Evolution of proviral DNA HIV-1 tropism under selective pressure of maraviroc-based therapy

Objectives: To evaluate the evolution of HIV-1 coreceptor tropism in proviral DNA of patients during maraviroc-based therapy. Methods: Fourteen heavily high active antiretroviral therapy (HAART)-treated patients with a CCR5 Troﬁle proﬁle were monitored over a 24 month period from the start of maraviroc therapy. Whole-blood samples were obtained at different timepoints, and coreceptor tropism was determined for proviral DNA from the V3-loop region sequence using the Geno2Pheno algorithm [false positive rate (FPR): 20%]. Results: At the start of maraviroc treatment, 13/14 patients were viraemic (median: 4.33 log copies/mL). Concordance in R5 tropism (R5/R5) was observed between circulating HIV-RNA (Troﬁle) and HIV-DNA provirus in 10/14 patients (median FPR ¼ 54.0%), while 4 patients showed a CXCR4-tropic R5/X4 variant in their provirus (FPR: 5.8%, 5.7%, 16.6% and 1.1%, respectively). All R5/R5 patients showed a stable HIV-1 DNA coreceptor usage. Two out of four R5/X4 patients showed a tropism shift in their archived provirus and, after 6 months a prevalence of R5-tropic virus was detected in DNA. The other two R5/X4 patients harboured the 11/25 genotype, and maintained X4 tropism in provirus during the study. Virological response did not reveal differences in RNA decay and CD4 + cell recovery in patients with discordant tropism. Conclusions: A relatively good correlation between RNA and DNA tropism was observed at baseline. Proviral DNA tropism remained stable over 24 months of maraviroc-based therapy, indicating that determination of proviral DNAV3 sequence could be used in tropism prediction in clinical practice. The data also conﬁrm the im-portance of the 11/25 rule in predicting viral tropism.


Introduction
The HIV type 1 (HIV-1) virion requires a primary receptor, CD4, and a secondary chemokine receptor (such as CCR5 or CXCR4) to start cell infection. 1 Over recent years the CCR5 entry coreceptor has become the target of a new class of inhibitors, with a novel drug, maraviroc, approved for clinical use. Because maraviroc targets the human coreceptor and not viral proteins, it was originally assumed that resistance acquisition by HIV would be slower. However, viral use of CCR5 in the presence of the pharmacological coreceptor blocker, and the emergence of preexisting minority CXCR4-tropic variants have been described. 2 Previous studies, both in vitro and in clinical trials, have described viral strains carrying different maraviroc-resistance-associated mutations within the third variable (V3) region of the viral gp120 gene. 3,4 In these clinical trials, about half of the individuals with therapy failure showed a coreceptor switch from CCR5 (R5) to CXCR4 (X4). The detectable presence of dual mixedtropic or X4 viruses has been associated with therapeutic failure under treatment with CCR5 antagonists. 5,6 The emergence of viral variants using the CXCR4 coreceptor is considered worrying by many clinicians because of its association with accelerated disease progression 7 -9 and reduced survival in untreated individuals.
Since maraviroc activity is restricted to patients harbouring exclusively R5-tropic variants, a phenotypic analysis for coreceptor usage is mandatory before prescription. 10 The Trofile test, based on a recombinant phenotypic assay, is the most commonly used, but genotypic methods have also recently been proposed and validated. 11,12 Coreceptor usage can be determined by bioinformatic prediction models based on analysis of the sequence of the HIV gp120 V3-loop region, which is the principal determinant of tropism. 13,14 Such genotypic approaches have been developed for both plasma (amplification of viral RNA) and peripheral blood mononuclear cells (PBMC; amplification of proviral DNA). This approach is particularly important to determine viral tropism in patients with undetectable viral load, and could represent a valid tool to investigate the prevalence of virus tropism in archived viral genomes. It is important to note that tropic shift can occur in patients receiving highly active antiretroviral therapy (HAART) for 5 years with undetectable viral load. 15 Previous studies have found a concordance between HIV-RNA and HIV-DNA, 16 -18 even if the percentage of X4-tropic viruses proved enriched in cell-associated provirus. 17,18 Herein we present a study on V3-loop sequencing using whole blood of treatment-experienced patients who started maravirocoptimized therapy and were monitored for 24 months. Our goal was to evaluate the evolution of provirus tropism in PBMC HIV-DNA of patients under the pressure of a CCR5 antagonist and its role as a tool to predict virological response.

Methods
The present study is part of an ongoing multicentre observational study on the use of new antiretroviral inhibitors in long-term treatment-experienced patients with HIV-1 (the NIA-ISS study). Patients had documented resistance or intolerance to the three main drug classes [nucleoside reverse transcriptase inhibitors (NRTIs), nonnucleoside reverse transcriptase inhibitors (NNRTIs) and protease inhibitors (PIs)] and were prescribed maraviroc (150 mg twice daily) in a context of salvage regimens. All patients included in this study had HIV coreceptor usage examined in plasma samples using the enhanced sensitivity Trofile assay (ESTA; Monogram Biosciences, South San Francisco, CA, USA), which is one of the most common diagnostic tests to determine HIV-1 coreceptor usage.
Clinical course and immunological and virological status were evaluated every 3 months, and biological samples were collected. Analysis of the V3 loop from whole blood was performed at entry (before initiation of maraviroc-based therapy) and at two or more subsequent timepoints. The study was conducted in compliance with the requirements of the sites' Institutional Review Boards/Ethics Committees.

Protease and reverse transcriptase resistance testing
RT-PCR and sequencing were performed at clinical sites using commercial kits, according to the manufacturer's' instructions. The genotypic susceptibility score (GSS) was calculated using the Stanford HIV Resistance Database, 19 and was assigned to each drug included in the salvage regimen (0 for drugs to which the virus was considered to be resistant, 0.5 for partially resistant, 1 for no resistance). The sum of the scores for the individual drugs provided the global GSS of the antiretroviral regimen. Based on treatment history, when patients were naive for a new drug with a new mechanism of action (i.e. enfuvirtide, raltegravir and maraviroc) the drug was assumed to be fully active and GSS¼1 was assigned. 20 The number of drugs with an individual GSS score of 1 (fully active) was recorded for each regimen.

Viral amplification and HIV-1 gp120 V3 region sequencing
For sequencing the HIV-1 gp120 V3 domain, HIV-1 DNA was extracted from 200 mL of whole blood using a QIAamp DNA Blood extraction kit (Qiagen, Hilden, Germany), following the manufacturer's instructions.
The HIV-1 env fragment (2.8 kb) including the V3 region was amplified by PCR using GP-160 primers and the Trugene core reagents kit (Siemens Healthcare Diagnostics Inc., Deerfield, IL, USA). The conditions for amplification were: one cycle at 948C for 2 min, 37 cycles (948C for 30 s, 608C for 30 s and 688C for 2 min) and a final step at 688C for 7 min. PCR-amplified products were sequenced in both 5 ′ and 3 ′ directions using the V3-F2 forward primer Cy5.5 GTA-CAATGTACACATGGAAT (HXB2 nt 6959 -6980) and the V3-RE2 reverse primer Cy5 AAAATTCCCCTCCACAATTA (HXB2 nt 7352 -7371) 21 and the Trugene core reagents kit. The sequencing conditions were: one cycle at 948C for 5 min, 30 cycles (948C for 20 s, 548C for 20 s and 708C for 1.5 min) and a final step at 708C for 5 min. Sequencing products were visualized on the OpenGene TM System automated sequencer and analysed using the sequence analysis software, with the limit for recognizing minority variants in a nucleotide mixture set to 20%.

Genotypic determination of HIV-1 coreceptor usage
Tropism prediction was performed with the clonal geno2pheno (G2P) prediction algorithm, 22 with the false positive rate (FPR) set to 20%. The determination of the FPR at 20% was based on recommendations of European Guidelines for HIV-1 tropism, which suggest increasing the false-positive rate to 20% when only one sequence from a DNA sample can be generated for tropism analysis. 12 The reported FPR was used as quantitative input. The FPR indicates the probability of falsely classifying an R5 variant as X4. The sample tropism was expressed as the proportion of non-R5 sequences within the sample's viral population. Patients harbouring ,20% non-R5 variants in PBMC were classified as having non-R5 virus. The presence of positively charged amino acids (arginine or lysine) at positions 11 and/or 25 was also evaluated. 23,24 Statistical analysis Differences were evaluated using the Fisher's exact test for categorical variables and Wilcoxon test for continuous variables. Significance levels were set at 0.05. All the analyses were performed using SPSS software, version 17.0 (SPSS Inc., Chicago, IL, USA).

Patients
The population characteristics are reported in Table 1 HIV-RNA log copies/mL) and a median CD4 cell count of 209 cells/mm 3 .

Antiretroviral therapy
Before starting maraviroc, regimens were based on a median of 4 drugs (range 1 -6). Three patients were taking only lamivudine, and 9/14 were taking an NRTI-based regimen plus PIs. Five patients were taking enfuvirtide. All patients reported resistance to at least one drug of the main antiretroviral classes (NRTIs, NNRTIs and PIs).
The introduction of maraviroc was accompanied by marked changes in treatment; most of the new regimens included drugs from four different classes with a median of 5 drugs (range 3 -7). Eight patients (57.1%) had co-administration of raltegravir; PIs and/or NRTIs were present in 71.4% of regimens and NNRTIs were present in 42.9%. Four patients (28.6%) were also taking enfuvirtide. Overall, the median GSS score, including maraviroc, was 3.75 (range 1.5-5.0) and the number of fully active drugs was 3 (1 -5; Table 2).

Virus tropism in plasma and whole blood at baseline
Based on phenotypic HIV-RNA Trofile analysis, all patients harboured a CCR5-tropic virus and were thus eligible for maraviroc.
V3-loop sequences were obtained from 14 of the 17 proviral DNA samples available at baseline, a success rate of 82.4% for amplification and sequencing. Only patients whose V3 regions were sequenced at baseline were included in the study. According

Evolution of virus tropism in proviral HIV-DNA under maraviroc-based HAART
Coreceptor usage was determined at two or more timepoints during a follow-up period of 24 months. The median interval between two subsequent analyses was 7 months (range 4 -18 months), and the median number of timepoints for V3-loop sequencing was 4 per patient (range 2-4).
All patients harbouring R5-tropic virus at baseline maintained a stable V3-loop proviral genotype (less than 50% variation with respect to baseline levels, Table 2). The fluctuations in FPR are indicated in Figure 1. Of the four patients harbouring The number of fully active drugs is given in parentheses.

Virological response
Virological response did not differ between patients harbouring R5-or X4-tropic virus. After 3 months of maraviroc-based therapy, most patients had undetectable (,50 HIV-RNA copies/mL) plasma HIV-RNA (69.2%). No difference in decay of plasma RNA was observed between patients harbouring R5-or X4-tropic virus (P ¼ 0.188; Figure 2). At month 6, one patient (patient 054) had virological failure (plasma RNA . 4.0 log copies/mL), and at month 12 exited the study. His tropism profile in DNA did not change over time, and from the V3-loop sequence analysis we could exclude resistance to maraviroc. Viral suppression persisted in most patients, and after 24 months of maraviroc-based therapy only three patients had detectable HIV-RNA plasma levels (P,0.002). All viraemic patients had predominance of the R5-tropic variant in archived DNA.
A positive trend in CD4-cell recovery was observed throughout the study period (P ¼ 0.007; Figure 3). The median cell recovery was +79 CD4 cells/mm 3 (range 248 to +218) after 3 months and +271 after 24 months. No differences in CD4 recovery were observed between patients harbouring X4 variants and patients with R5-tropic provirus.

Discussion
In this study viral tropism prediction was analysed using the web-based tool, geno2pheno[coreceptor], which predicts coreceptor usage from the V3 region of HIV-1 gp120. This bioinformatic software has been validated by many clinical studies that found a high concordance with the Trofile test. 25 -30 Although this study did not aim to compare geno2pheno[coreceptor] and Trofile for the prediction of coreceptor usage, our baseline data showed 71.4% concordance between the two approaches, which is consistent with data from analogous studies. 17,31,32 According to G2P analysis, four patients harboured an X4 variant. A higher prevalence of X4 variants in PBMCs than in plasma has already been observed by other authors, for patients in both early 17 and chronic phases of HIV infection, 18 and this is consistent with the observation that the archived virus population may not correspond to the most prevalent variant in plasma. 33  Baroncelli et al.
CCR5 usage in proviral DNA was stable over time in all individuals harbouring R5-tropic variants. In most of our patients we observed highly conserved HIV-DNA V3-region sequences during the 24 months on maraviroc-based therapy. Switches over time in HIV-DNA tropism under suppressive HAART are rare; 11,35 our findings argue against a selective pressure towards X4 evolution under suppressive therapy, thereby confirming that changes in tropism occur with low frequency in maraviroc-treated patients. 36 Moreover, changes in the cellular reservoir could tend to be less frequent in patients with undetectable viraemia. The two patients carrying the 11/25 genotype harboured the X4-tropic virus throughout the follow-up period. It has been observed that the base changes at position 11 and 25 are neither necessary nor sufficient for a phenotypic switch, 37 and our data confirms that the 11/25 rule is one of the most important genotypic predictors of X4-tropic virus. 38 Conversion from X4 to R5 tropism in HIV-DNA was observed in the other two R5/X4 patients with a prevalence of X4-tropic provirus at baseline. The conversions occurred in the first 6 months of maraviroc-based therapy, and the R5 variant remained prevalent in the following months. Why and under which circumstances the X4 R5 shift occurs is not well understood. However, consistent with our findings, previous studies reported that conversion from X4 to R5 variants is generally observed shortly after the initiation of antiretroviral therapy, before complete suppression of plasma viraemia. 15,39 This small study confirms the efficacy of maraviroc-based therapy in multidrug-experienced HIV patients. 6,40 In our study, most patients introduced to maraviroc-based therapy had a good virological response, and plasma viraemia was undetectable as early as month 3. The low number of HIV-DNA X4-tropic patients did not permit statistical analysis, but the presence of X4 variants in archived virus did not seem to influence immunovirological response, as already observed in a previous study. 41 One patient with R5-tropic provirus exited the study because of virological failure (probably due to scarce adherence, data not shown), and no variation in the HIV-DNA V3-loop sequence was recorded. During the clinical monitoring, 3 patients (all with baseline R5-tropic provirus) remained viraemic and R5-tropic during follow up. Unfortunately we were unable to verify the DNA-RNA tropism concordance because no plasma samples were available to analyse RNA tropism.
The major limitation of our study is the small number of patients. However the longitudinal analysis of proviral HIV-DNA V3 loop demonstrated stability of the V3 region under maravirocselective pressure. In aviraemic patients, the evaluation of HIV-DNA coreceptor usage might be important to investigate whether administration of maraviroc selects for HIV variants that prefer the CXCR4 coreceptor. Our results suggest that DNA V3-loop sequencing could be a useful tool to predict shift in virus tropism, and this study suggests that determination of the HIV-1 V3 loop in DNA and RNA populations may provide different yet complementary information.