High prevalence of integrase mutation L74I in West African HIV-1 subtypes prior to integrase inhibitor treatment.

OBJECTIVES
HIV-1 integrase inhibitors are recommended as first-line therapy by WHO, though efficacy and resistance data for non-B subtypes are limited. Two recent trials have identified the integrase L74I mutation to be associated with integrase inhibitor treatment failure in HIV-1 non-B subtypes. We sought to define the prevalence of integrase resistance mutations, including L74I, in West Africa.


METHODS
We studied a Nigerian cohort of recipients prior to and during receipt of second-line PI-based therapy, who were integrase inhibitor-naive. Illumina next-generation sequencing with target enrichment was used on stored plasma samples. Drug resistance was interpreted using the Stanford Resistance Database and the IAS-USA 2019 mutation lists.


RESULTS
Of 115 individuals, 59.1% harboured CRF02_AG HIV-1 and 40.9% harboured subtype G HIV-1. Four participants had major IAS-USA integrase resistance-associated mutations detected at low levels (2%-5% frequency). Two had Q148K minority variants and two had R263K (one of whom also had L74I). L74I was detected in plasma samples at over 2% frequency in 40% (46/115). Twelve (26.1%) had low-level minority variants of between 2% and 20% of the viral population sampled. The remaining 34 (73.9%) had L74I present at >20% frequency. L74I was more common among those with subtype G infection (55.3%, 26/47) than those with CRF02_AG infection (29.4%, 20/68) (P = 0.005).


CONCLUSIONS
HIV-1 subtypes circulating in West Africa appear to have very low prevalence of major integrase mutations, but significant prevalence of L74I. A combination of in vitro and clinical studies is warranted to understand the potential implications.


Introduction
Drug resistance is common amongst individuals with virological failure (VF) of first-line NNRTI-based ART regimens under conditions of infrequent viral load monitoring. 1,2 Second-generation integrase inhibitors such as dolutegravir are now recommended for first-line HIV treatment regimens, 3 following an increase in pre-treatment drug resistance to NNRTI-based regimens globally, including Nigeria. [4][5][6] A number of studies have shown that pre-existing integrase resistance, 27 as assessed using standard lists of mutations derived largely from subtype B data, is rare across globally dominant subtypes. 7,8 As dolutegravir-based ART is rolled out globally, a wider range of HIV-1 subtypes will be exposed and the effects of integrase polymorphisms and subtype diversity on the clinical efficacy of these agents are currently not well understood. Although polymorphisms are generally thought to have little impact on viral phenotype, this is not always true, particularly when comparing B with non-B subtypes. 9 A good example is G118R in integrase, a polymorphism that confers significant integrase strand transfer inhibitor (INSTI) resistance. 10 Two recent Phase 3 trials of the long-acting injectable secondgeneration integrase inhibitor cabotegravir and the injectable second-generation NNRTI rilpivirine, FLAIR in ART-naive participants and ATLAS in ART-experienced participants, 11 found noninferiority of long-acting injectables compared with oral therapy. However, three participants treated with the long-acting injectable drug experienced VF. All three were infected with HIV-1 subtype A1 and were from Russia. All three had L74I in integrase at both baseline and at VF. At VF the major integrase mutation Q148R occurred in two and G140R in one. 11 L74 is in the catalytic core domain, which carries out the integrase strand transfer reaction. It is part of a hydrophobic cluster of residues that includes resistance-associated mutations T97 and F121 near the active site. 12 In the Stanford Resistance Database (https://hivdb.stanford.edu) 13 L74I is reported to be observed in 3%-20%, depending on subtype. The L74M variant has been included as a minor mutation for the first-generation INSTI raltegravir in the IAS-USA drug resistance mutations list (https://www.iasusa.org/wpcontent/uploads/2019/09/27-3-111.pdf), but the L74I variant is not recognized as a resistance-associated mutation. 14 The Stanford Resistance Database includes L74I in combination with other integrase mutations. 13 L74I and L74M are assessed together and combined prevalences are often reported 7 as they have both been shown to enhance integrase inhibitor resistance when present with major INSTI mutations. A recent report suggested that methionine at residue 74 was in closer proximity to T97 and F121 as compared with leucine at position 74 in a modelled subtype C integrase and, of note, L74F was found to contribute to high-level dolutegravir resistance when combined with major mutations G140S and Q148H. 15 We studied a Nigerian cohort of people living with HIV in whom the West African CRF02_AG and G subtypes account for the majority of infections. 16 We aimed to determine the prevalence of INSTI resistance, as well as the prevalence and dynamics of L74I, in this setting.

Patients and methods
Study participants were selected from an HIV-positive second-line treatment cohort at the University of Abuja Teaching Hospital (UATH) in Abuja, Nigeria. Informed consent was obtained from all participants and ethics approval for virological testing was obtained from the Nigeria National Research Ethics Committee of Nigeria (NHREC/01/01/2007). Ethics approval was also obtained from the ethics board of UCL, UK. The Institute of Human Virology Nigeria (IHVN) database was used to identify people living with HIV (PLWH) aged >15 years who had attended UATH and received a first-line ART regimen of two NRTIs and one NNRTI, followed by a second-line ART regimen of two NRTIs and one PI (lopinavir or atazanavir). Participants were included in the study if they: (i) had experienced first-line VF, defined as HIV RNA >1000 copies/mL at least 6 months after first-line ART initiation; (ii) had a stored plasma sample that was obtained during first-line VF, prior to switching to a second-line regimen; and (iii) had a whole HIV genome sequence successfully generated from the first-line VF sample. If participants had subsequent stored plasma samples from first-line or second-line VF these were also included where possible. CD4 cell count and HIV-1 RNA quantitation were performed at the IHVN laboratories. For next-generation sequencing, manual nucleic acid extraction was done at University College London (UCL) using the QIAamp Viral RNA Mini Kit, (QIAGEN, Hilden, Germany) with a plasma input volume of 0.5-1.5 mL. The first strand of cDNA was synthesized using SuperScript IV reverse transcriptase (Invitrogen, Waltham, MA, USA), followed by NEBNext second-strand cDNA synthesis (E6111, New England Biolabs GmbH, Frankfurt, Germany). Sample libraries were prepared as per the SureSelect XT automated target enrichment protocol (Agilent Technologies, Santa Clara, CA, USA) with in-house HIV baits. Whole-genome deep sequencing was performed using the Illumina Miseq platform (San Diego, CA, USA).
Poor-quality reads were identified and removed using TrimGalore v0.6.4. 17 A set of 170 HIV-1 subtypes/circulating recombinant forms (CRFs) were downloaded from the Los Alamos database (https://www.hiv.lanl.gov) and the trimmed reads were compared with this database using BLAST to identify the closest reference. 18 Trimmed reads were mapped to the closest reference genome using the Burrows-Wheeler aligner. 19 Duplicate reads were removed from the BAM files with Picard (http://broadinstitute.github.io/ picard) and a consensus sequence of nucleic acids with a minimum wholegenome coverage of 20% was generated with BCFtools using a 50% threshold. 20 Consensus sequences were used to determine the codon usage at integrase position 74 and subtype was assigned using REGA v3.0. 21 The presence and frequency of IAS-USA resistance-associated mutations in integrase, in addition to L74I, was assessed. Mutations were included if they were present at over 2% frequency within the read mixture at that position and present on at least two reads. A threshold of 2% is supported by a study evaluating different analysis pipelines, which reported fewer discordances over this cut-off. 22 In addition, inspection of the mean read depth across regions of interest in the present study showed that a cut-off of <2% would not include sufficient reads to provide accurate assumptions regarding resistance mutations. An in-house custom script was used to identify SNPs at each position by BLAST analysis of individual HIV pol against the HXB2 reference genome. SNPs were identified and then translated to codons across all regions. Statistical analysis was performed in Stata v13.1 (StataCorp LLC, College Station, TX, USA) and SAS v9.4 (SAS Institute, Cary, NC, USA).

Results
Overall, 115 participants had a total of 163 plasma samples that yielded HIV pol sequences. Two participants had a sample obtained prior to receiving any ART, 72 participants had samples from firstline ART only (two had multiple first-line samples), 14 participants had samples from second-line ART only (four had multiple secondline samples) and 27 participants had samples obtained during first-line and second-line ART (Table S1, available as Supplementary data at JAC Online). The median sequencing coverage of all samples at the whole-genome level was 583 reads per base (IQR = 136-1313), 595 (IQR = 125-1324) for the pol gene and 735 (IQR = 162-1593) for the integrase L74 codon. Detailed coverage data are shown in Table S2. Four participants had major IAS-USA integrase resistance-associated mutations detected at low levels (2%-5% frequency). Two had Q148K and two had R263K minority variants (one of whom also had L74I). Another six participants had minor IAS-USA integrase resistance-associated mutations. Four had T97A [this was the consensus in three of them and a minority variant in one (who also had L74I)], another participant had an E138K minority variant and another had a G140A minority variant. The most common integrase polymorphism was E157Q, which was detected in 12 participants (6 of these also had L74I).
Forty-six (40%) of the participants had L74I detected at >2% frequency in at least one plasma sample. The participant characteristics of those with and without an L74I mutation were similar, with an overall median age of 30 years, two-thirds were female and the median CD4 cell count at first-line VF was 142 cells/mm 3 (IQR = 64-246) (see Table 1).
Thirty-three participants had more than one plasma sample and therefore intra-host changes could be evaluated. Eight participants had L74I as the majority allele (>85% frequency) in every sample ( Table 2). Four had low-level minority variants of less than 10% frequency and in two of these L74I was not detected in the preceding plasma samples. Two participants had reversion to low (<2%) or no L74I in subsequent samples.
Codon usage at the integrase 74 position was examined using consensus sequences. All sequences with L74I at consensus (27%, 31/115) had the trinucleotide ATA at this position. The remaining consensus sequences were all L74L. The trinucleotides CTG or TTG were the consensus codon in 60% (69/115), requiring two nucleotide changes to result in an amino acid substitution from leucine to isoleucine, and 13% (15/115) had either TTA or CTA, requiring only one nucleotide change to mutate to L74I. Synonymous changes in codon usage were noted in two of the participants from whom more than one sample was available. One changed from TTA to CTA (both requiring one further nucleotide substitution to code for I) and the other from CTG to CTA (from requiring two nucleotide substitutions to code for I to just one). There was no significant subtype difference in the number of substitutions required to mutate from L74L to L74I (P = 0.13).

Discussion
As we enter an era when dolutegravir-based ART will be used for tens of millions of PLWH, it is vital to understand determinants of treatment failure and drug resistance. To date, information on integrase sequences by next-generation sequencing in subtypes dominating in West Africa has been extremely limited. Recently, a series of VF patients in a trial of the integrase inhibitor cabotegravir were found to harbour L74I (a polymorphic mutation that is weakly selected for by INSTI therapy) in addition to other major integrase inhibitor drug resistance mutations. This prompted us to evaluate not only major integrase inhibitor mutations, but also L74I.
Reassuringly, we found that major resistance mutations to integrase inhibitors were very rare in this study population that had extensive NRTI and NNRTI resistance following VF with NNRTI-based regimens. One individual out of 115 had both L74I and the signature dolutegravir mutation R263K detected by next-generation sequencing, 23-26 though R263K was a minority variant. Missing data: CD4 cell count, 1 (L74I, 0; no L74I, 1); and HIV-1 RNA, 2 (L74I, 1; no L74I, 1).
Integrase L74I in West African HIV strains

JAC
We found over a third of the integrase inhibitor-naive HIVpositive participants in our study had the integrase L74I mutation and that L74I was more common in HIV-1 subtype G than CRF02_AG. This is the first time that such a high prevalence of L74I mutations has been reported in West African G and AG subtypes, with potential implications for the effectiveness of dolutegravir, which is now being rolled out as part of the first-line treatment in this setting, where there is less frequent viral load monitoring and less access to genotypic resistance testing. Bouzidi et al.
Uniquely, in this study we were able to assess the frequency of viral variants with L74I in longitudinal samples from multiple individuals. In most cases there was no change in variant frequencies, consistent with L74I being transmitted between individuals following a founder effect and L74I reverting rarely, even during second-line ART failure. This would be consistent with a lack of fitness cost of L74I in the absence of drug pressure.
A limitation of this study was that our patient group were mainly ART-experienced and as such there may be a different prevalence of L74I in treatment-naive individuals. Furthermore, although L74I was associated with VF in two studies including the long-acting injectable cabotegravir, it is not known whether L74I contributed to VF or what the impact on dolutegravir might be.
In vitro studies are needed to determine whether L74I facilitates high-level INSTI resistance in non-B subtypes and clinical studies are necessary to determine whether L74I at baseline impacts clinical or virological outcomes on integrase inhibitors, even when short-term outcomes in cross-sectional studies appear favourable. Transparency declarations C.A.S. has received honoraria from Gilead Sciences, ViiV Healthcare and Janssen-Cilag for membership of Data Safety and Monitoring Boards and Advisory Boards, and for preparation of educational materials. R.K.G. has received speaker fees for ad hoc consulting from Gilead and ViiV Healthcare. All other authors: none to declare.