OUP user menu

Performance of the VERSANT® HIV-1 Resistance Assays (LiPA) for detecting drug resistance in therapy-naive patients infected with different HIV-1 subtypes

Inge Derdelinckx, Kristel Van Laethem, Bart Maes, Yoeri Schrooten, Kirsten De Schouwer, Stéphane De Wit, Katrien Fransen, Sergio García Ribas, Michel Moutschen, Dolores Vaira, Georges Zissis, Marc Van Ranst, Eric Van Wijngaerden, Anne-Mieke Vandamme
DOI: http://dx.doi.org/10.1016/S0928-8244(03)00240-2 119-124 First published online: 1 November 2003

Abstract

In this study we evaluated the performance of the VERSANT® HIV-1 Resistance Assays (LiPA) in detecting drug resistance in therapy-naive HIV-infected patients diagnosed in Belgium in 2000. We compared the results with population sequencing and found concordance to be in line with previous studies in treatment-experienced patients (86.87% for reverse transcriptase (RT); 92.77% for protease (PRO)). Discordance was mainly due to indeterminate reactions on LiPA (8.45% for RT; 6.85% for PRO) and minor discordances (4.13% for RT; 0.25% for PRO). Major discordances were rare (0.46% for RT; 0.12% for PRO). Indeterminate reactions were significantly associated with strains belonging to non-B subtypes.

Keywords
  • Human immunodeficiency virus type 1
  • Resistance testing
  • Subtypes
  • Line probe assay

1 Introduction

Resistance testing is considered useful for guiding treatment decisions in human immunodeficiency virus (HIV) infected patients. Despite conflicting scientific evidence, guidelines recommend the performance of resistance tests at treatment failure [15]. Although most resistance guidelines do not prefer genotyping over phenotyping [15], population sequencing is the most frequently used resistance test at the moment. However, both population sequencing and the regular phenotypic assays are expensive and lack sensitivity to detect minor resistant species [6]. Other resistance methods, such as the VERSANT® HIV-1 Resistance Assays (LiPA), also known as line probe assays, are expected to be less expensive and have been shown to be more sensitive in detecting minor resistant variants [6]. LiPA is a resistance test based on reverse hybridization of biotinylated amplified HIV gene fragments to short oligonucleotide probes for wild-type (WT) or mutant (MT) sequences at key resistance codons, immobilized on a nitrocellulose strip [7]. An enzymatic reaction demonstrates the presence of this hybridization reaction leaving a brown-purple precipitate. Base mismatch causes an absence of hybridization. This way information is obtained on a limited number of key mutations in the reverse transcriptase (RT) and protease (PRO) region of the HIV. Evaluation of this test in various clinical settings is necessary before it can be used in clinical practice.

Convincing evidence of the impact of drug resistance testing on first-line therapy response is lacking, although several reports show a slower treatment response in patients with some level of drug resistance [8,9]. Recent guidelines state that resistance testing should strongly be considered when the treatment-naive patient may have been infected with a resistant strain, especially if it concerns a recent infection [1].

HIV-1 drug resistance tests in general have been developed based on genetic information of the B subtype, the dominant subtype in the Western world. However, recent surveillance studies point out that non-B subtypes are spreading globally [10]. Considering also the increased availability of antiretroviral drugs in developing countries, where infections are mainly caused by non-B subtypes, the performance of resistance tests in non-B subtypes is valuable information. It has already been demonstrated that the performance of resistance and viral load tests is not uniform for the different subtypes [11,12]. Since samples for this study were collected in Belgium, where subtype B constitutes only around 50% of infections [10,13], it provided an opportunity to evaluate resistance testing techniques in non-B subtypes compared to strains belonging to subtype B.

In this study we evaluated the utility of the VERSANT® HIV-1 Resistance Assays (LiPA) in detecting drug resistance in therapy-naive patients diagnosed in Belgium in 2000. We compared the results with standard sequencing results and investigated the influence of the subtype of the virus on this performance.

2 Materials and methods

2.1 Samples

Samples were collected retrospectively from treatment-naive patients diagnosed in 2000, in Belgium. This was done through collaboration with four Belgian AIDS Reference Laboratories (ARL) and AIDS Reference Centres (ARC). Only samples with a viral load above 1000 copies ml−1 were included.

2.2 VERSANT® HIV-1 Resistance Assays (LiPA)

Extraction/RT-PCR was performed using the QIAamp® viral RNA mini kit and Qiagen® OneStep RT-PCR kit (Qiagen/Westburg, Leusden, The Netherlands) according to the manufacturer's instructions. Tests were performed using the VERSANT® HIV-1 RT Resistance Assay (LiPA) and the VERSANT® HIV-1 Protease Resistance Assay (LiPA) (Bayer HealthCare LLC, Tarrytown, NY, USA; available for research use only, not for use in diagnostic procedures) according to the manufacturer's instructions. The latter kit comprises two separate strips (PRO 30–84 and PRO 90). The hybridization and the color reaction were performed on the semi-automated Auto-LiPA system (protocol HIV version 5) for the majority of the strips. Some samples were processed manually. All strips were read visually. Information was thereby obtained for the resistance-associated mutations M41L, T/A/N69D, K70R, L74V, V75T, M184V/I, T215D*/S*/A*/Y/F, Q151M, K/R103N, V/I106A and Y181C/I in RT and D30N, M46I, G48V, I50V, I/A54V, V/I82F/A/T, I84V and L90M in PRO (where * stands for a revertant amino acid). In cases where there was no visible line for any probe for a particular codon, the result was ‘indeterminate’.

2.3 Cycle sequencing

Sequencing was performed using the ViroSeq HIV-1 genotyping system (version 2) (Celera Diagnostics, Alameda, CA, USA) according to the manufacturer's instructions. Upon failure of this method, an in-house assay was used, based on an outer PCR with primers AV150 and RT2, followed by a nested PCR using primers RVP5 and RVP3 or M13USP-A35 and M13RSP-NE-(1)35 for PRO or RT, respectively. The sequencing reaction was performed with primers USP, RSP, AV36 and AV44 for the RT gene and RVP5 and RVP3 for the PRO gene as well as the dye terminator technology (ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction kit, Applied Biosystems, Foster City, CA, USA) and analyzed using ABI Prism 377 (Applied Biosystems) [14]. For one codon in one sample the consensus amino acid could not be determined based on the electrofluorograms and this was scored as an indeterminate reaction.

2.4 Comparison between LiPA and sequencing results

For each codon evaluated on the LiPA strip, a comparison was made between LiPA and sequencing results. Concordance was defined as the detection of the same result with both techniques. For codon 69 on RT and codon 82 on PRO, the detection by LiPA of T/A/N69 or V/I82, respectively, was considered concordant with the detection of one of these amino acids on sequencing. Minor discordance was defined as a mixed result by one method and a homogeneous result by the second, or a mixed result by both methods, with one amino acid in common. Major discordance was defined as the case where one technique detected a completely different amino acid compared to the other technique. Indeterminate reactions were considered as discordant results.

2.5 Subtyping

Sequences were aligned with PRO and RT reference sequences of group M subtypes available in the Los Alamos Database (http://www.hiv-web.lanl.gov) and identified as either belonging to subtype B or not. In case only PRO was available, the BLAST was done based on this region only.

2.6 Statistics

Where appropriate, mean values were reported with their standard deviation (SD). Odds ratios (OR) and Cochran–Mantel–Haenszel (CMH) statistics were used where appropriate to describe associations. All calculations were made with the statistical software package SAS (The SAS Systems for Windows V8, Release 8.2, 1999, SAS Institute Inc., Cary, NC, USA).

3 Results

3.1 Patient population

A total of 115 samples were collected from therapy-naive patients, newly diagnosed in 2000. The mean viral load was 5.0 log copies ml−1 (SD: 0.6 log copies ml−1, range 3.2 to >5.9 log copies ml−1). Patients were predominantly male (58.6%), black (44.8%), with a mean age at diagnosis of 36.9 years (SD: 10.6) and predominantly infected with HIV by heterosexual transmission (57.7%).

All samples were processed as described above to obtain a resistance profile by means of VERSANT® HIV-1 Resistance Assays (LiPA) as well as sequencing. This yielded a result for 100 (86.96%) and 103 (89.56%) samples for sequencing of RT and PRO, respectively. LiPA yielded an interpretable result for 111 (96.52%), 109 (94.78%) and 106 (92.17%) samples for RT, PRO 30-84 and PRO 90 strips of LiPA, respectively. A combined result with both techniques was available in 99, 99 and 95 samples, again for RT, PRO 30–84 and PRO 90, respectively.

3.2 Concordance between sequencing and LiPA

A total of 1089 codon positions on RT and 788 on PRO were retained for concordance evaluation. Overall, concordance was found in 86.87% of codons (946/1089) for RT, and 92.77% (731/788) for PRO (Figs. 1 and 2). Discordance was mainly due to indeterminate reactions on LiPA (92/1089 or 8.45% for RT; 54/788 or 6.85% for PRO) and in one case due to an indeterminate reaction on sequencing of RT (1/1089 or 0.09%). When codon positions with an indeterminate reaction were not considered, concordance was calculated to be 94.98% (946/996) for RT and 99.59% (731/734) for PRO. The rate of concordance/discordance/indeterminate reactions varied from codon to codon (Figs. 1 and 2). On RT, the highest rate of concordance was noted for codons 103 (96/99, 96.97%) and 151 (95/99, 95.96%), followed by codons 69 (90/99, 90.90%) and 106 (90/99, 90.90%). The lowest concordance was attained for codons 41 (67/99, 67.67%) and 181 (78/99, 78.78%). All concordant codon positions on RT were WT by both methods, except for three cases where a MT was detected: M41L (1), M184V (2). On PRO all concordant reactions were WT.

Figure 1

Concordance of the VERSANT® HIV-1 RT Resistance Assay (LiPA) with population sequencing. This figure shows the percentage of indeterminate reactions (IND), major (MDisc) and minor discordances (mDisc) and concordance (Con) for every codon evaluated on the LiPA strip for RT compared with sequencing results.

Figure 2

Concordance of the VERSANT® HIV-1 Protease Resistance Assay (LiPA) with population sequencing. This figure shows the percentage of indeterminate reactions (IND), major (MDisc) and minor discordances (mDisc) and concordance (Con) for every codon evaluated on the LiPA strip for PRO, compared with sequencing results.

3.3 Minor discrepancies

Minor discordances were particularly found on the RT strip (45/1089 or 4.13%) and less on the PRO strips (2/788 or 0.25%) (Figs. 1 and 2). Forty-four of 47 minor discordances were due to LiPA showing a mixture while sequencing detected the WT only. In two cases, sequencing detected a mixture of the WT and a polymorphic amino acid not detectable by LiPA. In one case, a different mixture was detected with the two methods, where LiPA could not detect one of the amino acids detected by sequencing. For RT, the proportion of discordance due to minor discrepancies was highest for codon 151 (2 of 3 or 66.66%), codon 215 (12 of 16 or 75.0%), codon 106 (7 of 9 or 77.77%) and codon 184 (12 of 14 or 85.7%).

3.4 Major discrepancies

Major discrepancies were rare, with five evaluated codon positions in RT (5/1089 or 0.46%) and one in PRO (1/788 or 0.13%) (Figs. 1 and 2). One case in RT involved the detection of WT for codon 69 with LiPA and of the polymorphism 69N with sequencing. The other four cases on RT all involved codon 41. In PRO the mutation L90M, detected by sequencing in one sample was not detected by LiPA.

3.5 Indeterminate reactions

For the PRO strip, 54 out of 57 (94.73%) of all discordant reactions were due to indeterminate reactions (Figs. 1 and 2). In 20% of samples, information on at least one codon on PRO 30–84 and PRO 90 combined was lacking. For all these codons, sequencing detected the WT amino acid. Indeterminate reactions were even more common for RT. Sixty of 111 samples (54.05%) had at least one indeterminate reaction present in RT. Most samples (45) had one position without information, 13 had two positions and two samples had four positions that could not be evaluated due to an indeterminate reaction. For some codons in RT indeterminate reactions were the main reason for discordance (codon 70 with 6/10 or 60%; codon 103 with 2/3 or 66.66%; codon 69 with 6/8 or 75%; codon 41 with 26 of 32 or 81%; codon 181 with 19/21 or 90.4%; codon 74 with 12 of 13 or 92.30%; codon 75 with 12 of 12 or 100%). All indeterminate reactions were WT on sequencing, apart from one indeterminate reaction at codon 181 in RT where a MT amino acid was detected by sequencing.

3.6 Prevalence of resistance

Six out of 100 samples showed RT resistance, as detected by sequencing, based on the IAS-USA mutation list and footnotes (March 2003). Three samples carried the V118I mutation, one sample carried the M184V mutation, one sample carried the M184V mutation in combination with the Y181C mutation and one sample exhibited three zidovudine-related mutations (M41L, T215D/G and L210W). One out of 103 (1.0%) PI sequences carried a primary PI mutation (L90M), and 96/103 (93.2%) of samples showed at least one secondary PI mutation.

Of the 10 mutations detected by sequencing, only three were detected concordantly by LiPA. Four mutations are not evaluated with the LiPA technique (three times codon 118 and one time codon 210 in RT). In three cases, sequencing and LiPA were discordant: mutation Y181C was indeterminate on LiPA, mutation L90M in PRO was WT on LiPA (major discordance) and mutation T215D/G in RT was detected with LiPA as being a mixture of WT and revertant (T/D/S) (minor discordance).

3.7 Subtype distribution

Of the 103 samples with sequencing data available, the majority belonged to the non-B subtype (74/103, 71.84%), thus reflecting the predominance of non-B subtypes among newly diagnosed patients in Belgium in recent years [10]. Of the non-B subtypes, almost half (34/74 or 45.9%) were pure subtypes (Subtype C: N=15, Subtype A: N=11, Subtype D: N=4, Subtype G: N=3, Subtype F1: N=1). Seventeen samples were circulating recombinant forms (CRF; CRF_01: N=6, CRF_02: N=11) and 23 samples were recombinant without being classified as a CRF.

3.8 Performance and association with subtype

Non-B subtypes had a higher likelihood of carrying at least one indeterminate reaction. This association was stronger for the RT strip (6/28 versus 28/73 of samples carrying at least one indeterminate reaction for B and non-B subtypes, respectively; OR=5.89, confidence interval (CI): 2.13–16.32) than it was for the PRO 30–84 strip (3/28 versus 22/71; OR=3.74, CI: 1.02–13.71). For the PRO 90 strip, five samples had an indeterminate reaction, four belonging to subtype non-B, but this was not statistically significant. Non-B subtype strains also produced a higher number of indeterminate reactions compared to subtype B strains. This association was significant for RT (CMH Statistics, Row Mean Scores Differ: P=0.0005), as well as for all strips combined (CMH, Row Mean Scores Differ P=0.0012). The correlation between the number of indeterminate reactions on the combined PRO 30-84 and PRO 90 strips and subtype was not significant (CMH, Row Mean Scores Differ: P=0.06).

Five of six major discordances were found in a non-B subtype strain. The majority (32/47 or 68.08%) of minor discrepancies occurred with strains belonging to the non-B subtype, but this was in line with the high prevalence of non-B subtypes in this cohort.

4 Discussion

The VERSANT® HIV-1 Resistance Assays (LiPA) are genotypic assays based on the reverse hybridization principle. So far, the performance of LiPA has predominantly been studied in treated patients. Since resistance testing in untreated patients might prove to be relevant, it is necessary to evaluate this technique also in this setting. In addition, the performance of LiPA in non-B HIV-1 strains is insufficiently known. In this study we investigated the performance of VERSANT® HIV-1 Resistance Assays (LiPA) in a cohort of therapy-naive patients diagnosed in Belgium in 2000. The concordance with sequencing was studied, as well as the correlation of its result with the subtype of the strain.

In general, an interpretable result was obtained for the majority of the samples. Compared to sequencing, LiPA was successfully performed in a larger number of samples (96.52% compared to 86.96% for RT; 94.78% and 92.17% for PRO 30–84 and PRO 90, respectively, compared to 89.56% for PRO).

The overall concordance (at codon level) was 86.87% for RT, and 92.77% for PRO. If indeterminate scores were excluded, this concordance rose to 94.98% for RT and 99.59% for PRO. This was in line with previously published data for samples from therapy-experienced patients [7,1520]. The concordance rate was higher for PRO than for RT.

Discordance was mainly due to indeterminate reactions on LiPA, with no reaction at either the WT or the MT line of one particular codon. The occurrence of these hybridization failures is most probably due to the variability of the HIV. One base pair mismatch, at the codon of interest or at one of the flanking codons, results in a missing signal. This is a very common problem with this technique, with more than half of the samples in our study having an indeterminate reaction for at least one codon in RT. These figures are in line with the recent study of Stürmer et al., who found 39.4% of samples from mainly treatment-experienced patients to be carrying at least one mutation in RT. The authors concluded that ‘LiPA might still work better in untreated patients’. However, in our study we also found the number of indeterminate reactions too high to be acceptable. Indeterminate reactions were noticed to a lesser extent for PRO, where only 20% of samples scored indeterminate for at least one codon. In all cases but one, the indeterminate position turned out to be WT on sequencing. This reflects the low prevalence of resistance in the samples and suggests that the flanking sequences are mainly responsible for the mismatch. However, since LiPA is designed to evaluate only key resistance mutations, overlooking even one mutation could have a significant impact. The rate of hybridization failure we observed for RT seemed to be somewhat lower, 8.44%, in our study compared to 16.25% as reported by Servais et al. [19], using the previous version of the strip. On the other hand, we scored a higher rate of hybridization failure for PRO, 6.85% in our study compared to 3.97% in the study of Servais et al. However, direct comparisons cannot be made, since the patient population was different in the two studies.

Minor discrepancies occurred in 4.13% and 0.25% of RT and PRO codons, respectively, which was much less than the number of indeterminate reactions. They are of particular interest, since the sensitivity of the LiPA strip for minor species is supposed to be higher compared to sequencing [6]. Minor discrepancies could therefore merely result from the higher sensitivity of the LiPA technique. To differentiate the presence of minor resistance variants with non-specific reactions, further study is needed. Two minor discrepancies were based on the detection of a mixture on sequencing and not on LiPA, but in these cases the other amino acids detected by sequencing were not evaluated on the LiPA strip. Significantly more minor discrepancies were noted for the RT strip compared to the PRO strips. Sensitivity and specificity of the probes could account for this phenomenon.

Very few major discrepancies (0.46% for RT; 0.13% for PRO) were present. This was in line with other previously published reports [7,1520].

The LiPA strips evaluate only a limited number of mutations. When compared with the list of mutations published by the IAS-USA, LiPA detects 6 of 19 (31.6%) NRTI, 3 of 15 (20%) NNRTI, and 10 of 11 (90.9%) primary PI-related mutations. Only one secondary PI mutation is detected. As a consequence, of the 10 resistant codons detected in the study, four were not evaluated with the LiPA strips (three for RT 118 and one for RT 210). It should be mentioned, however, that controversy exists whether isolated mutations at codon 118 of RT should be considered as a marker of transmitted resistance. Together with indeterminate reactions, minor and major discordances, this led to only three (33.3%) mutations being detected by LiPA in full concordance with sequencing.

When comparing strains belonging to B and non-B subtypes, the performance of LiPA was significantly worse in strains belonging to the non-B subtypes. This was particularly true for both the presence (OR: 5.89; CI: 2.13–16.32) and number (CMH, Row Mean Scores Differ: P=0.0005) of indeterminate reactions on RT, and of the presence (OR: 3.74; CI: 1.02–13.71) of indeterminate reactions on the PRO 30–84 strip. Although the recent update of the LiPA strip was also based on strains belonging to the non-B subtypes, the original probes were designed on a database predominantly consisting of B strains [7]. Since hybridization failure can be caused by one basepair mismatch, LiPA is particularly vulnerable in scoring the genotype of HIV, a virus known to have high sequence variability. This implies that constant updating of the probes is necessary in the light of the ever-evolving epidemic. In Belgium, as well as in other European countries, the prevalence of non-B subtypes is rising [10,13,21]. Also, with treatment becoming available in the developing countries, where the majority of strains belong to non-B subtypes, there is a growing need for validation of resistance testing for non-B subtypes.

In conclusion we can state that the performance of the VERSANT® HIV-1 Resistance Assays (LiPA) in treatment-naive patients is concordant with sequencing in the majority of evaluated codon positions. Major discordant reactions are rare. Minor discordances could be due to a higher sensitivity of the technique to minor resistant variants, although non-specific reactions cannot be ruled out. Indeterminate reactions are a serious drawback of this technique, with more than half of the samples lacking information on at least one codon. With LiPA about two-thirds of the resistance detected with sequencing was missed. The presence and number of indeterminate reactions is also linked to the subtype of the virus, with strains belonging to the non-B subtypes carrying more indeterminate positions compared to B subtypes.

Acknowledgments

The AIDS Reference Laboratories are supported by the Ministry of Social Affairs through a fund within the Health Insurance System. This work was supported in part by the Katholieke Universiteit Leuven (Grant nr. VIS/01/014). I.D. has benefited from a grant by the ‘Instituut voor de aanmoediging van de Innovatie door Wetenschap en Technologie in Vlaanderen’ (2001–2002). Innogenetics, Ghent, provided the VERSANT® HIV-1 Resistance Assays (LiPA) and materials for the processing of the samples.

References

  1. [1].
  2. [2].
  3. [3].
  4. [4].
  5. [5].
  6. [6].
  7. [7].
  8. [8].
  9. [9].
  10. [10].
  11. [11].
  12. [12].
  13. [13].
  14. [14].
  15. [15].
  16. [16].
  17. [17].
  18. [18].
  19. [19].
  20. [20].
  21. [21].
View Abstract