As more countries in RLS roll-out DTG in first-line regimens as recommended by the WHO [7], the demand for INSTI resistance testing will substantially increase and a population level increase in INSTI resistant mutations is expected [22]. Common methods of HIVDR testing use an approach of genotyping the PR and RT genes separately from the IN gene. This is preferred because of more efficient amplification of shorter gene fragments. However, because of the need to genotype two separate fragments, such an approach doubles the workload and increases cost of genotyping, adding pressure on an already strained health care system. We designed a HIVDR method that effectively counteracts challenges associated with generating separate gene fragments. This simplifies the genotyping process, providing drug resistance profiles for all relevant viral genes at a low cost, with shorter turnaround times. These advantages over current methods of HIV-1 genotyping make the single-assay method ideal for use in RLS.

In this study we validate our method using 7 blinded remnant EQA samples from QCMD and apply the method to genotype remnant plasma samples from a previous CAPRISA study. QCMD is an international programme that offers quality assessment in molecular diagnostics to ensure laboratory proficiency testing and competence. Assessment of HIVDR genotyping is based on sequence alignment against all participants sequences submitted in the programme. We observed 100% concordance in detection of all drug resistance mutations, demonstrating the IDR method as reliable for HIVDR genotyping of the HIV-1 PI, RT and IN genes. With only 3/7 non-subtype C sequences (i.e. one HIV-1 A and two HIV-1D), further assessment of the method to non-subtype C sequences is warranted.

Using the IDR method, we did not detect any major primary INSTI resistance mutations. Only one of the 78 sequences had an INSTI accessory mutation E157Q. E157Q is a common polymorphic mutation observed in INSTI-naïve patients, with an estimated frequency of 0.5–2.3% across different subtypes [32,33,34]. When present alone, E157Q causes only potential low-level resistance to first-generation INSTIs (i.e. elvitegravir and raltegravir), with no resistance to DTG, cabotegravir and bictegravir [30]. However, it causes intermediate resistance to all INSTIs when it occurs together with R263K (a common INSTI mutation at ART failure), whilst decreasing DNA binding activity [35, 36]. As expected, we identified similar proportions of PI, NRTI and NNRTI drug resistance mutations as compared to previous CAP103 genotypes, with M46I, M184IV and K103NS being the most common mutations detected, respectively.

The two discordant sequences (IDR036 and IDR094) resulting in different choice of subsequent ART regimens had the M184V mutation which causes high-level resistance to 3TC and emtricitabine (FTC), whilst increasing viral susceptibility to zidovudine (AZT) and TDF [30]. Sequence IDR036 had multiple RT mutations, with discordances observed in both NRTI and NNRTI mutations. CAP103 detected TAMs M41L and K70R which were not detected in IDR, whilst IDR detected A62V and K65R mutations which were not detected in CAP103. K70R alone causes intermediate resistance to AZT, with M41L playing a minimal role in increasing AZT resistance in the absence of the T215Y mutation [37]. K65R detected by IDR causes high-level resistance to TDF and intermediate resistance to abacavir and 3TC/FTC, even in the absence of other NRTI mutations [30]. Addition of mutations A62V, D67N and K219Q (to K65R) results in high-level resistance to all NRTIs except AZT, thus a predicted AZT-based regimen would be recommended on the basis of the IDR sequence, whereas a predicted TDF-based regimen would be recommended on the basis of the CAP103 sequence. Detection of the mixture F227FL in IDR sequence had no significant impact on other NNRTIs and choice of DTG in subsequent ART regimen.

Sequence IDR094 had 3 discordant NRTI mutations, all of which were due to detection of nucleotide mixtures in CAP103 sequence. Detection of K65KR alone resulted in high-level TDF resistance thus a preferred AZT + 3TC + DTG subsequent regimen for CAP103, as opposed to TLD for IDR. Despite K219KN causing potential low-level resistance to AZT, presence of M184V reduced overall AZT resistance to susceptible. The reason for such discordances is not clear but could be explained by several reasons. Heterogenous distribution of HIV-1 variants in cells (arising from rapid evolution of HIV quasispecies), and absolute number of viral variants obtained during viral RNA extraction at any given time, could result in discordances associated with mixed bases [38]. Other factors to consider include, primer binding preference and location, general sequence quality, and technical errors introduced during PCR from Taq polymerase misincorporation [38].

Amplification and sequencing of larger gene regions by Sanger sequencing is met with challenges in obtaining successful PCR amplification and complete sequence coverage. In this study, we obtained ~ 91% (87/96) amplification success among samples with VLs ≥ 1,000 copies/mL (Fig. 1). Amplification success was improved by using the two-fragment approach which amplified an additional 9 samples. The 7 amplicons with failed sequencing had relatively lower band intensities observed in gel electrophoresis. The two sequences excluded from final analysis (IDR074 and IDR076) showed very high sequence similarity (> 98%) after repeating both samples from RNA extraction stage, with no evidence of epidemiological linkage. This suggested potential sample mix-up.

In efforts to provide genotyping results at the shortest time possible, this method reduces genotyping time from ~ 3 to ~ 2 days, saving at least one working day. In addition to providing timely results, it means more samples can be processed over time increasing the capacity of genotypic testing. We used Platinum Taq enzyme because it achieves full-length amplification of the ~ 2.9 kb pol gene of interest, without need for more expensive long-range enzymes. Also, we deliberately designed the method to use 8 sequencing primers, to make cycle sequencing reaction setup easier for laboratory operators working with standard 96-well plate formats. With this setup, sequencing primers are added in the 8 rows and samples in the 12 columns (Additional file 1: Figure S4), achieving coverage of all mutations of interest in the HIV-1 pol gene.

With cost remaining one of the major limiting factors to HIVDR genotyping in RLS [39], we estimated the cost per sample at ~ US$43–$US49 to genotype complete PR, RT, and IN genes in one fragment, or using the two-fragment approach (i.e., genotyping PR and RT separate from IN), respectively (Additional file 1: Tables S7 and S8). These estimates did not include labour and instrument maintenance costs, as well as assay accreditation/validation costs, which tend to vary by region. However, given that common in-house genotyping assays cost between US$48–US$155 to genotype the PR and RT genes only, with commercial assays ranging between US$155 and US$276 as described previously [39, 40], our method provides a cheaper option whilst genotyping not only the PR and RT genes, but also the IN gene.

There are some limitations to consider. Firstly, the majority of samples processed using the IDR method were HIV-1C samples, the most prevalent subtype accounting for almost half of all HIV infections globally and predominant in RLS [41]. However, we demonstrated successful sequencing of HIV-1A and HIV-1D subtypes in 3 of the 7 EQA samples processed. Secondly, prediction of subsequent ART regimens was based on levels of resistance and research evidence, but did not account for other clinical considerations that would typically guide treatment decisions, such as age, weight, co-infections (e.g. Hepatitis B status and tuberculosis), co-morbidities (e.g. renal impairment), and drug contraindications. Thirdly, we mostly compared sequence pairs in the PR and RT genes due to the parent (CAP103) study not having IN gene sequence data, although we would not expect a difference in concordance if paired IN sequences were included. Lastly, use of remnant samples meant that we could not process and compare ~ 17% of samples (19/115) due to low plasma volumes available for RNA extraction. Also, this potentially affected amplification success rates as RNA tends to degrade with repeated freeze thaw cycles.

In conclusion, we developed a simple, labour efficient and affordable HIVDR genotyping method for detecting mutations in the HIV-1 PR, RT and IN genes, and demonstrated high concordance with EQA samples. Despite discordances in two sequences resulting in differences in choice of subsequent regimens, recent data from NADIA trial (96-weeks follow up) showed TDF to be superior to AZT when administered with DTG, suggesting both patients would still benefit from switching to TLD. The lower cost, shorter turnaround time, coverage of all genes of interest, and ease of use, shows several advantages of this method over common in-house assays, making it ideal and relevant for use in monitoring HIVDR in RLS.