INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The human immunodeficiency virus (HIV) population in an infected patient includes genetically distinct viral variants called "quasispecies" that evolve from the initial inoculum (8). The propagation of certain novel mutant quasispecies is a function of existing environmental selective pressures created by drug therapy and immunologic responses of the host. Under this specific selection pressure, only viral strains that contain mutations that confer a survival advantage will propagate and emerge as a major quasispecies (U. Dietrich, H. Ruppach, S. Gehring, H. Knechten, M. Knickmann, H. Jager, E. Wolf, R. Husak, C. E. Orfanos, H. D. Brede, H. Rubsamen-Waigmann, and H. von Briesen, Letter, AIDS 11:1532-1533, 1997). HIV genomic mutations have been loosely categorized as primary and secondary mutations by their direct effects on drug susceptibility and the mechanism of resistance (10). Primary mutations generally decrease the level of drug binding to target enzymes and are selected for early as resistance mutations accumulate. They usually confer a severalfold decrease in drug activity and are relatively inhibitor specific. Secondary mutations usually have less discernible effects on viral resistance alone but can contribute to drug resistance by increasing the fitness of viral strains that already contain primary mutations. This compensatory effect of secondary mutations may spread and contribute to resistance to other drugs of the same class. In contrast to (acquired) mutations associated with drug resistance, naturally occurring polymorphisms exist as well (13; G. Myers, B. Korber, B. H. Hahn, K.-T. Jeang, J. W. Mellors, F. E. McCutchan, L. E. Henderson, and G. N. Pavlakis, Human retroviruses and AIDS: a compilation and analysis of nucleic acid and amino acid sequences, [http://hiv-web.lanl.gov/]). Such polymorphic variants apparently represent strains with enough viral fitness or replication capacity to exist in the absence of drug-induced selective pressure. Numerous polymorphisms have been described in the viral protease gene in HIV-infected patients who are treatment naive (11).

Evaluation of the genetic composition of HIV has evolved from a traditional epidemiological tool into an important clinical asset in the management of HIV infection. Genotypic analysis of HIV enables the identification of individual and combinations of nucleotide substitutions that are known to confer resistance to specific antiretroviral agents. Whether a patient has failed treatment with multiple drug combinations or is just beginning drug therapy, genotypic analysis can help identify drugs less likely to be therapeutically effective. This technology can help "individualize" drug combinations to attain maximum virus suppression and patient longevity.

The HIV-GenotypR method (GenotypR; Specialty Laboratories, Inc., Santa Monica, Calif.), the HIV PRT GeneChip assay (GeneChip; Affymetrix, Santa Clara, Calif.), and the HIV-1 RT Line Probe Assay (LiPA; Innogenetics, Alpharetta, Ga.) are three genotyping assays currently in use. Each system uses different methods to identify nucleotide mutations within the viral genome.

The GenotypR uses the ABI 377 DNA sequencing system (Applied Biosystems Inc.), which consists of an electrophoresis and detection unit, reagents, and a computer software program for sequencing analysis. The ABI 377 DNA sequencing system uses a modification of the Sanger dideoxynucleotide chain terminator chemistry, producing DNA amplicons labeled with four different fluorescent dyes, one specific for each nucleotide. DNA samples undergo electrophoretic separation on a polyacrylamide gel, and an argon laser excites the fluorescence-labeled nucleotides. The nucleotides are identified one by one as they migrate past the laser detector. The digital signal that represents the DNA sequencing data is analyzed by specialized software and is displayed as a graphic representation called an "electropherogram" or as a linear string of one-letter nucleotide codes. The single-letter code is then compared to a consensus sequence of wild-type HIV obtained from GenBank to identify the presence of drug resistance mutations at the regions of interest.

GeneChip consists of probe arrays, reagents, a fluidics station, a confocal laser scanner, and computer software. The probe array is a silicon-glass chip that contains over 16,000 unique oligonucleotide probes complementary to the viral reverse transcriptase (RT) and protease genes. The probes are laid out in a precise location or a grid pattern on the array and are available for hybridization with fluorescein-labeled target nucleic acids. Nucleotide and mutation identification is dependent upon the hybridization of labeled viral nucleic acid fragments to these oligonucleotide probes. The probe array is housed in a hollow plastic cartridge, through which RNA product and wash buffers are cycled. On the basis of the patterns of hybridization of the target sequences to the oligonucleotide probes, HIV genomic sequences and point-specific mutations are simultaneously identified.

LiPA consists of nitrocellulose strips that contain immobilized oligonucleotide probes, strip troughs, trough holding trays, and reagents. The LiPA test strip for the RT gene contains 33 oligonucleotide probes immobilized as 20 parallel bands. The probes contain both wild-type and single-base changes of the RT gene known to confer resistance to specific nucleoside RT inhibitors. Like GeneChip, mutation identification is dependent upon the hybridization of viral sequences to these probes; however, LiPA does not perform viral sequencing. It searches only for mutations in a preselected group of RT gene codons. Biotinylated target sequences hybridize to the probes attached to specific locations on the strip and form colored precipitate bands (14). The locations of these colored bands on the strip determine the presence of specific wild-type codons, mixtures of codons, and mutant codons.

We evaluated and compared each system with respect to (i) the concordance of the total nucleotide mutations identified and (ii) the identification of mutations with specific practical relevance to patients' antiretroviral drug programs. We describe further the practical feasibility of the assays, sources of assay data variation, and clinical applications of each assay.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

HIV genotyping assays basically use a two-step procedure: (i) PCR to amplify RNA or DNA fragments to sufficient quantities for mutation detection and (ii) mutation identification via nucleotide sequencing or hybridization of labeled nucleic acid fragments to oligonucleotide probes. Plasma samples with greater than 1,000 copies of viral RNA per ml are generally required for accurate results. Samples with "undetectable" viral loads (<400 copies/ml) are not suitable for genotyping. Accurate genotyping relies on laboratory quality assurance, skilled technique, and expert clinical interpretation of the mutations identified. The interpretation of resistance-associated mutations is dependent upon an up-to-date, expanding genomic database of sequence variations that correlate with phenotypic drug resistance.

Study population and specimen collection. Twenty-four plasma samples from 22 HIV type 1 (HIV-1)-positive patients (confirmed by enzyme-linked immunosorbent assays and Western blot assays) were collected from August 1998 through March 1999. Patients were recruited with informed consent at the University of Wisconsin Hospitals and Clinics, Madison, Wis. (n = 19), and at the Mayo Clinic, Rochester, Minn. (n = 3). Thirteen of the patients enrolled in the study were failing their first triple-therapy regimen, as defined by a rising "detectable" viral load (>400 copies/ml), despite combination antiretroviral therapy (4). Sixteen patients had received prior drug therapy, including one patient who was off therapy at the time of sample collection. Six patients were treatment naive at the time of sample collection. Plasma from two treatment-naive patients was collected and genotyped a second time after both were failing an initial three-drug regimen. The plasma viral load of each patient was determined by the Roche-Amplicor HIV Monitor Test (12) and ranged from 1,027 to more than 10⁶ copies/ml. Upon collection, each sample was spun and separated into three plasma aliquots (one for each assay) and frozen to 20°C. GenotypR was performed at Specialty Laboratories, Inc., while GeneChip and LiPA were both performed at the Mayo Clinic.

Target preparation. Viral RNA isolation was conducted by the Qiagen (Qiagen Inc., Valencia, Calif.) extraction protocol at Specialty Laboratories, Inc., and the RNAzol B (Biotech Laboratories, Inc., Houston, Tex.) nucleic acid extraction protocol was conducted at the Mayo Clinic. Target cDNA was generated by reverse transcription-PCR with specific primers from the HIV-1 pol gene, as described by the manufacturers. The cDNA amplicons were then subjected to DNA sequencing or probe hybridization, depending upon the assay used.

GenotypR by DNA sequencing. For DNA sequencing, amplification of viral RNA was performed with RT and by various nested PCR methods. Briefly, reverse transcription was done in one cycle of 10 min at 25°C, 30 min at 42°C, and 5 min at 99°C with the primer NE1 (5'-CCT ACT AAC TTC TGT ATG TCA TTG ACA GTC CAG CT-3'). The first PCR was done in 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 60 s with primers NE1 and pol A (5' TTG GTT GCA CTT TAA ATT TTC CCA TTA GTC CTA TT 3'). These primers flank the first 266 amino acids (800 bp) of the HIV polymerase gene. Conditions similar to those used for the first PCR were used for the second PCR with primers pol A and Brev (5'-GGT GAT CCT TTC CAT CC-3') or primers NE1 and B (5'-GGA TGG AAA GGA TCA CCG-3'). These primers generated fragments of 157 (471 bp) and 113 amino acids (339 bp), respectively. The amplified products were visualized by agarose gel electrophoresis, purified on a Qiagen QIAquick 8 PCR Purification Kit, and stored at 20°C for DNA sequencing.

GeneChip assay. In GeneChip, target HIV nucleic acids were converted into fluorescein-labeled, fragmented RNA copies by an in vitro transcription reaction in a mixture that consisted of 5× transcription buffer, 5× HIV PRT labeling mix containing fluorescein-UTP, MgCl₂, dithiothreitol, T3 or T7 RNA polymerase, RNA guard, and 1 to 6.5 µl of the PCR cDNA template. The mixture was incubated at 37°C for 60 min to denature the cDNA and allow bidirectional transcription to occur. The cRNA obtained from this reaction was then fragmented by adding concentrated MgCl₂, and the mixture was incubated at 94°C for 30 min. A hybridization mix was created by combining the cRNA transcripts with 5× SSPE (1× SSPE is 0.18 M NaCl, 10 mM NaH₂PO₄, and 1 mM EDTA [pH 7.7])-Triton X-100 solution and a control oligonucleotide.

LiPA system. LiPA is based on the principle of reverse hybridization. After reverse transcription of HIV RNA, which was done according to the manufacturer's specifications, target viral sequences were amplified with biotinylated nested PCR primers (14). The amplification mixture contained 5 µl of cDNA, 4.5 µl of 10× Taq buffer, 0.3 µl of each deoxynucleoside triphosphate at a concentration of 25 mM, 1 µl (10 pmol) of each biotinylated primers, 38 µl of H₂O, and 0.2 µl (1 U) of Taq (Stratagene, La Jolla, Calif.). The annealing temperature was set at 57°C, extension was at 72°C, and denaturation was at 94°C. Each step was 1 min, the outer PCR was performed for 40 cycles, and the nested reactions were performed for 35 cycles.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Details of the treatment history of each patient with HIV infection and measurements of each patient's viral load at the time of genotyping are presented in Table 1, along with a list of mutations discordantly identified by the three genotyping systems. The analysis of the results is described below.

The performances of HIV genotyping by GenotypR and GeneChip were compared for the viral open reading frames represented in both systems; 40 codons in the RT and protease genes were evaluated. LiPA identified mutations in seven codons within the RT gene alone, six of which represent known drug resistance-associated mutations. All samples were genotyped by all three assays, and the results were subjected to pairwise comparisons. In comparisons that involved LiPA, only the six resistance-associated codons evaluated by LiPA were considered.

A total of 960 codons (40 codons per patient sample × 24 samples) were evaluated by using GenotypR and GeneChip (Table 2). Of these codons, 927 were identified identically (concordant codons) by both methods (96.6% concordance). Of the 33 mutant codons that were not identified identically by both methods (discordant mutant codons) (3.4% discordance), 20 were identified only by GenotypR and 13 were identified only by GeneChip. Of the mutations identified by GenotypR alone, eight corresponded to secondary mutations and natural polymorphisms (RT gene, D67N and K219Q; protease gene, L33F, M36I, V77I, and L90M) associated with patients' current failing drug regimens. The other 12 mutations identified by GenotypR alone correlated with past drug regimens. Among the mutations identified by GeneChip alone, five mutations corresponded to patients' current failing drug regimens (RT gene, K65R and L74V; protease gene, L10I, L10R, and K20R). The L74V mutation in the RT gene, a primary mutation conferring resistance to didanosine, was found in a patient failing a didanosine-based regimen. The other eight mutations identified by GeneChip alone included five mutations related to previous drug regimens and three mutations unrelated to current or past therapies.

A total of 138 RT codons (6 codons per patient sample × 23 patients) were evaluated by LiPA and compared to the same codons evaluated by GenotypR (Table 3) and GeneChip (Table 4). A total of 127 codons were identified identically by LiPA and GenotypR and also by LiPA and the GeneChip (92% concordance for both pairs). Among the mutations confirmed by both LiPA and GeneChip, GenotypR failed to identify the L74V mutation in the RT gene for a patient failing a didanosine-based regimen. Among the mutations confirmed by both GenotypR and GeneChip, LiPA failed to identify the primary mutations T215F and M184V in the RT gene for two patients failing current drug regimens that containing zidovudine and lamivudine, respectively. All mutations confirmed by both GenotypR and LiPA were also identified by GeneChip. Two mutations identified by LiPA alone, L74V and R70K in the RT gene, were associated with patients' previous drug regimens. The 11 discordant mutant codons (8.0% discordance) between LiPA and GenotypR represented six mutations identified only by GenotypR and five mutations identified only by LiPA (Table 3). Similarly, the 11 discordant mutant codons between the LiPA and the GeneChip represented seven mutations identified only by GeneChip and four mutations identified only by LiPA (Table 4).

Because no single assay may be capable of identifying all drug resistance-associated mutations present in a heterogeneous viral population, we reanalyzed the data using the total number of mutations identified by each assay as an independent reference or "gold standard" for comparison (Table 5). By using GenotypR as the reference method, GeneChip identified 85.4% and LiPA identified 85.7% of all mutations identified by GenotypR. Similarly, GeneChip identified 88.1% and LiPA identified 88.9% of the mutations associated with resistance to patients' current treatment programs as identified by GenotypR. GeneChip identified 100% of the primary mutations identified by GenotypR.

By using GeneChip as the reference method, GenotypR identified 90% and LiPA identified 84.1% of all mutations identified by GeneChip. Similarly, GenotypR identified 92.2% and LiPA identified 88.9% of the mutations associated with resistance to patients' current treatment programs as identified by GeneChip. GenotypR identified 96.2% of the primary mutations identified by GeneChip.

By using LiPA as the reference method, GenotypR identified 87.8% and GeneChip identified 90.2% of all mutations identified by LiPA. Similarly, GenotypR identified 94.1% and GeneChip identified 100% of the mutations associated with resistance to patients' current treatment programs as identified by LiPA. For two patients, mutations K70R and L74V in the RT gene were identified by LiPA alone. Both of these mutations correlated with the patients' previous drug programs and most likely represent minority quasispecies.

For the treatment-naive patients, GenotypR, GeneChip, and LiPA identified mainly wild-type virus. Certain polymorphisms and secondary mutations were identified by GenotypR (A71A/V and V77I in the protease gene) and GeneChip (L10I and V77I in the protease gene). LiPA identified the K70R mutation in the RT gene of one treatment-naive patient, the first mutation to appear with zidovudine treatment and a possible transmitted mutation.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The 96.6% concordance of codon identification between GenotypR and GeneChip and the 92% concordance between LiPA and the other two assays lend support to the premise that the three assays are comparable in terms of mutation identification. The few discordant mutations present generally corresponded to secondary mutations and natural polymorphisms with unrecognized clinical significance. The discordant mutations most commonly identified were D67N in the RT gene, which was missed six times by GeneChip, and L10I/R in the protease gene, which was missed five times by GenotypR. Intrinsic assay sensitivity could be a factor, but LiPA could neither confirm nor refute the presence of these mutations (which were not included on the test strip), raising the additional possibility of false-positive calls. These mutations did not appear to significantly alter the drug resistance profiles of the isolates from the patients. GeneChip identified all primary mutations that corresponded to patients' current drug regimens as identified by either GenotypR or LiPA. GenotypR and GeneChip identified slightly more mutations than LiPA (among the six codons evaluated with all three systems). With few exceptions, the rate of discordance between the three assays for the detection of mutations that corresponded to patients' current failing drugs was minimal.

By using each assay as a reference for comparison, at least 84% of the total mutations identified by one assay were confirmed by another assay (Table 5). The <16% rate of discordance among the assays was mainly seen for secondary mutations and polymorphisms of unrecognized clinical significance. Mutations associated with patients' current failing therapy were more consistently identified by each assay, with a resulting >88% concordance. By using GenotypR as the reference method, the percentages of overall mutations identified by GeneChip and LiPA were similar. By using GeneChip as the reference method, GenotypR identified a higher percentage of both total and current treatment-associated mutations than LiPA. In a comparison of GenotypR and GeneChip, each assay identified at least 96% of current treatment-associated primary mutations identified by the other assay. By using LiPA as the reference method, GeneChip identified more mutations than GenotypR.

No genotyping assay is 100% sensitive for the detection of mutations present in all quasispecies, and these results indicate that each assay has the capacity to identify mutations not detected by one or both of the other assays. The number of total codons analyzed by LiPA must be considered within the context of the other assays. The reduced number of total codons analyzed by LiPA enabled each missed mutation to have a more profound impact on the degree of concordance. Nevertheless, when taken together, each system demonstrated a similar capacity to identify clinically relevant mutations associated with contemporaneous drug therapies. Mutations associated with resistance to failing treatment regimens are present in the predominant circulating viral quasispecies (R. W. Shafer and D. Stevenson, Primer on HIV resistance [http://hivdb.stanford.edu]). In this study these predominant mutations were more readily identified among the three assays. Because discrepancies in mutation identification can be found between genotyping assays, ultimate decisions regarding antiretroviral therapy should also incorporate information regarding patient clinical and virologic response to therapy and a thorough ascertainment of treatment history.

A general understanding of fundamental assay technology, protocol feasibility, data interpretation, and sources of data variation all play a role in the selection of the optimal genotyping assay for the clinical microbiology laboratory. Identification of the limitations of each assay is important for the clinician to appropriately interpret and apply test results to patient care (Table 6).

General observations. The ability to sequence the entire RT and protease genes enables GenotypR and GeneChip to identify essentially any referenced nucleotide mutation present in the region of interest. GenotypR is the only method at present able to identify insertions, deletions, and frame shifts implicated in resistance to specific drugs. In this study, we used the first-generation LiPA method, which identifies mutations in primary codons that correspond to resistance to four antiretroviral drugs. LiPA strips that identify mutations that correspond to resistance to the nonnucleoside RT inhibitors and protease inhibitors have recently become available. The relevant quality control procedures for these three assays have recently been described (1-3).

Assay feasibility. The automated hybridization and wash cycles performed by the GeneChip fluidics station along with the automated laser scanner and computer software data analysis allow efficient sample handling and shortened processing time. Laboratory output can be maximized while minimizing sample mishandling. An automated reagent dispenser with an aspiration manifold is available for use with LiPA; however, it is not included with the standard assay and poses an additional cost. Manual performance of LiPA wash and rinse cycles requires additional effort and may require additional training of technicians.

Data interpretation. The identification of colored precipitate bands on the LiPA strip enables more simplified, objective, and consistent data interpretation. Bands specific for wild-type and mutant codons enable identification of true viral mixtures. Both sequencing with the sequencer from Applied Biosystems Inc. and GeneChip produce a complete list of mutations identified in the RT and protease genes but require additional expert interpretation to define clinically relevant mutations.

Sources of data variation. There are numerous potential sources of data variation between the three assays. Patient samples contain numerous heterogeneous viral strains variably present as quasispecies. The prevalence of each strain is a product of the specific environmental selective pressures, both immunologic and drug induced (7). Variable levels of some of these viral quasispecies may not be consistently identified by each assay. LiPA may provide enhanced identification of some of these low-prevalence strains and assist in the detection of emerging drug-associated mutations (14). For our patients, discrepancies in mutation identification occurred predominantly among secondary mutations and polymorphisms. The extraction procedures for any assay can be technically demanding for the inexperienced technician and pose an enduring risk of sample contamination. The automated systems used in GeneChip and the optional LiPA reagent dispenser can minimize errors due to sample handling and maximize product output.

Conclusion. GenotypR, GeneChip, and LiPA are three systems currently in use for HIV genotyping. Each presents a different approach to mutation identification. Overall, the total mutations identified among the codons evaluated were concordant by the three assays. There was a higher concordance rate among the assays when mutations associated with patients' current failing regimens were identified. Most mutations that were not concordantly identified by the assays were secondary mutations and polymorphisms with minimal clinical significance. Among the three assays, GenotypR identified the largest number of mutations and polymorphisms. GeneChip identified all primary mutations associated with patient's current drug regimens, and LiPA was the only one that identified a possible transmitted zidovudine resistance mutation. Mutations associated with failing treatment regimens were occasionally missed by each assay. Nevertheless, for the vast majority of isolates in patient samples genotyped, each assay correctly identified the most relevant mutations associated with patients' current treatment programs. Aside from equipment costs, each system can readily be adapted to most clinical microbiology laboratories and used efficiently by the experienced technician (5).