fetal Calf Serum (R10 media) supplemented with 10 IU/mL IL-2. The antiretroviral drug raltegravir was added at a concentration of 4uM to prevent viral replication. The cells were stimulated with 10 μg/mL of peptide pools (Gag ARFP, Env ARFP, and Pol ARFP) and DMSO. The initial cell count per condition ranged from 12 to 45 million PBMCs. Media replenishment was performed on days 3 and 7 by replacing half of the culture medium with fresh R10 media containing 10 IU/mL IL-2. After 10 days, cells were washed, replated in R10 media with 10 U/mL IL-2, and rested for at least 6 hours at 37°C before stimulation. For restimulation, cells were incubated with the same peptide pools (10 μg/mL) in the presence of protein transport inhibitors (GolgiPlug, BD Biosciences, 1 μg/mL; GolgiStop, BD Biosciences, 0.7 μg/mL) and co-stimulatory antibodies targeting CD28 (BD Biosciences) and CD49d (BD Biosciences). After a 16-hour incubation at 37°C, cells were washed and stained for surface markers using antibodies against CD3 (Pacific Blue, BioLegend), CD4 (PerCP-Cy5.5, BioLegend), and CD8 (BV-605, BioLegend). Following surface staining, cells were fixed, permeabilized, and stained intracellularly with antibodies targeting TNF-α (PE-Cy7, BD Biosciences), IFN-γ (APC, BD Biosciences), and IL-2 (PE, BioLegend). Flow cytometry was performed using a BD FACS LSR Fortessa flow cytometer, with at least 100,000 events collected within the lymphocyte gate. Data were analyzed in FlowJo software (version 10.10.0) to quantify cytokine-producing antigen-specific T cells.

In order to identify epitopes, an expansion assay was performed using Gag ARFP or Env ARFP peptide pools, where cells were cultured for 10 days before undergoing a 16-hour restimulation with individual Gag ARFP or Env ARFP peptides. This assay followed the same culture conditions and methodology described above.

Polyfunctionality of antigen-specific T cells was analyzed using FlowJo for intracellular cytokine staining and Boolean gating. Data were processed and exported for further visualization. Statistical analyses were performed using GraphPad Prism (version 10.4.1).

High-resolution HLA class II typing was carried out by the Johns Hopkins Hospital Immunogenetics Laboratory. To assess peptide-MHC binding, in silico predictions were performed using the Immune Epitope Database (IEDB; http://www.iedb.org) between September 11 and 12, 2024. The IEDB-recommended 2.22 prediction algorithm was used, which applies a consensus approach that integrates NN-align, SMM-align, CombLib, and Sturniolo methods when available for the specific allele, or defaults to NetMHCIIpan when not (10, 11). Binding predictions were generated for the most immunogenic peptides. MHC class II alleles with a predicted percentile rank below 30 were included and are presented in Table 2 in order of increasing percentile rank.

An interferon-gamma (IFN-γ) ELISpot assay was performed using individual peptides spanning Gag ARFP and Env ARFP. There were 14 individual Gag ARFP peptides and 3 individual Env ARFP peptides. Each peptide was used at a final concentration of 10 μg/mL and incubated with 200,000 CD8-depleted PBMCs/well in RPMI media supplemented with 10% fetal bovine serum for 20 hours at 37°C, following the manufacturer’s instructions. The plates were incubated for 20 hours with the individual Gag ARFP and Env ARFP peptides. Two replicates were performed for each peptide.

ELISpot plates were analyzed by an independent investigator using the AID iSpot Spectrum and vendor-supplied software. The software quantified IFN-γ spot-forming units (SFUs) per well. SFU/million cells was calculated by adjusting the number of spots per well with the appropriate dilution factor. A positive response was defined as a mean SFU of ≥30.

Supernatant samples were collected on day 7 after stimulation with Gag ARFP, Env ARFP, or Pol ARFP peptide Pools. RNA was extracted using the Zymo Quick-RNA Viral Kit according to the manufacturer’s instructions. Briefly, 200 μL of supernatant was mixed with DNA/RNA Shield and processed through Zymo-Spin IC Columns with successive washes using Viral Wash Buffer and ethanol. RNA was eluted in DNase/RNase-Free water and stored at -80°C until further processing.

Complementary DNA (cDNA) was synthesized using qScript cDNA SuperMix (Quanta Biosciences). Reverse transcription was performed under standard cycling conditions (25°C for 5 min, 42°C for 30 min, 85°C for 5 min, and 4°C hold). The resulting cDNA was used for HIV RNA quantification as previously described (12) to assess latency reversal following ARFP peptide stimulation. The limit of detection for HIV RNA was 300 copies/ml.

HIV-1 alternate reading frame proteins (ARFPs) have been shown to elicit CD8+ T cell responses. However, it is not known whether they are also recognized by CD4+ T cells. To investigate this, CD8+ T cell-depleted PBMCs from 13 CPs and 6 HNDs were stimulated with Gag, Pol or Env ARFP ARFP peptide pools for 10 days, followed by a 16-hour restimulation with the same peptide pools. Cytokine production was assessed using intracellular cytokine staining and flow cytometry. Figure 1A illustrates the ARFP Gag peptide pool response in CP9, where PBMCs were cultured for 10 days with either ARFP peptides or DMSO. Restimulation with Gag ARFP peptides resulted in a marked increase in IFN-γ+ TNF-α+ CD4+ T cells (0.22%) compared to DMSO-cultured cells (0.05%). A similar response was observed in CP97, where restimulation with the Env ARFP peptide pool led to an increase in IFN-γ+ TNF-α+ CD4+ T cells (0.37%) compared to DMSO-cultured cells (0.02%).

Following restimulation with Gag ARFP peptides, significant increases in IFN-γ+ TNF-α+ CD4+ T cells were observed in 7 out of 13 CPs (median 0.06%, range 0.0%–0.19%) compared to cells cultured with DMSO for 10 days (median 0.01%, range 0%–0.02%). Env ARFP peptides elicited IFN-γ+ TNF-α+ CD4+ T cell responses in 2 of 13 CPs, with response frequencies of 0.29% and 0.31%, compared to 0.01% and 0.02% in cells cultured with DMSO for 10 days. In contrast, neither Gag nor Env ARFP peptides were recognized by CD4+ T cells from any of the 6 HNDs ( Figure 1B ). Additionally, none of the 13 CPs exhibited significant IFN-γ+ TNF-α+ CD4+ T cell responses upon restimulation with the Pol ARFP peptide pool ( Supplementary Figure 1 ).

To further investigate the polyfunctional capacity of CD4+ T cells, we assessed the co-expression of IFN-γ, TNF-α, and IL-2 following restimulation with Gag and Env ARFP peptide pools. A significant increase in T cells recognizing all three cytokines was observed in the 7 Gag ARFP responders and 2 Env ARFP responders. In the Gag ARFP responders, the median percentage of IFN-γ+ TNF-α+ IL-2+ CD4+ T cells was 0.04% following 10-day culture with Gag ARFP, compared to 0% for cells cultured with DMSO. Similarly, the two Env ARFP responders exhibited increases in polyfunctional responses, with 0.22% and 0.13% of IFN-γ+ TNF-α+ IL-2+ CD4+ T cells following 10-day culture with Env ARFP, compared to 0.02% with DMSO ( Figure 1C ).

The next objective was to define the specific ARFP peptides targeted by CD4+ T cells. To assess this, we cultured CD8+ T cell-depleted PBMCs from 4 CPs that responded to Gag ARFP peptide pool stimulation and 2 CPs that responded to Env ARFP peptide pool stimulation over 10 days. We then restimulated these cells with individual peptides from Gag ARFP and Env ARFP for 16 hours, and then we measured cytokine production by intracellular cytokine staining and flow cytometry analysis. Figure 2A illustrates this response in CP9, where PBMCs cultured with Gag ARFP peptides for 10 days exhibited a marked increase in IFN-γ+ TNF-α+ CD4+ T cells (0.16%) upon restimulation with Gag ARFP peptide 11, compared to DMSO-cultured cells (0.01%).

Across CPs that responded to Gag ARFP or Env ARFP peptide pools, a few individual peptides elicited IFN-γ and TNF-α production, with variability in the number of recognized peptides across participants ranging from 1–2 individual peptides ( Figures 2B, C ).

To determine whether ARFP-specific T cells exist within the circulating effector T cell pool, an ELISpot assay was performed in three CPs using 14 individual Gag ARFP peptides and 3 individual Env ARFP peptides. In contrast to the cytokine-producing memory T cell responses observed in expanded cultures, no effector IFN-γ responses were detected across all three CPs in response to the individual peptides suggesting that ARFP-specific CD4+ T cells were present at a low frequency ( Supplementary Figure 2 ).

Given the identification of immunogenic ARFP peptides, we next investigated whether these peptides are predicted to bind MHC-II alleles in CPs. High-resolution class II HLA typing was performed in five CPs who exhibited responses to either Gag ARFP or Env ARFP peptides. Using an in silico binding prediction algorithm, we identified multiple class II alleles with high predicted binding affinity for the most immunogenic ARFP peptides. Each CP exhibited unique MHC-II alleles capable of binding at least one ARFP peptide ( Table 2 ).

To evaluate whether ARFP peptides contribute to HIV latency reversal, viral RNA was quantified from culture supernatants collected on day 7 from 6 CPs following stimulation with Gag, Env, or Pol ARFP peptide pools. Latency reversal, indicated by an increase in HIV RNA, was observed in response to Gag ARFP peptides in 2 of 6 participants, with viral RNA levels reaching 31,613.7 copies/mL and 64,484.5 copies/mL. Similarly, Env ARFP peptides induced latency reversal in 2 of 6 participants, with RNA levels of 698.1 copies/mL and 143,464 copies/mL. In contrast, stimulation with Pol ARFP peptides failed to induce latency reversal in any of the 6 CPs ( Figure 3 ). The lack of correlation between the magnitude of ARFP-specific T cell responses and latency reversal could be explained by the fact that not all ARFP-specific T cells are latently infected.