Kinetics of HIV-1 RNA and resistance-associated mutations after cessation of antiretroviral combination therapy

Objective: To study the kinetics of HIV-1 RNA and drug-induced mutations after cessation of antiretroviral therapy (ART).

Design and methods: Successive plasma samples from 26 patients were tested for HIV-1 RNA by PCR and for mutations associated with drug resistance by sequencing of the pol gene.

Cited Here...: After cessation of ART the phase of undetectable virus (< 50 copies/ml), ranging from 6 to more than 29 days, was followed by a rapid viral increase, which slowed down before a plateau corresponding to pre-treatment levels or higher was reached in most cases (14/19 patients). In one patient virus was still undetectable at 4 weeks. Also, a significantly larger number of primary protease inhibitor (PI)-associated mutations reverted to wild-type, as compared with secondary PI-, and primary reverse transcriptase inhibitor (RTI)-associated mutations. During the rapid viral increase no mutations disappeared, which instead happened during the slower viral increase preceding the viral plateau level.

Conclusion: After discontinuation of ART large individual variations were found for the time period until HIV-1 became detectable in plasma, possibly due to differences in the HIV-1 specific immunity. The more rapid loss of primary PI mutations suggests that they might cause a more impaired viral fitness than primary RTI mutations. However, the persistence of drug mutations during the initial viral load increase indicates that mutated strains may still replicate efficiently.

Introduction

The use of potent antiretroviral therapy (ART) has improved substantially the clinical outcome of HIV-1 infection ^[1,2]. In most optimally treated patients, combination ART maintains the HIV-1 RNA load below the level of detection ^[3,4]. However, after therapy withdrawal a rapid viral rebound is observed in most patients ^[5-7] to a postulated individual viral load set-point, which is not altered by ART ^[8].

Insufficient viral suppression selects for HIV-1 strains with mutations in the genome, which cause reduced drug sensitivity ^[9]. Various characteristic primary and secondary drug-related mutations have been identified both in the pol gene ^[10,11] and at the cleavage sites for the viral protease ^[12,13]. Primary mutations often appear earlier during treatment failure, while secondary mutations accumulate over time and are seen as compensatory changes ^[14].

So far, the way in which these mutations affect viral fitness has not been characterized. Because viral mutant strains with resistance to antiretroviral agents are seldom seen in the major viral population before initiation of therapy ^[14], these viral variants are likely to have a diminished ability to replicate in a drug-free environment. However, other resistance-associated mutations have also been described in the major viral population before ^[11] and after discontinuation of ART ^[15,16] and do not seem to hamper viral replication. Especially in multi-experienced patients who do not have any attractive therapeutic options left, the indirect antiretroviral effect caused by decreased viral fitness could possibly be of clinical importance. It is therefore important to characterize further the impact of different constellations of mutations on the viral replication in vivo.

To elucidate these aspects further, we studied the kinetics of the HIV-1 RNA levels after cessation of potent ART and compared them with pre-treatment levels. In addition, the evolution of the protease and reverse transcriptase (RT) genes in these patients was followed closely.

Methods

Patients

Twenty-six consecutive patients on triple ART (six females, 20 males), who discontinued their treatment between September 1998 and September 2000, were included in the study. All patients were followed up at the Clinics of Infectious Diseases at the Karolinska Institutet, Stockholm, Sweden (Table 1). The last modification of their drug regimens had been made between 7 and 831 days (median, 235 days) before cessation of ART, and thereafter all patients had received at least triple antiretroviral combination therapy (Table 2). The reasons for this discontinuation were adverse events (n = 16) [gastrointestinal (n = 4), allergic (n = 3), neurological (n = 3), liver enzyme elevations (n = 3), myopathy (n = 2), nephrolithiasis (n = 1)], virological treatment failure (n = 6) and miscellaneous (n = 4), e.g. drug holiday or other social reasons. Lipodystrophy was not registered as a cause for cessation of ART (Table 2). Ethical permission was obtained from the ethical committee at Karolinska Institutet (98/098).

Table 1

Table 2

RNA extraction

Viral RNA was extracted from plasma using the Qiagen Viral RNA Kit (Qiagen GmbH, Hilden, Germany), according to the manufacturer's recommendations, and was reverse transcribed in to cDNA (Moloney murine leukaemia virus RT; Boehringer GmbH, Mannheim, Germany). A large part of the pol gene was amplified by PCR by using the outer primers: Prot1+, 5′-GCT GTT GGA AAT GTG GRA ARG (HXB2, 2024-2044); and RT4-, 5′-CCT GWA TAA ATY TGA CTT GCC (HXB2, 3346-3367). The thermal cycling parameters were: 95°C for 4.5 min and then 30 cycles at 95°C for 30 sec, 50°C for 30 sec and 72°C for 90 sec, and finally at 72°C for 3.5 min.

A second round of PCR was performed using 5 μl of the first round PCR product under identical conditions. The primers used were: Prot-Bio, 5′-BIO-GCA GGA GCM GAW AGA CAR GG (HXB2, 2214-2234); and Prot-M13, 5′-CGA CGT TGT AAA ACG ACG GCC AGT GCC ATC CAT TCC TGG CTT TA (HXB2, 2584-2603) for the protease gene amplification and RT-M13 and RT-Bio ^[11] for the RT gene amplification. The sizes of the amplification products were checked by agarose gel electrophoresis.

Sequencing reaction and sequence analysis

The amplified DNA was separated using magnetic beads (Dynabeads M-280 Streptavidin, Dynal, Oslo, Norway). The RT and protease genes were sequenced by the Sanger method using ALF express (Pharmacia Biotech, Uppsala, Sweden) and the Cy5 Autoread sequencing kit (Pharmacia Biotech) using the provided universal primer for sequencing of the biotinylated strand and the primer RT-minus (5′-AGG CTG TAC TGT CCA TTT AT-3′) for the negative RT strand. The gene sequences were analysed with DNAsis software package (Hitachi Software Engineering, San Bruno, California, USA) and were related to the HIV-1_HXB2 sequence (Genebank accession number, K03455). Details of primary and secondary mutations in the pol gene associated with reduced sensitivity to antiretroviral treatment were obtained from the literature ^[10] and web resources (http://hivdb.stanford.edu/hiv). The mutations were also classified as conservative or non-conservative, as described by Fontenot et al. ^[17].

Analyses of T-cell subsets and HIV-1 RNA levels

Determinations of T-cell subsets were performed using routine flow cytometry. Quantification of plasma HIV-1 was performed using a commercial kit (Amplicor HIV-1 Monitor Test, Ultra sensitive, Roche Diagnostic Inc., Branchburg, New Jersey, USA).

Statistical analysis

Linear regression, paired t test and contingency table tests were used. A value of P < 0.05 was considered to be statistically significant. The doubling time was calculated by the formula:EQUATION where: T, doubling time; A, end-value; I, start-value; d, days.

Equation U1

Results

Dynamics of HIV-1 RNA

Following discontinuation of therapy the HIV-1 RNA levels of each patient were monitored at two to eight time points (median, 5.5) in a median of 214 days (range, 12-437 days). Short observation times were consequences of a rapid reintroduction of ART. The delay between ART cessation and the first assessment of the viral load ranged between 2 and 22 days (median, 7 days; mean, 7.3 days). We could divide the patients into two major groups. In 13 patients (B, D, F, L, M, N, Q, R, S, T, V, W, Y) (Fig. 1a) the viral load altered from undetectable HIV-1 RNA (< 50 copies/ml) to high levels (median, 85 000 copies/ml; range: 1500-750 000 copies/ml). Twelve patients had detectable HIV-1 levels at cessation of therapy. Here, we could define two subgroups. In nine patients (A, C, E, I, J, K, O, P, Z) (Fig. 1b) the viral load changed from detectable levels (50-35 000 copies/ml) to high levels (median, 310 000 copies/ml; range, 52 000-750 000 copies/ml), and in three patients (G, H, X) (Fig. 1c) it remained stable within 1 log₁₀ (1500, 66 000 and 34 000 copies/ml, respectively), during the entire follow-up (G, 398 days; H, 437 days, X, 79 days). In only one of the patients (U) did the virus remain undetectable for longer than 29 days after cessation of therapy (Fig. 1a).

Fig. 1

The initial doubling time (up to 43 days) of the rebounding virus was calculated based on the baseline HIV-1 RNA level in patients whose virus load changed > 1 log₁₀ (n = 22). The doubling times differed inter- and intra-individually. Both the inter- and intra-individual analysis showed a tendency to a gradually increasing doubling (Y) time with time, which was described by the regression: Y = 1.51 + 0.059 days (R = 0.474) (Fig. 2).

Fig. 2

A presumed viral set-point prior to initiation of any therapy was available in 19 patients (log₁₀ HIV-1 RNA copies/ml: mean, 5.29; median, 5.02; range, 4.15-6.60) and was determined on average 854 days (range, 244-1403 days; median, 909 days) before cessation of therapy. Within 43 days (minimum, 8 days; median, 16 days) after discontinuation of therapy 14 out of 19 patients had approached the prior set-point level within 0.5 log₁₀ and in the further observation period 10 of these patients exceeded the viral set-point. Four patients never approached the set-point, although their viral load increased notably (observation period; patient B, 12 days, E, 24 days, D, 39 days and L, 148 days, respectively) (data not shown). Only one patient (U) had no detectable HIV-1 RNA 29 days after cessation of ART.

Kinetics of resistance associated mutations

For 21 patients sequencing data were available for the pol gene at or close to the time-point of discontinuation of therapy. Thereafter, one to six samples (median, three) were sequenced, obtained up to 245 days (minimum, 7 days; median, 31.5 days) after cessation of therapy.

In the protease gene 11 primary mutations (one D30N, two M46L, three M46I, four V82A, one L90M) associated with resistance to PI were detected prior to cessation of therapy (Fig. 3a). Ten of these mutations were still detectable at a median of 8 days after discontinuation of ART, but all had disappeared after a median of 22 days. Only one M46I mutation was still detectable after 16 days.

Fig. 3

When the 21 secondary protease mutations were followed (two L10F, two L10I, one K20R, one V32I, seven M36I, two I54V, two A71T, one A71V, two V77I, one N88D) two patterns emerged (Fig. 3b). The eight mutations at amino acid positions L10I, K20R, V32I, I54V, A71V, N88D plus one of two L10F mutations, which were still detectable after a median of 8 days after stopping ART, had all disappeared after a median of 21 days. However, none of two V77I and two A71T mutations and only one of seven M36I mutations were affected by the cessation of ART (median observation time, 34 days). Also, of the 12 secondary protease mutations at amino acid positions L63P (n = 8) and I93L (n = 4), which are sometimes acknowledged as natural polymorphisms, none changed during the observation period (median, 30.5 days) (Fig. 3c).

In the RT gene, we detected 16 and four, respectively, primary mutations associated with nucleoside analogue RT inhibitors (NRTI) and non-nucleoside RT inhibitors (NNRTI) resistance (five M41L; one L74V; four M184V; six T215Y; three K103N; one G190A). Here, six out of 16 primary NRTI mutations and two out of four primary NNRTI mutations disappeared during a median observation period of 16 and 25.5 days, respectively (Fig. 3d). Two out of 10 secondary NRTI mutations (five D67N, three K70R, two K219Q) and one out of two NNRTI secondary mutations (one L100I, one K101E) disappeared in the median observation period of 28 and 14.5 days, respectively (Fig. 3e).

Contingency table analyses showed significant differences after cessation of therapy during the above mentioned time periods between the outcome of primary and secondary PI mutations (c², P < 0.05), and between the outcome of primary PI and primary RT mutations (c², P < 0.01). No significant difference was found between mutations classified as conservative or non-conservative (data not shown) ^[17].

Association of HIV-1 RNA and resistance-associated mutations

In our patients the resistance-associated mutations did not seem to interfere with the doubling time (data not shown). However, when relating the sequencing data to the dynamics of HIV-1 RNA, we could observe three distinct patterns.

The first pattern consisted of a rapidly increasing HIV-1 RNA load despite persistence of the resistance-associated mutations. Four patients (F, Q, R, S) had undetectable (< 50 copies/ml) and two patients (C, E) had low HIV-1 RNA levels (≤ 500 copies/ml) and evidence of primary and/or secondary mutations in the major plasma viral population prior to cessation. When ART was discontinued the virus levels increased by 2 log₁₀ (patients C, E) or 3 log₁₀ (patients F, Q, R, S), even though all mutations remained. In a further patient Z, the HIV-1 RNA increased within 1 month of cessation of ART from 35 000 to 275 000 HIV-1 RNA copies/ml with all mutations (M41L, T215Y, K103N) still present.

The second pattern consisted of less pronounced changes in the HIV-1 RNA load (< 1 log₁₀) and a concomitant disappearance of resistance-associated primary and/or secondary mutations. In this group, three patients (H, I, X) had detectable viral load (> 1000 copies/ml) and multiple drug-related mutations. After cessation of therapy the HIV-1 RNA load stabilized in the observation time (I, 36 days, H, 437 days, X, 79 days) within 1 log₁₀ of the baseline, after the loss of the following mutations: I, K20R, V32I, M46L, A71V, V82A, M41L, K103N; H, I54V, V82A, L90M; X, M184V, 215Y. Also, in three of the six patients who had an initial rapid increase with mutated virus, the second phase was associated with disappearance of the M184V (patient C), the L10F plus V82A (patient E) and the L10I mutations (patient S), respectively.

The third pattern was represented by eight patients (A, B, G, J, L, M, N, O) who had no drug-related mutations and low (< 2500 copies/ml) or undetectable (< 50 copies/ml) virus levels before cessation of ART. Thereafter, the viral load increased immediately in all of them, except in patient G who remained stable within 0.5 log₁₀, and no drug-related mutations were observed.

CD4 T-lymphocyte count data were available for 10 patients. Between 14 and 104 days after cessation of ART (median, 76 days) the CD4 cell count had decreased by, on average, 192 cells × 10⁶/l (95% confidence interval, 90-294 × 10⁶/l;P < 0.01) in agreement with previous studies ^[7,18-20].

Discussion

Four phases were distinguished in the dynamics of HIV-1 RNA levels after cessation of ART. During the first phase with undetectable virus, a gradual increase in virus replication is likely to occur as a result of decreasing drug concentrations and inefficient residual HIV-1 specific immune responses. Thus, both the HIV-1 specific cellular and humoral immunity ^[21,22] have been reported to decline during successful therapy. Our patients exhibited different lengths of this first phase starting from 6 days, but after 28 days all except one patient had detectable virus. During the second phase there was an exponential increase in the HIV-1 RNA levels as a result of immunological inability to contain the infection and possibly due to availability of an excess of new target cells ^[23]. Although we did not analyse the HIV-1 specific immune responses, it is possible that the length of the first phase and/or the doubling time during the second phase correlates with the degree of anti-HIV-1 specific immunity at cessation of therapy ^[24,25]. At the end of the second phase we found that the increase in viral load slowed down, possibly representing an increasing HIV-1 specific immunity; this then led to the fourth phase, that of a steady state between virus replication and clearance.

All but five patients reached their presumed pre-therapy viral set-point, which is in agreement with results published earlier ^{[5,6,8,19,26]}. However, in two patients the HIV-1 RNA levels remained below the limit of detection for more than 3 (patient D) or 4 weeks (patient U). As an early initiation of ART at seroconversion, intermittent low increases of viral load, or successive treatment interruptions may lead to at least a transient control of viral replication after cessation of ART ^[24,27,28] we investigated whether such factors were present in these two patients. However, no such factors were found in either of them.

We studied also the kinetics of drug-related mutations after cessation of ART. Here, the inter-individual differences did not follow any pattern and in contrast with Izopet et al.^[29] we found no correlation between clinical status ^[30] and the kinetics of the mutations. However, analysis of amino acid positions associated with reduced drug sensitivity revealed a clear pattern. After cessation of therapy significant differences were thus found in the outcome between primary PI- and secondary PI-, and between primary PI- and primary RTI-associated mutations.

In the protease > 90% of the primary mutations converted to wild-type within 1 month and in the only patient who did not convert the observation time (16 days) was within this range. In contrast, only half of the secondary mutations disappeared during the same time period. Here, all except one mutation at positions M36I, A71T and V77I, and the mutations L63P and I93L, which are also acknowledged as natural polymorphisms ^[11,31,32] did not change. The rapid disappearance of the primary and some secondary PI-associated mutations is likely to illustrate the impact of these amino acid changes on viral fitness.

Moreover, it was recently reported that virus from all patients with phenotypic resistance became susceptible to PI within 6 weeks of discontinuing therapy ^[33]. The seemingly slower disappearance of the resistance measured by the phenotypic method, as compared with our data, may represent the influence of minor resistant quasispecies, which are not detectable by direct sequencing.

In the RT around one-third of all primary mutations disappeared. Here, in particular, some thymidine analogue-related mutations remained for a longer time, in our study longer than 245 days, whereas M184V and the NNRTI-associated mutations disappeared more rapidly. Other groups have reported similar findings ^[16,33-35]. The rapid disappearance of primary PI mutations in particular again stresses the importance of performing resistance testing while patients are still on therapy or soon after discontinuation of therapy ^[15]. Also, the fact that NRTI mutations did not disappear as frequently and as fast as PI and possibly NNRTI mutations might indicate that such mutations have a lower impact on viral fitness than mutations associated with the drugs that inhibit the enzymes directly.

No correlation was found between the kinetics of HIV-1 RNA and the dynamics of specific constellations of mutations after cessation of therapy. However, distinct patterns were seen. An initial rapid increase from, in most cases, undetectable or low viral load to high viraemia levels appeared with the mutated virus. Thereafter, and also in patients who had a significant viraemia before cessation of therapy, the drug-related mutations started to disappear concomitantly with a phase of a higher viral doubling time. The first pattern obviously represents the withdrawal of the inhibitory drug effect and during this phase there was no obvious difference in doubling time, or in the viral kinetics, between patients who at cessation had resistant virus and those who had wild-type virus. The data thus suggest that despite the potentially decreased replication potential, the mutated viral strains were initially capable of producing high numbers of viral progeny at an efficiency similar to that of wild-type virus. Thus, it is possible that the replication of mutated virus contributes to the CD4 T-cell decrease, which occurs simultaneously with the increase in HIV-1 load after cessation of therapy as reported by others ^[33,36].

After the initial rapid increase, and also in patients who had a significant plasma viral load at cessation of therapy, a more limited doubling time of the virus levels was seen concomitant with disappearance of mutations in some patients. The wild-type strains thus often outgrow mutant strains, most probably due to differences in viral fitness. The concomitant viral increase (< 1 log₁₀) suggests that a decreased viral fitness, presumably associated with the mutations detected in our study, causes an indirect antiretroviral effect in this range. A parallel increase in viral load and viral fitness has been reported after cessation of therapy ^[33]. Thus, the impact of decreases in viral fitness on viral replication in vivo could possibly be estimated by analysing the kinetics of viral load after cessation of drug therapy, taking into consideration the different phases of viral rebound.