HIV-1 Phenotypic Susceptibility to Lopinavir (LPV) and Genotypic Analysis in LPV/r-Naive Subjects With Prior Protease Inhibitor Experience

The relationship between phenotypic susceptibility to lopinavir (LPV) and genotypic pattern was investigated in LPV-naive, protease inhibitor (PI)-experienced subjects. Protease sequences of 100 HIV isolates with ascertained susceptibility (determined by Antivirogram) to LPV were analyzed (VircoGen). Two different thresholds (2.5- and 10-fold) were used for defining reduced susceptibility. Mutations were classified as LPV/r (the actual formulation of LPV that combines LPV with low-dose ritonavir) mutations according to the International AIDS Society-USA. Thirty-four isolates showed reduced LPV susceptibility (2.6- to 75.9-fold). Fold resistance to LPV correlated with the number of total and LPV/r mutations (Spearman coefficient = 0.62 and 0.74, respectively; P < 0.001). Current PI therapy (P = 0.002) and indinavir administration (P < 0.001), >5 LPV/r mutations (P < 0.0012), and detection of L10FIRV, K20MR, M46IL, I54VL, A71VT, G73SA, V82AFTS, I84V, and M90L were associated with LPV resistance in univariate analysis. Factors independently associated with LPV resistance were K20MR (odds ratio [OR], 13.9; 95% confidence interval [CI], 1.3-145.1; P = 0.028 ), I54VL (OR, 131.7; 95% CI, 10.5-1654.7; P < 0.001), G73SA (OR, 19.2; 95% CI, 1.4-273.7; P = 0.029), and I84V (OR, 177.5; 95% CI, 6.0-5232.5; P = 0.003) mutations and >9 protease mutations (OR, 18.6; 95% CI, 1.6-213.0; P = 0.019). Sixteen of 34 and 18 of 34 isolates with reduced LPV susceptibility showed >10-fold or <10-fold LPV resistance, respectively. Linear regression analysis demonstrated that each additional LPV mutation and I54VL accounted for much of the fold resistance to LPV (adjusted R² = 0.70). In conclusion, for PI-experienced patients requiring salvage therapy, switching to LPV should be based on the number of baseline mutations and the presence of mutation 54.

Virologic failure in patients treated for HIV-1 infection is commonly associated with the emergence of drug-specific resistance mutations. In addition, a troublesome cross-resistance within the same class of medications complicates the therapeutic options for subjects who have treatment regimen failure. Cross-resistance is particularly common among the protease inhibitors (PIs), making the sequential use of these agents frequently problematic.

The resistance profile of lopinavir (LPV), the most recently approved PI, has yet to be clearly defined. The actual formulation of LPV combines lopinavir with low-dose ritonavir (LPR/r), resulting in a medication that reaches such high plasma levels that viral replication is strongly restricted and resistance is hindered. ¹ In fact, due to the relative nature of HIV resistance, the increased LPV exposure that is obtained in vivo may overcome a certain grade of resistance; therefore, in vitro resistance to LPV/r does not always coincide with clinical resistance.

In HIV isolates obtained from individuals for whom regimens with other PIs fail, mutations have been identified at 11 positions (10, 20, 24, 46, 53, 54, 63, 71, 82, 84, and 90) in the protease gene that correlate with >2.5-fold resistance to LPV, ² and it has been suggested that the accumulation of ≥8 of these mutations substantially reduces the response to LPV-based therapy. ³ The K20MR and F53L mutations possibly indicate >10-fold resistance to LPV, whereas mutations at positions 10, 54, 63, 71, 82, and 84 are associated with relatively low levels of LPV resistance (4- and 10-fold). ² Nevertheless, the presence of mutations at codons 10, 54, and 82 when combined with 4 additional resistance mutations in baseline isolates from PI-experienced patients who subsequently received LPV/r-based salvage therapy was significantly associated with an increased risk of virologic failure. ⁴ Similarly, univariate analysis demonstrated that the detection of mutations 10, 20, 54, and 82 in patients with prior multiple PI exposure was significantly associated with virologic failure of LPV/r, whereas mutations 46 and 36 were almost irrelevant for failure. ⁵ In this same study, stepwise logistic regression analysis showed that mutations 10, 20, 46, 47, and 54 were each independently associated with an increased risk of unsuccessful LPV/r combination treatment. Finally, patients for whom LPV/r therapy failed were most often found to develop mutations at codons 10, 33, 46, 54, and 82. ⁶

As a result of these observations, in the most recent inventory of drug resistance mutations in which primary and secondary mutations in the protease gene now read as major and minor, the International AIDS Society-USA listed 26 mutations at 16 protease positions that are associated with reduced susceptibility to LPV. ⁷ However, because the extent to which each individual mutation impacts LPV is poorly defined, all these mutations are considered without distinction. Moreover, even the number of mutations necessary to confer resistance to LPV is debatable. In fact, recent observations indicate that as few as 4 mutations can confer >10-fold resistance to LPV. ⁸

In this study, we further investigated the relationship between phenotypic susceptibility to LPV and the genotypic pattern. For this purpose, HIV-1 protease sequences were analyzed from subjects with prior PI experience whose virus isolates were shown to be either susceptible or resistant to LPV.

PATIENTS AND METHODS

The patients included in this study participated in the Italian PhenGen prospective, multicenter, randomized study comparing genotype with virtual phenotype and real phenotype for prediction of virologic response to a new regimen administered to patients whose previous antiretroviral therapy had failed. Subjects in the PhenGen study had a plasma viral load between 2,000 and 200,000 copies/mL during therapy. Patients with ≥1 previous failures could be included if their total antiretroviral experience did not exceed 6 antiretroviral drugs. Plasma samples obtained at the time of entering the PhenGen study were used for assessment of either phenotypic or genotypic drug resistance. Samples were analyzed by Virco (Virco NV, Mechelen, Belgium) using the Antivirogram phenotype and the Virco genotyping system (VircoGen). An expert advisory committee suggested a new therapeutic approach based on clinical information plus the resistance profile of the genotypic or phenotypic resistance assays.

The preliminary results of the parent study have been reported elsewhere. ⁹

For this substudy, only patients enrolled in the phenotype arm and previously exposed (either earlier or at time of resistance evaluation) to ≥1 PIs were considered. A total of 100 PI-experienced HIV-positive patients were included; at the time of testing, all patients were naive to amprenavir and LPV.

The results of phenotyping and the concurrent sequence of the HIV-1 protease region were available for all patients in this substudy. Phenotypic data were expressed as fold resistance and calculated by dividing the 50% inhibitory concentration of LPV for the recombinant virus from the patient by the 50% inhibitory concentration for the wild-type reference strain. According to the manufacturer's instructions, viruses with a >2.5-fold resistance value had a reduced susceptibility to LPV; however, according to the US Food and Drug Administration's label, a 10-fold value was considered as a clinical threshold above which the probability of response to the LPV/r combination is reduced. Therefore, viruses with a fold resistance value between 2.5 and 10 were classified as intermediate resistant, and viruses for which this value was >10 were indicated as resistant to the LPV/r combination.

Genotype data were determined by population sequencing and reported as sequence changes with respect to a consensus B. Mutations at all positions in the protease gene were recorded; mutations were classified as LPV/r resistance-associated mutations according to the International AIDS Society-USA Drug Resistance Mutations Group. ⁷ For purpose of analysis, mixtures of mutant and wild-type viruses were considered mutants, as were viruses with multiple amino acid substitutions including those indicating LPV/r resistance.

Statistical Analysis

To compare susceptible and resistant isolates for categorical variables, the χ² test or Fisher exact test was used as appropriate; for continuous variables, the t test and Mann-Whitney U test were used for normal and nonnormal distributions, respectively. Normality was tested by means of the Levene test, and correlations were analyzed by means of the Spearman coefficient.

LPV fold resistance was log transformed, and univariate and multivariate linear regression models were used to identify resistance-associated variables. Univariate and multivariate logistic regression models were used to identify variables associated with LPV resistance (defined as LPV fold resistance >2.5). Variables included in the model were previous exposure to individual PIs, exposure to individual PIs at the time of susceptibility testing, and individual mutations. The likelihood ratio tests were used to assess the significance of variables included in the models.

RESULTS

There were 66 males and 34 females (mean age, 39.7 years; range, 21-73 years) with a median CD4 cell count of 345.5/μL (range, 7-1010/μL) and a median plasma viral load of 4.27 log₁₀ copies/mL (range, 2.68-5.72 copies/mL). At the time of resistance testing, 40, 32, and 28 patients were classified as Centers for Disease Control and Prevention stages A, B, and C, respectively. ¹⁰

According to the selection criteria, all patients had been exposed to ≥1 PIs (median number, 1; range, 1-3), but all were amprenavir and LPV naive. The results of testing for susceptibility to LPV in the study population are shown in Figure 1. Viral isolates were mostly susceptible to LPV, but quantifiable (>2.5-fold) reduced susceptibility to the drug was found in HIV-1 isolates from 34 of 100 patients. Among these 34 isolates, the mean LPV fold resistance was 17.01 (median, 8.9; range, 2.6-75.9). Unlike LPV-susceptible viral isolates of which a minor proportion (41%) demonstrated resistance to other PIs (mostly 70% to a single PI), all 34 isolates with decreased LPV susceptibility were simultaneously resistant to ≥1 PIs (P < 0.001). In particular, most (68%) of these isolates were resistant to 4 or 5 PIs.

Figure 1

Overall, a median of 9 mutations (range, 0-19) were detected within the protease domain, whereas the median number of LPV/r-associated mutations was 3 (range, 0-8). Figures 2 and 3 show the correlation between LPV fold resistance and the number of total protease and LPV/r-associated mutations, respectively. A Spearman coefficient of 0.62 was found for the correlation of LPV fold resistance with the total number of protease mutations (P < 0.001); the Spearman coefficient was equal to 0.74 when the correlation between LPV fold resistance and the number of LPV/r-associated mutations (P < 0.001) was assessed. Using the existing list of LPV mutations, ⁷ the overall frequency of LPV/r resistance-associated mutations was as follows: L10F/I/R/V, 54%; K20M/R, 11%; L24I, 3%; V32I, 1%; L33F, 4%; M46I/L, 28%; F53L, 2%; I54V/L, 19%; L63P, 67%; A71V/T, 45%; G73S, 9%; V82A/F/T/S, 24%; I84V, 9%; and L90M, 37%. Mutations at codons 47 and 50 were never detected. Moreover, in order of decreasing prevalence, additional mutations not associated with LPV/r resistance were found at positions 77 (V77I, 50%), 36 (M36I, 32%), 93 (I93L, 27%), 30 (D30N, 14%), 88 (N88DS, 14%), and 48 (G48V, 8%).

Figure 2

Figure 3

Previous and current exposure to PIs, the number of mutations detected within the protease domain, and individual mutation distribution among isolates susceptible and resistant to LPV are shown in Table 1. Univariate analysis revealed that current PI therapy was significantly associated with LPV resistance (P = 0.002): at the time of testing, 37 (56.0%) of 66 subjects with susceptible viral isolates and 30 (88.2%) of 34 with reduced susceptibility to the LPV/r combination were receiving a PI-based regimen. Similarly, current indinavir administration was significantly associated with resistance to the LPV/r combination (odds ratio [OR], 7.5; P < 0.001). Other variables associated with LPV resistance were >9 total protease mutations (OR, 27,6; P < 0.001), accumulation of >5 LPV mutations (OR, 39.7; P < 0.001), and presence of the L10FIRV, K20MR, M46IL, I54VL, A71VT, G73SA, V82AFTS, I84V, and M90L mutations (Table 1). Mutations at positions 24, 32, and 53 were never detected in patients with a susceptible viral isolate; however, these mutations were also rare among subjects from whom a virus with reduced susceptibility to LPV was derived. Conversely, the L63P mutation was almost regularly detected independent of the viral isolate phenotype (OR, 2.5; P = 0.25).

Table 1

At multivariate analysis (Table 2), the variables independently associated with LPV resistance were mutations K20MR (OR, 13.9; 95% confidence interval [CI], 1.3-145.1; P = 0.028), I54VL (OR, 131.7; 95% CI, 10.5-1654.7; P < 0.001), G73SA (OR, 19.2; 95% CI, 1.4-273.7; P = 0.029), and I84V (OR, 177.5; 95% CI, 6.0-5232.5; P = 0.003) and the detection of >9 protease mutations (OR, 18.6; 95% CI, 1.6-213.0; P = 0.019).

Table 2

Atypical mutations at positions for LPV resistance were detected in 33 isolates. These atypical mutations involved positions 20, 33, 63, 73, and 82 and were equally distributed among susceptible isolates and those with decreased LPV susceptibility (P = 0.30). However, the I54T/A mutation was detected in only 4 viral isolates with reduced susceptibility to LPV (P = 0.011). Moreover, additional mutations at positions 30, 36, 48, 77, 88, and 93 did not seem to influence susceptibility to the drug (P = 0.50). Finally, polymorphisms and natural mutations principally involved the positions 13,15, 35, 37, 41, 62, 64, 69, and 72.

Overall, of the 34 viral isolates with reduced susceptibility to LPV, 18 (52.9%) were classified as intermediate resistant to the LPV/r combination (median fold resistance, 4.05; interquartile range, 3-6.4), and 16 (47.0%) were classified as resistant (fold resistance value, >10) (median fold resistance, 22.95; interquartile range, 16.1-41.7). PI exposure did not influence the level of resistance to LPV/r (P = 0.96). Similarly, neither the number of total mutations within the protease domain nor the number of LPV mutations or the number of mutations associated with resistance to other PIs appeared to influence the level of resistance (P = 0.17, P = 0.11, and P = 0.96, respectively). Figure 4 demonstrates the distribution of LPV mutations between the viral isolates with >10-fold or <10-fold resistance to LPV. Only the presence of the 82AFTS mutation seemed to significantly impact the level of resistance to the LPV/r combination (P = 0.001). In addition, among mutations associated with resistance to other PIs, the G48V mutation increased the resistance to LPV/r (P = 0.03), whereas, even if not statistically significant, the 30N and 88DS mutations were detected only in viral isolates with <10-fold resistance to LPV. At linear regression analysis, each additional LPV mutation increased the LPV resistance by a factor of 1.3 (95% CI, 1.2-1.4), and the presence of mutation I54VL increased LPV fold resistance by 4.2 (95% CI, 2.7-6.4). These two variables alone could account for a remarkable portion of the LPV fold resistance (adjusted R² = 0.70).

Figure 4

DISCUSSION

A great deal of data from both retrospective ^11,12 and prospective studies ^13-16 has demonstrated the utility of baseline resistance testing to foresee virologic response to treatment in drug-experienced patients. Although phenotypic tests are not broadly performed because they are limited to certain centralized laboratories, genotype analysis is now within the clinicians' reach, but the appropriate interpretation of mutational patterns is problematic.

In this study, we evaluated the protease sequences of 100 HIV-1 isolates with ascertained phenotypic susceptibility to LPV. Due to the high plasma LPV trough levels, ¹ the clinical threshold above which the probability of response to the LPV/r combination decreases is usually greater than the biologic threshold. ¹⁷ Therefore, two different cutoff values (2.5-fold and 10-fold) were considered for identifying HIV isolates with reduced susceptibility to LPV/r. Overall, decreased susceptibility (>2.5-fold) to LPV was observed in 34% of HIV isolates, while 16% displayed >10-fold reduced LPV susceptibility that could extend up to 75.9-fold resistance. All 34 isolates with reduced susceptibility to LPV were also resistant to other PIs.

Viral isolates were derived from PI-experienced subjects who were LPV naive. Naturally, the intent of this analysis was to contribute to the early identification of those individuals who will not benefit from shifting to the LPV/r combination that currently constitutes the backbone of most salvage regimens and to try to define the debated allocation of LPV within the highly active antiretroviral regimen. In fact, data have been presented that demonstrate a sustained virologic response in antiretroviral-naive patients, ^18,19 and its use as salvage therapy has also been reported in numerous other studies. ^20-22 Actually, there are reasons for and against each application. As reduced susceptibility to LPV results from the accumulation of mutations associated with resistance to other PIs, ⁷ it is reasonable to administer LPV to patients with little or no PI experience. On the other hand, similar to other PIs, the use of LPV is not without complications because it may elevate triglyceride and cholesterol levels, ^21,23 thus subjecting patients to therapy that in the long run could provoke serious damage.

In our study, current PI therapy and administration of indinavir were both associated with reduced susceptibility to LPV (P = 0.002 and P < 0.001, respectively). As multiple protease mutations are needed to confer at least 10-fold resistance to LPV, ²⁴ this result was not unexpected; resistance to indinavir generally requires the stepwise accumulation of multiple mutations ²⁵ that emerge under the selective pressure of the drug. Moreover, in univariate analysis, the presence of the L10FIRV, K20MR, M46IL, I54VL, A71VT, G73SA, V82AFTS, I84V, and M90L mutations was associated with LPV resistance. Our study could not verify whether the presence of these mutations would actually impair the response to the LPV/r combination in vivo; however, some of these same mutations have been detected more frequently in patients with virologic failure of LPV/r than in those who responded to treatment. ^22,26 Moreover, 4 of these mutations (K20MR, I54VL, G73SA, and I84V) were also confirmed to be independently associated with LPV resistance by multivariate analysis; thus, we can reasonably attribute a concrete responsibility for LPV phenotypic resistance to these mutations. Our results also confirmed previous observations ^4,22,26 claiming that the number of LPV mutations is the most relevant variable in determining decreased LPV susceptibility. In our setting, the presence of >5 LPV mutations was significantly associated with reduced LPV susceptibility (OR, 39.7; 95% CI, 11.8-132.7; P < 0.001); this figure corresponds to that of Kempf et al. ⁴ who observed the highest response to LPV-based regimens in patients whose virus isolate showed an LPV mutation score of ≤5.

Perhaps the most interesting findings to emerge from this study were that the linear regression analysis demonstrated that each additional LPV mutation increased the LPV resistance by 1.3-fold and that the presence of the I54VL mutation increased LPV fold resistance by 4.2. These two variables alone might be held responsible for the high levels of LPV resistance. We are aware that these results refer to a limited population sample; however, the coefficient 1.3 emerging from our data is similar to the finding of Kempf et al. ² who reported that the 50% inhibitory concentration of LPV increased by an average of 1.74-fold per mutation in isolates with ≥3 mutations. If we multiply these coefficients by the number of LPV mutations identified in isolates from PI-experienced patients, the cutoff figure for LPV clinical resistance is reached. It is foreseen that the relative contribution of each single mutation to LPV resistance will be different ²; in our setting, mutation 54, which has already been implicated in the virologic failure of LPV-including regimens, ^4,22 increased the 50% inhibitory concentration of LPV by 4.2-fold.

Finally, among mutations associated with resistance to other PIs, the G48V mutation seemed to further reduce LPV susceptibility, and the 30N and 88DS mutations, although not statistically significant, were observed in viral isolates with <10-fold resistance to LPV. Masquelier et al. ²² also reported a better virologic response in subjects treated with LPV/r whose virus isolates harbored the 30N and 88D mutations; the impairment of viral replication capacity induced by these mutations ^27,28 most likely accounts for these observations.

Our results coincide with those of other studies investigating LPV resistance; however, a particular pattern of mutations representative of LPV resistance has yet to be identified. The existing list of LPV mutations might depict a simple census of major and minor mutations in the protease that contribute to all PI resistance; in fact, all 34 isolates in our study with decreased susceptibility to LPV were simultaneously resistant to other PIs. The only consensus reached thus far is that the higher the number of mutations, the greater the resistance.

In conclusion, in PI-experienced patients, such as those in our study who were prospective candidates for salvage therapy, the decision regarding the use of LPV depends on the number of baseline mutations and also on the presence of mutation 54. Therefore, on the basis of our results, the use of LPV/r might be reserved for those patients with <5 mutations and consequently with slight PI experience-before the accumulation of mutations induced by other PIs that would impair its efficacy.

Acknowledgments

The authors thank all the participants in the PhenGen study. They also thank Ms. Paulene Butts for the helpful review of the manuscript and Ms. Antonella Lagioia, Mr. Antonio Spinelli, and Ms. Eliana Cinori for technical assistance.

APPENDIX

The following individuals were participants in the PhenGen study: G. Pastore, N. Ladisa, and P. Maggi (Bari); G. Carosi, F. Castelli, C. Torti, and L. Tomasoni (Brescia); A. Mandas, F. Pigliaru, and S. Manca (Cagliari); G. Angioni, S. Angioni, and G. Abeltino (Cagliari); P. Bellissima and S. Bonfante (Caltagirone); F. Ghinelli and L. Sighinolfi (Ferrara); F. Mazzotta, S. Lo Caputo, and P. Pierotti (Firenze); A. Gallo and L. Saracino (Frosinone); M. Toti, T. Carli, and E. Donati (Grosseto); A.M. Orani and P. Perini (Lecco); A. Scasso and M. De Gennaro (Lucca); G. Todaro and G. Magaraci (Messina); M. Moroni, A. D'Arminio Monforte, and T. Bini (Milano); A. Cargnel, C. Atzori, and V. Micheli (Milano); A. Lazzarin and N. Gianotti (Milano); A. Chirianni and M. Gargiulo (Napoli); N. Abrescia and M. D'Abbraccio (Napoli); C. Izzo and T. Pizzella (Napoli); G. Marani Toro and G. Parruti (Pescara); D. Dionisio and A. Vivarelli (Pistoia); A. Antinori and G. Liuzzi (Roma); N. Pasquale and V. Tozzi (Roma); F. Resta and G. Buccoliero (Taranto); G. Di Perri, A. Sinicco, S. Bonora, and S. Audagnotto (Torino); F. Branz and P. Delle Foglie (Trento); P. Grossi and C. Basilico (Varese); and A. Poggio and V. Mondino (Verbania)