Minority Variants of Drug-Resistant HIV

Abstract

Minor drug-resistant variants exist in every HIV-infected patient. Since these minority variants are usually present at very low levels, they cannot be detected and quantified using conventional genotypic and phenotypic tests. Recently, several assays have been developed to characterize these low-abundance drug-resistant variants in the large genetically complex population present in every HIV-infected individual. The most important issue is, what results generated by these assays can predict clinical or treatment outcomes and might guide patient management in clinical practice. Cutoff-values for the detection of these low-abundance viral variants that predict increased risk of treatment failure should be determined. These thresholds may be specific for each mutation and treatment regimen.

In this review we summarize the attributes and limitations of the currently available detection assays and review the existing information about both acquired and transmitted drug resistant minority variants.

Introduction

HIV replicates at high rates with estimates of 10¹⁰–10¹¹ virions being generated daily in untreated patients [1]. RNA-virus polymerases have high error rates that are not subject to host cell proofreading mechanisms. RNA-viruses, including HIV, thus average approximately one mutation per genome per replication cycle[2, 3].HIV genetic diversity is also facilitated by several recombination events per cycle between the diploid genomes in each virion[4, 5]. With this combination of high rates of replication, mutation and recombination, every possible mutation and many double mutations are likely generated in each untreated individual on a daily basis [6–8]. Drug-resistant mutants are thus present in all infected subjects before the initiation of therapy and this fact underlies the basis for the need for combination therapy for HIV. The principle of the need for combination therapy based on population size and mutation rate was first documented in the 1950’s for tuberculosis [9, 10]and has been most thoroughly confirmed with the development of antiviral chemotherapy for HIV-1 infection, hepatitis B and hepatitis C.

The recent interest in minority drug-resistant variants for HIV has been driven by the development of more sensitive and precise assays to detect and quantify minority variants in the large genetically complex population of variants present in HIV-infected subjects. The practical implications of these new assays are not whether they confirm the preexistence of drug-resistance variants, a fact that was already known. The issue is, what results generated by these assays can inform treatment strategies and decisions. For example, is there a threshold level of drug-resistant minority variants that compromises one treatment regimen and dictates another.

In this review we will summarize the attributes and limitations of the various assays to detect drug-resistant variants and the assay measurements that predict clinical or treatment outcomes and might guide patient management.

Methods to detect minority variants

Mutations associated with drug resistance in clinical practice are generally detected by direct sequencing of the pol gene from the population of HIV-RNA in plasma[11]. The interpretation of standard resistance assays is limited by the inability to detect minority viral populations at levels below 20–30%[12–15]. Several assays have been developed to characterize drug-resistant HIV variants of lower abundance, [15].These assays utilize two main approaches to detect minority species: point mutation assays and newer sequencing technologies. Alternative procedures based on phenotypic detection of minority variants have also been developed.

Before describing the specific approaches, it should be stressed that all assays for minority variants are subject to certain common limitations. First, the specimen assayed must have a population size sufficient to permit the representative presence of minority variants. An amplification assay that uses the nucleic acid extract derived from 10 µl of plasma with 5000 copies per ml will not provide informative data about variants present as a few percent. Second, polymorphisms selectively associated with drug resistance may skew the sensitivity of primers and probes. For example, the T215Y reverse transcriptase mutation conferring zidovudine resistance is often associated with variants at codons 210 and 214. Third, primers and probes designed to amplify commonly utilized laboratory strains may not perform as well with patient-derived variants and even less well with the variety of different subtypes circulating globally. Finally, population-sequencing technologies that do not interrogate clones, as do single genome sequencing (SGS) or ultra-deep pyrosequencing (UDPS), cannot determine whether mutations are present on different variants in the population or are co-linear within a given sequence. Table 1 summarizes the most common technologies for detecting minority variants that have been applied to HIV-infection with an indication of their performance characteristics and their potential attributes and limitations. Online supplement 1 describes in detail the different assays.

Table 1.

Methods to detect minority variants of drug-resistant HIV-1

Assays	Attributes	Limitations	Methods	Max Sensitivity*	References
Point mutation assays (with or without additional ligation step and/or PCR)	High sensitivity/specificity Moderately labor intensive Easy interpretation Additional ligation step improves specificity Additional PCR-amplification improves sensitivity	False positive results at lower limits Polymorphisms can reduce sensitivity Single-allele analyses No genetic linkage	LiPA	2%–10%	[17, 18]
			AS-PCR	0.01%–<1%	[19–27]
			MultiCode-RTx		[30, 31]
			OLA	<1%–12%	[32–37]
			LigAmp		[38, 39]
			RCA		[40, 41]
Sequencing assays	Extended sequence analyzed Genetic linkage if clonal	Labor intensive Cost Recombination	SGS	<1%–10%	[42, 43]
			CSA		[12, 15, 44, 45]
			HTA		[47, 48]
			SSA		[49]
			UDPS	< 1%	[50–54]
			DNA microarray	0.01%–<1%	[13, 56]
			PASS	0.01%–0.1%	[57]
Phenotypic assays	High sensitivity for co-receptor tropism Additional phenotypic information Possibility of identify novel DRM Can discern reversal of resistance	Low sensitivity for drugs with viral targets Cost Turnaround time	ES trofile assay TyHRT	0.3%–<1%	[58, 61–64] [39, 60]

PCR: polymerase chain reaction, DRM: drug resistance mutation, LiPA: line probe assay, AS-PCR: allele-specific PCR, OLA: oligonucleotide ligation assay, RCA: rolling circle amplification, UDPS: ultra-deep pyrosequencing, SGS: single genome sequencing, CSA: clonal sequence analyses, HTA: heteroduplex tracking assay, SSA: successive specific amplification, PASS: parallel allele-specific sequencing, TyHRT: phenotypic assays based on Saccharomyces cerevisiae Ty1, ES: enhanced-sensitivity (trofile assay).

^*Sensitivity depends upon the assay as well as the number of copies being interrogated.

Point mutation assays

Point mutation assays are generally highly sensitive and specific for detecting a selected minority drug resistance mutation (DRM) (Table 1). While some of them depend entirely on differential hybridization to alternative viral variants, an additional ligation-step and/or PCR-amplification can improve their specificity and sensitivity. These assays are relatively inexpensive, and moderately labor-intensive. The results are generally easy to interpret. This methodology is especially suitable for epidemiological studies where dynamics and persistence of the most common resistance mutations are analyzed, for example, following the use of nevirapine for the prevention of mother to child transmission (PMTCT). They cannot completely address the number and complexity of DRMs, following treatment failure for example.

In addition to the limitations common to all assays described above, their major limitations are the ability to detect only a single point mutation at a time and the reduced ability to detect alternative polymorphisms in the codon of interest or relevant mutations in nearby codons. Moreover, point mutation assays may be prone to false positives at very low percent minority readouts; therefore, a careful determination of a biologically relevant lower level of detection is important, and these are likely to be different for different resistance mutations (and may be as high as 2%)[16]. Validating an assay with pure mixtures of wild type and mutant alleles may not reflect the diversity and complexity of polymorphisms in a clinical specimen. In addition, validation profiles established with laboratory strains or clade B patient-strains, may not apply to field studies, for example for PMTCT in sub-Saharan Africa.

Sequencing assays

In contrast to point mutation assays, sequencing methods permit analyses of the entire sequence context, and with sequencing of molecular clones, the genetic linkage of each detected mutation. They are less susceptible to primer and probe polymorphisms and do not need individual cutoff assessment for each mutation to avoid false positive results. On the other hand, for the detection of minority variants, sequencing techniques tend to be more labor and cost intensive than point mutation assays. Sequencing assays like SGS or UDPS represent the “gold standard” by which point mutation assays of minority variants in clinical specimens must be validated.

Phenotypic assays

Phenotypic assays assess drug susceptibility by determining the effect of different inhibitors on the replication of viral isolates or recombinant vectors carrying patient-derived viral domains. In routine practice their ability to detect minor drug-resistant variants is limited[58]. Recently, highly sensitive phenotyping assays were developed to detect relevant, low frequency drug-resistant virus variants in clinical specimens[59, 60]. These methods provides phenotypic selection of drug-resistant variants, which can be further characterized by DNA sequencing, leading the opportunity to discover new resistance mutations, but do not provide quantification of minority variants.

Chemokine receptor utilization provides a unique situation in which the detection of any variants utilizing X4 can predict treatment failure with CCR5 inhibitors like maraviroc[61]. A phenotypic assay that can detect the presence of X4 utilizing variants down to 0.3% of the population with 100% sensitivity has proven as sensitive and specific as any approach for the prediction of the utility of this class of drugs [61–64].

Natural history of drug resistance mutations in the absence of therapy

The existence of minority variants with DRMs in the absence of drug exposure is a common theme in microbiology. Luria and Delbruck first described the rare generation of resistant variants in bacteria in 1943 [66]. The clinical relevance of rare drug-resistant populations of M.tuberculosis formed the rationale for combination therapy in the 1950’s[9, 10].A very high population size[1],characterized by massive viral turnover and subsequent rapid evolution[67], characterizes HIV-1 infection in vivo. The extraordinary genetic diversity gives rise to the coexistence within one infected individual of numerous genetic variants derived from a clonal or closely related oligoconal inoculum[6]. This population of genetic variants within an individual was designated a “quasispecies” by Manfred Eigen [68].

The diversity of this quasispecies in a patient increases over time [69] suggesting that the complexity of resistant viral variants expands over time in untreated patients [8]. Variants carrying any possible DRM are likely to coexist in untreated patients [6, 7].Most DRMs, however, confer a fitness cost [70] and remain minority species because of the superior replication capacity of wild type virus in the absence of the selective pressure of drug therapy [6]. The presence of low-frequency mutations conferring drug resistance has been well documented in drug naïve patients who were originally infected with wild type virus[8, 21, 44, 71]. The dynamics of the decay of wild type virus and the emergence of resistant HIV-variants during nevirapine monotherapy in previously untreated patients was used to calculate the prevalence of Y181C mutations before treatment[72]. The prevalence of the resistance mutation Y181C before any NNRTI treatment was calculated to be as high as 0.1–1%[72].Low levels of DRMs confer the primary rationale for combination therapy of chronic viral infections.

Acquired drug resistance mutations

Acquired drug resistance is generated from a background of transmitted drug susceptible virus. The wild type virus and the progressively more resistant viruses that evolve under the ongoing selective pressure of drug treatment are archived in the latent reservoir[73, 74]. The discontinuation of ART by patients in whom viral suppression was not achieved results in the rapid replacement of the drug-resistant population by wild type virus over a period of weeks [75–78]. The better fitness of the wild type virus, which can be measured by assays of replication capacity, accounts for this rapid replacement and is often reflected by a higher level of plasma HIV RNA after the replacement of the drug-resistant virus with wild type virus [79].

The strategy to interrupt ineffective treatment to permit the reemergence of wild type virus has proven to be ineffective to address drug resistance [80]. Persistence of drug-resistant virus has been well documented not only in long-lived cellular reservoirs [73, 74], but also in blood plasma as minority species several months up to years after treatment discontinuation when conventional sequencing reports a complete reversion to wild type virus[20, 21, 47, 81–84].

Long-term persistence of resistant variants at low levels is potentially important in patients stopping and restarting ART or in women and infants using limited doses of nevirapine to prevent mother-to-child transmission of HIV-1 in resource-limited countries. Using ultra-sensitive assays, nevirapine resistance mutations were identified in 65 to 87% of women and infants between 6 and 36 weeks even after single dose nevirapine administration [85–87]. These data confirm that emergence of nevirapine-resistant variants after single dose nevirapine administration is underestimated using conventional population sequence analyses.

These DRMs decline rapidly over time but can persist at low levels in blood plasma of some women and infected children up to 12 to 24 months after exposure before returning to basal background levels[39, 85, 86, 88–90].Any of three nevirapine-resistance mutations (K103N, Y181C, and G190A) was found even in the latent cellular reservoir [Wind-Rotolo et al, CROI 2008, Abstract 634]. DRMs in cellular DNA declined in one study from 52% at 6 weeks to 4% at 12 months post partum [86]. The short duration of nevirapine-exposure during PMTCT may limit the accumulation of additional DRMs that arise with more prolonged exposure [91].

Minority resistant subpopulations can be also found in subjects undergoing structured treatment interruption (STI) of successful ART regimens[19, 92, 93]. Resistant variants are most likely selected during periods of increasing HIV-replication (viral rebound) while components of combination regimens are decaying with different half-lives and drug levels may be suboptimal. Persistence of lamivudine-sensitive virus despite high drug concentrations in patients receiving zidovudine/lamivudine dual therapy suggests that residual low-level replication may occur also in patients under ART because of the impaired fitness of M184V and M184I [94].

Transmitted drug resistance mutations

Since the first report of primary infection with a zidovudine resistant virus in 1993 [95], numerous reports have described transmission of drug-resistant viral variants[96, 97]. In contrast to acquired drug resistance, transmitted DRMs are usually not associated with a reduced viral replication capacity[98, 99]. Transmission of drug-resistant virus generally appears to be less efficient than of wild type virus, perhaps due to its diminished fitness in the absence of the antiretroviral drugs [99]. The drug-resistant virus that does get transmitted tends to be the subset of drug-resistant virus that is as fit as most wild type virus [98, 100, 101]. In the absence of drug pressure the stability of transmitted resistance mutations varies markedly [102], and transmitted M184V mutation, which significantly impairs viral fitness, can quickly revert back to wild type [103].

Also in contrast to acquired drug resistance the monoclonal or oligoclonal transmission of drug-resistant virus results in a pure population without an archived population of wild type virus that can readily emerge in the absence of treatment. With a transmitted virus with high replication capacity and without an archived population of wild type virus, transmitted drug resistance is likely to persist for long durations in blood [98] and in semen [104] of infected individuals, providing a prolonged “window of opportunity” for secondary transmission.

While the impact of transmission of drug resistance on the natural course of disease is still a matter of debate [105], responsiveness to initial ART in patients infected with drug-resistant virus is suboptimal [96, 106].

The development of more sensitive assays to detect drug-resistant viruses as minority variants has resulted in the identification of these in a proportion of acutely and recently infected individuals [25, 107–109]. These reports raise a fundamental question of whether transmission of drug-resistant HIV variants has been underestimated when measured by standard genotypic assays[110, 111].

The conundrum raised by these observations is the incompatibility of these reports of the presence of minority drug-resistant variants transmitted in pol with the very detailed descriptions of numerous clones of env as measured by SGS and UDPS in acutely infected patients [43]. These latter studies strongly argue that monoclonal and oligoclonal populations in acutely infected individuals are not consistent with the observations with the allele specific assays that describe a proportion of individuals with minority drug-resistant variants. These apparently conflicting observations require reconciliation. Several potential explanations can be generated: 1.) The allele specific assays generate levels of false positive values with wild type sequences, 2.) A process involving recombination between env and pol could be occurring during acute infection but requires documentation, 3.) Occasionally DRM soccur early after infection and expand as a relatively substantial proportion of the population,4.) Infection with a drug-resistant virus occurs in which there is reversion in some of the progeny early in the course of infection, causing the initial resistance mutation to decline to a minority. This last hypothesis, however, would be unlikely because transmitted viruses (with exception of M184V) are generally not associated with lower replicative fitness. Other explanations for the reports of transmission of minority variants and monoclonal or oligoclonal env transmission might be hypothesized. Confirmation of these reports of transmitted minority drug-resistant variants will require SGS or UDPS.

Therapeutic consequences of minority variants

Therapy-experienced patients

Several studies demonstrated that drug-resistant HIV-variants are present at low frequencies in therapy-experienced patients for prolonged durations not only after viral failure [20, 21, 42, 47, 81–83],but also after discontinuation of suppressive ART in patients originally infected with wild type virus[19, 92, 93]. These drug-resistant strains, stored as minority populations and often missed by standard genotyping may be the cause of viral failure[82, 83, 112–114].

A higher proportion of minor NNRTI-resistant variants was detected in NNRTI-experienced patients compared to NNRTI-naïve patients using two different methods, SGS and AS-PCR[114]. A reduced response to an efavirenz-containing regimen was significantly associated with low-level K103N variants at frequencies of 0.5%–1%, but interestingly, not if the frequency was below 0.5%. In contrast, no association with virologic failure was found for Y181C minority variants.

In 7 of 20 NNRTI-experienced patients who developed a K103Nmutation, UDPS has revealed additional mutations in minority populations that were predicted to confer resistance to the new-generation NNRTI etravirine [50]. In contrast, in13 treatment-naïve patients with transmitted K103N, UDPS could not detect additional major NNRTI mutations.

NNRTI-based combination therapy is recommended as the first-line regimen for adults in resource-limited settings; however, a major concern has been raised by the use of nevirapine in PMTCT. Because of the long half-life of nevirapine, which can be found in sub-therapeutic plasma levels three weeks after a single dose [115],the selective pressure conferred by this short regimen is prolonged. Thus, the fast development and persistence of resistance has been well documented and has been associated with reduced efficacy of subsequent treatment [116, 117]. Some studies, however, have found that the likelihood of viral suppression and the clinical outcome are not compromised if nevirapine-based regimen is started more than 12 months after delivery [116–121].

A significantly higher rate of virologic failure with nevirapine-based therapy was also found in infants exposed to nevirapine peripartum[116]. This is most likely explained by the longer half-life of nevirapine in infants[122].

In contrast to NNRTI, where a single point mutation can confer high-level drug resistance, substantial resistance to protease inhibitors (PI) requires the progressive accumulation of multiple mutations. Some studies [83, 112] suggest that minority variants may play a role in evolution of resistance also with PI-based regimen. The selection of these variants, however, may occur slowly, depending on the context in which they arise.

Therapy-naïve patients

The clinical importance of minority DRMs in the setting of drug-naïve patients is not clearly defined. Several distinct situations occur, including an inaccurate or denied history of treatment. Cases of super infection with both wild type and drug-resistant virus have been reported [123–125].

Several studies have suggested that pre-existing minor DRMs can rapidly emerge with treatment in treatment-naïve patients[16, 52, 126–131].

More than twice as many subjects with DRMs were detected in a cohort of chronically infected, ART-naïve patients, using UDPS compared to standard genotyping[52].Detection of pre-existing, minor NNRTI-resistant variants was associated with increased risk of virological failure in subjects starting an NNRTI-based first-line regimen. Similarly, four subjects with PI-resistant minority variants experienced virological failure with PI-containing regimen.

Using a modified AS-PCR, the percentage of chronically infected patients with resistant strains nearly doubled compared to conventional bulk-sequence analysis [16].In this study, an individual resistance mutation assay cutoff for qPCR was established, that ranged from a lower level of 0.4% up to 2% for mutation K70R.

Since low viral load diminishes the discriminatory ability of detection assays, the limit of detection for minority variants is defined by the virus load for each sample [19, 126].A rapid selection of the pre-existent low frequency DRMs (frequency range 0.07–2%) in plasma samples before and during early virological failure in 4 patients was observed despite good adherence and adequate drug plasma level [126].

In contrast, no significant difference between the outcomes of first-line therapy in acutely and recently infected patients carrying minority variants has been reported [132, 133]. These last studies detected M184V mutation as the most common DRM. M184V is known to confer a high fitness-cost to the virus, reducing its replicative capacity. Most of the patients received ART-regimen with high genetic resistance barriers, including two NRTI and one boostered PI, possibly masking the putative effect of this DRM on the virological response.

An almost 9-fold increase in the detection of primary NNRTI-resistance(44% versus 5%)was observed using AS-PCR compared with standard genotypingin pre-treatment plasma samples from drug-naïve chronically infected patients with and without virological failure[130].In this study a low interpretation threshold was used for Y181C (0.03%) and K103N (0.001%–0.003%) and the level of detected mutants were all <1%.The detection of pre-existing minority Y181C variants significantly increased the risk of failure to an initial efavirenz-based regimen in patients with optimal compliance to ART. This increased risk has been observed in subjects with HIV-RNA levels >100,000 copies/ml but also in patients with lower viral load. Interestingly only 7% of the patients with baseline Y181C showed emergence of Y181C at failure and 38% eventually failed with K103N.

By multiplying the proportion of virus with K103N in reverse transcriptase and the viral load, a threshold of >2000 copies/ml of DRMs was associated with an increased risk of viral failure after administration of efavirenz-based treatment (Goodman, HIV Drug Resistance Workshop 2009, Abstract). The copy number of DRMs correlated with failure better than the proportion. This raises the critical question of whether viral load of DRM is a better predictor of treatment failure. Treatment failure has been frequently shown to correlate with viral load, a precedent well documented with the chemotherapy for tuberculosis.

Conclusions and future perspectives

Drug-resistant minority variants exist in every patient. They are more prevalent in patients with prior drug exposure. The practical question is what assay and what result can guide a management decision to improve treatment outcome, since we know that the major rationale for combination antiretroviral therapy is to contend with pre-existing drug-resistant variants.

Several assays have been developed to detect minor mutants at levels as low as 0.5 to 0.01%. The results generated by these highly sensitive assays then raise a series of additional questions: 1.) What is the true sensitivity and specificity of this assay result for the clinical specimen being tested? 2.) What result predicts diminished response to a particular regimen?

One issue that has been inadequately considered is the distinction between the proportion of drug-resistant mutants in the population and the magnitude of the drug-resistant mutants in a population. Studies have shown that many regimens are less successful with higher viral loads, which may be the result of a higher load of drug-resistant virus[63, 134]. Even the trial of the dual nucleoside regimen of zidovudine/lamivudine showed a substantial proportion of undetectable viral loads in subjects with low baseline levels of HIVRNA[135].

We need to understand at what point a minority resistant viral population may become clinically relevant. More studies are needed in order to determine a cutoff-value for the detection of these low-abundance viral variants, below which the risk of failure declines. This threshold may be specific for each mutation and treatment regimen.