Deep sequencing of the HIV-1 polymerase gene for characterisation of cytotoxic T-lymphocyte epitopes during early and chronic disease stages

Background Despite multiple attempts, there is still no effective HIV-1 vaccine available. The HIV-1 polymerase (pol) gene is highly conserved and encodes cytotoxic T-lymphocyte (CTL) epitopes. The aim of the study was to characterise HIV-1 Pol CTL epitopes in mostly sample pairs obtained during early and chronic stages of infection. Methods Illumina deep sequencing was performed for all samples while Sanger sequencing was only performed on baseline samples. Codons under immune selection pressure were assessed by computing nonsynonymous to synonymous mutation ratios using MEGA. Minority CTL epitope variants occurring at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ge$$\end{document}≥ 5% were detected using low-frequency variant tool in CLC Genomics. Los Alamos HIV database was used for mapping mutations to known HIV-1 CTL epitopes. Results Fifty-two participants were enrolled in the study. Their median age was 28 years (interquartile range: 24–32 years) and majority of participants (92.3%) were female. Illumina minority variant analysis identified a significantly higher number of CTL epitopes (n = 65) compared to epitopes (n = 8) identified through Sanger sequencing. Most of the identified epitopes mapped to reverse transcriptase (RT) and integrase (IN) regardless of sequencing method. There was a significantly higher proportion of minority variant epitopes in RT (n = 39, 60.0%) compared to IN (n = 17, 26.2%) and PR (n = 9, 13.8%), p = 0.002 and < 0.0001, respectively. However, no significant difference was observed between the proportion of minority variant epitopes in IN versus PR, p = 0.06. Some epitopes were detected in either early or chronic HIV-1 infection whereas others were detected in both stages. Different distribution patterns of minority variant epitopes were observed in sample pairs; with some increasing or decreasing over time, while others remained constant. Some of the identified epitopes have not been previously reported for HIV-1 subtype C. There were also variants that could not be mapped to reported CTL epitopes in the Los Alamos HIV database. Conclusion Deep sequencing revealed many Pol CTL epitopes, including some not previously reported for HIV-1 subtype C. The findings of this study support the inclusion of RT and IN epitopes in HIV-1 vaccine candidates as these proteins harbour many CTL epitopes. Supplementary Information The online version contains supplementary material available at 10.1186/s12985-022-01772-8.


Introduction
The HIV/AIDS pandemic has been a global crisis for four decades [1][2]. At the end of 2020, there were 37.6 million people living with HIV globally, 1.5 million new HIV-1 infections, and 690 000 AIDS-related deaths [3]. South Africa has 7.9 million people living with HIV-1 (PLWH), making it the country with the highest number of infections in the world [4][5]. This highlights the need to better understand HIV-1 natural immune responses in order to bolster the efforts of developing effective HIV-1 vaccines.
During early HIV-1 infection, effective cytotoxic T-lymphocyte (CTL) immune responses play an important role in the control of viraemia, by contributing to suppression of the viral load (VL) to a set-point [6]. CTL responses target and kill virusinfected cells, via recognition of viral peptide epitopes and releasing cytokines and cytotoxic granules that facilitate cell killing [7][8][9]. Viral peptide epitopes are presented on the surface of infected cells by the human leukocyte antigen type I (HLA I) trans-membrane proteins encoded by HLA-A, B and C alleles [10][11][12]. The targeted epitopes undergo immune selection pressure, which leads to the emergence of viral escape mutations [13][14]. This results in evasion of the immune system, loss of immune control and ultimately progression to AIDS [7,10]. The majority of epitopes that play a role in the control of viraemia have been extensively studied and reported for HIV-1 Gag [15][16]. However, some studies show that HIV-1 Pol also harbours important CTL epitopes that contribute to the control of HIV-1 viraemia [17][18][19], hence pol is included in some HIV-1 vaccine candidates [20][21][22][23][24][25]. There are limited data on the characterisation of HIV-1 Pol CTL epitopes in early and chronic HIV-1 disease stages. The aim of this study was to characterise and assess the evolution of CTL epitopes encoded by HIV-1 pol during early and chronic stages of HIV-1 infection, using deep sequencing methods.

Study population
This was a retrospective study that used stored plasma samples that were collected from individuals in the Tshwane district of South Africa. Participants enrolled in this study had a con rmed diagnosis of early or chronic HIV-1 infection and were antiretroviral therapy (ART) naïve. They were identi ed in a study that screened for early HIV-1 infection in individuals who had a negative rapid test at point-of-care facilities, and diagnosis was con rmed through HIV-1 VL, antibody (enzymelinked immunoassay), p24 antigen, Western blot and avidity testing [26]. Most of the participants were followed up, and thus samples at two time-points were available (at baseline and follow-up) for analysis.

Nucleic acid extraction and ampli cation of HIV-1 pol
Total nucleic acids were manually extracted from plasma samples using the QIAamp UltraSens Virus kit (Qiagen, Hilden, Germany). Samples with a VL > 1000 copies/ml were extracted from a plasma input volume of 500 µl, which was adjusted to 1 ml using phosphate-buffered saline (PBS). Samples with a VL ≤ 1000 copies/ml were extracted directly from 1 ml plasma input volume to increase nucleic acid yield. Extraction was performed according to the manufacturer's instructions except for the rst centrifugation step, which was optimized to 800 relative centrifugal force (RCF) instead of the recommended 1200 RCF. This modi cation facilitated a more e cient dissolution of the pellet. Nucleic acids were eluted in a volume of 60 µl and all the eluates were stored at -70 ºC immediately after extraction.
The complete HIV-1 pol gene was ampli ed in all samples using an in-house nested PCR method, employing the SuperScript™ III One-Step RT-PCR System with Platinum™ Taq High Fidelity DNA Polymerase (Invitrogen, Carlsbad, California, USA). Each PCR reaction was performed in a 50 µl reaction, which included 5 µl of extracted RNA / DNA template, 2X reaction mix containing 2.4 mM magnesium sulphate (MgSO4), 0.4 mM of each deoxynucleotide triphosphate (dNTPs), 10 mM sense primer, 10 mM anti-sense primer, 5 units (U) of enzyme mix for ( rst round) or 1U of Platinum Taq polymerase (for second round), and nuclease-free water. The rst-round amplicon was used as template for the second-round ampli cation (www.lifetechnologies.com) (Supplementary Table 1). One or two extracted HIV-1-negative controls were included in each experiment to assess potential contamination.

Sanger sequencing and sequence analysis
Sanger sequencing was performed on PCR amplicons, but only for baseline samples. Sequencing was performed commercially in six overlapping regions covering the entire pol ORF (Inqaba Biotechnical Industries, Tshwane, South Africa) Table 1). Sequences were edited in CLC Main Workbench 21 software (Qiagen, Hilden, Germany) and consensus sequences were generated, viewed in BioEdit 7.2.5 (https://bioedit.software.informer.com/download/) and aligned using the online version of the MAFFT program (https://mafft.cbrc.jp/alignment/server/). The HIV-1 HXB2 reference sequence (K03455.1) was used for nucleotide numbering. Phylogenetic analysis was performed on the multiple alignment and the ratio of nonsynonymous to synonymous (dN-dS) mutations within the HIV-1 Pol region was computed using MEGA software (https://www.megasoftware.net/). A dN-dS ratio ≥ 5 was used to identify codons under high selection pressure and these sites were subsequently mapped to CTL epitopes documented in the Los Alamos HIV database [27].
Illumina sequencing and analysis of consensus sequences Sequencing of PCR amplicons was performed at Inqaba Biotechnical Industries (Pretoria, South Africa), a commercial NGS service provider. Brie y paired-end libraries (2 × 300 bp) were prepared using the NEBNext® Ultra™ II DNA library prep kit for Illumina® (New England Biolabs, USA) and sequencing was performed on an Illumina MiSeq instrument (Illumina, USA).
Some samples (n = 12) that were viewed to have poor quality scores were sent for repeat sequencing to the National Institute for Communicable Diseases (NICD), Sequencing Core Facility, Johannesburg, South Africa. Multiplexed paired-end libraries (2 x 300 bp) were prepared using the Nextera™ XT DNA Library Sample Preparation kit (Illumina Inc., California, United States) according to the manufacturer's instructions and sequencing was carried out on an Illumina MiSeq instrument (Illumina Inc., California, United States) Illumina sequencing data was analysed in the PASeq program (https://paseq.org/) using the feature that excludes APOBEC-induced hypermutations, and consensus sequences were generated. The quality of the Illumina reads was assessed in PASeq and CLC Genomics Workbench 21 (Qiagen, Hilden, Germany). PASeq-generated Illumina consensus sequences were viewed in BioEdit 7.2.5 software and aligned to HIV-1 reference sequences obtained from the GenBank database, in the MAFFT program available online. Sanger and Illumina consensus sequences obtained from baseline samples were compared through phylogenetic analysis in MEGA software and were observed to correctly cluster together ( Supplementary Fig. 1).

Analysis of Illumina sequencing reads for minority variants
Deep sequencing reads were mapped to an HIV-1 subtype C reference (AY162225.1) in CLC Genomics. Minority variants were analysed using a low-frequency variant detection tool in CLC Genomics, setting the cut-off for signi cance at 1% and the minimum frequency for variants at 5%. Minority variants were assessed in all participant samples and only nonsynonymous variants were mapped to HIV-1 CTL epitopes in the Los Alamos HIV database, to identify the epitope they fall within. Thus, in this study, minority variant epitopes were de ned as epitopes existing at proportions above 5% but less than 20%. The Los Alamos HIV database was also used to predict human leukocyte antigen (HLA) allotypes that may be involved in presenting the identi ed epitopes to CTLs. Evolution within Pol CTL epitopes was assessed by comparing the proportion of minority variants between baseline and follow-up samples, and in early and chronic HIV-1 disease stages.

Data analysis
Descriptive statistics was used to present median values and interquartile range (IQR) for age, HIV-1 VL and sequencing depth. Fisher's exact test was used to assess if there was an association between the distribution of mutations and HIV-1 disease stage for epitopes identi ed through Sanger sequencing. Two sample t-test was used for comparing the proportions of minority variant epitopes among the different regions of the Pol protein. A p-value of ≤ 0.05 was considered statistically signi cant. All statistical tests were performed on the STATA 16.0 software package (StataCorp LP, College Station, TX, USA). The statistical analyses of dN-dS ratios were computed on the MEGA programme using the HyPhy test for selection. Due to the small sample size, codons with dN-dS ratios of ≥ 5 were considered for assessment of epitopes under immune selection pressure. GraphPad Prism v.9.1.2 (GraphPad Software, San Diego, California, USA) was used to Loading [MathJax]/jax/output/CommonHTML/jax.js prepare graphs for the frequently recognised minority variant epitopes among the different regions of the Pol protein, and also for showing the dynamics of these epitopes in sample pairs obtained during early and chronic HIV-1 infection stages.

Demographics
This study enrolled 52 participants with HIV-1 infection, 34 of whom had paired plasma samples. Their median age was 28 years (IQR: 24-32 years), and majority were females (92.3%). There were 15 participants (28.9%) with early HIV-1 infection and 37 (71.1%) with chronic HIV-1 infection. The median HIV-1 VL at baseline was 2.8 x 10 4 copies/mL (IQR: 8.6 x 10 3 -9.1 x 10 4 copies/mL). The median interval between baseline and follow-up sampling was 4 weeks (IQR: 3 -8 weeks) ( Table 1).  The complete HIV-1 pol gene was successfully ampli ed in all project samples using the in-house nested PCR method. Phylogenetic analysis of Illumina consensus sequences showed that most study sequences clustered with HIV-1 subtype C strains. All baseline and follow-up sequences for each participant correctly clustered together (Fig. 1), showing accurate ampli cation and sequencing of viral genes from the same participant.
Pol CTL epitopes based on Sanger consensus sequences Eight CTL epitopes, with amino acid residues under high selection pressure, were identi ed. The majority of the identi ed epitopes were located within RT. The distribution of escape mutations was comparable between early and chronic HIV-1 disease stages (Table 2). However, these escape mutations were identi ed mostly in participants with chronic HIV-1 infection (Supplementary Table 2). One or more escape mutations were observed in each epitope. Two other codons were identi ed to be under high selection pressure but could not be mapped to the reported CTL epitopes in the Los Alamos HIV database (Table 2). Pol CTL epitopes based on analysis of minority variants The median sequencing depth (for baseline and follow-up) was 23052X (IQR: 13011-49753X) ( Table 1). Minority variant analysis identi ed 65 frequently targeted Pol CTL epitopes. There was a signi cantly higher proportion of epitopes in RT (n = 39, 60.0%) compared to IN (n = 17, 26.2%) and PR (n = 9, 13.8%), p = 0.002 and < 0.0001, respectively. However, no signi cant difference was observed between the proportion of minority epitopes in IN versus PR, p = 0.06. The majority of these epitopes were identi ed in participants with chronic HIV-1 infection compared to those with early HIV-1 infection.
Some epitopes were only identi ed in either early or chronic HIV-1 infection, whereas others were identi ed in both stages of disease (Fig. 2). Different distribution patterns of minority variant epitopes were observed. Some variant epitopes increased or decreased between baseline and follow-up, while others remained constant between these two time-points (Fig. 3, Supplementary Table 3). There were minority variants (2 in early and 26 in chronic HIV-1 infection) that could not be mapped to the reported CTL epitopes in the Los Alamos HIV database (Supplementary Table 3).
Pol CTL epitopes and predicted HLA class I presentation HLA class I allotypes that possibly present the identi ed epitopes were predicted from the Los Alamos HIV database. Some participants recognised more epitopes than others. This was observed for participants 8047 and 6743 who each   HLA alleles in boldface are those that have been identi ed (reported) in South Africa or southern Africa [11,[28][29][30].

Discussion
To our knowledge, this is the rst study to assess the evolution of HIV-1 Pol CTL epitopes in samples obtained during early and chronic disease stages in a setting where HIV-1 subtype C is predominant. We identi ed epitopes within all three regions of Pol, unlike previous studies that have only characterised epitopes within RT and PR [16, 29,33]. This highlights the advantage of using a PCR method that ampli es the complete pol gene. Baseline and follow-up sequences for each participant correctly clustered together, indicating accurate ampli cation and sequencing of viral genes belonging to the same participant. It was expected that the majority of sequences would cluster with HIV-1 subtype C as this is the most common subtype in the southern African region [34][35]. A small number of non-subtype C strains were detected, which could have been introduced through migration, and this has also been reported in previous South African studies [36][37].
Many epitopes were identi ed through the analysis of minority variants obtained by deep sequencing compared to majority variants obtained by Sanger sequencing. This shows the advantage of using deep sequencing for characterisation of Pol CTL epitopes as this method can detect variants that exist below 20% of the total virus population, which are not readily detected by Sanger sequencing [38][39]. Some identi ed minority variant epitopes have never been reported for HIV-1 subtype C and were possibly missed by previous studies that employed Sanger sequencing. The use of Sanger sequencing in most studies has most likely underestimated the presence of CTL epitopes in Pol. Other studies have also shown increased detection of CTL epitopes when using next-generation sequencing (NGS) methods [40][41]. This shows an advantage in the use of NGS for characterising HIV-1 CTL epitopes in future research studies.
In this study, CTL epitopes were identi ed in all regions of Pol, but mostly in RT and IN. Some of these epitopes such as GKKAIGTVL (PR 68-76) and IAMESIVIW (RT 375-383) ( Table 3) have been previously reported to induce CTL responses [42][43]. These data indicate that all three regions of Pol are immunogenic and may play a role in the control of viraemia and the establishment of a viral set-point during early HIV-1 infection, in agreement with previous studies that assessed CTL responses against the HIV-1 proteome [42][43][44]. In a study that assessed CTL responses during acute HIV-1 infection, Kim et al showed that the larger proportion of targeted epitopes were within Pol as compared to other regions of the HIV-1 proteome, and some of these epitopes stimulated strong CTL responses [42]. Ojwach  HIV-1 infection. CTL epitopes encoded by pol have also been reported to play a role in the control of viraemia in long-term non-progressors (LTNPs) and elite controllers (ECs) [47][48].
HLA class I allotypes play an important role in the control of HIV-1 viraemia [29]. Several HLA class I alleles that encode allotypes predicted to present epitopes identi ed in this study are known to be present in the South African population, including HLA-B*58:01 and HLA-A*02 that present SAAVKAACW (IN 123-131) and , respectively [28,30,49]. Past gene association studies showed that alleles within the HLA-B group correlate with greater viral control than alleles in the HLA-A and HLA-C groups [28][29]50]. Payne et al found that HLA-B*58:01 was associated with slower HIV-1 disease progression and was protective against HIV-1 infection [51], which could explain why some participants such as 8047 and 6743 recognised more epitopes, as they also recognised an epitope reported to be presented by HLA-B*58:01.
A hierarchy in epitope presentation may be an explanation for recognition of more epitopes during chronic HIV-1 infection.
Immunodominant epitopes could be preferentially presented earlier during the course of infection, followed by presentation of subdominant epitopes later [52]. Immunodominant epitopes are often located in variable domains whereas subdominant epitopes usually map to more conserved regions of the HIV-1 proteome [18,53]. The HIV-1 pol gene is more conserved than gag and env [54] and therefore harbours more subdominant epitopes [18]. Examples of subdominant responses being more effective in controlling viraemia have been shown in past studies [55][56].
The evolution of Pol CTL epitopes was monitored by assessing minority variants in sample pairs. Minority variant epitopes that increased in proportion from baseline to follow-up may have replaced the wild-type epitopes over time and facilitated immune escape [57]. Fitness cost is the likely explanation for minority variant epitopes that decreased in proportion over time, and these may have been out-competed by new variants [57]. The proportion of some minority variant epitopes remained constant between the two time-points, and these could also represent variants that do not replicate e ciently and are hence maintained at lower levels within the HIV-1 quasispecies [40,57]. Some variant sites were located outside of reported epitopes (Supplementary Table 3), and these may represent new unreported epitopes or could highlight important adjacent sites that may affect epitope processing and presentation [13,58]. Viral evolution was also assessed by comparing sequences from early and chronic HIV-1 stages. Data from this study showed that evolution within CTL epitopes began quite early following infection, as epitopes with escape mutations were detected in participants who had early HIV-1 infection [59]. Many epitopes with escape mutations were detected in participants in the chronic HIV-1 stage, indicating that viral evolution of HIV-1 results from immune selection pressure that occurs throughout the course of infection and diverse viral populations evolve that constantly evade immune responses [57,60].
All three expressed proteins (PR, RT and IN) of Pol harbour CTL epitopes and this highlights the importance of including pol in the design of a vaccine candidate [2,20]. Some researchers have included RT and IN in the design of HIV-1 vaccines [22][23]53]; this is supported by the ndings of our study as the majority of epitopes mapped to these two proteins. Conserved epitopes have been shown to provide cross-clade protection against HIV-1 infection [2,18]. Some of the conserved Pol epitopes such as IETVEPVKL (RT 5-12) and SVPLDEGFRK (RT 117-126) ( Table 3), that were previously observed in a study that looked at using a vaccine candidate with conserved immunogens were also identi ed in our study [18]. The immune responses directed towards epitopes in conserved regions are usually subdominant but may provide protection against multiple HIV-1 subtypes [17,53]   Availability of data and materials All data generated or analysed during this study are included in this manuscript and its supplementary information les.    were variants that increased in proportion over time to become majority variants and were thus maintained above 20% (black dotted line). (iv) Some minority variants that were detected at follow-up existed as majority variants at baseline. The median interval between baseline and follow-up sampling was 4 weeks (IQR: 3 -8 weeks).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.