Prevalence and dynamics of the K65R drug resistance mutation in HIV-1-infected infants exposed to maternal therapy with lamivudine, zidovudine and either nevirapine or nelfinavir in breast milk

Abstract

Background

K65R is a relatively rare drug resistance mutation (DRM) selected by the NRTIs tenofovir, didanosine, abacavir and stavudine and confers cross-resistance to all NRTIs except zidovudine. Selection by other NRTIs is uncommon.

Objectives

In this study we investigated the frequency of emergence of the K65R mutation and factors associated with it in HIV-1-infected infants exposed to low doses of maternal lamivudine, zidovudine and either nevirapine or nelfinavir ingested through breast milk, using specimens collected from the Kisumu Breastfeeding Study.

Methods

Plasma specimens with viral load ≥1000 copies/mL collected from HIV-infected infants at 0–1, 2, 6, 14, 24 and 36 weeks of age and maternal samples at delivery were tested for HIV drug resistance using Sanger sequencing of the polymerase gene. Factors associated with K65R emergence were assessed using Fisher's exact test and the Wilcoxon rank-sum test.

Results

K65R was detected in samples from 6 of the 24 infants (25%) who acquired HIV-1 infection by the age of 6 months. K65R emerged in half of the infants by 6 weeks and in the rest by 14 weeks of age. None of the mothers at delivery or the infants with a positive genotype at first time of positivity had the K65R mutation. Infants with K65R had low baseline CD4 cell counts (P = 0.014), were more likely to have DRMs earlier (≤6 weeks versus ≥14 weeks, P = 0.007) and were more likely to have multiclass drug resistance (P = 0.035). M184V was the most common mutation associated with K65R emergence. K65R had reverted by 3 months after cessation of breastfeeding.

Conclusions

A high rate of K65R emergence may suggest that ingesting low doses of lamivudine via breast milk could select for this mutation. The presence of this mutation may have a negative impact on future responses to NRTI-based ART. More in vitro studies are, however, needed to establish the molecular mechanism for this selection.

Introduction

The introduction of antiretroviral chemoprophylaxis for prevention of mother-to-child transmission (PMTCT) has led to significant reductions in paediatric HIV acquisition.^1,2 Despite this success, a significant number of infections still occur; in 2014, ∼220 000 paediatric infections occurred worldwide with 99% in sub-Saharan Africa.¹ Subsequent care, treatment and survival of the infected infants are issues of concern, with treatment success being hampered by drug resistance.^3–12 Previous studies have documented the emergence of both acquired and transmitted drug resistance variants in infants.^3,4,7–9,11 Acquired drug resistance occurs due to incomplete suppression of viral replication during administration of antiretroviral drugs to infants or by ingestion of maternal drugs through breast milk. The latter has been shown to result in drug resistance mutation (DRM) patterns including multidrug resistance, defined as resistance to more than one drug within NRTIs or NNRTIs, and multiclass resistance, defined as resistance to two or more classes of antiretroviral drugs, such as resistance to NRTIs and NNRTIs.³ Emergence of multidrug resistance and multiclass resistance has significant implications as they could limit future treatment options for the HIV-infected infants.

One multidrug resistance mutation of concern is K65R as it causes the loss of drug susceptibility of HIV-1 to all NRTIs with the exception of zidovudine.^13–17 The K65R mutation is usually selected by tenofovir disoproxil fumarate, didanosine, abacavir and, to a lesser extent, stavudine usage.^14,16,18 Its occurrence is relatively infrequent, but there has been a significant rise with increased use of tenofovir.^16,17 Little information exists on selection of K65R with lamivudine usage, while the use of zidovudine has been shown to inhibit its emergence.^16,19

In the present study, we investigated the frequency of K65R emergence and factors associated with it in HIV-infected infants exposed to low doses of zidovudine, lamivudine and either nevirapine or nelfinavir via breast milk in the Kisumu Breastfeeding Study (KiBS) conducted in 2003–09.

Methods

Study participants

KiBS was a Phase IIb PMTCT single-arm and non-randomized trial that aimed to assess the safety and efficacy of a triple-antiretroviral drug regimen consisting of zidovudine, lamivudine and either nevirapine or nelfinavir from 34 weeks of gestation to 6 months post-partum for PMTCT among HIV-infected breastfeeding mothers. Single-dose nevirapine was administered to infants within 72 h of birth. Details of the study findings have previously been described.²⁰ Both mothers and infants were followed for up to 18 months after cessation of breastfeeding. The current analysis focuses on the period up to 9 months post-partum. Infants were initiated on ART based on CD4 cell counts with the regimen being determined by a clinical team, as there were no standard guidelines for ART for infants in Kenya at that time.

Specimen collection and laboratory testing

Whole blood samples were collected from the infants in EDTA-treated anti-coagulant Vacutainers (Becton Dickinson, Franklin Lakes, NJ, USA). Haematological and virological laboratory tests were performed at 0–1, 2, 6, 14, 24 and 36 weeks of age. Roche DNA-PCR, version 1.5 (Roche Diagnostics Systems, Branchburg, NJ, USA) was used to diagnose HIV-1 infections in infants at 14, 24 and 36 weeks with dried blood-spot specimens. For those infants who were HIV-1 DNA PCR positive using dried blood spots, PCR was performed retrospectively on previously collected specimens to determine the timing of HIV infection. CD4 cell counts and viral load were assessed during multiple study visits using FACSCalibur (Becton Dickinson, Franklin Lakes, NJ, USA) and the Roche Amplicor HIV-1 Monitor Test, v 1.5 (Roche Diagnostics Systems, Branchburg, NJ, USA), respectively.

HIV-1 genotyping and viral subtyping

HIV drug resistance genotyping was conducted retrospectively on plasma specimens collected from infants with viral loads of ≥1000 copies/mL at 14 weeks of age, using the ViroSeq™ HIV-1 Genotyping System (Applied Biosystems, Foster City, CA, USA). For those infants with detectable DRMs at 14 weeks of age, progressively earlier samples (6, 2 and 0–1 weeks) were then analysed to determine the time when the HIV DRM first emerged. In addition, drug resistance testing was conducted on all infant samples collected at 24 and 36 weeks of age by use of either ViroSeq or an in-house assay previously described.²¹

Analysis of DRMs was performed with the ViroSeq HIV-1 genotyping analysis software, version 2.6, and with the Stanford HIVDB algorithm (Stanford University, CA, USA) together with the International AIDS Society-USA mutation list-2011 update.^22,23 To determine the extent of genetic diversity, sequences obtained were aligned using the Se-Al 1.0 multiple sequence alignment program, and subjected to phylogenetic analysis using parsimony (PAUP*) 4.0b software along with HIV-1 reference sequences from an HIV database (Los Alamos database). HIV-1 inter-subtype recombination was analysed using Simplot software, version 3.5.1.²⁴

Statistical analysis

Fisher's exact test was used to compare categorical variables and the Wilcoxon rank-sum test was used to compare continuous variables to determine factors associated with K65R emergence.

Ethics statement

This study was approved by the Ethical Review Committee of the Kenyan Medical Research Institute (KEMRI), as well as the Institutional Review Board of the US CDC, Atlanta, GA, USA. All mothers provided written informed consent that included parental consent for their infants.

Results

Based on the previously published data^3,20 a total of 32 infants acquired HIV-1 during the study period, 24 (75%) of whom were infected by the age of 6 months. Among the 24 infants, 16 developed DRMs during the 6 month breastfeeding period. Table 1 shows that among the 16 infants that had DRMs detected, 17% and 13% already had multidrug resistance mutations and multiclass resistance mutations at the first time of DRM emergence, respectively. The K65R mutation was detected in 6 (25%) of the 24 infants who acquired HIV-1 infection by the age of 6 months; these 6 were among the 16/24 who had detectable DRMs (6/16, 37.5%). Of these 16 infants, 10 had a genotype at the first time of positivity and none had K65R mutation. Of the remaining six, only two had the K65R mutation at the next earliest timepoint (Table 1). Most infants had documented HIV-1 infection by the age of 2 weeks, and the median time of K65R detection from these infants was at the age of 14 weeks (Table 1). In addition, the K65R mutation was only detected from specimens collected during the breastfeeding period, at which time none of the infants was on ART for their own treatment.

Patterns of K65R mutation evolution in lamivudine-exposed infants

To study the evolution of the K65R mutation in these infants, we genotyped plasma specimens collected from the 16 infants at 0–1 week and up to 9 months of age (Figure 1). The K65R mutation was first detected in three infants by 6 weeks of age, while it was detected in the remaining infants at 14 weeks of age (Table 2). Of the six infants, four had earlier genotypes, three of which had WT strains and one had only the M184V mutation (Table 2). Three of the mothers had an undetectable viral load at delivery, which was sustained through the breastfeeding period. Of the remaining three, two had sufficient viral load at delivery and yielded a WT genotype.

The most common mutation detected in infants with K65R was M184I/V [5 of 6 (83%)]. Some infants also had accompanying NNRTI mutations, which included K101E, K103N, Y181C and G190A. The K65R mutation was undetectable by 6 months of age in four infants and by 9 months in the rest (Table 2). Four of the infants initiated treatment by 6 months [two on PI regimens (lopinavir/ritonavir/abacavir/zidovudine) and two on NNRTI regimens (nevirapine/lamivudine/zidovudine)]. None of these infants had achieved viral suppression by 9 months, but had persisting reverse transcriptase inhibitor mutations with no re-emergence of the K65R mutation. Sequences analysed in this study have been submitted to GenBank and their accession numbers are HM164112-HM164123, HM164127-HM164128 and HM164130-HM164131.

Factors associated with selection of the K65R mutation

Table 3 summarizes factors associated with K65R mutation, considering all infants who acquired HIV-1 infection by 6 months of age. Infants with K65R were more likely to have developed DRMs by 6 weeks of age compared with those without the mutation (P = 0.007). They were also likely to have multiclass resistance (P = 0.035), as well as low baseline CD4 cell counts (P = 0.014). Maternal regimen (nevirapine based versus nelfinavir based), viral load and CD4 cell counts at delivery and infant's baseline HIV-1 RNA level and timing of infection (before or after 1 week of age) were not significantly associated with K65R mutation emergence (Table 3).

Subtype distribution

Four (75%) of six infants with the K65R mutation were infected with subtype A viral strains, which were also the predominant variant among all the HIV-1-infected infants in this study. The remaining two infants were infected with an A/D recombinant.

Discussion

We report a high incidence of K65R mutations in infants exposed to sub-optimal concentrations of antiretroviral drugs through breast milk. Our previous study showed the occurrence of DRMs in breastfeeding infants by 6 month of age and illustrated that these mutations were acquired, rather than transmitted, through ingestion of sub-optimal dosages of the maternal regimen in breast milk.³ None of the infants genotyped at first time of positivity or mothers with a detectable viral load at delivery had the K65R mutation. The mothers of these infants were ART naive at treatment initiation and remained exclusively on the study regimens.^3,20 NRTI mutations were the most commonly detected mutations in these infants. A notable finding was the high incidence of K65R in infants exposed to regimens infrequently known to select for this mutation. We postulated that this mutation could have been selected by lamivudine. In our previous assessment of drug levels in these infants and their mothers we showed that lamivudine and nevirapine were the only drugs detected in infants through ingestion of breast milk in quantities sufficient to have therapeutic effect.²⁵ In that study we reported drug levels in breast milk and infant's blood at 2, 6, 14 and 24 weeks post-partum; zidovudine was detected at below the quantitative limit (bql) in the infant samples and a low of 9 ng/mL (IQR bql–26) in breast milk. Lamivudine, on the other hand, was detected at high levels in the infant plasma (an estimated daily intake of 182 μg/kg, ∼2% of the daily recommendation) and an average of 23 ng/mL (during the assessment period of 2–24 weeks), which is just slightly above the IC₅₀ of lamivudine for WT strains (0.6–21 ng/mL).^25,26 This led to our postulate that the emergence of these mutations was due to exposure to sub-optimal levels of lamivudine. A few studies have also reported the occurrence of K65R in patients treated with a lamivudine/zidovudine-containing regimen, with an estimated prevalence of <0.5% from the Stanford University HIV Drug Resistance Database²³ and <6% from individual studies.^27–32

K65R has a low genetic barrier and occurs through a point mutation involving a single purine–purine transition (AAA to AGA for non-subtype C viruses), which would explain its fast selection.¹⁶ However, the frequency of K65R occurrence in the general population is usually low and this may be attributed to such factors as impaired viral replicative capacity of K65R mutants, reduced viral fitness in the presence of M184V, counter-selection with thymidine analogue-associated mutations (TAMs) and high potency of available regimens.^16,19,33,34 The high incidence observed in our study may be attributed to the continuous exposure of the virus to low doses of lamivudine ingested in breast milk. In a previous study, we showed that lamivudine and nevirapine were transferred from mothers to infants via breastfeeding and this led to the occurrence of acquired resistance to both drugs.^3,25 This was also reported in the PEPI-Malawi study, which showed that infants whose mothers were given post-partum PMTCT were likely to have multiclass drug resistance associated with the maternal regimen. In that study 11/37 infants developed resistance during breastfeeding, 7 of whom had the K65R mutation.⁹ The mothers, however, were on a stavudine-containing regimen, which has been associated with selection of the K65R mutation. As reported in other studies,^9,35 we observed a reversion of the K65R mutation within 2–3 months of its emergence. This reversion could be a result of a reduction in the fitness of the K65R mutant, as reflected in its low replicative capacity, especially when it co-emerges with M184V or TAMs. The M184V mutation emerged in 5 (83%) of 6 infants with K65R and had not reverted by 9 months. An additional explanation could be the fact that these infants were weaned at 5.5 months and hence were no longer exposed to the low doses of lamivudine. This corresponded to the pharmacokinetic data, which showed plasma levels below the quantitative limit at 24 weeks. Four of the infants had initiated treatment at 6 months, but had not achieved viral suppression by 9 months. In addition, they had persisting reverse transcriptase inhibitor mutations, but with no re-emergence of the K65R mutation.

We further investigated factors associated with the K65R mutation. A low baseline CD4 cell count, early occurrence of any DRM and the presence of multiclass resistance were significantly associated with K65R emergence. Most of the infants were likely to have been infected prenatally, at delivery or 1 week post-partum. Since the majority of the infants with K65R mutations had first acquired mutations by 6 weeks of age, it is likely that cumulative accumulation of mutations could have been averted through immediate treatment initiation. In addition, the development of multiclass resistance in these infants further supports the need for early diagnosis and timely treatment of HIV-infected infants.³⁶

In agreement with other studies was the absence of TAMs in almost all of the infants in whom K65R emerged.^14,16,19 TAMs are known to act antagonistically with K65R, restoring the catalytic activity of the reverse transcriptase enzyme initially lost due to the presence of this mutation, and for this reason they rarely co-emerge in the same genome.^16,19 In two instances, occurrence of TAMs followed reversion of K65R. We also observed the joint occurrence of NNRTI mutations K103N, G190A, K101E and Y181C with the K65R mutation. Previous studies have reported a synergistic fitness interaction between K65R and G190A/Y181C and a negative association with K103N.^16,37 This uncommon occurrence of K103N together with K65R in these infants may highlight the potential risk of unusual combinations of mutations as a result of exposure to sub-optimal doses of antiretroviral drugs in breast milk.

The findings of this study highlight the challenges in PMTCT strategies following paediatric HIV infection. The recent WHO guidelines recommending ‘test and treat’ strategies for HIV-infected infants are likely to reduce the occurrence of DRMs, including K65R. However, logistical and financial challenges in resource-limited settings leading to late HIV diagnosis may result in findings similar to those observed in this study. Currently, early diagnosis of HIV in infants is still a challenge in Kenya, and furthermore only a small percentage of HIV-infected infants (21%) are on therapy.³⁸ As a result there is need to strengthen early infant diagnosis programmes in Kenya to ensure timely access of HIV-1-infected infants to care and treatment.

There are limitations in this study. First, we were unable to conduct in vitro experiments to elucidate the molecular mechanism of K65R selection by lamivudine. Second, the follow-up period was short and insufficient to assess the long-term impact of the K65R mutation. Third, because of the small numbers we were not able to assess potential confounders of the predictors of K65R occurrence.

Finally, the use of conventional sequencing-based genotyping technique with sensitivity <20% may have limited the detection of minority K65R mutants and hence underestimated the prevalence of K65R in these infants. It is also possible that the K65R mutation could have persisted with time, but at low frequency.

In conclusion, a high incidence of K65R was observed among HIV-infected breastfeeding infants whose mothers were taking zidovudine, lamivudine and either nevirapine or nelfinavir. We postulate that continuous exposure to low dosages of lamivudine through breast milk likely selected for this mutation. Future studies are, however, needed to elucidate the mechanism of K65R selection with sub-optimal doses of lamivudine. The emergence of the K65R mutation in these infants may have a negative impact on the infant's future regimen selection and treatment outcomes. The current recommendation for early diagnosis, immediate treatment and using a PI-based regimen is likely to prevent the emergence and accumulation of such mutations acquired through exposure to the maternal regimen.

Funding

This research was supported by the President's Emergency Plan for AIDS Relief (PEPFAR) through a cooperative agreement between the US CDC and the Kenya Medical Research Institute under the terms of Grant Award Number-5U19C1000323-05. The antiretroviral drugs used in this study were provided by GlaxoSmithKline and Boehringer Ingelheim.

Transparency declarations

None to declare.

Kenya Medical Research Institute and US CDC investigators were involved with all aspects of study design, data collection and analysis, interpretation of findings, report writing and the decision to submit the manuscript.

Disclaimer

The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the US CDC. Use of trade names is for identification purposes only and does not constitute endorsement by the US CDC or the Department of Health and Human Services.

Acknowledgements

We are grateful to the study participants, the KiBS, the KEMRI/CDC HIV research laboratory, Kenya Medical Research Institute and Kenya Ministry of Health, whose participation made this study possible. This paper is published with the permission of the Director of KEMRI.

References