Search tips
Search criteria 


Logo of jvirolPermissionsJournals.ASM.orgJournalJV ArticleJournal InfoAuthorsReviewers
J Virol. 2005 March; 79(6): 3329–3338.
PMCID: PMC1075714

Selection of Resistance in Protease Inhibitor-Experienced, Human Immunodeficiency Virus Type 1-Infected Subjects Failing Lopinavir- and Ritonavir-Based Therapy: Mutation Patterns and Baseline Correlates


The selection of in vivo resistance to lopinavir was characterized by analyzing the longitudinal isolates from 54 protease inhibitor-experienced subjects who either experienced incomplete virologic response or viral rebound subsequent to initial response while on treatment with lopinavir-ritonavir in Phase II and III studies. The evolution of incremental resistance to lopinavir (emergence of new mutation[s] and/or at least a twofold increase in phenotypic resistance compared to baseline isolates) was highly dependent on the baseline phenotype and genotype. Among the subjects demonstrating evolution of lopinavir resistance, mutations at positions 82, 54, and 46 in human immunodeficiency virus protease emerged frequently, suggesting that these mutations are important for conferring high-level resistance. Less common mutations, such as L33F, I50V, and V32I together with I47V/A, were also selected; however, new mutations at positions 84, 90, and 71 were not observed. The emergence of incremental resistance contrasts greatly with the low incidence of resistance observed after initiating lopinavir-ritonavir therapy in antiretroviral-naive patients, suggesting that partial resistance accumulated during prior protease inhibitor therapy can compromise the genetic barrier to resistance to lopinavir-ritonavir. The emergence of incremental resistance was uncommon in subjects whose baseline isolates contained eight or more mutations associated with lopinavir resistance and/or displayed >60-fold-reduced susceptibility to lopinavir, providing insight into suitable upper genotypic and phenotypic breakpoints for lopinavir-ritonavir.

Virologic rebound during antiretroviral (ARV) therapy is often associated with the emergence of drug resistance, which can compromise the efficacy of ongoing and future therapy. Resistance to human immunodeficiency virus (HIV) protease inhibitors (PIs) occurs via the accumulation of primary and secondary mutations that are, respectively, located within and outside of the enzyme active site, sometimes accompanied by mutations in one or more gag cleavage sites (6, 18, 24, 28, 38). The patterns of mutations selected by most currently available PIs have been characterized (6, 19, 24, 28, 32). In general, individual viral mutations produce relatively modest changes in phenotypic susceptibility to the members of the PI class. However, more than 20% of the 99 amino acids comprising the HIV protease homodimer have been shown to mutate in response to drug pressure (15, 27). Consequently, the genotypic patterns in patients with PI-resistant HIV are highly complex. The primary mutation(s) selected by different PIs may be distinct; however, secondary mutations tend to be common to the PI class, potentially limiting the success of subsequent PI therapy following failure on any PI-containing regimen (6, 19, 24, 28, 32).

The development of in vivo resistance to PIs traditionally has been characterized by examining the selection of mutations during drug failure in previously ARV-naive patients (6, 18, 28, 32). However, high rates of virologic suppression achieved in this population with potent combination regimens, combined with the potential for a low incidence of resistance to the PI upon virologic failure, may limit this approach (23). The genotypic correlates of reduced susceptibility to PIs have been characterized by using statistical analysis of large databases (22, 31, 40). Furthermore, the examination of the virologic response to PI-containing regimens with respect to baseline phenotype and/or genotype can provide information on the mutation patterns that produce clinical resistance (or cross-resistance) and on the appropriate clinical breakpoints above which partial (lower breakpoint) or nearly complete (upper breakpoint) loss of drug activity is observed (10, 11, 21). However, the selection of incremental resistance upon failure of PI-experienced patients to a subsequent PI-containing regimen has not been widely characterized. Examination of the effect of baseline genotype and phenotype upon the development of incremental resistance in PI-experienced patients, compared to ARV-naive patients, may help to provide greater understanding of the conditions producing selective pressure in vivo.

Lopinavir-ritonavir (LPV-RTV) is a pharmacokinetically boosted PI regimen that produces potent and durable suppression of plasma HIV RNA in both PI-naive and PI-experienced patients (2, 29, 39). HIV strains resistant to LPV have been selected by using in vitro passage (5, 26). Furthermore, several studies of the effect of genotype on in vitro LPV susceptibility or response to LPV-RTV have produced LPV-RTV algorithms that are predictive of reduced response in PI-experienced patients (22, 31, 40). However, to date, the development of resistance to LPV has not been observed among ARV-naive patients treated with an LPV-RTV-based combination regimen for up to 4 years in clinical trials (14, 23), and only recently have isolated anecdotes of primary resistance in previously untreated patients been reported (7, 9). This apparently high barrier to resistance is consistent with the observation that the mean plasma LPV trough levels exceed the serum-adjusted 50% inhibitory concentration (IC50) of LPV for wild-type HIV substantially and suggests that considerably reduced susceptibility to LPV is required for a compromised virologic response (3, 16). In multiple PI- and nucleoside reverse transcriptase inhibitor (NRTI)-experienced, nonnucleoside reverse transcriptase inhibitor (NNRTI)-naive patients, maximal virologic response to treatment with LPV-RTV plus efavirenz (EFV) and NRTIs was observed in patients with baseline viruses containing up to five protease mutations associated with LPV resistance and/or displaying up to 10-fold reduced susceptibility to LPV (lower clinical breakpoint). Although virologic response rates differed between patients with baseline viral isolates displaying <40-fold and >40-fold reduced susceptibility to LPV, the ability to define an upper breakpoint for LPV-RTV activity in that study was limited by the relatively small number of patients with high-level baseline resistance and was obscured by the concomitant activity of EFV, which produced an initial response in every subject (21). In the absence of any development of LPV resistance in PI-naive patients, we have characterized the genotype and phenotype of longitudinal rebound isolates from 54 PI-experienced patients with plasma HIV RNA levels of >400 copies/ml while on treatment with LPV-RTV in two Phase II studies and one Phase III study. These results provide information on the patterns of mutations leading to LPV-RTV resistance and illustrate a novel approach to the estimation of an upper clinical breakpoint for this boosted PI regimen.


Viral isolates.

Samples were collected from subjects enrolled in two Phase II studies (M97-765, n = 70; M98-957, n = 57) (2, 21) and one Phase III study (M98-888, n = 148). All subjects were NNRTI naive but either single or multiple PI experienced prior to entering these studies. During the studies, subjects were treated with LPV-RTV in combination with either nevirapine (NVP) (single PI-experienced patients from M97-765 and M98-888) or EFV (multiple PI-experienced patients from M98-957) and two NRTIs (at least one NRTI was new).

Determination of phenotype and genotype.

Samples from subjects who had at least two consecutive plasma HIV RNA values of >400 copies/ml without a documented treatment interruption of LPV-RTV during therapy were retrospectively selected for resistance testing. Baseline samples were also analyzed for each of the above subjects. Phenotypic susceptibility to commercially available ARVs was determined by using either the PhenoSense HIV assay (ViroLogic, Inc.) or the Antivirogram method (version 3.0; Virco Inc.) (12, 33). Viral RNA containing the protease (PR) and reverse transcriptase (RT) gene sequences from each subject was isolated and amplified by PCR and incorporated into recombinant viruses for phenotypic susceptibility testing. Phenotypic data were expressed as the factor of change in the IC50, which was calculated by dividing the mean IC50 against each viral isolate by the corresponding IC50 against the wild type (WT) virus. Genotype was also determined at ViroLogic, Inc. or Virco, Inc. by population sequencing with an ABI automated sequencer and is reported as sequence changes with respect to the sequence of the pNL4-3 laboratory strain. Both phenotype and genotype assays used in this study assess only the impact of the PR and first 300 amino acids of the RT gene as well as the p6/p1 and p1/p7 gag cleavage sites. The effect of the alterations in the genes beyond these regions was not studied.

Definition of evolution of LPV resistance and NNRTI resistance.

Evolution of incremental LPV resistance was defined as displaying >2.5-fold reduced susceptibility to LPV in the rebound sample compared to the WT virus as well as satisfying one or both of the following criteria: (i) emergence of a new primary mutation in the PR gene associated with PI resistance [D30N, V32I, G48V, I50V, V82(A/F/T/S), I84V, and L90 M]; (ii) emergence of a new secondary mutation [L10(F/I/R/V), K20(M/R), L24I, L33F, M36I, M46(I/L), I47(A/V), I54(A/V/L/S), A71(V/T), G73(S/A), V77I, and N88D] accompanied by an increase greater than or equal to twofold in the LPV IC50 between baseline (pre-LPV-RTV treatment) and rebound. The time from the nadir HIV RNA level until genotype was summarized for patients who did or did not exhibit evolution of LPV resistance. For patients with evolution of LPV resistance, the first genotype demonstrating resistance was used, and for patients without evidence of resistance, the latest available genotype was used. The mean time from the nadir HIV RNA level was compared between groups by a one-way analysis of variance. Selection of additional NNRTI resistance was defined as emergence of the A98G, L100I, K101(/E/Q/P), K103(N/S), V106(A/M), V108I, Y181C, Y188(L/C/H), G190(A/E/S), and/or M230L mutations in the RT gene or at least a 2.5-fold increase in IC50 to NVP or EFV between baseline and rebound. Mixtures of mutant and WT amino acids were scored as mutant.

Analysis of genotypic and phenotypic predictors of evolution of LPV resistance.

The effect of the baseline genotype (number of LPV-associated mutations) and phenotype on the selection of additional LPV resistance was assessed by logistic regression analysis. For the purposes of this analysis, LPV-associated mutations were defined as the 11 mutations (LPV mutation score) previously identified as associated with reduced in vitro LPV susceptibility (22) plus the major mutations reported by others to be associated with LPV resistance (5, 17, 26, 31, 40). This combined list includes L10(F/I/R/V), K20(M/R), L24I, V32I, M46(I/L), I47(A/V), G48V, I50V, F53L, I54(L/S/T/V), L63P, A71(I/L/V/T), V82(A/F/T), I84V, and L90M.


Baseline viral isolates.

To characterize the patterns of HIV evolution in response to LPV-RTV in vivo, we analyzed the genotype (PR and RT genes) and phenotypic susceptibility of longitudinal plasma viral isolates from 54 PI-experienced (41 single PI-experienced and 13 multiple PI-experienced) subjects who had plasma HIV RNA levels of >400 copies/ml during LPV-RTV-based therapy in three Phase II and III studies (Table (Table1).1). Both baseline and rebound genotypes were available from all 54 subjects. Phenotypic results were available from 54 subjects at rebound but from only 45 subjects (35 for amprenavir [APV]) at baseline. Baseline isolates from the other nine subjects were not submitted for phenotypic testing because the rebound isolates from each were sensitive to LPV (IC50 of less than onefold change, compared to WT levels) and the baseline genotypes revealed no evidence of LPV resistance. Most (69%, or 31 of 45) of the baseline isolates displayed ≥2.5-fold reduced susceptibility to two or more PIs, and 42% (19 of 45) displayed ≥10-fold decreased susceptibility to four or more of the six PIs tested (Table (Table2).2). The median factor of change in LPV IC50 was 6 (range, 0.4 to 167) compared to WT HIV, and 19 of 45 isolates demonstrated ≥10-fold reduced susceptibility to LPV. These 19 isolates were highly cross-resistant to RTV, nelfinavir (NFV), and indinavir (IDV) (92-, 50- and 26-fold, respectively, median change in IC50) but displayed lower cross-resistance to both saquinavir (SQV) (median, sevenfold change in susceptibility) and APV (median, threefold change in susceptibility). Consistent with the phenotypic results, most of the 54 baseline isolates contained multiple mutations associated with PI resistance (15). The median number of mutations associated with reduced susceptibility to LPV (LPV mutation score) was four (range, zero to nine). Only five baseline isolates (9%) had no PI mutation. The most commonly observed mutations were at positions 63, 10, 82, 71, 54, 46, 36, 90, and 77, in order of decreasing frequency from 89 to 20%. Isolates from 36 of 54 subjects (67%) had one or more primary mutations at positions 82, 84, 90, and 32 together with position 47. Since the subjects were NNRTI naive, the prevalence of NNRTI resistance was low (6 of 54). Three of these isolates displayed the K103N RT mutation with or without V106A and demonstrated at least 20-fold reduced susceptibility to NVP and EFV. Baseline isolates from two of six subjects exhibited moderately decreased susceptibility (9- and 12-fold) to NVP and EFV but no major NNRTI mutations. In one subject, a mixture at positions 103(K/N) and 181(Y/C) was detected without an observable change in susceptibility.

Clinical studies of PI-experienced patients for the analysis of incremental resistance development
Susceptibility of baseline isolates to protease inhibitors

Selection of incremental LPV resistance and NNRTI resistance.

Among these 54 subjects, the selection of resistance to LPV-RTV and NNRTIs differed significantly (Fig. (Fig.1).1). Selection of NNRTI resistance was observed in 45 of 49 (92%) subjects who remained on NNRTI therapy and had baseline isolates with IC50s <10-fold above WT. In contrast, only 19 of 54 (35%) subjects, including 14 of 41 (31%) single PI-experienced subjects and 5 of 13 (39%) multiple PI-experienced subjects demonstrated the selection of incremental LPV resistance. No significant difference was observed in the time from the nadir HIV RNA level until the time of genotype among patients with evolution of LPV resistance (32 weeks) compared to patients without evolution of LPV resistance (29 weeks; P = 0.68). All 19 subjects with evolution of incremental LPV resistance either developed or retained NNRTI resistance at rebound or discontinued the NNRTI due to adverse events.

FIG. 1.
Disposition of PI-experienced subjects failing subsequent therapy with LPV-RTV.

Genotypic analysis of HIV selected by LPV-RTV in PI-experienced patients.

The changes between the genotypes of the baseline and latest available rebound isolates from the 19 subjects demonstrating evolution of LPV resistance, relative to the WT pNL4-3 strain, are shown in Fig. Fig.22 and Table Table3.3. All 19 baseline isolates had at least four mutations associated with resistance to the PI class (15), including at least one primary mutation at the following positions: 32 (together with position 47), 82, 84, and/or 90. The most common baseline mutations associated with PI resistance were at positions 63 (18 subjects), 10 (17 subjects), 82 (13 subjects), 71 (13 subjects), and 54 (11 subjects). Mutations at positions 36, 46, 77, and 90 were present in seven to nine subjects. Following rebound, one to six new mutations associated with PI resistance (15) were observed. The median number of PI mutations after rebound was 8 (range, 6 to 12). The most common mutations to emerge at rebound included M46I/L [emerged in 10 of 13 subjects, or 73%, without M46I/L at baseline], I54V (6 or 8 subjects, or 75%) and V82A (4 of 6 subjects, or 67%), followed by L33F (6 of 18 subjects, or 33%) and K20R/M/T (4 of 16, or 25%). The I50V mutation emerged in two subjects with prolonged periods of detectable viral load (>9 months). In addition, a number of other PI mutations, including L10F/I/V, L24I, V32I, M36I, I47V/A, F53L, and G73S, also emerged in one to three subjects at rebound. Furthermore, changes from one amino acid to another at positions 88 (N88S to N88G) and 47 (I47V to I47A) were observed in one subject each. In contrast, no additional new mutations were observed at positions 63, 71, 84, and 90.

FIG. 2.
Mutations present in baseline or rebound isolates among 19 patients selecting additional LPV resistance. White bars, mutations at given amino acid positions that appeared in the baseline samples; black bars, mutations that emerged during LPV-RTV therapy. ...
Phenotypic and genotypic characterization of isolates from 19 subjects

Statistical analyses of large databases have identified noncanonical mutations in HIV PR associated with reduced susceptibility to LPV (31, 40). A minority of the rebound isolates from subjects demonstrating evolution of incremental LPV resistance (5 of 19) also contained one or more of these mutations that were not present at baseline, including K43T and L89I (one subject each) and K55R and I72V (two subjects each). In contrast, none of the 35 rebound isolates from subjects in which incremental LPV resistance was not observed contained any of the noncanonical mutations not present at baseline, even though one or more were frequently present at baseline (data not shown).

Phenotypic analysis of isolates with evolution of incremental LPV resistance.

Phenotypic analysis was performed on 51 longitudinal samples from the above 19 subjects with evolution of incremental LPV resistance. Figure Figure33 shows the median and interquartile range (IQR) of the factors of increase in the IC50s of the isolates from baseline and the last available visit. For these 19 subjects, most baseline isolates demonstrated moderately reduced susceptibility to LPV with a median LPV IC50 increase of 6.2-fold compared to WT (IQR, 2.7- to 20-fold). Following rebound, the LPV susceptibility of the isolates decreased longitudinally in all subjects. By analogy to previous findings with other PIs in ARV-naive patients (6, 19, 24, 28, 32), this incremental loss in susceptibility correlated with the stepwise accumulation of new mutations (Table (Table3).3). The rebound isolates from the last available study visit displayed a median of a 4.3-fold increase in IC50 compared to baseline (from 6.2- to 43-fold, compared to WT) (Fig. (Fig.33).

FIG. 3.
Median and IQR of changes (n-fold) in susceptibility to LPV and five other PIs of isolates from baseline and last study visit among 19 subjects with evolution of LPV resistance. B, baseline sample; R, last available rebound sample.

The median (IQR) susceptibilities of the above patient isolates to other PIs are shown in Fig. Fig.3.3. Following virologic rebound on LPV-RTV therapy, the rebound viruses either retained (if cross-resistant at baseline) or developed significant cross-resistance to RTV, NFV, and IDV. In contrast, a lower degree of cross-resistance to APV and SQV was observed. The median (IQR) increases in susceptibility to APV and SQV at the last available visit among all 19 subjects were 5.4-fold (2.4- to 21-fold) and 3.9-fold (1.2- to 63-fold), respectively, compared to WT. In the subset of 11 subjects whose rebound isolates did not contain I50V, I84V or I47V/A in combination with V32I, the rebound isolates displayed either full or slightly reduced susceptibility (<6.3-fold) to APV. Moreover, among another subset of 12 subjects not previously exposed to SQV, no significant loss in SQV susceptibility was observed between the baseline and rebound (median [IQR] increase in SQV susceptibility, from 1.9-fold [0.7- to 3.3-fold] at baseline and 1.4-fold [0.7- to 3.2-fold] at rebound). Notably, the evolution of I47V to I47A in conjunction with the appearance of a V82V/I mixture in a background of other PI mutations (Table (Table3,3, subject 6) decreased LPV susceptibility from 4.5-fold to 209-fold, compared to WT, but had no effect on the susceptibility to SQV (baseline IC50, 0.5-fold increase; rebound, 0.7-fold increase).

Characterization of NNRTI resistance.

High-level NNRTI phenotypic resistance was observed in the 45 subjects without baseline resistance who remained on NNRTI therapy at rebound (34 NVP-treated and 11 EFV-treated subjects). Compared to WT, the median changes in susceptibility of the rebound isolates to NVP and EFV were 89-fold and 148-fold, respectively, among subjects treated with each respective drug. Each of the 45 subjects developed one or more mutations associated with NNRTI resistance (K103N, Y181C, Y188C, G190A, V106A, and K101E). Among those mutations, K103N was the most common in both NVP- and EFV-treated subjects.

Genotypic and phenotypic predictors of incremental LPV resistance.

In order to investigate the factors predicting the selection of LPV resistance, we analyzed the likelihood of emergence of incremental LPV resistance during virologic failure with respect to baseline genotype and phenotype (Fig. (Fig.4).4). The probability of incremental LPV resistance emergence was highest (11 of 13 subjects, 85%) in subjects with four to six baseline LPV-associated mutations (see Materials and Methods). In contrast, the likelihood was low if less than two (0 of 14 subjects, 0%), two to three (1 of 9 subjects, 11%), or more than seven (1 of 4 subjects, 25%) LPV-associated mutations were present at baseline. The incidence of incremental LPV resistance was intermediate in subjects with six to seven (6 of 14 subjects, 43%) baseline LPV-associated mutations. In addition, the probability of selection of LPV resistance differed significantly among the subjects whose baseline viruses contained (19 of 39 subjects) versus those whose baseline lacked (0 of 15 subjects) at least one primary PI mutation at the following positions: 32 (together with position 47), 82, 84, and/or 90 (P < 0.001). Finally, among the subjects with at least one primary mutation, a mutation at position 32 was statistically significantly associated with emergence of incremental LPV resistance: 4 of 4 subjects with a V32I mutation developed incremental LPV resistance, versus 15 of 35 subjects without a V32I mutation (P = 0.047). No other mutation was significantly associated with selection of LPV resistance in this data set.

FIG. 4.
Proportion of patients with rebound selecting LPV resistance with respect to the number of baseline LPV mutations (a) and baseline LPV susceptibility (b).

Similarly, analysis of the relationship of baseline phenotype with the selection of LPV resistance indicated that the probability of selection of incremental LPV resistance was highest in subjects with an intermediate level of reduced LPV susceptibility at baseline; for example, the predicted probability of incremental resistance was above 65% for viruses with 4.2- to 22-fold reduced LPV susceptibility (Fig. (Fig.5).5). A substantial drop in selective pressure during rebound on LPV-RTV therapy was observed in subjects with >60-fold reduced baseline LPV susceptibility (Fig. (Fig.4b4b and and5).5). Logistic regression analysis indicated that the probabilities (95% confidence interval [CI]) of incremental selection of LPV resistance in subjects with 40-, 60-, and 80-fold baseline LPV IC50s were 46% (95% CI, 25 to 72%), 31% (95% CI, 11 to 63%), and 20% (95% CI, 5 to 56%), respectively (Fig. (Fig.5).5). Among subjects with more than four baseline PI mutations, incrementally resistant HIV emerged in 13 of 19, 2 of 4, and 1 of 6 subjects with <40-fold, 40- to 60-fold, and >60-fold reduced LPV susceptibility, respectively, at baseline.

FIG. 5.
Logistic regression model of predicted probability (and 95% CI) of incremental LPV resistance by baseline LPV phenotypic susceptibility among the 54 patients with viral rebound.


In this study, we characterized the development of resistance in single or multiple PI-experienced patients with viral rebound while they received LPV-RTV along with NVP or EFV plus NRTIs. These subjects had widely ranging levels of reduced PI susceptibility at baseline (prior to LPV-RTV) and evolution of additional PI resistance was observed in 19 of 54 (35%) subjects. The time from the nadir HIV RNA level until genotype was similar for patients with (32 weeks) or without (29 weeks) evolution of LPV resistance, suggesting that the results were not due to differences in exposure to viral replication. Concomitantly, NNRTI resistance was commonly observed (>90% of subjects) during rebound. Previously, we have observed that no resistance to LPV developed in ARV-naive patients receiving LPV-RTV plus stavudine-lamivudine for up to 4 years (14, 23). The present results contrast greatly with those findings, suggesting that the barriers to resistance development to LPV-RTV present in ARV-naive patients are substantially compromised in PI-experienced patients with preexisting PR mutations from prior virologic failure. This contention is further supported by analysis of the incidence of evolution with respect to baseline genotype. Thus, none of the subjects with zero to one baseline PR mutations, whose viral isolates resemble WT virus in the PR gene, experienced evolution of LPV resistance. In contrast, as the number of baseline mutations increased to four or more, the genetic barrier to resistance decreased, and the probability of evolution increased substantially. The differences in the overall probability of PI-resistance development in ARV-naive or PI-experienced patients treated with LPV-RTV may have implications for overall therapeutic strategies wherein the avoidance of resistance is a priority.

The baseline (pre-LPV-RTV) sequences of the viruses from these PI-experienced subjects were highly heterogeneous. Consequently, the addition of a wide variety of new PI mutations upon rebound is unsurprising. Mutations at positions 82, 54, and 46 emerged frequently and were present at a high level of prevalence after rebound, suggesting that these mutations are important for conferring high-level resistance. Each of these mutations has previously been associated with reduced susceptibility to LPV (22). In contrast, new mutations at positions 84, 90, and 71 were not observed at rebound, despite being present at relatively high frequency at baseline. The lack of selection of these mutations is consistent with the absence of increased cross-resistance to SQV between baseline and rebound. Several mutations of relatively low prevalence in PI-experienced patients also appeared upon rebound from LPV-RTV therapy. The L33F mutation, which emerged in seven subjects, is associated with resistance to the PI class (1, 28). The I50V mutation, commonly observed during failure on APV therapy (24), has previously been shown to be selected by and to produce resistance to LPV in vitro (26, 31, 34). Its appearance in the rebound isolates from two subjects indicates that LPV-RTV can also select I50V; however, in both subjects, the mutation appeared only after prolonged periods of detectable viral load.

The association of a baseline V32I mutation with the emergence of LPV resistance and the appearance of the unusual I47A mutation upon rebound are both notable. Two viruses out of a data set of >1,900 mutant isolates that contained V32I and I47A along with M46I displayed much greater than predicted in vitro resistance to LPV (31). In vitro selection with LPV produced passaged viruses containing V32I and I47V, which evolved to V32V and I47A, accompanied by changes in both p1/p6 and p7/p1 gag cleavage site sequences (5). The incidence of I47A among samples being submitted for genotypic HIV resistance testing increased concomitantly with the wide availability of LPV-RTV (20). Finally, a strain with V32I and I47A emerged in a previously therapy-naive patient receiving LPV-RTV monotherapy (9). Taken together, these results suggest that the V32I and I47A mutations may constitute a significant pathway for LPV resistance. Residues 32 and 47 occupy adjacent positions in the symmetry-related S2 and S2′ subsites of the HIV PR active site. The P2 isopropyl group and the terminal dimethylphenoxy moiety of LPV are in van der Waals contact with the side chains of isoleucine 47 and 47′ (37). Evolution of I47V to I47A creates a small but significant amount of unoccupied volume, decreasing the van der Waals interaction and lowering the affinity of the drug for the PR active site (5). Further work will be required to provide an understanding of the maintenance of activity of SQV, which also occupies the S2 and S2′ subsites, against the mutant with I47A together with V32I.

The present results provide insight into exploratory algorithms for LPV resistance, shown in Table Table4,4, which are based either on the correlation of genotype and phenotype in the isolates from PI-experienced patients (22, 31, 40) or the correlation of baseline genotype with virologic response in PI-experienced patients (17). In general, these algorithms have been shown to be predictive of the virologic response to LPV-RTV therapy in PI-experienced patients (8, 17, 21, 25, 35). All of the canonical PI mutations (15) emerging upon rebound in PI-experienced subjects in this study are contained in at least one algorithm, including L10I, K20M/R, L24I, L33F, M46I/L, G48V, I54V, and V82A, which appear in three or more of the four algorithms (Table (Table4).4). No PI mutation emerged that has not previously been associated with resistance to LPV-RTV. The emergence of several noncanonical mutations correlating with reduced susceptibility to LPV in algorithms based on large databases (31, 40) and recently demonstrated to be of higher prevalence in isolates from PI-experienced patients (41) suggests that these mutations emerged due to selective drug pressure. Deciphering the mechanism by which these ancillary mutations may contribute to resistance requires further study.

Algorithms for describing resistance to LPV-RTVa

In general, increased phenotypic resistance to LPV upon rebound also produced increased resistance to RTV, IDV, and NFV. Cross-resistance to these PIs is consistent with the relatively high correlation between the LPV phenotype and the susceptibility to these agents in isolates from a panel of PI-experienced patients (22). Cross-resistance to APV was noted in a portion of the rebound isolates, most notably among isolates containing the I50V and/or L33F mutations. However, none of the isolates from patients naive to SQV developed significant resistance to SQV, suggesting that RTV-boosted SQV, as well as boosted APV regimens, may be useful for salvage of treatment failure on LPV-RTV therapy. Notably, the accumulation of the I47A mutation did not appear to substantially increase cross-resistance, and the rebound isolates remained completely susceptible to SQV (data not shown), in accord with observations by Friend et al. (9).

The incidence of incremental resistance to LPV with respect to baseline genotype and phenotype provides insight into defining an upper clinical breakpoint for LPV-RTV. Selection was most common in subjects whose baseline isolates displayed four or more mutations (enough to overcome the genetic barrier) but <40-fold reduced susceptibility to LPV. If baseline susceptibility was reduced by >60-fold, the emergence of additional resistance was uncommon. Taken together, these results indicate that selective pressure in viruses with >60-fold reduced susceptibility is usually insufficient to require the emergence of additional mutations during failure, suggesting that LPV-RTV was only marginally active in these individuals. All patients were NNRTI naive and therefore experienced an initial virologic response prior to rebound. However, patients with higher baseline LPV resistance (>60-fold) generally experienced rapid viral rebound, suggesting little or no antiviral activity of LPV in these patients, while patients with lower baseline LPV resistance generally demonstrated a longer duration of viral suppression and a more gradual increase in viral load. Previous studies have suggested that the response to LPV-RTV is relatively unaffected in patients with up to a 10-fold change in baseline susceptibility or up to five mutations associated with reduced susceptibility to LPV (21). In that study, response was further attenuated in the small number of subjects with >40-fold reduced baseline susceptibility. The present results are consistent with those findings and suggest that suitable lower (initial point at which lowered response relative to WT virus is expected) and upper (point beyond which incremental drug activity is unlikely) phenotypic breakpoints for LPV-RTV are 10- and 60-fold, respectively. Suitable genotypic breakpoints, based on the number of LPV-associated mutations, appear to be between five and six mutations and between seven and eight mutations, respectively. The phenotypic upper breakpoint is consistent with a pharmacological model for PI activity in vivo that predicts that the average inhibitory quotient (ratio of trough concentration of drug to human serum-adjusted IC50) for LPV-RTV in patients with WT HIV is 67 (13). Patients with >60-fold reduced susceptibility would be likely to have plasma levels around the IC50 for the mutant virus, and selective pressure for additional resistance would be expected to be reduced, particularly if further evolution requires a compromise in viral fitness.

The use of selection of resistance as a criterion for defining upper clinical breakpoints may be broadly applicable and advantageous over the use of virologic response to salvage therapy as an endpoint. Since such therapy is always in combination, discerning the activity (or lack thereof) of an individual drug with respect to baseline resistance is complicated by partial activity of the other agents in the regimen. Consequently, large numbers of patients and sophisticated statistical methodologies are generally required (4). By using the selection of incremental resistance as a criterion, the undetermined activity of the accompanying drugs can be ignored, provided that the drug of interest is not affected by resistance to the other agents (e.g., if it is the only drug in the regimen to target a particular viral protein). The detection of selective pressure is also likely to be a sensitive gauge of the remaining activity of the drug that is substantially compromised by preexisting resistance and, thus, may be ideal for defining an upper breakpoint where differentiation of partial activity from no activity is desirable.

The present study has several limitations. The ability to define the mutation pattern for LPV-RTV by using the selection of resistance in PI-experienced patients is complex because of the heterogeneous nature of the baseline viruses. Mutations that emerged may have evolved during prior therapy at low prevalence and then emerged following switching to LPV-RTV. Consequently, although the present work reveals combinations of mutations that produce high-level resistance to LPV, it does not provide definitive information regarding the patterns of mutations that emerge in ARV-naive patients. Characterizing those patterns is hampered by the extremely low incidence of PI resistance in treatment-naive patients receiving LPV-RTV-based combination therapy, even during periods of detectable plasma HIV RNA (7, 14, 23). The present study is also limited by the fact that only mutations encompassed by the PCR fragments amplified for testing (the PR, RT, and a small part of the gag genes) were assessed. The effects of genetic alterations in response to therapy outside of this region, which could contribute to reduced susceptibility, are unknown (30).

Definition of the upper clinical breakpoints for LPV-RTV in the present analysis is limited by the small number of subjects and requires validation with larger cohorts. It is also partially limited by the pharmacokinetic interactions between LPV-RTV and the NNRTI class. Both NVP and EFV lower LPV trough concentration levels through hepatic induction. Some, but not all of the subjects in recent studies were receiving an additional capsule of LPV-RTV, which mostly compensates for the negative drug-drug interactions (16, 36). Therefore, the upper breakpoints estimated in this analysis may be slightly higher for other combination regimens where LPV-RTV is used at the standard dose without an NNRTI. In addition, higher doses of LPV-RTV (or the codosing of additional RTV) might be expected to exhibit partial activity against viruses with >60-fold reduced susceptibility (3). Finally, the general method of defining an upper breakpoint by using the selection of resistance as a criterion may be less suitable if the incremental mutations produce a large loss in viral fitness. In such cases, the residual activity of the drug in question may be underestimated.

In summary, we have described the evolution of HIV in PI-experienced patients failing subsequent therapy on regimens containing LPV-RTV. The risk of evolution of LPV resistance was highly dependent on the baseline genotype and phenotype, and the study provides both insight into the genetic barrier to resistance to LPV-RTV and an estimate of the upper clinical breakpoints for this regimen. Finally, these findings provide information on key mutations and mutation patterns in HIV PR that can define LPV-RTV resistance.


The assistance of Kent Stewart in structural analysis of mutations in HIV protease and of Yolanda Lie and Nick Hellman (Virologic) and Kurt Hertogs and Brendan Larder (Virco) in obtaining the phenotypes and genotypes of the isolates described in this study is gratefully acknowledged. In addition, we thank the investigators and the study site coordinators for studies M97-765, M98-957, and M98-888 as well as the subjects for participating in these studies. Finally, we thank Steve Deeks (UCSF) for an advance copy of an unpublished manuscript.


1. Ait-Khaled, M., A. Rakik, P. Griffin, C. Stone, N. Richards, D. Thomas, J. Falloon, and M. Tisdale. 2003. HIV-1 reverse transcriptase and protease resistance mutations selected during 16 to 72 weeks of therapy in isolates from antiretroviral therapy-experienced patients receiving abacavir/efavirenz/amprenavir in the CNA2007 study. Antivir. Ther. 8:111-120. [PubMed]
2. Benson, C. A., S. G. Deeks, S. C. Brun, R. M. Gulick, J. J. Eron, H. A. Kessler, R. L. Murphy, C. Hicks, M. King, D. Wheeler, J. Feinberg, R. Stryker, P. E. Sax, S. Riddler, M. Thompson, K. Real, A. Hsu, D. Kempf, A. J. Japour, and E. Sun. 2002. Safety and antiviral activity at 48 weeks of lopinavir/ritonavir plus nevirapine and 2 nucleoside reverse-transcriptase inhibitors in human immunodeficiency virus type 1-infected protease inhibitor-experienced patients. J. Infect. Dis. 185:599-607. [PubMed]
3. Bertz, R., J. Li, M. King, D. Kempf, D. Podzamczer, C. Flexner, C. Katlama, D. Havlir, S. Letendre, J. J. Eron, L. Weiss, J. Gatell, A. Simon, K. Robinson, and S. Brun. 2004. Lopinavir inhibitory quotient predicts virologic response in highly antiretroviral-experienced patients receiving high-dose lopinavir/ritonavir, abstr. 134, p. 125. Abstr. 11th Conf. Retrovir. Opportunistic Infect., San Franciso, Calif. Foundation for Retrovirology and Human Health, Alexandria, Va.
4. Brun-Vezinet, F., D. Costagliola, M. Ait Khaled, V. Calvez, F. Clavel, B. Clotet, R. Haubrich, D. Kempf, M. King, D. Kuritzkes, R. Lanier, M. Miller, V. Miller, A. Phillips, D. Pillay, J. Schapiro, J. Scott, R. Shafer, M. Zazzi, A. Zolopa, and V. DeGrutolla. 2004. Clinically validated genotype analysis: guiding principles and statistical concerns. Antivir. Ther. 9:465-478. [PubMed]
5. Carrillo, A., K. Stewart, H. L. Sham, D. W. Norbeck, W. E. Kohlbrenner, J. M. Leonard, D. J. Kempf, and A. Molla. 1998. In vitro selection and characterization of human immunodeficiency virus type 1 variants with increased resistance to ABT-378, a novel protease inhibitor. J. Virol. 72:7532-7541. [PMC free article] [PubMed]
6. Condra, J. H., W. A. Schleif, O. M. Blahy, L. J. Gabryelski, D. J. Graham, J. C. Quintero, A. Rhodes, H. L. Robbins, E. Roth, M. Shivaprakash, D. Titus, T. Yang, H. Teppler, K. E. Squires, P. J. Deutsch, and E. A. Emini. 1995. In vivo emergence of HIV-1 variants resistant to multiple protease inhibitors. Nature 374:569-571. [PubMed]
7. Conradie, F., I. Sanne, W. Venter, and J. Eron. 2004. Failure of lopinavir-ritonavir (Kaletra)-containing regimen in an antiretroviral-naive patient. AIDS 18:1084-1085. [PubMed]
8. De Luca, A., A. Clingotani, S. D. Giumbenedetto, A. Bacarell, V. Tozzi, F. Forbici, A. Antinori, and C. Perno. 2003. Validation and implemetation of clinically relevant interpretation rules of HIV-1 genotypic resistance for lopinavir/ritonavir in salvage therapy, abstr. 41, p. 19. Abstr. First European HIV Drug Resistance Workshop, De Luca, Luxembourg. Virology Education, Utrecht, The Netherlands.
9. Friend, J., N. Parkin, T. Liegler, J. N. Martin, and S. G. Deeks. 2004. Isolated lopinavir resistance after virologic rebound of a ritonavir/lopinavir-based regimen. AIDS 18:1965-1966. [PubMed]
10. Gonzalez de Requena, D., O. Gallego, L. Valer, I. Jimenez-Nacher, and V. Soriano. 2004. Prediction of virological response to lopinavir/ritonavir using the genotypic inhibitory quotient. AIDS Res. Hum. Retrovir. 20:275-278. [PubMed]
11. Harrigan, P. R., K. Hertogs, W. Verbiest, R. Pauwels, B. Larder, S. Kemp, S. Bloor, B. Yip, R. Hogg, C. Alexander, and J. S. Montaner. 1999. Baseline HIV drug resistance profile predicts response to ritonavir-saquinavir protease inhibitor therapy in a community setting. AIDS 13:1863-1871. [PubMed]
12. Hertogs, K., M. P. de Bethune, V. Miller, T. Ivens, P. Schel, A. Van Cauwenberge, C. Van Den Eynde, V. Van Gerwen, H. Azijn, M. Van Houtte, F. Peeters, S. Staszewski, M. Conant, S. Bloor, S. Kemp, B. Larder, and R. Pauwels. 1998. A rapid method for simultaneous detection of phenotypic resistance to inhibitors of protease and reverse transcriptase in recombinant human immunodeficiency virus type 1 isolates from patients treated with antiretroviral drugs. Antimicrob. Agents Chemother. 42:269-276. [PMC free article] [PubMed]
13. Hickman, D., S. Vasavanonda, G. Nequist, L. Colletti, W. M. Kati, R. Bertz, A. Hsu, and D. Kempf. 2004. Estimation of serum-free 50-percent inhibitory concentrations for human immunodeficiency virus protease inhibitors lopinavir and ritonavir. Antimicrob. Agents Chemother. 48:2911-2917. [PMC free article] [PubMed]
14. Hicks, C., M. S. King, R. M. Gulick, A. C. White, Jr., J. J. Eron, Jr., H. A. Kessler, C. Benson, K. R. King, R. L. Murphy, and S. C. Brun. 2004. Long-term safety and durable antiretroviral activity of lopinavir/ritonavir in treatment-naive patients: 4 year follow-up study. AIDS 18:775-779. [PubMed]
15. Hirsch, M. S., F. Brun-Vézinet, R. T. D'Aquila, S. Hammer, V. A. Johnson, D. R. Kuritzkes, C. Loveday, J. W. Mellors, B. Clotet, B. Conway, L. M. Demeter, S. Vella, D. M. Jacobsen, and D. D. Richman. 2000. Antiretroviral drug resistance testing in adult HIV-1 infection: recommendations of an international AIDS society-USA panel. JAMA 283:2417-2426. [PubMed]
16. Hsu, A., J. Isaacson, S. Brun, B. Bernstein, W. Lam, R. Bertz, C. Foit, K. Rynkiewicz, B. Richards, M. King, R. Rode, D. J. Kempf, G. R. Granneman, and E. Sun. 2003. Pharmacokinetic-pharmacodynamic analysis of lopinavir-ritonavir in combination with efavirenz and two nucleoside reverse transcriptase inhibitors in extensively pretreated human immunodeficiency virus-infected patients. Antimicrob. Agents Chemother. 47:350-359. [PMC free article] [PubMed]
17. Isaacson, J., D. Kempf, V. Calvez, I. Cohen-Codar, D. Descamps, E. Guilevic, B. Bernstein, E. Sun, J. Chauvin, and R. Rode. 2002. Quantitative estimate of the effect of individual baseline mutations in HIV protease on the virologic response to lopinavir/ritonavir therapy in heavily antiretroviral-experienced patients, abstr. 559, p. 260. Abstr. 9th Conf. Retrovir. Opportunistic Infect., Seattle, Wash. Foundation for Regroviorology and Human Health, Alexandria, Va.
18. Ives, K. J., H. Jacobsen, S. A. Galpin, M. M. Garaev, L. Dorrell, J. Mous, K. Bragman, and J. N. Weber. 1997. Emergence of resistant variants of HIV in vivo during monotherapy with the proteinase inhibitor saquinavir. J. Antimicrob. Chemother. 39:771-779. [PubMed]
19. Jacobsen, H., M. Hanggi, M. Ott, I. B. Duncan, S. Owen, M. Andreoni, S. Vella, and J. Mous. 1996. In vivo resistance to a human immunodeficiency virus type-1 proteinase inhibitor: mutations, kinetics, and frequencies. J. Infect. Dis. 173:1379-1387. [PubMed]
20. Kagan, R., M. Shenderovich, K. Ramnarayan, and P. N. R. Heseltime. 2003. Emergence of a novel lopinavir resistance mutation at codon 47 correlates with ARV utilization, abstr. 49. Abstr. XII Int. HIV Drug Resistance Workshop, Los Cabos, Mexico. Antivir. Ther. 8:S54.
21. Kempf, D. J., J. D. Isaacson, M. S. King, S. C. Brun, J. Sylte, R. B., B. Bernstein, R. A. Rode, and E. Sun. 2002. Analysis of the virologic response with respect to baseline viral phenotype and genotype and protease inhibitor-experienced HIV-1-infected patients receiving lopinavir/ritonavir therapy. Antivir. Ther. 7:165-174. [PubMed]
22. Kempf, D. J., J. D. Isaacson, M. S. King, S. C. Brun, Y. Xu, K. Real, B. M. Bernstein, A. J. Japour, E. Sun, and R. A. Rode. 2001. Identification of genotypic changes in human immunodeficiency virus protease that correlate with reduced susceptibility to the protease inhibitor lopinavir among viral isolates from protease inhibitor-experienced patients. J. Virol. 75:7462-7469. [PMC free article] [PubMed]
23. Kempf, D. J., M. S. King, B. Bernstein, P. Cernohous, E. Bauer, J. Moseley, K. Gu, A. Hsu, S. Brun, and E. Sun. 2004. Incidence of resistance in a double-blind study comparing lopinavir/ritonavir plus stavudine and lamivudine to nelfinavir plus stavudine and lamivudine. J. Infect. Dis. 189:51-60. [PubMed]
24. Maguire, M., D. Shortino, A. Klein, W. Harris, V. Manohitharajah, M. Tisdale, R. Elston, J. Yeo, S. Randall, F. Xu, H. Parker, J. May, and W. Snowden. 2002. Emergence of resistance to protease inhibitor amprenavir in human immunodeficiency virus type 1-infected patients: selection of four alternative viral protease genotypes and influence of viral susceptibility to coadministered reverse transcriptase nucleoside inhibitors. Antimicrob. Agents Chemother. 46:731-738. [PMC free article] [PubMed]
25. Masquelier, B., D. Breilh, D. Neau, S. Lawson-Ayayi, V. Lavignolle, J. M. Ragnaud, M. Dupon, P. Morlat, F. Dabis, and H. Fleury. 2002. Human immunodeficiency virus type 1 genotypic and pharmacokinetic determinants of the virological response to lopinavir-ritonavir-containing therapy in protease inhibitor-experienced patients. Antimicrob. Agents Chemother. 46:2926-2932. [PMC free article] [PubMed]
26. Mo, H., L. Lu, T. Dekhtyar, K. D. Stewart, E. Sun, D. J. Kempf, and A. Molla. 2003. Characterization of resistant HIV variants generated by in vitro passage with lopinavir/ritonavir. Antivir. Res. 59:173-180. [PubMed]
27. Molla, A., G. R. Granneman, E. Sun, and D. J. Kempf. 1998. Recent developments in HIV protease inhibitor therapy. Antivir. Res. 39:1-23. [PubMed]
28. Molla, A., M. Korneyeva, Q. Gao, S. Vasavanonda, P. J. Schipper, H.-M. Mo, M. Markowitz, T. Chernyavskiy, P. Niu, N. Lyons, A. Hsu, G. R. Granneman, D. D. Ho, C. A. B. Boucher, J. M. Leonard, D. W. Norbeck, and D. J. Kempf. 1996. Ordered accumulation of mutations in HIV protease confers resistance to ritonavir. Nat. Med. 2:760-766. [PubMed]
29. Murphy, R. L., S. Brun, C. Hicks, J. J. Eron, R. Gulick, M. King, A. C. White, Jr., C. Benson, M. Thompson, H. A. Kessler, S. Hammer, R. Bertz, A. Hsu, A. Japour, and E. Sun. 2001. ABT-378/ritonavir plus stavudine and lamivudine for the treatment of antiretroviral-naive adults with HIV-1 infection: 48-week results. AIDS 15:F1-9. [PubMed]
30. Nijhuis, M., N. M. v. Maarseveen, P. Schipper, I. W. Goedegebuure, G. Heilek-snyder, N. Cammack, and C. A. B. Boucher. 2004. Novel HIV drug resistance mechanism leading to protease inhibitor (PI) resistance in response to a high genetic barrier PI in vitro, abstr. 36. Abstr. XIII Int. HIV Drug Resistance Workshop, Costa Adeje, Canary Islands, Spain. Antivir. Ther. 9:S42.
31. Parkin, N. T., C. Chappey, and C. J. Petropoulos. 2003. Improving lopinavir genotype algorithm through phenotype correlations: novel mutation patterns and amprenavir cross-resistance. AIDS 17:955-961. [PubMed]
32. Patick, A. K., M. Duran, Y. Cao, D. Shugarts, M. R. Keller, E. Mazabel, M. Knowles, S. Chapman, D. R. Kuritzkes, and M. Markowitz. 1998. Genotypic and phenotypic characterization of human immunodeficiency virus type 1 variants isolated from patients treated with the protease inhibitor nelfinavir. Antimicrob. Agents Chemother. 42:2637-2644. [PMC free article] [PubMed]
33. Petropoulos, C. J., N. T. Parkin, K. L. Limoli, Y. S. Lie, T. Wrin, W. Huang, H. Tian, D. Smith, G. A. Winslow, D. J. Capon, and J. M. Whitcomb. 2000. A novel phenotypic drug susceptibility assay for human immunodeficiency virus type 1. Antimicrob. Agents Chemother. 44:920-928. [PMC free article] [PubMed]
34. Prado, J. G., T. Wrin, J. Beauchaine, L. Ruiz, C. J. Petropoulos, S. D. Frost, B. Clotet, R. T. D'Aquila, and J. Martinez-Picado. 2002. Amprenavir-resistant HIV-1 exhibits lopinavir cross-resistance and reduced replication capacity. AIDS 16:1009-1017. [PubMed]
35. Rice, H., J. Nadler, J. Schaenman, T. Hawkins, C. Cohen, R. Rode, D. Kempf, and A. Zolopa. 2003. Genotypic predictors of response of lopinavir/ritonavir in clinical practice, abstr. 153. Abstr. XII Int. HIV Drug Resistance Workshop, Los Cabos, Mexico. Antivir. Ther. 8:S169.
36. Saez-Llorens, X., A. Violari, C. O. Deetz, R. A. Rode, P. Gomez, E. Handelsman, S. Pelton, O. Ramilo, P. Cahn, E. Chadwick, U. Allen, S. Arpadi, M. M. Castrejon, R. S. Heuser, D. J. Kempf, R. J. Bertz, A. F. Hsu, B. Bernstein, C. L. Renz, and E. Sun. 2003. Forty-eight-week evaluation of lopinavir/ritonavir, a new protease inhibitor, in human immunodeficiency virus-infected children. Pediatr. Infect. Dis. J. 22:216-224. [PubMed]
37. Stoll, V., W. Qin, K. D. Stewart, C. Jakob, C. Park, K. Walter, R. L. Simmer, R. Helfrich, D. Bussiere, J. Kao, D. Kempf, H. L. Sham, and D. W. Norbeck. 2002. X-ray crystallographic structure of ABT-378 (lopinavir) bound to HIV-1 protease. Bioorg. Med. Chem. 10:2803-2806. [PubMed]
38. Tessmer, U., and H. G. Krausslich. 1998. Cleavage of human immunodeficiency virus type 1 proteinase from the N-terminally adjacent p6* protein is essential for efficient Gag polyprotein processing and viral infectivity. J. Virol. 72:3459-3463. [PMC free article] [PubMed]
39. Walmsley, S., B. Bernstein, M. King, J. Arribas, G. Beall, P. Ruane, M. Johnson, D. Johnson, R. Lalonde, A. Japour, S. Brun, and E. Sun. 2002. Lopinavir-ritonavir versus nelfinavir for the initial treatment of HIV infection. N. Engl. J. Med. 346:2039-2046. [PubMed]
40. Wang, D., and B. Larder. 2003. Enhanced prediction of lopinavir resistance from genotype by use of artificial neural networks. J. Infect. Dis. 188:653-660. [PubMed]
41. Wu, T. D., C. A. Schiffer, M. J. Gonzales, J. Taylor, R. Kantor, S. Chou, D. Israelski, A. R. Zolopa, W. J. Fessel, and R. W. Shafer. 2003. Mutation patterns and structural correlates in human immunodeficiency virus type 1 protease following different protease inhibitor treatments. J. Virol. 77:4836-4847. [PMC free article] [PubMed]

Articles from Journal of Virology are provided here courtesy of American Society for Microbiology (ASM)