Background
HIV drug resistance remains a major threat to the success of scaling up of combination antiretroviral therapy (cART) in developing countries, especially in Africa where about 60% of HIV-infected individuals were on cART in 2012 [
1]. The increased availability of cART has led to remarkable treatment results in most programs in sub-Saharan Africa (SSA), leading to a significant reduction of new infections [
2]. Despite this success, some patients with incomplete adherence to treatment protocols might fail therapy and develop resistant viruses [
3]. One challenge of cART in SSA has been the inability to routinely monitor virological parameters such as viral load (VL) and drug resistance, which are important to guide patient management and choose the right salvage treatments [
4,
5]. Recent WHO recommendations stipulate that VL monitoring be utilized as the preferred method to determine treatment failure in SSA [
6]. Within the past years, VL monitoring has in fact been successfully implemented as part of care in South Africa [
6].
Sequencing is the most accurate method to track the emergence of DRM and is typically done by Sanger sequencing. However, cost and availability of infrastructure to perform these assays still remain major obstacles [
7]. As a consequence, most patients failing therapy remain on first-line therapy for long periods causing an accumulation of mutations, which might lead to the emergence of high-level resistance [
8,
9]. This is particularly true for patients who started treatment long before VL was more widely available. Samples from such long term cART patients provide a good source to track previous DRM which can be found only in the cellular compartment.
Genotypic drug resistance studies of subjects with unsuppressed VL have been performed in several SSA countries including South Africa and suggest a high level of DRM as well as some subtype specific mutations [
10‐
16]. For example, previous studies among South Africans reported that infections with HIV-1 subtype C may lead to a rapid emergence of certain non-nucleoside reverse transcriptase (NNRTI) or nucleotide reverse transcriptase inhibitor (NRTI) DRM, which are less common among other subtypes. Specifically, V106M (NNRTI), K65R (NRTI) and N348I are more common among subtype C strains compared to other group M strains [
17‐
19].
The South African treatment protocols are consistent with the 2016 WHO guidelines, which recommend a combination of two NRTIs and one NNRTIs as a first-line regimen and two-NRTI and one protease inhibitor (PI) as second line. According to the April 2014 South African revised guidelines, a preferred first-line regimen could be: tenofovir (TDF)/lamivudine (3TC) + efavirenz (EFV)/nevirapine (NVP), or alternatively abacavir (ABC) + 3TC + EFV; or stavudine (d4T) + 3TC + EFV. For second line treatment, TDF + 3TC + lopinavir/ritonavir (LPV/r); or 3TC + AZT + LPV/r could be used.
Despite the benefits of cART, a significant number of patients still experience treatment failure, associated with the development of DRM due to incomplete adherence to therapy or the presence of transmitted DRM [
20,
21]. Given the changing treatment strategies and options, genotypic and phenotypic drug resistance patterns will continue to evolve. Although considerable data on virologic failure in patients with unsuppressed VL on first or second line treatment from South Africa are available, these studies have been mostly done in urban settings [
15,
22,
23]. On the contrary, rural settings (making up almost 40% of the South African population) still remain largely understudied. Such rural populations often face significant challenges such as distance to treatment sites and access to virological monitoring. In addition, cultural beliefs often increase stigma and could impact adherence and prevalence of DRM [
24,
25]. Two recent studies in Uganda and South Africa reported that the intensity of virological monitoring, frequency and pattern of DRM is influenced by the location (urban or rural) of the treatment site [
26,
27]. Such results illustrate that inadequate infrastructure and low per capita income common in rural settings might influence the successful implementation of cART. It is therefore imperative that frequent monitoring of DRM is implemented in order to better understand the relationship between drug resistance development and therapy failures in these rural settings.
The use of integrated proviral DNA rather than RNA isolated from circulating virus to monitor HIV drug resistance in patients failing treatment has been considered as an alternative method for drug resistance genotyping [
28]. Sequencing of DNA is also an efficient method to study the archived and persistent viral reservoirs in patients with low or suppressed viral loads [
29]. With the current focus on HIV persistence, it is important that studies investigating persisting viral populations during suppression are performed to better understand the viral dynamics and alternative future treatment options.
In this study, we describe the drug resistance profile of HIV-1 infected individuals experiencing unsuppressed VL attending two rural HIV treatment centres; the HIV/AIDS Prevention Group Wellness Clinic (HAPG) in Bela-Bela and the Donald Frazer Hope Clinic (DFHC) in Vhufhuli, both located in the Limpopo Province of northern South Africa. Our findings show a high prevalence of NNRTI and NRTI, but low level PI mutations. DRM were more prevalent in this rural setting than reported for a nearby urban setting in Pretoria. Furthermore, by comparing the DRM (NRTI and NNRTI) in viral RNA as well as proviral DNA among circulating viruses in plasma and peripheral blood compartments in 35 subjects, we found that the DRM in these two compartments were mostly similar with a slight dominance of thymidine analogue mutations (TAMs) in DNA, a result that necessitates further studies with a larger cohort.
Discussion
This study was performed among patients failing treatment in two treatment centres in rural South Africa: the HAPG Wellness Clinic which started cART administration in 2000, 4 years before the roll-out of antiretroviral therapy in the public health sector in South Africa; and the Donald Fraser Clinic which started providing treatment in 2004. Although both sites have been implementing VL and CD4+ T-cell count monitoring, according to national guidelines, and as part of HIV management for several years, the prevalence of DRM was still high. The M184V mutation was the most prevalent DRM, observed in 52% of study subjects. This coincided with the extensive use of 3TC, surpassed only by the use of TDF and EFV in the population. Interestingly, TDF mutations were one of the least common DRM in the study despite its frequent usage, supporting its inclusion as the main backbone during first line treatment.
Although currently used cART is effective for all group M subtypes, the pattern of drug resistance differs between subtypes [
10,
36]. Our study confirmed the dominance of K65R and V106M, two mutations which are most common among subtype C strains [
36,
37]. In fact, very recently in a cohort of more than 3700 subjects failing cART in several West and Central Africa countries, these two DRM were found to be more prevalent among subtypes C strains compared to CRF02_AG, CRF06cpx, and subtypes A and G [
17,
18,
36]. This dominance has been attributed to the genomic sequence of subtype C, which requires only a single nucleotide change at these codon positions to transition to a drug resistant amino acid residue [
37]. Similarly, we speculate that the occurrence of the E138A/K/Q mutation could be due to the codon usage of subtype C RT at this position which is GAA versus GAG in other group M strains. In fact, in the Stanford Drug Resistance database, subtype C strains from drug exposed patients have a higher prevalence of the E138A/K/Q mutation compared to other group M strains (8–9.5% vs. 2–5%). Further verification of such subtype-specific differences in the pattern of resistance mutations would clearly be of great significance for treatment management and also for a better understanding of the overall resistance mechanism.
The high occurrence of both K103N and V106M could be a reflection of the wide use of EFV and NVP as part of the first line therapy. In addition, the high levels of V106M mutations could also be attributed to subtype-specificity, since this mutation had been found most frequently among subtype C [
17,
36]. It seems quite clear that NNRTIs should not be used for second line treatment in rural South African populations. This conclusion is supported by the fact that a significant number of the patients we studied already show resistance to newer NNRTIs (ETR and RPV), which have not even been introduced in South Africa.
The selection of TAMs such as D67N, M41L and T215Y could be attributed to treatment with the thymidine analogues AZT and d4T, which also cause cross resistance to other NRTIs [
8]. Similarly, patients harbouring TAMs also had high level resistance to NNRTIs (EFV and NVP) and 3TC supporting the point that they were likely failing thymidine containing ART at the point when TDF replaced AZT in the national guidelines [
38]. TAMs have been reported to be more common among CRF06cpx strains than other HIV-1 subtypes or CRFs [
10,
36]. In this study, there was a noticeable absence among subjects of the L210W and very low occurrence of the M41L DRM, both components of the Type 1 TAMs pathway (M41L, L210W and T215Y). Although, this could likely be related to the reduced usage of d4T, recent reports suggest that compared to other dominant HIV-1 genotypes (subtypes A, CRF02_AG and CRF06cpx), subtype C show the lowest prevalence of M41L DRM [
36]. The absence of M41L and L210W was initially reported from a subtype C cohort in Botswana where the most common TAM pathway was D67N, K70R and T215Y [
39]. In our study, we observed that the four TAMs (D67N, K70R, T215Y and K219Q/E) which make up the Type 2 TAM pathway, occurred more frequently and correlated well with the usage of AZT. Although we could not state categorically the order of emergence of these mutations as this was a cross sectional study, the mutations D67N and K70R were the most frequent single occurring TAMs. Both AZT and d4T have been reported to cause TAMs at an equal rate [
40].
The majority of participants in this study were adults >18 years old. This is not surprising given that treatment of children and adolescents with cART has not been vigorously pursued in SSA until recently. Out of the 13 subjects <18 years old with unsuppressed virus, DRM could be detected in 70% (6 were TAMs and 3 were either NNRTI or other NRTI). DRM has been reported to be highly prevalent among African children with unsuppressed VL [
41]. In Uganda and South Africa, DRMs were higher in children from rural compared to urban clinics indicating a better management and/or easier adherence in urban clinics [
26,
27]. Due to the low number of children and adolescents in this study, we could not independently confirm this. With the increasing usage of cART in children and adolescents in SSA, more longitudinal studies are required to better understand the pattern of DRM among this group.
The observed high levels of DRM in these rural settings could be linked to several factors. First, the availability of transportation to treatment centers is generally significantly reduced in rural areas compared to urban settings. Secondly, shortage of cART supplies are common occurrences in rural areas. Finally, cultural differences or stigma might affect the open administration of cART especially among women, a phenomenon that is more strongly present in rural areas. Another factor that likely influenced the high rates of DRM in one of the study sites (HAPG) was the inadequate planning and poor follow-up by health authorities. In 2009, at the HAPG privately run facility in Bela-Bela, patients’ healthcare providers were switched as patients were transferred from HAPG to a newly opened nearby Public Health Clinic (PHC). Records dating from the beginning of cART therapy at HAPG in 2004 showed good patient follow-up and less treatment failure. However, after the massive patient transfer in 2009 to the PHC and their return to the HAPG private clinic in 2012, the rate of virologic failure increased significantly among returning patients. Despite the high prevalence of DRM observed in this study, the number of patients with detectable VL, but no DRM was similar to previous reports (16% vs. 19–36% respectively) in some urban centres in South Africa [
22,
36,
42,
43].
There are few reports on the prevalence and pattern of DRM in rural South Africa [
26,
30,
44]. A recent study compared the frequency of DRM from rural KwaZulu-Natal and urban Pretoria, and found a clear difference between these two sites [
26]. In our current study, we confirmed the findings of Rossouw et al. Notably, the number of DRMs at the two sites in rural Limpopo were generally higher than in neighboring urban Pretoria (Table
2). Only two DRMs were substantially more common in Pretoria as compared with the Limpopo population. The mutations V106A/M and Y181C were 13.3 and 4.8% more common in Pretoria than in Limpopo, respectively. Both mutations are associated with exposure to NVP [
45]. At the time of sequencing, the Pretoria subjects were much more likely to be exposed to NVP than were the subjects in Limpopo (32.4% vs. 5%). Although subjects in these rural settings would be predicted to respond to the ART regimens that are currently in use, we suggest that the WHO recommendation that includes a PI in the second line treatment and close monitoring be strictly observed in order to reduce the transmission of DRM. This recommendation is supported by the very low level of PI DRMs observed in this study.
We compared DRM from both plasma (RNA) and cellular (proviral DNA) compartments from patients experiencing unsuppressed viral load who had received treatment for varying periods, ranging from 1 to 10 years. Plasma is routinely used for drug resistance testing, because viral RNA reflects the current replicating virus in blood. The prevalence of DRM in RNA and DNA was mostly similar with the exception of the TAMs, which were more prevalent in DNA than RNA. The persistence of some TAMs in DNA could be explained by the formerly common administration of NRTI (AZT and d4T) regimens. Of note is the fact that d4T is gradually being phased out and was not included in any of the current treatment regimens received by the study subjects. Three viral RNA and two proviral DNA derived sequences from patients who had never been treated with ddI, ABC, or TDF harboured a K65R mutation.
It is important to point out that our analyses were done using routine population-based Sanger sequencing that does not allow the detection of minority viral variants which comprise less than 20% of the viral population in a patient sample [
46,
47]. Recent methods such as next generation sequencing would most likely detect minor drug resistant variants at levels as low as 1% [
47,
48]. Considering the fact that undetected drug resistant minority variants may persist when cART is discontinued or changed, it will be worthwhile to perform comparative RNA/DNA drug resistance studies using more sensitive next generation sequencing [
47]. Although advances in cART has improved the lives of HIV infected patients, current regimens have failed to eliminate latent reservoirs, and it is acknowledged that eradication of the virus is not possible with current cART regimens alone [
49]. Furthermore, the efficacy of switching drugs for patients experiencing a rebound in VL will depend mainly on the number of DRM persisting in the proviral reservoir, following previous therapeutic failures [
50,
51].
Overall, these results clearly show that patients receiving cART, but experiencing therapeutic failure, harbour diverse drug-resistant variants, which in some cases persisted in the cellular blood compartment. Detection of this persisting variants can be improved by using more sensitive methods such as next generation sequencing. The proviral variants observed in this study closely resemble the population found in the plasma compartment (viral RNA) with the exception of some persistent TAMs; confirming the known fact that proviral DNA constitutes a reservoir for drug-resistant variants which might replenish plasma pool during suboptimal therapy.
Two limitations should be noted in our investigation: first, this was a cross-sectional study in which data on adherence to treatment was not available; as a result, it was not possible to associate the observed virologic failure to poor adherence. Second, information on prior exposure to NVP in the early days of mother-to-children transmission (MTCT) prevention efforts was also not available. The use of nevirapine monotherapy could have potentially impacted on the observed level of NNRTI mutations particularly K103N.
In conclusion, this study suggests that there is a high level of DRM in rural South Africa and confirms recent observations from KwaZulu-Natal that DRMs are more prevalent in rural than urban areas. It also confirms that long term ART with no virological monitoring could lead to an accumulation of DRMs especially of the NRTI class. Although plasma and PBMCs presented largely identical DRMs, the later compartment may act as an archive of drug resistant variants, especially TAMs, an observation that needs further studies with a larger sample size including subjects with suppressed virus. Such studies could shed more light on the persistence of DRMs and their impact on HIV treatment and cure.
Authors’ contributions
DMT, PB, DR and MLH designed the study; EME, CM, LM, KMG were involved in sampling and performing experiments; PJ interpreted the data and participated in writing the manuscript. All authors read and approved the manuscript.