Persistence of primary drug resistance among recently HIV-1 infected adults

Objectives: Primary, or transmitted, drug resistance is common among treatment naive patients recently infected with HIV-1, and impairs response to anti-retroviral therapy. We previously observed that patients with secondary resistance (developed in response to anti-retroviral treatment) who chose to stop an anti-retroviral regimen experience rapid overgrowth of drug resistant viruses by wild-type virus of higher pol replication capacity. We sought to determine if primary drug resistance would be lost at a rapid rate, and viral pol replication capacity would increase, in the absence of treatment.

Methods: We tracked drug resistance phenotype, genotype, viral pol replication capacity (single cycle recombinant assay incorporating a segment of the patient pol gene [pol RC]), plasma HIV-1 RNA, and CD4 T cell counts in the absence of treatment among patients in early HIV-1 infection.

Results: Six of 22 patients had evidence of primary drug resistance to at least one class of drug; three resistant to protease inhibitors, three resistant to non-nucleoside reverse transcriptase inhibitors, and four resistant to nucleoside reverse transcriptase inhibitors. All six patients maintained evidence of drug resistance for the period of observation. Among patients with baseline primary drug resistance pol RC did not increase over time.

Conclusion: The selection environment of early infection is determined by immune pressure, and stochastic events, not viral pol replication capacity. In contrast to secondary resistant infections that are rapidly overgrown when therapy is stopped, primary drug resistance persists over time. Surveillance and clinical detection of primary resistance is feasible in the first year of infection.

Introduction

Primary resistance is common among adults recently infected with HIV-1 in North America and Europe. For example, current estimates of primary genotypic drug resistance to at least one compound vary from 23% in San Francisco in 1996-2001 [¹], to 12% in 10 other North American cities in 1999-2000 [²], to 11% in Geneva in 1996-1998 [³], to 11% in a study of 15 European countries and Israel in 1996-2002 [⁴]. Drug resistance mutations may limit anti-retroviral treatment options, lengthen time to virologic suppression during anti-retroviral treatment, or be associated with subsequent virologic failure of treatment [^1,2]. Drug resistance has been associated with durably lowered viral pol replication capacity, and elevated CD4 T cell counts during virologic failure of a continuing protease inhibitor (PI)-based regimen [^5-8].

Secondary, or acquired, resistance evolves from drug-sensitive variants in order to allow viral replication in the presence of anti-retroviral compounds. Acquisition of mutations into the dominant variant may occur in steps, creating a diverse population of drug sensitive and partially resistant variants. Viruses of diverse drug susceptibility and fitness remain archived, or replicate at low levels during treatment [^9,10], re-emerging when therapy is halted [⁸]. By contrast, primary resistance may be established by a small pool of drug resistant viruses. In this setting, evolution of drug susceptible virus may require selection for drug sensitive variants that are associated with greater in vitro replicative efficiency [¹¹] and fitness [^12,13]. This process of mutational reversion and selection may require long periods of time if immune selection early in infection poses challenges for the evolving virus.

To this end, we conducted a longitudinal study of viral pol replication capacity, genotypic and phenotypic drug resistance, plasma HIV-1 RNA level, and CD4 T cell counts, in a cohort of persons in early HIV-1 infection, with and without primary drug resistance.

Methods

Subjects

All specimens were drawn from subjects enrolled in a study [^1,14] of acute and early HIV-1 infection conducted at a university-based HIV clinical research program. All participants gave written, informed consent using protocols approved by the Committee on Human Research, University of California, San Francisco. All subjects were anti-retroviral treatment naive. We identified individuals with and without primary drug resistance (genotypic or phenotypic) during recent infection who did not elect to receive anti-retroviral therapy, and had been observed for at least 6 months. We chose all six patients with drug resistance who fit these criteria. We chose a subset of 16 patients without drug resistance, who fit these criteria, as a comparison group for an approximate 3 : 1 study design. The patients without resistance were randomly chosen from a group of patients with similar less sensitive enzyme linked immune absorbent assay optical density (LS EIA OD) [¹⁵] scores. Patients were followed until they initiated anti-retroviral therapy, or were lost to follow-up.

Criteria for recent HIV-1 infection

We identified persons with acute and early HIV-1 infection as previously described [¹⁶].

Drug resistance testing

The presence of mutations associated with drug resistance was evaluated using population-based sequencing of protease and reverse transcriptase reading frames using the TRUGENE assay (Visible Genetics, Toronto, Canada) [^16-18].

Phenotypic resistance testing

Phenotypic drug resistance testing was performed using a recombinant virus based assay (Phenosense, ViroLogic, Inc., South San Francisco, California, USA) [¹⁹]. Phenotypic cut-offs were defined as the 99^th percentile in fold change 50% inhibitory concentration (IC₅₀) among wild-type viruses as recently described [²⁰].

Viral replication capacity

HIV-1 pol replication capacity (pol RC) was assessed via a modification of the Phenosense phenotypic drug susceptibility assay (ViroLogic, Inc., South San Francisco, California, USA) [¹⁹].

Statistical analysis

To assess changes in phenotypic drug resistance, and viral pol RC over time we used mixed effects models, with a random effect specified for the individual (intercept) [^21,22]. All statistical analyses were performed in the SAS System version 8.2 (SAS Institute, Cary, North Carolina, USA).

Results

Baseline characteristics

Twenty-two patients were observed for a median 12 months (interquartile range [IQR]), 8-13.2 months), involving 145 assays, for an average of seven time points per patient. The median LS EIA OD was 0.10 (IQR, 0.03-0.30), indicating that study entry occurred within approximately 60 days of seroconversion in the majority of subjects [^15,23]. At baseline, six of 22 individuals had evidence of carrying a virus of reduced phenotypic drug susceptibility (Table 1) to at least one of the following drug classes: PI (3/22), nucleoside reverse transcriptase inhibitor (NRTI; 4/22), non-nucleoside reverse transcriptase inhibitor (NNRTI; 3/22). Genotypic drug resistance markers were evident in all individuals with phenotypic drug resistance. Infections involved resistance to the NNRTI class alone in two individuals, resistance to NRTI alone in one individual, and resistance to two or more classes of drugs in three individuals. At baseline, the median pol RC was 85% of control (IQR, 28-106%), median HIV-1 RNA was 3.82 log₁₀ copies/ml (IQR, 3.47-4.28 log₁₀ copies/ml), and median CD4 T cell count was 595 × 10⁶ cells/l (IQR, 540 × 10⁶-714 × 10⁶ cells/l). Baseline pol RC tended to be lower for those with genotypic evidence of drug resistance (38% versus 110%; P = 0.07, Wilcoxon Two-Sample Test).

Table 1

Case descriptions of three individuals with diminishing drug resistance over time

Three patients demonstrated a decline in phenotypic resistance to one or more NRTI while maintaining at least some evidence of genotypic or phenotypic resistance to this class of drug. At baseline, patient A had phenotypic (Fig. 1) and genotypic (Table 1) resistance to three drug classes. Baseline phenotypic resistance to zidovudine and lamivudine was intermediate. Manual review of the baseline sequencing pherogram revealed no nucleotide mixtures at these codons. Patient A showed reversion to a phenotypically lamivudine sensitive virus after 15 months. Phenotypic susceptibility to lamivudine emerged with conversion from M184V to M184M, and zidovudine susceptibility was partially restored with T215Y conversion to T215C (Fig. 1). Patient A experienced temporary increases in pol RC as M184V and then T215Y were lost.

Fig. 1

Patient B experienced a decline in zidovudine phenotypic resistance as the D67N mutation converted to a D67N/D mixture, then resolved to a D67D drug-sensitive residue (Fig. 1). Patient B carried a K70K/R mixture that resolved to a dominant K70R resistance mutation. Patient D experienced no change in genotypic resistance markers but had a late decline in zidovudine resistance (Fig. 1) to below the phenotypic cut-off level (0.25 log₁₀ fold-change in IC₅₀).

No loss of phenotypic and genotypic resistance to a PI or NNRTI in five individuals

There was no loss of phenotypic or genotypic resistance to a PI or an NNRTI during the period of observation for patients A, C, D, E and F (Fig. 1). The observed increase in NNRTI resistance for patient A may be due to the loss of NRTI resistance during the same time period. Increases in NNRTI susceptibility have been associated with NRTI resistance [^24,25].

Changes in pol RC

Among those without primary drug resistance, pol RC declined (-0.98 pol RC percentage points per month; P = 0.02). Among those with primary drug resistance, pol RC did not change significantly over time, either by single drug class [PI (P = 0.76) or NRTI (P = 0.82), or NNRTI (P = 0.57) phenotypic resistance at baseline], or by resistance to any class (PI, and/or NRTI, and/or NNRTI phenotypic resistance at baseline; P = 0.99).

Discussion

All six recently infected persons with primary drug resistance maintained some evidence of genotypic drug resistance over a median 1-year period of observation. Unlike patients with secondary resistance who stop a partially suppressive regimen [⁸], patients with primary drug resistance did not undergo rapid conversion to a wild-type drug sensitive virus [²⁶] of higher viral pol replication capacity. Previous reports have demonstrated persistence or a delay in loss of phenotypic and genotypic resistance among persons who had recently acquired a multi-drug resistant variant [^26,27]. In early HIV-1 infection, immune responses, such as neutralizing antibody [²⁸], or cytotoxic T lymphocytes [²⁹], but not pol RC, may be the dominant selection pressures.

Two patients experienced a decline in the level of phenotypic resistance to an NRTI or a decline of one or more resistance mutations, although at least one genotypic marker of resistance was retained in all patients with primary resistance. Patient A (Fig. 1) experienced loss of phenotypic resistance to lamivudine, and a decline in the level of phenotypic resistance to zidovudine. The T215C mutation was retained, which indicates prior resistance to NRTI, and hastens re-appearance of zidovudine resistance via the T215Y mutation [³⁰]. Patient B experienced a decline in phenotypic resistance to zidovudine, but retained the K70R mutation, and phenotypic resistance to zidovudine. In both cases, some evidence of current or prior drug resistance to NRTI was retained, although the degree of resistance in the detected virus appeared to wane over time.

Phenotypic and genotypic resistance to NNRTI and PI did not decrease over the period of observation. The degree of baseline phenotypic resistance to NNRTI and PI were as great as resistance to NRTI yet resistance to NNRTI and PI was observed to persist (Fig. 1). The likelihood of loss of resistance was not related to degree of baseline phenotypic resistance, or to baseline pol RC levels. However, PI resistance patterns may be complex, and require co-evolution within Gag, for conversion to a drug sensitive isolate. The complexity of this pathway may impede conversion to drug susceptibility [^7,31]. Of the patients presenting with NNRTI drug resistance (Table 1) only the K103N mutation was present, which has not been associated with decrease in viral pol replication capacity [³²]. Hence, there may have been no selection advantage to the loss of K103N.

In our cohort of patients with primary drug resistance, pol RC did not appear to be a strong fitness determinant. Patient C bore both the M184V and T215Y mutations and had very low pol RC at baseline (2% of control). Over a 9-month observation period, there was no loss of either mutation, each of which lowers viral fitness [^12,33]. However, our follow-up time was limited. Greater lengths of time may be required for conversion to drug susceptibility, and evolution of higher pol RC among persons recently infected with a drug resistant form of HIV-1 who elect to defer treatment [^26,27].

A drug resistant variant observed during early infection will persist, either as the dominant species, or as an archived variant in the latent pool. Patients with primary genotypic or phenotypic resistance are at increased risk of poorer virologic response to a first anti-retroviral regimen [^1,2]. Primary drug resistance will persist and remain detectable via genotypic and phenotypic tests for a year or longer after acquisition of infection. Resistance testing performed early in infection will allow the clinician and patient to consider regimens most likely to lead to optimal virologic suppression once treatment is elected.

Acknowledgements

Kits for some genotypic testing were provided by Visible Genetics (Toronto, Canada). ViroLogic, Inc. performed the replication capacity, and phenotypic drug resistance assays used in this study.

Sponsorship: J.D.B is a Burroughs Wellcome Fund Fellow in the Program in Quantitative Biology, University of California, San Francisco. Support for this project was provided by the National Institute of Allergy and Infectious Diseases (AI41531), the University of California San Francisco Center for AIDS Research (NIH P30 MH59037), the Centers for Disease Control and Prevention (R18/CCR 920936), and the J. David Gladstone Institutes.