Genotypic tropism testing: evidence-based or leap of faith?

Tropism testing in antiretroviral therapy HIV-1 tropism is largely determined by the specific chemokine co-receptor used in conjunction with the CD4 receptor for entry into target cells. In the early 1990s, the discovery of co-receptor use significantly advanced our understanding of virus–host interactions and disease pathogenesis in HIV infection. Work over the following decade identified a novel class of antiretroviral drugs that, uniquely among available agents, act on the host cell rather than the virus. Although HIV-1 can use several co-receptors for entry, the vast majority of HIV-1 strains can be categorized as [1]: R5: using only CCR5 as a co-receptor X4: using only CXCR4 as a co-receptor Dual-tropic or R5/X4: capable of using both CCR5 and CXCR4, although one co-receptor may be favoured. Throughout infection, the detection of only R5 virus in plasma is most common. CXCR4-using variants are more likely to be detected in patients with advanced disease and low CD4 cell counts, as either R5/X4 or mixed populations of R5 and X4 strains. The detection of exclusively X4 virus in clinical samples is rare [2]. These properties have been exploited in the development of antiretroviral agents that target the interaction between the virus and the CCR5 co-receptor, including small-molecule CCR5 antagonists and monoclonal antibodies. Maraviroc is the only small-molecule CCR5 antagonist currently licensed for use in patients infected with R5 HIV-1 [3], and is approved for treatment-experienced patients in Europe and both treatment-naive and treatment-experienced patients in the USA. The development of vicriviroc was halted in July 2010 due to its modest performance in phase 3 trials of treatment-experienced patients (Table 1) [4] and in a phase 2 trial of treatment-naive patients receiving vicriviroc or tenofovir/emtricitabine with ritonavir-boosted atazanavir [5]. Promising results have recently been reported from phase 1 and phase 2 studies of the dual CCR5/CCR2 antagonist TBR-652 [6,7]. In addition, PRO140, a monoclonal antibody that targets the CCR5 co-receptor, is currently in phase II clinical trials [8]. PRO140 is administered by subcutaneous injection every other week and offers promise as a potent and durable treatment option. Clinical trials of CCR5 antagonists have confirmed the specificity of the antiviral effect for R5 virus (Table 1) [4,9–15]. Since these agents appear to induce little viral load suppression in patients harbouring CXCR4-using variants as the dominant viral species in plasma [12,15], a tropism test is essential prior to CCR5 antagonist use in order to optimize the selection of treatment candidates.Table 1: Pivotal randomized clinical trials of CCR5 antagonists.Which assay for tropism testing? HIV-1 tropism may be observed phenotypically or genotypically, and both methods have their own advantages and drawbacks. Phenotypic tropism tests (PTT) typically produce a recombinant virus containing HIV-1 envelope sequences amplified from the patient's plasma, and observe the virus ability to infect reporter cell lines that express the CD4 receptor with either CCR5 or CXCR4 [16]. The most widely applied phenotypic test to date has been the commercial Trofile assay (Monogram Biosciences, San Francisco, California, USA), which was used to screen treatment-naive and treatment-experienced patients entering pivotal clinical trials of CCR5 antagonists (Table 1) [4,9,11–14]. The Trofile assay showed a lower limit of sensitivity of approximately 10% for consistent detection of CXCR4-using virus in a clonal mixture of R5 and X4 variants [16]. In summer 2008, a modified version of the test known as the enhanced sensitivity Trofile assay (ESTA) superseded the original Trofile as a screening tool. ESTA has a nominal lower limit of sensitivity of 0.3% for detecting CXCR4-using virus within clonal mixtures [17], but sensitivity with clinical samples varies [18]. ESTA was found to more accurately identify patients likely to show a virological response to maraviroc in a post-hoc re-analysis of the MERIT trial of maraviroc versus efavirenz in treatment-naïve patients [11,19], which used the original Trofile assay to screen patients for inclusion (Table 1). ESTA also showed a marginal benefit over Trofile in a post-hoc re-analysis of the ACTG 5211 trial of vicriviroc in treatment-experienced patients [15]. There are a number of factors limiting the use of ESTA in routine patient care. Testing is only performed in a central laboratory in California, and is expensive (list price US$1960, per test and transport cost, although it may be available through agreements with ViiV Healthcare at a lower price) and labour-intensive, with a turnaround time of about 4 weeks. The test has a relatively high failure rate [20] reflecting its complexity and stringent sample collection, storage and transport requirements [16]. A minimal volume of 3 ml of plasma is recommended, which often poses a problem for retrospective testing of stored samples and in children. In addition, there is a minimum viral load requirement of 1000 copies/ml for reliable amplification [16], thus excluding this approach in a proportion of patients, such as those with early virological failure or those with an undetectable viral load who may be seeking to switch treatment for tolerability reasons. To circumvent this limitation, use of proviral DNA recovered from peripheral blood mononuclear cells (PBMCs) is being explored. Whereas the preliminary findings indicate that proviral DNA is suitable for testing by ESTA [21], clinical validation data are awaited. HIV-1 tropism may also be inferred from a genotypic analysis of HIV-1 envelope proteins. Genotypic tropism testing (GTT) offers the advantage of platform portability, low cost (approximately 150 and up to 250 Euro per test), and rapid turnaround. The tests are based on observations that there are certain quantifiable differences between the sequences of R5 and X4 envelopes, which may be incorporated into a predictive algorithm. The determinants of tropism are thought to lie mainly (if not exclusively) in the HIV-1 gp120 V3-loop region, and the amino acid sequence of this region can therefore be used to infer likely co-receptor usage [22–24]. A number of algorithms based on bioinformatic theory, such as geno2pheno[co-receptor] (http://coreceptor.bioinf.mpi-inf.mpg.de/) and position-specific scoring matrices (PSSMs) (http://indra.mullins.microbiol.washington.edu/webpssm/), have been developed to facilitate this prediction. These compare the amino acid sequence of the test sample with sequences known to be associated with X4 or R5 tropism and the results provide an indication of how likely the virus is to use the CXCR4 co-receptor. The algorithms have become increasingly sophisticated over time; for example, the geno2pheno[co-receptor] algorithm allows clinical data (most importantly the nadir CD4 cell count, but also viral load and CD8 cell count) to be included in the model, which may improve predictions [25–27]. Since genotypic resistance testing is already routinely available in many centres, GTT is likely to be more broadly available, in addition to being faster and cheaper than PTT. As well as plasma RNA, proviral DNA amplified from PBMC can be used to determine tropism using genotypic analysis, thus removing the viral load threshold, and making all patients potentially eligible for testing [26–28]. Clinical data in support of this approach are becoming available from small cohort studies [29,30]. Is there sufficient evidence to support genotypic tropism testing in routine practice? One obvious limitation of GTT is that in its common applications it uses population (‘bulk’) sequencing to produce a consensus sequence from dominant strains within the viral quasispecies. The technique has a lower limit of detection of approximately 10–20%, and is therefore unable to detect CXCR4-using variants if present at low frequency. This limitation can be only partially circumvented by increasing the number of replicate sequences produced from each sample. As also true of PTT, the sensitivity of GTT for low-frequency variants is influenced by the size of the input and declines with diminishing levels of viral load. Many studies evaluating GTT have compared its performance using Trofile or ESTA as a reference, which has obvious limitations as neither Trofile nor ESTA can be regarded as a gold-diagnostic standard. Initial comparative studies did not look promising [31]. Subsequently, concordance between phenotypic and genotypic analyses has increased as the genotypic tests have evolved to the extent that good to excellent agreement with Trofile is now commonly seen [32,33], and good to excellent concordance with ESTA has also been shown in some studies [20,26,34]. However, the most relevant analysis is not measuring concordance between tropism assays, but rather observing how effective an assay is at predicting virological responses to CCR5 antagonist use. The availability of stored plasma samples from patients enrolled in clinical trials of maraviroc has allowed two retrospective evaluations of GTT in a large number of patients. Importantly, these analyses allow both a comparison of the tropism results given by PTT and GTT, and a comparison of clinical outcomes. The results have demonstrated that GTT, performed using plasma RNA to generate triplicate bulk V3-loop sequences from each patient, and interpreted according to predefined parameters, is similarly predictive of virological outcomes to Trofile in treatment-experienced patient [35], and equally predictive to ESTA in treatment-naive patients [36]. The first analysis used 1916 screening plasma samples from treatment-experienced patients within the MOTIVATE-1 and MOTIVATE-2 trials [9] and the sister safety study A4001029 (Table 1) [12]. Importantly, samples from patients who entered in the MOTIVATE trials were screened as R5 by Trofile, whereas those from patients who entered A4001029 were screened as CXCR4-using (‘non-R5’). Prediction of tropism by GTT using the geno2pheno[co-receptor] and PSSM interpretation algorithms was found to have relatively poor sensitivity but high specificity for predicting non-R5 virus relative to Trofile. However, among 1164 patients with clinical outcome data, the phenotypic and genotypic tests had similar ability to predict long-term virological responses [35]. The observation was confirmed in sub-analyses, including evaluation of virological responses according to number of active drugs in the background regimen. Using a similar approach with over 700 plasma samples from treatment-naive patients enrolled in the MERIT trial (Table 1), the post-hoc analysis previously performed with ESTA [11,19] was repeated using GTT [36]. As also seen with ESTA, the genotypic method was found to improve the screening of drug-naïve patients relative to the original Trofile. By excluding patients with non-R5 virus that had been classified as R5 by Trofile, and who would have been ineligible for study entry had the newer assay – either ESTA [11,19] or GTT [36] – been available, the number of virological failures on maraviroc was reduced. Virological responses (proportion with viral load <400 copies/ml and proportion with viral load <50 copies/ml at week 48) fell within the predefined noninferiority thresholds compared to efavirenz. Importantly, in the two retrospective analyses of GTT, patients with discordant PTT/GTT results did not show a large preferential direction of response [35,36]. An increasing number of prospective cohort studies in both treatment-naive and treatment-experienced patients starting maraviroc also indicate that GTT performed by bulk V3-loop sequencing and interpreted using the geno2pheno[co-receptor] algorithm is reliable in terms of positive predictive value [29,30,37]. These findings have led to the consensus that GTT is suitable for use in routine clinical practice, and guidelines which endorse the role of GTT (performed in triplicate by bulk sequencing and interpreted according to defined parameters) have been recently developed by national and European consensus panels (given below) [38–40]. Recommendations regarding the use of HIV-1 tropism testing in routine clinical practice (adapted from reference [39]): HIV-1 tropism testing should be performed prior to CCR5 antagonist therapy using a validated phenotypic or genotypic method. GTT offers a more easily accessible, rapid and inexpensive method for tropism diagnostics than phenotypic testing and is the likely preferred option in many settings. Laboratories undertaking genotypic tropism testing should do so under quality assurance schemes and according to the prevailing consensus about preferred methodology for sampling, testing, and interpretation. In treatment-naive patients, tropism testing should be performed immediately prior to the start of therapy whenever CCR5 antagonist use may be considered in the first-line regimen (unlicensed indication in Europe). In treated patients experiencing virological failure, tropism testing should be performed and the results should become available at the same time as those of drug-resistance testing to ensure all available therapeutic options may be considered. In treated patients with suppressed viraemia for whom a switch to a CCR5 antagonist is considered (e.g. due to toxicity), tropism testing may be performed using either PBMC-derived proviral DNA from a current sample, or plasma-derived RNA from a stored sample collected immediately before viral load suppression. The clinical utility of either approach should be monitored closely as supporting evidence is limited. Detection of CXCR4-using virus at any time should be considered long-lasting. No specific recommendations can be made about the longevity of R5 predictions in patients with ongoing virus replication, although a 90-day cut-off has been commonly applied. In patients with a high risk of emergence of CXCR4-using virus (based on a low nadir CD4 cell count) the test should be repeated as near as possible to the start of CCR5 antagonist therapy. Methodological considerations One important consideration is that current PTT and GTT methodologies do not use viable virus for assessing tropism. The assays cannot therefore differentiate between replication-competent and therefore clinically relevant strains, and defective variants that might have a different co-receptor usage but not necessarily impact on virological responses to the antagonist. In this aspect, the assays differ from the original methodology for assessing tropism, which used virus isolates recovered from PBMC to infect immortalized T-cell lines such as H9 or MT-2 cells [41]. Whereas producing good-quality V3-loop sequences may be achieved easily in laboratories with experience of genotypic resistance testing, it is important that the methodological approach to GTT should follow the prevalent consensus. Evidence indicates that geno2pheno[co-receptor] provides reliable discrimination between R5 and X4 sequences when the assay interpretative parameter, called the false-positive rate (FPR), is set between 5 and 10%. Using data from the MOTIVATE studies in treatment-experienced patients, a loss of antiviral activity with maraviroc was observed at an FPR of 5.75% using the geno2pheno[co-receptor] algorithm and at a cut-off of -4.25 using the PSSM algorithm [42,43]. Nearly complete loss of maraviroc activity was seen in patients with FPRs below 2%. Analysis from the MERIT trial of treatment-naive patients confirmed the utility of the 5.75% cut-off [36]. In addition, to improve sampling of the viral quasispecies and sensitivity for the detection of CXCR4-using virus, replicate testing – typically triplicate – is recommended, whereby samples undergo three separate polymerase chain reaction (PCR) amplifications followed by separate sequencing of the three PCR products [33]. Three separate results are therefore obtained for each sample, and if any sequence is identified as X4, presence of CXCR4-using variants is reported. What do we still need to clarify? Although there is an increasing body of evidence now available in support of GTT, there are still some important practical aspects for users to consider from a methodological as well as clinical perspective. For instance, the cost-effectiveness of replicate testing for samples with a high viral load is not uniformly endorsed, and the caveat remains that prospective data for GTT are limited to small observational cohorts [29,30,37]. Tropism assays should ideally detect X4 variants at levels that may cause virological failure but not exclude patients who are likely to benefit from CCR5 antagonist use. The geno2pheno[co-receptor] interpretation algorithm, for example, allows users to select the FPR, and therefore both the sensitivity for detecting CXCR4-using virus and the probability of classifying an R5 virus falsely as X4. The optimal cut-offs may differ depending on the individual circumstances of the patient. Thus, a lower FPR might be suitable for patients with severely limited therapeutic options, or in the presence of a strongly supportive background regimen, when the presence of a small proportion of ‘unrecognized’ X4 variants may be acceptable. Conversely, a greater degree of certainty of excluding X4 virus would be preferred in patients with a large range of available agents, in order to prevent unnecessary sacrifice of active agents. It is hoped that use of cut-offs could provide a tropism result similar to that seen for resistance tests. Thus, using predefined higher and lower cut-off points, samples with no X4 variants or with X4 variants suggested only above the higher cut-off would fall into a ‘green’ category predictive of a good response to CCR5 antagonists; those with X4 variants suggested below the higher and above the lower cut-off would fall into a ‘yellow’ category predictive of an intermediate response; and those with X4 variants suggested below the lower cut-off would fall into a ‘red’ category predictive of a poor response. For patients falling into the ‘yellow’ category, the clinician would need to fully consider the results of the test in the context of the individual patient's clinical situation before deciding whether a CCR5 antagonist is appropriate. For example, considerations would have to the number of other active agents in the regimen and the of the regimen, as well as the of There is a category of patients in whom is when R5 results are obtained as the presence of low-frequency X4 strains is more This patients with evidence of either current or typically by a low nadir CD4 cell available, the nadir CD4 cell may be used with the geno2pheno[co-receptor] algorithm to improve the ability to predict likely presence of X4 virus in patients sequences are A benefit of GTT is the ability to determine tropism in patients with a low or undetectable viral by sequencing proviral DNA recovered from PBMC between GTT tropism from plasma RNA and proviral DNA has been reported to be and some studies indicate that in X4 variants to be more commonly found in DNA compared with RNA It is important to that the evidence in support of tropism from proviral DNA to treatment is than that available for plasma evidence that the risk of virological is low in treated patients with a viral load below copies/ml who an with maraviroc following an R5 from proviral GTT [29,30]. In small cohort the reported risk of to be low within switch that did not a ritonavir-boosted (Table cohort of patients who maraviroc with background therapy with a plasma HIV-1 RNA load below copies/ml following an R5 genotypic tropism obtained from proviral these preliminary findings are clinical evaluation of proviral GTT is within using tropism predictions from proviral DNA there is currently an to the response to An approach for patients with suppressed viraemia is testing of a stored plasma sample collected immediately before viral load suppression. from small cohorts of patients have suggested that tropism on treatment although clinical validation is What for genotypic tropism testing? some limitations to be as the evidence in support of GTT we are becoming increasingly in its ability to phenotypic methods and in its in different settings. that have led to the development and of and quality assurance we are the time when GTT is in routine the of patients suitable for CCR5 antagonist use. Whereas both the geno2pheno[co-receptor] and PSSM have shown a good positive predictive value for responses to maraviroc there is a and clinical evidence in support of the geno2pheno[co-receptor] different bioinformatic may improve the of clinical outcomes to the genotypic methods such as which can generate of clonal sequences from a sample, are being developed with sensitivity for detecting low-frequency variants of HIV compared with bulk sequencing which at may the clinically relevant cut-offs for X4 variants and improve our understanding of X4 A consideration is whether it may be possible in the to identify patients harbouring non-R5 virus in plasma in whom a degree of virological activity of CCR5 antagonist therapy might be which enrolled treatment-experienced patients with non-R5 virus by Trofile, did not found evidence of a virological benefit of maraviroc over (Table 1) [12]. In addition, the trial found that patients as non-R5 virus by ESTA showed a poor virological response to vicriviroc [15]. Thus, whereas the that CCR5 antagonists may activity in some patients with dominant non-R5 virus in plasma remains clinical validation data are currently there is evidence to indicate that the geno2pheno[co-receptor] interpretation well a range of different HIV-1 are to fully the of on both performance of tropism tests and their predictive value for clinical responses to CCR5 In as the available data are not GTT in routine practice is both to an a of In an a prospective randomized trial would the about which approach to tropism testing is the most In the of we that GTT is an approach for routine clinical The the European on Tropism Testing and the of the European on Tropism Testing for that have the of this The is based partially on the of a that in in with support by from support to a of the and and the and are for its ViiV ViiV and Biosciences, ViiV Biosciences, ViiV