Human immunodeficiency virus type 1 diversity in patients initiating treatment in Northern South Africa

Background: Clinical trials carried out to determine when to start antiretroviral therapy (ART) have been showing that early ART initiation results in better health outcomes which are primarily indicated by suppressed viral loads and increase in CD4+ cell counts to healthy human ranges. This in turn is expected to reduce HIV transmission thereby, leading to a decreased spread of the virus and eventual elimination of the virus from the human population. Findings from these trials resulted in the near global adoption of universal test and treat program. The program stated that in the absence of contraindications, people living with HIV (PLWH) should start treatment as soon as they are diagnosed with HIV. However, these outcomes could differ in real world settings where some of the support provided in clinical trial settings are not available. Furthermore, there is historical data that showed variations in HIV prevalences and prevalences of HIV drug resistance within provinces and sub-regions of provinces in South Africa and to have an accurate perspective of the impact of universal test and treat (UTT) implementation, these geographical differences need to be investigated. This study hypothesized that UTT will result in improved health outcomes within a prospective, observational setting in Northern South Africa. Objective: The main objective of the study was investigating treatment outcomes, pretreatment and acquired drug resistance and viral diversity in a cohort of HIV infected individuals entering universal test and treat program in Northern South Africa. Study design and approaches: The study protocol was approved by the Human and Clinical Trials Research Ethics Committee of the University of Venda (Project Number: SMN/15/MBY/23/07/10). Permission to carry out the study from public health care centers was granted by the Limpopo Provincial Department of Health. This was an observational and prospective study which recruited participants from four public HIV treatment centers within Capricorn and Vhembe districts of Limpopo province, Northern South Africa. All participants were over 18 years old and written informed consent was obtained prior to inclusion into the study. Phlebotomy was carried out by a registered nurse and identifiable participant information was concealed using unique codes generated during study initiation. Participants were recruited from April 2016 to March 2017 and followed up every 3 months from June 2016 to July 2018. To determine treatment outcomes for people in UTT, demographic, socioeconomic, and clinical data was collected using a structured questionnaire which was then captured in a spreadsheet. Missing and ambiguous data entries were clarified with participants. The data captured was age, sex, marital status, highest level of education, occupation, income level, probable place of infection, probable year of infection, current anti-retroviral (ARV) regimen, start date of treatment, place of residence, other diagnosis at start of treatment, clinical HIV stage as defined by World Health Organization (WHO), other medications taken with HIV treatment, cluster of differentiation 4 (CD4)+ cell count, percentage CD4, hemoglobulin, viral load in counts per millimeter of blood. During initial and follow up visits, three milliliters of whole blood was collected into each of two ethylenediaminetetraacetic acid (EDTA) coated tubes. One tube was shipped to a commercial laboratory for viral load measurements which was done on a COBAS Ampliprep/COBAS TaqMan analysis platform (Roche, USA). The platform uses reverse transcription polymerase chain reaction (RT-PCR) with DNA probe to quantify viral copies present within a sample. The second tube was shipped to the University of Venda laboratory for CD4+ cell count measurement using a Becton Dickinson FACSPresto system (BD Biosciences, San Jose CA). The system employs immunofluorescence and spectrophotometry techniques to quantify CD4+ cells and hemoglobulin within a test sample. The CD4+ cell count measurements and viral load measurements were also captured within the same spreadsheet that the demographic, socioeconomic, and clinical data was captured. The captured data was cleaned and imported into Rstudio for statistical analysis. The treatment outcomes investigated were viral suppression which was defined as reduction in baseline viral load to less than 50 viral copies per milliliter of blood after starting ART and immunological response which was defined as an increase in baseline CD4+ cell count by 50 CD4+ cells after 6 months on ART. Participants were also categorized as early ART initiators which was defined as those with baseline CD4+ cell counts greater than 500 cells per microliter blood (cells/μL) and late ART initiators or those with baseline CD4+ cell counts less than 500 cells per micro-liter blood (cells/μL). The study inclusion criteria was people 18 year and older who had never been on ART. Demographic, socioeconomic, and clinical data collected while initiating study was summarized and presented as a table. Kaplan Meier curves were used to estimate viral suppression and immunological response. Hazard ratios for demographic, socioeconomic, and clinical data against the two outcome variables namely viral suppression and immunological response were determined using cox regression analysis. To study pre-treatment drug resistance (PDR), total ribonucleic acid (RNA) was extracted from plasma using QIAamp Viral RNA mini kit (Qiagen, Netherlands). Complimentary deoxyribonucleic acid (DNA) was synthesized using cloned avian myeloblastosis virus (AMV) reverse transcriptase (RT) (Invitrogen, USA). The first 1600 nucleotides of the pol gene spanning the complete protease gene and about 1000 nucleotides of reverse transcriptase gene were amplified by nested polymerase chain reaction (PCR) using FastStart Taq DNA Polymerase kit (Sigma-Aldrich, USA). Expected PCR amplicons identified by gel electrophoresis; and samples with desired length were selected for next generation sequencing (NGS). PCR amplicons were purified using AMPure XP magnetic beads (Beckman Coulter, USA). Sequencing libraries were prepared from amplicons by normalizing the concentration of each sample’s amplicon to 0.2 nanogram per microliter. These normalized amplicons were the fragmented using transposase enzyme within the Nextera XT DNA library preparation kit (Illumina, San Diego, California, U.S.A.). Sequencing primers were then ligated to both 5-prime and 3-prime ends of the fragmented amplicons or DNA libraries through PCR. The concentration for each sample’s DNA library was determined using Qubit dsDNA High sensitivity florescent dye assay kit (Life technologies, U.S.A.). The libraries were normalized to four nanomolar prior to pooling the normalized libraries, denaturing, and diluting the denatured or single stranded libraries to 1.8 picomolar. The denatured libraries were then spiked with 25% PhiX and loaded onto a sequencing cartridge (Illumina, San Diego, California, U.S.A.) and ran on a MiniSeq sequencing system. The sequencing reads generated after the run were demultiplexed and post-sequencing quality check report was generated. The reads were then exported from the machine and validated using FastQC program. https://www.bioinformatics.babraham.ac.uk/projects/fastqc/. The reads were then imported into Geneious Prime 2020 (Biomatters, 2020) where the sequences were filtered and trimmed to remove reads shorter than 50 nucleotides and left-over sequencing adapters. The remaining high-quality reads were assembled to generate contigs which were mapped to a reference sequence and consensus sequences were extracted. Variants at each nucleotide position was called at >5% threshold to account for minority variants and >20% threshold. These variants were queried for surveillance drug resistance mutations (SDRM) using Calibrated Population Resistance (CPR) Tool from the Stanford drug resistance database. Univariate and multivariate logistic regression analysis was done to determine associations between SDRM and predictor variables namely, sex, age, CD4+ cell count and viral load. To study acquired drug resistance (ADR), only samples collected within the first 12 months after initiating ART were considered. DNA was extracted from isolated peripheral blood mononuclear cells (PBMC) using QIAamp blood DNA midi kit (Qiagen, Netherlands). pol gene was amplified and sequenced. Sequencing for follow amplicons was prioritized based on if the baseline amplicons for each follow up sample had been successfully sequenced and analysis of the baseline sequences did not have drug resistance mutations within its >20% threshold. Variant calling was done at >1%, >5% and >20% viral thresholds and drug resistant mutations were queried from the Stanford drug resistance database at each of the thresholds. Wilcoxon method was used to determine correlations between viral load and drug resistance mutations in Rstudio and phylogenetic tree was constructed using Interactive Tree of Life (iTOL) https://itol.embl.de/. To determine HIV diversity in the study population, baseline pol gene sequences were subtyped using Molecular Evolutionary Genetics Analysis Version 7.0 (MEGA7), Context-based Modeling for Expeditious Typing (COMET), Subtype Classification Using Evolutionary Algorithms (SQUEAL) and REGA HIV-1 Subtyping Tool Version 3.0. The sequences were screened for recombination using Recombinant Identification Program (RIP) and jumping profile Hidden Markov Model (jpHMM). Subtypes were assigned based on consensus achieved with all four of the subtyping tools. Recombination was inferred based on detection of mixed subtypes by recombinant identification program (RIP) or jumping profile hidden markov. Amplification of a 9084 base-pair HIV sequence was done in two overlapping fragments whereby the first 4516 nucleotides of the 9084 HIV sequence was amplified in a separate nested PCR reaction to the next 4680 nucleotides of the 9084 HIV sequence. Expanded long template enzyme (Roche, Germany) was used for amplification and then sequenced using Illumina MiniSeq sequencing system. Sequence reads were imported into Geneious Prime 2020 (Biomatters, 2020), filtered and trimmed and de novo assembled to obtain full length sequences. The full-length sequences and the extracted HIV genes were subtyped using phylogenetic means using MEGA 7 and COMET. Recombination was detected using RIP, jpHMM and Simplot. Consensus subtype was determined by detection of the subtype by two or more tools. The recombination pattern was illustrated using the recombinant HIV-1 Drawing tool (https://www.hiv.lanl.gov/content/sequence/DRAW_CRF/recom_mapper.html). Results: To determine treatment outcomes for people in UTT, a total of 548 participants were recruited. However, fourteen were excluded as they had started ART prior to recruitment. The study population was made up of 71% (378/534) female. The median age for the study population was 35 (IQR: 29 – 44). Seventy one percent (381/534) had baseline viral loads greater than one thousand copies per ml. The population had 75.3% (402/534) late ART initiators (LAI). There were eleven individuals within the cohort at 24 months. Dataset for the viral suppression analysis, excluded eighty-four of the participants who did not have baseline viral load measurements, 24 participants who had suppressed baseline viral loads (less than 50 copies/ml) and 180 participants who were never followed up. Individuals were censored on the first follow up viral load that dropped to fifty copies/mL. The dataset for viral suppression analysis comprised of 246 individuals whose baseline and follow up viral loads were captured. Censoring occurred at the first viral load that dropped to or below 50 copies/mL after starting ART. Fifty percent (122/246) achieved viral suppression after approximately 6 months of ART and 73.6% (181/246) after approximately 12 months and 97% after approximately 24 months. Multivariable cox regression analysis showed that viral suppression in males was slower compared to females (AHR=0.642; CI: 0.41 – 0.92; p value=0.025) and this difference was statistically significant. There was also a statistically significant positive association between viral suppression and hemoglobulin counts above 12 grams per deciliter (AHR = 2.307; 95% CI: 1.6067 – 3.312; p<0.001). Dataset for immunological response analysis compromised of 104 individuals derived from the 246 individuals within the viral suppression analysis. All 104 individuals had baseline and follow up CD4+ cell counts. Censoring occurred when an increase of baseline CD4+ cell count by 50 cells/μL was observed at or after 6 months of ART initiation. There was 11.5% (12/104) individuals who achieved an immunological response after approximately 6 months of ART use and 39.4% (41/104) after approximately 12 months and 97.5% (101/104) after approximately 24 months. There was negative associations between immunological response and baseline viral loads greater than 50 to 1000 copies/mL (AHR = 0.108; 95% CI: 0.0303 – 0.388; p<0.001) as well as baseline viral loads greater than 1000 copies/mL (AHR = 0.125; 95% CI: 0.0541 – 0.291; p<0.001). Both associations were statistically significant. There was also negative associations between immunological response and individuals who were classified as WHO clinical stage 1 (AHR = 0.173; 95% CI: 0.0460 – 0.655; p=0.0097) as well as clinical stage 3 (AHR = 0.123; 95% CI: 0.0155 – 0.977; p=0.0475). Both of these associations with immunological response were statistically significant. Whole blood was collected from 534 participants to determine diversity of drug resistance mutations for people starting ART. Low blood volume (less than 1 ml) was collected from 135 individual of the 534 recruited at baseline and blood was recollected from them within 4 days of initial collection. Recollection was successful for fifty-four out of the 135 individuals whereas 81 participants could not be reached. From the fifty-four participants, only 21 had 2 ml to 3 ml of blood collected. RNA was isolated from 420 blood samples from which 338 were successfully amplified and sequenced. Twenty of the 338 failed to generate sequencing reads and seventy-seven had small numbers of high quality assembled contigs and could not generate a continuous sequence. Surveillance drug resistance mutation (SDRM) prevalence at >20% threshold was 9.5% (23/241) and 12.9% (31/241) at >5% threshold. SDRM associated with non-nucleoside reverse transcriptase inhibitor (NNRTI) had a 7.5% prevalence within the >20% threshold and 8.7% prevalence within >5% threshold. The NNRTI SDRMs that chiefly contributed to this prevalence at >20% threshold were K103N (6.2%; 15/241), V106M (1.7%; 4/241) and P225H (1.2%; 3/241) whereas at the >5% threshold were K103N (7.9%; 19/241), V106M (2.5%; 6/241) and P225H (1.7%; 4/241). Nucleoside reverse transcriptase inhibitors (NRTI) SDRMs prevalence at >20% threshold was 3% and 4.6% at >5% threshold. NRTI SDRM at the >20% threshold were M184V (0.8%; 2/241) and K65R (1.7%; 4/241) and at the 5% threshold were D67G/N (1.2% 3/241), K70R/E (1.7%; 4/241), M184V (1.2%; 3/241) and K65R (2.5%; 6/241). Protease inhibitor SDRMs prevalence was at 0.4% at >20% threshold and 2.1% at >5% threshold. The PI mutation D30N was the only SDRM detected at >20% threshold whereas the mutations F53L, G73S and M46I were detected with a prevalence of 0.4% each at >5% threshold. Associations between surveillance drug resistance mutations and predictor variables namely sex, age, viral load and CD4+ cell count showed that there were decreased odds risk for males having SDRM compared to females (aOR= 0.73; 95% CI: 0.29 – 1.83; p value = 0.499) and increased odds risk for people with baseline viral load over one million copies/mL having SDRMs (aOR = 1.25; 95% CI: 0.11 – 14.55; p value=0.86). None of these associations with SDRMs were statistically significant. To determine diversity of drug resistance mutations for PLWH on ART, 203 baseline participants who were followed up at the four timepoints within the first year of ART with 155 followed up at 3 months timepoint, 111 at 6 months, 96 at 9 months and 94 at 12 months. Amplification was successful for 108, 67, 50 and 28 samples at 3, 6, 9 and 12 months, respectively. Follow ups amplicons were selected for NGS sequencing based on the absence of DRM at >20% threshold for their baseline sequences. This meant that 34, 46, 37 and 20 follow up amplicons for 3, 6, 9 and 12 months respectively were selected for Illumina sequencing. High-quality (Q score >=30) sequences were obtained from 29, 37, 33 and 15 samples at 3, 6, 9 and 12 months, respectively. Forty-six percent (52/114) of the 114 total follow up sequences had at least one DRM at >20% threshold with 55% (63/114) with at least one DRM at >5% threshold. The NNRTI DRM prevalence at >20% threshold was 6.9% (2/29), 8.1% (3/37), 9.1% (3/33) and 6.7% (1/15) at 3 months, 6 months, 9 months, and 12 months timepoints, respectively. The NNRTI DRM prevalence at >5% threshold was 6.9% (2/29), 8.1% (3/37), 12.1% (4/33) and 6.7% (1/15) at 3 months, 6 months, 9 months, and 12 months timepoints, respectively. The frequent NNRTI DRM at >20% threshold was K103N, P225H and E138A. The NNRTI mutation K101E was observed at >5% threshold. The NRTI DRM prevalence at >20% threshold was 3.4% (1/29) at 3 months, 2.7% (1/37) at 6 months, 6.1% (2/33) at 9 months, and 0% at 12 months. The NRTI DRM prevalence at >5% threshold was 3.4% (1/29) at 3 months, 2.7% (1/37) at 6 months, 9.1% (3/33) at 9 months, and 0% at 12 months. The commonly occurring NRTI DRM at 20% threshold was M184V, K70ER and at >5% threshold was T215I. Primary PI DRM prevalence was 0% at >20% threshold however, at >5% threshold the prevalence was 3.4% (1/29) and 3% (1/33) at 3 months and 9 months respectively and 0% at 6 months and 12 months. The PI mutations that contributed at >5% threshold was M46I and V82A. Correlation of viral loads among majority variants (at greater 20% threshold) and minority variants (less than 20% threshold) showed a slight difference in median viral loads at 3 months (median log10 copies per ml for majority variant = 2.08; IQR: 1.59 – 2.28; median log10 copies per ml for minority variant = 1.81; IQR: 1.30 – 2.31; p value = 0.55) with no difference in viral loads at 6, 9 and 12 months. The differences in viral loads between the groups were not statistically significant at any of the timepoints. Subtypes inferred through phylogenetic means showed that all sequences were HIV subtype C. To determine HIV diversity for people starting ART, 241 baseline sequences were screened for non-C subtypes using four subtyping tools. The four tools showed that 99% (238/241) of the sequences were HIV-1 subtype C. However, subtyping of three samples (AHDR-R244, AHDR-R118 and AHDR-S100) revealed non-C subtypes within their sequences. Phylogenetic analysis revealed that AHDR-R244 was a subtype G whereas COMET showed the sample to be a subtype G and C whereas SQUEAL inferred it to be a subtype C and REGA inferred it to be a subtype G and C. AHDR-R188 was inferred to be a subtype C when analyzed by phylogenetic means, COMET and REGA but was inferred as subtype A by SQUEAL analysis. AHDR-S100 was inferred to be a subtype C when analyzed by phylogenetic means, SQUEAL and REGA whereas it was inferred to be a recombinant of subtype F2 and C with COMET. The recombinant detection tool jpHMM detected AHDR-R244 was a recombinant of subtypes G and C whereas AHDR-R118 was a recombinant of subtypes A2 and C and AHDR-S100 was subtype C. The recombinant detection tool RIP detected AHDR-R244 as recombinant of G and C, AHDR-R118 as subtype C and AHDR-S100 as recombinant of subtype C and F2. Near full genome amplification and sequencing was successful for all three samples including AHDR-R114 which was used as a quality control sample. Screening of AHDR-R114 using all four subtyping tools inferred it was a subtype C sequence. Subtyping the HIV genes with 2 subtyping tools and recombination tools revealed that AHDR-R244 had a mix of subtypes C and G for gag and pol genes, subtypes G and H for env gene, subtypes B and C for vif, subtype C for vpu, subtypes G and C for rev, subtypes J and D for vpr, subtype C and recombinant CRF01_AE for tat, subtypes J and C for nef. Analysis with the three recombinant tools revealed that same subtypes were present within the HIV genes analyzed. The consensus of the complete genome was a recombinant with subtypes A1, C, G and recombinant CRF01_AE. Subtyping of HIV genes using phylogenetic means for AHDR-R118 and AHDR-S100 revealed that all the HIV genes for these two samples were subtype C however, there was a recombinant CRF88 detected within tat gene for AHDR-R118 and AHDR-S100. COMET analysis revealed that AHDR-R118 was a subtype C for all HIV genes whereas AHDR-S100 was subtype F2 for pol gene with the remaining genes being subtype C. RIP revealed that AHDR-R118 pol gene was a recombinant of subtypes A2 and C and the rest of the HIV genes were subtype C. RIP revealed that AHDR-S100 pol gene was a recombinant of subtypes B, C and F2 and env gene was a recombinant of subtypes C and F2 and recombinant CDF01_AE with the remaining HIV genes being a subtype C. The recombinant tool jpHMM revealed that AHDR-R118 had subtypes A2 and C within pol gene and subtypes C and J within tat gene with the rest of the HIV genes being a subtype C. The jpHMM tool revealed that AHDR-S100 had subtypes A2 and C within gag gene and subtypes C and J within tat gene with the rest of the HIV genes being a subtype C. Simplot revealed AHDR-R118 had subtypes C and D within env gene with the rest of the genes being a subtype C. Simplot revealed that AHDR-S100 was subtype C for all HIV genes. The consensus of AHDR-R118 was that it was a subtype C with subtype A2 within the pol gene whereas the consensus for AHDR-S100 was that it was a subtype C with subtype F2 within the pol gene. Conclusion: PLWH initiating UTT within Northern South Africa showed good response to treatment with moderate levels of PDR, no ADR and low non-subtype C variants. Seventy-five percent of people initiating UTT were initiating with a CD4+ cell count of less than 500 cells/𝜇l also known as late ART initiators. Immunological response was slow which could indicate that there are other factors that played a role in immunological recovery. Future investigations need to focus on whether late ART initiation continued as the norm during UTT and what factors played a role in initiating ART. In addition, there were moderate levels of PDR, which was driven by NNRTI resistance however, there has also been an increase NRTI PDR within Northern South Africa. None of the people on ART developed resistance to NRTI or PI by the end of first year of ART use. The introduction of integrase-inhibitor regimen could eliminate NNRTI resistance associated mutations though studies need to investigate whether NRTI resistance associated mutations will also decline with the current integrase-based regimen. The majority (99%) of viral subtypes were subtype C although one recombinant was detected in the study population. This indicates the diagnostic tools used in UTT should continue to be effective in detecting infecting HIV variants, at least in the study population.

Author dc:creator

• Ogola, Bixa O.

Advisors dc:contributor.advisor

• Bessong, Pascal O.
• Mavhandu-Ramarumo, Lufuno G,

• Universal test and treat
• Northern South Africa
• Treatment outcomes
• Pretreatment drug resistance
• Next-generation sequencing
• Viral diversity
• Viral suppression
• Drug resistance mutations

Statement dc:rights

• University of Venda

Language dc:language.iso

en

Repository record dc:identifier.uri

http://univendspace.univen.ac.za/handle/123456789/2705

OAI identifier oai:identifier

oai:univendspace.univen.ac.za:123456789/2705