INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

The randomized controlled clinical trial for the measurement of a clinical outcome remains the benchmark for testing the clinical efficacy of a therapeutic intervention. Such studies are expensive, and the search for alternative endpoints to clinical outcome that can act as surrogate markers for clinical endpoints continues with the aim of reducing the size and duration of the trials. This is especially the case for trials of antiretroviral drugs for the treatment of human immunodeficiency virus (HIV) type 1 (HIV-1) infection. The number of available drugs and the large numbers of possible combinations of these drugs make the use of surrogate markers a necessity. The recent development of assays for virus expression and virus replication have increased the pressure to seek alternatives to the clinical endpoint (8, 12, 14, 22, 26). As part of the investigation of surrogate markers for determining the therapeutic effect of antiretroviral drugs, the Medical Research Council of the United Kingdom has funded studies into the application of virological parameters. Although to date major phase III trials have relied upon clinical endpoints (7, 9, 10, 30), laboratory markers (including CD4 cell counts and p24 antigen [p24Ag] levels) have been extensively used in phase I/II trials. More recently, virological markers have been used to monitor the effects of antiretroviral therapy (21).

There is a clear need to validate laboratory endpoints against clinical outcome. To this end a cross-sectional study was undertaken with subjects previously enrolled in the Medical Research Council Concorde Trial (zidovudine monotherapy versus placebo) comparing measurements of virus expression by culture-based and non-culture-based techniques. While culture is labor intensive, somewhat insensitive, and not easily subjected to quantification, it has the advantage of providing a virus isolate for further study, although propagation of the isolate in culture is likely to subject the in vivo viral population to selective pressures in vitro (23). On the other hand, molecular biology-based methods for the quantification and characterization of viruses, although relatively new and unproven, have the advantage of high sensitivity and do not introduce the perturbations and selection pressures associated with viral culture.

Culture-independent molecular biology-based methods have been developed for measuring HIV RNA and DNA levels (6, 24, 26, 29) and have already been used for the monitoring of these parameters in the plasma of patients in clinical trials (1, 13). More recently, investigators have developed quantitative assays which enable the accurate identification of virus populations carrying mutations associated with drug resistance (15, 20). This cross-sectional study has allowed these methods to be compared with quantitative culture of virus from peripheral blood mononuclear cells (PBMCs) and plasma. It has also allowed comparison of the phenotypic drug resistance of cultured virus isolated from the subjects enrolled in the study with quantitative genotyping for the presence of resistance-associated codon mutations in the RT gene of the virus isolate, PBMC-derived proviral DNA, and RNA in plasma.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Patients. Participants in the Concorde Trial who had been on trial capsules for more than 1 year, who were still taking capsules, and who were classified as CDC group II or III by their clinician at the time that the sample was taken were eligible for participation in the present study.

Samples. Samples for quantification by culture were sent to the laboratory and were processed within 6 h of venipuncture. PBMCs were separated on Ficoll-paque gradients within 6 h of venipuncture. Serum samples for HIV RNA and p24Ag level determinations were separated after clotting within 6 h of venipuncture. PBMCs, plasma, and serum were stored at 80°C until they were analyzed.

Quantification by culture of provirus load in PBMCs and virus load in plasma. The provirus load in PBMCs was determined as described previously (2, 12). Tenfold dilutions of previously unfrozen PBMCs from the study subjects, ranging from 2 × 10⁶ to 2 × 10² cells, were cocultivated with 2 × 10⁶ phytohemagglutinin (PHA)-stimulated normal donor PBMCs in the presence of interleukin-2. Medium was changed after 24 h and every 3 to 4 days thereafter. Cultures were monitored over 28 days for the presence of p24Ag, and a concentration of >200 pg/ml was considered indicative of virus growth.

Assay of zidovudine resistance by culture. Virus was initially isolated from PBMCs by cocultivation of PBMCs at a ratio of 1:1 with PHA-stimulated PBMCs from healthy donors. Culture supernatants were monitored for reverse transcriptase (RT) activity, and those collected at the peak of activity (14 to 21 days) were used in the phenotypic assay for measuring drug sensitivity (4). Briefly, 1-ml volumes of supernatant or dilutions up to 10⁵ were mixed with 3 ml of 2 × 10⁶/ml PHA-stimulated PBMCs from healthy donors for 2 h at 37°C. The cells were washed twice and added in quadruplicate to microtiter wells containing zidovudine to give final drug concentrations of 0, 0.025, 0.25, 0.5, 1.25, and 6.25 µM. Half of the culture medium was changed on days 5, 7, and 10, and the RT activity was monitored in the supernatants harvested at these times. The virus titer (50% tissue culture infective dose [TCID₅₀]) in the wells containing no drug was estimated (by the Karber or Reed-Muench methods) and the sensitivity to the drug was assayed by linear regression analysis (logit/log) of the peak RT activity in the wells containing 100 TCID₅₀s. Resistance was defined as a 90% inhibitory concentration (IC₉₀) of zidovudine of >0.3 µM.

Quantification by PCR amplification of provirus load in PBMCs and plasma virus load in plasma. Serum HIV RNA levels were assayed by an immunocapture quantitative PCR as described previously (26). Test samples were quantified by comparison to a standard curve generated in parallel from a high-titer serum from an HIV-infected individual diluted in normal human serum. Test samples and dilutions of the standard were immunocaptured, reverse transcribed, and PCR amplified in duplicate. The PCR product from each amplification was assayed in duplicate.

Assay of zidovudine resistance by PMA. DNA from stored PBMCs and RNA from plasma and from culture supernatants was prepared as described above and was genotyped. Mutations at codons 41, 67, 70, 215, or 219 of the RT gene in pol were detected and quantified by a PCR-based point mutation assay (PMA) (15). Proportions of mutant sequence in excess of 2% (4% at codon 215) were considered to indicate the presence of virus bearing resistant genotypes at that particular codon.

Statistical methods. In quantifying virus levels in plasma and provirus levels in PBMCs by culture, the mean number of tissue culture infectious units per ml (plasma) and 10⁶ cells (PBMCs) was estimated from the number of positive wells (those with a p24Ag concentration of >200 pg/ml) among the replicates of each dilution, assuming a Poisson distribution for the number of infectious units per unit volume. Nominal infectious units (NIUs) are related to TCID₅₀ by the relation NIU = log(2) × TCID₅₀.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Baseline characteristics. Of the 97 participants included in the study, 97% were male and 88% were homosexual or bisexual males. The mean age was 34 years (standard deviations, 7.63 years). Of the 64 participants who had received zidovudine, 3 had received it for less than 1 year, 18 had received it for between 1 and 2.5 years, and 43 had received it for more than 2.5 years. The median CD4 cell count was 500 (interquartile range [IQR], 380 to 606). p24Ag was detected in the serum of 7% of the subjects.

View larger version (13K): [in this window] [in a new window]   FIG. 1.   (A) Relationship between proviral load in 74 subjects measured by culture of PBMCs and PCR amplification. (B) Relationship between plasma HIV-1 load in 74 subjects measured by culture and the serum HIV load measured by immunocapture PCR.

Correlation between phenotypic and genotypic resistance. Phenotypic drug resistance was assayed in virus isolated from PBMCs collected from 34 subjects. Genotypic drug resistance was assayed in cDNA generated from viral RNA in the tissue culture supernatants for 32 of the 34 available isolates. Comparison of the data for the 32 patients subjects whose samples were tested by both methods indicates that genotypic resistance was closely related to virus phenotypic resistance, expressed as IC₉₀s (in micromolar) of zidovudine (Table 3).

Ex vivo studies. HIV sequences were recovered from three ifferent types of specimens. Thus, proviral DNA from PBMCs, cDNA from reverse transcription of HIV-1 RNA in serum, and cDNA from HIV-1 RNA in tissue culture supernatants were separately analyzed by PMA. Measurements of each analyte for a particular subject gave broadly similar measures of genotypic resistance (Table 5). The GRSs for viruses cultured from study subjects correlated well with the IC₉₀s for the isolates (Fig. 2A), as did the GRSs for peripheral blood DNA (Fig. 2B) and serum RNA (Fig. 2C).

View larger version (10K): [in this window] [in a new window]   FIG. 2.   Relationship between phenotypic zidovudine resistance, expressed as IC90 of zidovudine, and genotypic resistance, expressed as GRSs for RNA in culture (A) (n = 32), DNA in PBMCs (B) (n = 35), and RNA in serum (C) (n = 33).

Prevalence of resistance to zidovudine. Genotypic resistance was detected in three study subjects who were reported not to have received zidovudine. Approximately 25% of the viral population from the first subject carried a mutation at codon 215. Samples taken before and after the cross-sectional survey confirmed the persistent detection of this genovar. For a second zidovudine-naive subject, an analysis of proviral DNA showed proportions of mutant sequence at codons 41, 70, 215, and 219 of 0, 33, 79, and 0%, respectively. No signal was detected at codon 67. However, the assay signals produced by PMA from this sample were unusually low, prompting further analysis by single-copy sequencing (28). The sequence showed that all five codons were those found in the wild type, but a number of probe target mismatches were revealed at or close to the codons of interest. These mismatches were probably responsible for the low signal strength and false detection of a mutation. Phenotypic resistance data were not available for these two subjects because virus could not be cultured from their PBMCs. The third subject, whose virus was also phenotypically resistant (IC₉₀ = 0.88 µM), carried a viral population exhibiting proportions of mutant sequence at codons 67, 70, 215, and 219 of 60, 71, 25, and 81%, respectively, when the DNA was examined and at the same codons of 100, 81, 24, and 90%, respectively, when the RNA in culture was examined. It was subsequently confirmed that this subject was misclassified and had in fact been receiving zidovudine for 2.9 years.

View larger version (16K): [in this window] [in a new window]   FIG. 3.   Correlation between duration of zidovudine therapy and GRS for 71 subjects receiving treatment and for whom resistance measurements were available.

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Correlation between duration of zidovudine therapy and IC90 for virus from 34 subjects receiving treatment and for whom resistance measurements were available.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The specimens analyzed in this study were predominantly from homosexual men who had been asymptomatic at the start of the Concorde Trial. Those on zidovudine therapy had been taking the drug for a median duration of 2.8 years (range, 0.9 to 3.8 years). At the time of sampling for the present study, the median CD4 cell count for the subjects was 500/mm³. p24Ag was detected in the serum of 7% of the subjects. Taken together, these characteristics of the subjects would predict a low mean virus load in the serum of the study group at the time of sampling (22), irrespective of any continuing antiviral effects of zidovudine therapy.

The rates of detection of PBMC-derived proviral DNA and cell-free viral RNA in serum by RT-PCR were higher (92 and 95%, respectively) than rates of detection by culture of PBMCs and plasma (78 and 22%, respectively). The detection rates by either method were lower than those reported by other workers (8, 12, 25, 27), although this probably reflects the clinical status of the subjects studied and the continuing antiviral effects of zidovudine rather than a lack of detection sensitivity of the methods used. The correlation between virus load in PBMCs assayed by culture and by PCR was good (Spearman rank correlation coefficient = 0.67; Fig. 1A). In contrast, a poor correlation was seen between the virus load in plasma assayed by culture and PCR (Spearman rank correlation coefficient = 0.36; Fig. 1B). This is likely to have resulted from inaccuracies in quantification by culture for samples with low infectious virus loads, an interpretation borne out by the fact that the median titer of infectious virus for the 17 subjects from whom virus was cultured from plasma was only 2.2 TCID₅₀s/ml (range, 1.4 to 41 TCID₅₀s/ml).

Overall, the results of the load studies indicate that the use of PCR-based quantification for the monitoring of viral load in clinical trials of antiretroviral therapies is probably the preferred method. The higher rates of detection by the PCR-based methods enable the greatest proportion of trial subjects to be monitored for virus load at any given time during therapy, and the finding of the higher titers may itself result in a higher degree of accuracy in the quantification of the virus load in plasma and the virus response to antiviral therapy. From a practical point of view, the use of PCR-based methods is preferred over the use of culture-based techniques. PCR-based methods are generally safer because most nucleic acid preparation methods result in the inactivation of infectious virus; furthermore, they do not generate viable virus. This avoids the requirement for the use of P3-level containment facilities. Also, to attain maximum sensitivity for virus isolation, the use of cultures of PBMCs from uninfected donors and the processing of samples from fresh clinical material in real time are needed. The donor cultures vary in their sensitivity to infection with HIV-1 and need to be maintained as a continuous supply of PHA-stimulated cultures in order to anticipate the receipt of samples for analysis. This is demanding on time and resources. On the other hand, genome amplification methods can use stored clinical material without an apparent loss of sensitivity, allowing batch testing of specimens. Overall, the time and staff resources needed to make virus load measurements genomically are considerably lower than those needed to determine virus load by culture. The anticipated use of microtiter-format PCR systems and automated handling equipment is likely to facilitate a further increase in sample throughput by genomic amplification methods. Although the use of culture-based methods will remain essential for assessing drug resistance and the phenotype of the virus, the lower detection rates seen with culture will need to be addressed. Samples with low virus loads are more likely to be the rule when investigating patients on combined antiviral therapy.

Thus, although culture-based methods for defining phenotypic drug resistance will continue to be required to establish in vitro and in vivo resistance patterns for newly developed antiretroviral compounds, defining an association of phenotypic drug resistance with changes in the viral genome sequence, as elegantly demonstrated for zidovudine by Larder and Kemp (18), will allow the detection of drug resistance by molecular biology-based genotyping methods, at least for monotherapy. The present study has compared a genotyping method (PMA) with determination of phenotypic drug resistance.

Table 3 and Fig. 2A, B, and C show the high degree of association between genotypic and phenotypic measures of resistance. The more stringent measure of IC₉₀ rather than the more commonly reported IC₅₀ was used to define phenotypic resistance since the IC₉₀ more clearly discriminated between drug-sensitive (IC₉₀, < 0.3 µM) and drug-resistant (IC₉₀, >0.3 µM) virus strains. A strong correlation was seen between the phenotypic and the genotypic measures (Table 3; Fig. 2). All phenotypically resistant strains in culture had detectable resistance-associated mutations at one or more gene loci in viral sequences derived from HIV RNA in the culture supernatant (Fig. 2A). No mutation levels above the arbitrary cutoff were detected in any drug-sensitive virus strain.

A good correlation of phenotypic and genotypic resistance was shown (Table 4) for each of the codons analyzed and was particularly marked when the effects of all five mutations are combined in the GRS. Genotypic resistance measurements made with virus RNA in culture, virus RNA in plasma, and proviral DNA in PBMCs showed good correlations with each other (Table 5). Again, this relationship was enhanced when the influence of all mutations was expressed as a GRS. The advantage of using a combined GRS is that it allows a single measure for a given viral population to define the degree of genomic resistance of that population by using data for five loci.

Not surprisingly, increasing levels of drug resistance, both phenotypic and genotypic, were observed as the time on therapy increased (Fig. 3 and 4; Table 6). It was interesting that isolates from a proportion of the subjects remained drug sensitive beyond 2.5 years of therapy (2 of 13 isolates by phenotyping; 4 of 71 isolates by genotyping). The significance of this remains to be determined.

As was the case with virus load measurements, the PCR-based method for the detection of genotypic drug resistance offered a number of advantages over culture-based phenotyping. The greater number of samples that could be amplified by PCR allowed a higher proportion of samples to be analyzed by genotyping methods. More importantly, the use of direct genotyping of viral nucleic acid sequences isolated without culture from clinical samples eliminated the disturbance to the observed virus population induced by isolation and culturing of the virus in the presence of the drug during phenotyping. The comparison of viral RNA sequences and phenotyping in culture indicated a close correlation such that no zidovudine-sensitive virus carried detectable resistance mutations at any codon (Fig. 2A). However, when the DNA sequences from the ex vivo PBMC samples used to set up the isolation were compared with the resistance phenotype, the relationship was less precise (Fig. 2B). The divergence became more apparent when the HIV sequences in the plasma of the subjects whose isolates had the resistance phenotype were compared (Fig. 2C), and in the comparison four phenotypically resistant viruses could not be detected in plasma. It seems likely that this represented a selection of minor variants in culture and that these variants differed from virus replicating in the patient at the time of sampling. If this were the case, it may be better to avoid the selection inherent in culture when measuring levels of viral resistance during therapy and to rely upon direct genotyping of the virus that is replicating in vivo.

The limitation of genotyping remains the indirect nature of the measurement. Mutations other than those described previously may contribute to the observed phenotypic resistance, and such mutations, even if they are detected by gene sequencing, can only be proven to be effective in causing phenotypic resistance by reintroduction of the RT sequences into a wild-type infectious molecular clone. In the present study it was reassuring to find that all phenotypically resistant virus strains contained a detectable mutation in at least one of the five codons previously associated with zidovudine resistance. Future studies of combination therapy may require the development of alternative laboratory methods for studying the phenomenon of drug resistance in treated patients. As the number of drugs used in a therapeutic regimen increases, the use of phenotypic resistance measurements becomes even more time-consuming and unwieldy. Furthermore, the increasing number of resistance-associated point mutations in combination therapies make the detection of point mutations inefficient, and although the use of automated sequencing techniques makes the analysis of large parts of the viral genome possible, the possibility that previously undescribed mutations will arise from drug interactions exists, such as the recently described multiple resistance mutation at codon 151. One possible solution to the analysis of drug resistance in combination therapies may lie in the use of recombinant virus assays (17) or the analysis of phenotypic resistance at the enzymatic level in cloned and expressed viral proteins (5).