Surveillance of HIV-1 pol transmitted drug resistance in acutely and recently infected antiretroviral drug-naïve persons in rural western Kenya

HIV-1 transmitted drug resistance (TDR) is of increasing public health concern in sub-Saharan Africa with the rollout of antiretroviral (ARV) therapy. Such data are, however, limited in Kenya, where HIV-1 drug resistance testing is not routinely performed. From a population-based household survey conducted between September and November 2012 in rural western Kenya, we retrospectively assessed HIV-1 TDR baseline rates, its determinants, and genetic diversity among drug-naïve persons aged 15–59 years with acute HIV-1 infections (AHI) and recent HIV-1 infections (RHI) as determined by nucleic acid amplification test and both Limiting Antigen and BioRad avidity immunoassays, respectively. HIV-1 pol sequences were scored for drug resistance mutations using Stanford HIVdb and WHO 2009 mutation guidelines. HIV-1 subtyping was computed in MEGA6. Eighty seven (93.5%) of the eligible samples were successfully sequenced. Of these, 8 had at least one TDR mutation, resulting in a TDR prevalence of 9.2% (95% CI 4.7–17.1). No TDR was observed among persons with AHI (n = 7). TDR prevalence was 4.6% (95% CI 1.8–11.2) for nucleoside reverse transcriptase inhibitors (NRTIs), 6.9% (95% CI 3.2–14.2) for non- nucleoside reverse transcriptase inhibitors (NNRTIs), and 1.2% (95% CI 0.2–6.2) for protease inhibitors. Three (3.4% 95% CI 0.8–10.1) persons had dual-class NRTI/NNRTI resistance. Predominant TDR mutations in the reverse transcriptase included K103N/S (4.6%) and M184V (2.3%); only M46I/L (1.1%) occurred in the protease. All the eight persons were predicted to have different grades of resistance to the ARV regimens, ranging from potential low-level to high-level resistance. HIV-1 subtype distribution was heterogeneous: A (57.5%), C (6.9%), D (21.8%), G (2.3%), and circulating recombinant forms (11.5%). Only low CD4 count was associated with TDR (p = 0.0145). Our findings warrant the need for enhanced HIV-1 TDR monitoring in order to inform on population-based therapeutic guidelines and public health interventions.

Introduction sampling method was used to recruit study participants aged 15-59 years from 3,302 randomly selected households. Those who volunteered to participate in the study were interviewed and tested for HIV using a serial rapid testing algorithm. CD4 cell count, recency of HIV infection, and viral load for those who tested HIV positive were measured using validated laboratory procedures. Of the 6,076 persons eligible for NHIPS, 1,457 (24.1%) were confirmed as HIV-1 infected [37]. We collected the baseline characteristics of HIV-1 infected persons from the NHIPS records, and drug-naïve cases with AHI/RHI infection (n = 93) were included in the study.

Ethical considerations
The NHIPS was approved by the Ethics Committee of the Kenya Medical Research Institute (KEMRI), Kenya; the Committee for the Protection of Persons, Saint-Germain-en-Laye, France; and the Division of HIV/AIDS Prevention (DHAP) under the Centers for Disease Control and Prevention (CDC), Atlanta, US. All the participants provided written informed consent to be involved in the study. For participants younger than 18 years of age but older than 15 years of age and not a mature minor, the study details were explained to at least one parent/guardian for parental written permission and then to the minor for written consent. Mature minors are defined in the Kenya National voluntary counselling and testing guidelines as persons under 18 who are married, pregnant, parents, engage in behaviour that put them at HIV transmission risk or are child sex workers. However, mature minors were strongly advised to speak with and involve their parents/guardians in the study participation process if they felt it was safe to do so. All minors, unless matured, had parental/guardian written permission and also provided written consent.

HIV status and duration of recency
HIV status was determined using a parallel testing algorithm of two HIV rapid tests: Uni-Gold TM HIV (Trinity Biotech, Ireland) and Determine TM HIV-1/2 (Abbott Laboratories 1 , Tokyo, Japan). The recency of HIV infection was established by two serological assays: the Limiting Antigen (LAg) (Maxim Biomedical, Inc., Rockville, USA) and BioRad Avidity Index (AI) (Bio-Rad Laboratories, WA, USA) enzyme immunoassays (EIA). Assay cut-offs for the duration of recency were 1.5 normalized optical density (ODn) and 30% AI, respectively, for the two serological assays. Any sample that was identified as recently infected from either assay, was included in the study. Persons with AHI were identified by negative result with Determine TM (Abbott Laboratories 1 , Tokyo, Japan) and positive result with nucleic acid amplification test (NAAT) [37]. CD4 cell count and viral load were measured using PIMA point-of-care CD4 Analyzer (Alere, Inc., Waltham, MA, USA) and COBAS AmpliPrep/COBAS TaqMan v2.0 (Roche Molecular Systems, Branchburg, NJ, USA), respectively, according to the manufacturers' instructions.

Phylogenetic analyses
HIV-1 subtyping was statistically done by maximum likelihood (ML) phylogenetic reconstruction in MEGA6; a bootstrap of 1000 replicates was used to generate the consensus evolutionary tree and assess the strength of cluster from each node. Nineteen known reference sequences from Los Alamos National Laboratory HIV sequence database (http://hiv-web.lanl.gov) of common HIV-1 subtypes in Kenya were included to deduce the subtypes of the ingroup. HIV-1 subtype K was used as an outgroup species. Multiple sequence alignment was achieved by MUSCLE program within MEGA6. The alignment was trimmed to a length of 1035 bp. According to the Bayesian Information Criterion scores of 24 different nucleotide substitution models, the Hasegawa-Kishino-Yano with discrete Gamma distribution (0.42) (for non-uniformity of evolutionary rates among sites: 5 rate categories) and Invariable sites (0.54) (HKY +G+I) was selected as the ideal model to describe the substitution pattern. All positions containing gaps or missing data were completely eliminated. Subtree-Pruning-Regrafting-Extensive (SPR level 5) method was used as the ML heuristic option for tree inference due to its effectiveness. The initial tree was automatically generated by BioNJ. A moderate branch swap filter was set in order to allow for a semi-stringent exhaustive optimization of the branch lengths and improvements in log likelihood values [44].

Statistical analyses
The overall TDR and drug-specific TDR prevalence were assessed in accordance with WHO standards [20] and the 95% CI computed based on Wilson's approach [45]. Associations of TDR with categorical (gender, recency of HIV-1 infection, and HIV-1 subtype) and continuous (participant's age, CD4 cell count, and viral load) variables were computed by Fisher's exact and Wilcoxon signed-rank tests, respectively. The overall TDR prevalence and patterns were descriptively compared to previous surveys in East Africa.

Population baseline characteristics
Of the 6,076 eligible persons, 1,457 (24.1% after adjustment) were confirmed as HIV-1 infected, of which 93 (6.4%) were either acutely (n = 7) or recently-infected (n = 86) and ARV drug-naïve by self-report. None of the persons included in the survey reported being aware of or exposed to prophylaxis for prevention of mother-to-child transmission (PMTCT). Table 1 shows the key baseline characteristics of the 93 persons with AHI/RHI included in the analysis.

TDR prevalence and pattern comparison
We further compared the observed TDR with those observed in other regions of East Africa ( Table 2).
Except for Entebbe and Mombasa (current updates), the observed TDR was relatively higher than those of selected regions within East Africa, published over the last decade [11,16,[26][27][28][29][30][31][32]. All the class-specific TDRMs were higher than those reported in rural parts of East Africa [16,30,31]. Whereas only NRTI TDRM was the highest when compared with TDR data in Mombasa [26], the contrast was true when compared with Kampala [29].

Clinical significance of observed TDRMs
According to the Stanford HIVdb v7.0, predicted resistance to the first-line drugs among the patients with TDR was 75% intermediate-to high-level resistance to efavirenz and nevirapine (recommended first-line NNRTI drugs), 25% to the cytosine analogues; lamivudine and emtricitabine and about 13% to tenofovir, abacavir, stavudine and didanosine (recommended NRTI drugs), and rilpivirine and etravirine (NNRTI drugs). None of the patients had intermediateto high-level resistance to zidovudine or PI drugs (Table 3 and Fig 1).

Discussion
We report the existence of a moderate level TDR in rural western Kenya, a level higher than the estimated prevalence (7.4%) in East Africa [7] and the overall weighted TDR prevalence (5.6%) in the high HIV prevalence sSA [27]. Global TDR estimates range from 8-22% [8][9][10][11][12][13]25,46]. The few independent studies in Kenya and East Africa have cited TDR prevalence of between 1-13% [16,26,27,32] and 0-19%, respectively [10,11,[27][28][29][30][31]33]. Our results herein, including TDR prevalence, reflect information and activities as of 2012. Wide-scale and long duration of ARV use in this region could have contributed to the observed TDR prevalence [5,6]. Current ARV coverage among adults in this region is high, estimated at 50-69% [37] having been available for about nine years. Due to lack of temporal data especially after ARV scale-up [10], it remains unknown whether the 9.2% represents an increased TDR in this region.  [27]. The observed moderate TDR level in Ndhiwa is challenging to ART program due to increased risk of virological failure associated with pre-treatment drug resistance (PDR) [24]. The Stanford HIVdb revealed that about 9 of every 100 persons with AHI/RHI in Ndhiwa have viruses with mutations that confer at least potential low-level resistance to at least one of the ARV drug classes. Moreover, this together with the observed TDR have epidemiological consequences for further transmissions [14,22] owing to the high HIV incidence in this region [38]. Compared to other regions in sSA, the rates of TDR in East Africa have been shown to be increasing, with a projected annual increase of 29% since ART rollout [7]. Comparatively, the findings of this study showed a moderate level of TDR that was higher than that reported from most of the regions in East Africa [11,[27][28][29][31][32][33] especially in studies from rural settings [16,[28][29][30][31]. This can partly be explained by the high prevalence and incidence in the study region as well as the more accurate means for assessing incidence [37]. Among the studies compared, majority assessed TDR among persons initiating ART with unknown time of infection. Studies have shown possible archiving of TDR mutations due to comparatively higher viral fitness of the wild-type strains and it is likely that these studies may have underestimated the TDR levels as confounded by the choice of population studied [47]. The observed TDR in Ndhiwa was however lower than that observed in the urban settings of Mombasa (13.2%) and Entebbe (19.2%). The difference in this may be explained by the difference in the epidemic in the two regions. For example, in Mombasa HIV transmission occurs substantially through injection drug use (IDU) [48] while that in Ndhiwa it is mainly heterosexual. IDU has been associated with a higher risk of primary HIV transmission in many parts of the world [49]. On the other hand, ART rollout in Uganda and Kenya started in 2000 and 2003, respectively [27]. Thus, the length of ART use in Uganda is higher than that of Ndhiwa, which may partly explain the nearly 2 times TDR difference observed in the two regions, i.e., 19.2% in Entebbe (Uganda) against 9.2% in Ndhiwa [27].
As reported from other studies, NNRTI mutations comprised a majority of the detected TDRMs [10,16,17,[25][26][27]29,30,32], with K103N/S predominating. The high occurrence of this mutation is due to its ease of transmission compared with PI-and NRTI-associated TDRMs [50] and the tendency to persist for long durations [12,51]. Moreover, the observed high level could be due to the previous use of nevirapine monotherapy or nevirapine tail in short-course triple ART PMTCT, which has been associated with increased risk of resistance [37,52]. Contrary to other studies in East Africa, we also observed a high frequency of M184V mutation [16,26,28,[30][31][32]. The levels of observed NRTI mutations may suggest a change from the predominantly NNRTI to a combined RTI TDR epidemic, with majority of NRTI being M184V mutation. M184V is however generally highly revertant. Although studies have suggested a low fitness and hence transmission of M184V mutation, studies using more sensitive genotyping assays have shown that this mutation may still be transmitted [53] as minority variants. However, it is also likely that the observed mutations may be transmitted from patients on treatment who received non-standardized regimen or those failing treatment with resistance strains. Limited information exists pertaining acquired drug resistance from this population. Thus, more studies may be needed to verify whether the HIV TDR variants that we identified were transmitted from persons on HAART.
Phylogenetic analyses revealed that among persons with AHI/RHI, HIV-1 subtype A prevailed as the most frequent strain in circulation. HIV-1 subtypes C and G, and each of the variants of intersubtype recombinant mosaics (AD and A2D) were marginally present. Cumulatively, the frequency of CRFs was high. Some recombinants might indicate multiple infections [54] or independent transmission networks [55], but further analyses are essential to validate these. Consistent with previous molecular epidemiological assessments, we found subtypes A and D to be the most predominant strains that drive the epidemics in Kenya [16,26,32,56] and East Africa [11,[28][29][30][31]33]. Intersubtype AD remained the most common HIV-1 recombinant [16,31,33]. Great HIV diversity can foil accurate diagnosis, effective treatment and vaccine design [57].
Low CD4 cell count was significantly associated with TDR, a plausible indication of a rapid disease progression associated with drug-resistant strains. TDR has been shown to have a profound impact on HIV-1 pathogenesis in recently infected individuals. Individuals infected with HIV-1 strains with TDR have a rapid depletion of CD4 cell count during the first year of infection [58]. The associations of TDR distribution with other variables were not statistically significant. Though one study reported an association between TDR prevalence and AHI (mean seroconversion period: 45 days) [12], in our study, we found no association of TDR with the AHI/RHI period. The absence of TDR mutations in persons with AHI in our survey could have resulted from missed minority-level archived mutations by DBS genotyping [59].
Shortcomings of the study included potential for misclassification (bias) associated with the serological assays for characterizing the age of infection. Thus, there is chance that a few chronic infections might have been misclassified as RHI. To minimize this, we used recent infection testing algorithm (RITA), based on two serological assays and surrogate markers to provide information on recency [18]. The two serological assays that we used have been observed to have a low false positive rate (FPR) in this population; BioRad (2.4) and LAg avidity (0.5%) [60]. This variation can be due to influence by HIV-1 subtype [61]. The low FPR (<2%) of LAg avidity was consistent with the characteristics of an acceptable assay [35]. Because RITA cannot eliminate all FPR, viral load measurements were included to reduce the FPR [18]. As compared to massive parallel sequencing, the population-based sequencing might have not detected "sentinel mutations" in quasispecies, thus, possibly underestimating the TDR in this population [4,[62][63][64].

Conclusions
Despite these limitations, we reveal the presence of a relatively higher TDR prevalence in rural western Kenya than in most regions within East Africa. We propose that the WHO-recommended programmatic actions for moderate TDR level be performed, including evaluation of early warning indicators and both regional and nationwide PDR survey to determine the causes of TDR and advise on choice of first-line treatment in case of persistent high-level PDR [34].