|Home | About | Journals | Submit | Contact Us | Français|
In a sample of 6,156 sequences from 4,183 persons, the top 30 patterns of protease inhibitor, nucleoside reverse transcriptase (RT) inhibitor, and nonnucleoside RT inhibitor mutations accounted for 55, 46, and 66%, respectively, of sequences with drug resistance mutations. Characterization of the phenotypic and clinical significance of these common patterns may lead to improved treatment recommendations for a large proportion of patients for whom antiretroviral therapy is failing.
The optimal regimens for the initial treatment of human immunodeficiency virus type 1 (HIV-1) infection have become increasingly well defined (6). However, the management of persons who develop HIV-1 drug resistance or who are infected primarily with a drug-resistant virus remains a clinical challenge. Genotypic testing for HIV-1 drug resistance is useful for selecting antiretroviral drugs for patients developing treatment failure, but the optimal means for interpreting genotypic tests is not known because many HIV-1 protease and reverse transcriptase (RT) mutations contribute to drug resistance.
To identify common combinations of drug resistance mutations and to determine their contribution to the burden of HIV-1 drug resistance, we examined HIV-1 protease and RT sequences from a clinic-based population tested at Stanford University Hospital. Identifying common patterns of drug resistance mutations is important for determining which mutation patterns should be examined for their phenotypic and clinical significance.
Between July 1997 and September 2003, 6,153 protease sequences and 6,156 RT sequences were determined for HIV-1 isolates from 4,183 persons in Northern California at the request of their physicians. Two thousand, nine hundred forty-seven persons had one sequence. Seven hundred ninety-eight persons had 2 sequences, 257 persons had 3 sequences, and 181 persons had 4 or more sequences. Forty-nine sequences from 32 persons belonged to a non-B subtype, including 27 subtype C sequences, 15 subtype A sequences, and 7 subtype D sequences.
Protease inhibitor (PI) resistance mutation patterns were defined by mutations at 14 nonpolymorphic protease positions associated with PI resistance: positions 24, 30, 32, 46, 47, 48, 50, 53, 54, 73, 82, 84, 88, and 90. Nucleoside RT inhibitor (NRTI) resistance mutation patterns were defined by mutations at 18 RT positions associated with NRTI resistance: positions 41, 44, 62, 65, 67, 69, 70, 74, 75, 77, 115, 116, 118, 151, 184, 210, 215, and 219. Nonnucleoside RT inhibitor (NNRTI) resistance mutation patterns were defined by mutations at 15 positions associated with NNRTI resistance: positions 98, 100, 101, 103, 106, 108, 179, 181, 188, 190, 225, 227, 230, 236, and 238.
Drug resistance mutations at polymorphic protease positions (positions10, 20, 33, 36, 63, 71, 77, and 93) were not used to define PI mutation patterns. Two common polymorphisms at NNRTI resistance positions, A98S and V179I (5), were not used to define NNRTI mutation patterns. Drug resistance mutations consisting of mixtures of wild-type and mutant variants were classified as mutant. Sequences with a mixture of more than one drug resistance mutation at the same position were excluded from analysis.
Drug susceptibility data obtained with the PhenoSense assay (ViroLogic, South San Francisco, Calif.) (4) on isolates matching specific mutation patterns were obtained from the Stanford HIV RT and Protease Sequence Database (5).
Of 6,153 isolates, 21.3% had no drug resistance mutations, 22.6% had mutations associated with resistance to one drug class (15.4% NRTI, 5.7% NNRTI, and 1.5% PI), 34.3% had mutations associated with resistance to two drug classes (20.8% NRTI and PI, 12.4% NRTI and NNRTI, and 1.1% NNRTI and PI), and 21.8% had mutations associated with resistance to three drug classes.
Among the 6,153 protease sequences, 2,934 (47.7%) had a mutation at one or more of the 14 nonpolymorphic PI resistance positions. Of these sequences, 139 (4.7%) had a mixture of two mutations at the same position and were excluded from analysis. The top 30 PI patterns accounted for 54.5% of the 2,795 mutant sequences (Table (Table1),1), and 523 additional patterns accounted for the remaining 45.5% mutant sequences (Fig. (Fig.1).1). Of these additional 523 patterns, 204 (677 sequences) had one of the top 30 patterns plus one additional drug resistance mutation.
The top 30 PI patterns were associated with decreased susceptibility (>2.5-fold increase in the 50% inhibitory concentration [IC50] of drug) to a median of five PIs (Table (Table1).1). These patterns included mutations at 12 of 14 nonpolymorphic PI resistance positions and had a median of 2.5 PI resistance mutations: 7 patterns with 1 mutation, 8 patterns with 2 mutations, 8 patterns with 3 mutations, and 7 patterns with 4 mutations. The remaining 523 patterns had a median of 4 nonpolymorphic PI resistance mutations (data not shown).
L90M was the most common mutation, occurring in 15 patterns and 28.1% of viruses. Each of the top 30 patterns was common in the first and second halves of the study (data not shown), except for G48V/I54V/V82A, which decreased from 18 sequences between 1997 and 2000 to 3 sequences between 2001 and 2003 (P < 0.001).
Among the 6,156 RT sequences, 4,517 (73.4%) had a mutation at 1 or more of the 18 NRTI resistance positions. Of these sequences, 428 (9.5%) had a mixture of two mutations at the same position and were excluded from analysis. The top 30 NRTI patterns accounted for 46.0% of the 4,089 mutant sequences (Table (Table2),2), and 1,090 additional patterns accounted for the remaining 54.0% of mutant sequences (Fig. (Fig.1).1). Of these additional 1,090 patterns, 250 (784 sequences) had 1 of the top 30 patterns plus 1 additional drug resistance mutation.
The top 30 NRTI patterns were associated with decreased susceptibility (>1.4-fold to >4.5-fold increase in IC50 depending on the drug) to a median of three of six NRTIs (Table (Table2).2). These patterns had a median of three NRTI resistance mutations: five with one mutation, eight with two mutations, five with three mutations, five with four mutations, five with five mutations, and one each with six and seven mutations. The remaining 1,090 patterns had a median of 5 NRTI-resistance mutations (data not shown).
M184V was the most common NRTI resistance mutation, occurring alone in 17.4% of mutant sequences and in 16 patterns. T215Y occurred with the next-highest frequency and was in 14 patterns. Mutations at positions 65, 75, 77, 115, 116, and 151 did not occur among the top 30 patterns. Each of the top 30 patterns was common in the first and second halves of the study (data not shown), although significant increases were observed in the patterns A62V/M184V (P < 0.001) and T69N (P = 0.001). K65R was in 1.6% of mutant sequences, occurring alone in 10 sequences and in combination with M184V in 18 sequences, in combination with Q151 M in 15 sequences, and in combination with M184V and Q151 M in 8 sequences. The frequency of K65R increased from 1.2 to 1.9% between the first and second halves of the study (P value not significant).
Among the 6,156 RT sequences, 2,658 (43.2%) had a mutation at one or more of the 15 NNRTI resistance positions. Of these sequences, 132 (5.0%) had a mixture of two mutations at the same position and were excluded from analysis. The top 30 NNRTI mutation patterns accounted for 66.2% of mutant sequences (Table (Table3),3), and 381 additional patterns accounted for the remaining 33.8% of mutant sequences (Fig. (Fig.1).1). Of these additional 381 patterns, 210 (503 sequences) had 1 of the top 30 patterns plus 1 additional drug resistance mutation.
The top 30 NNRTI patterns were associated with decreased susceptibility (>2.5-fold increase in IC50 of drug) to a median of three NNRTIs (Table (Table3).3). These patterns had a median of 2 NNRTI resistance mutations: 14 with 1 mutation, 13 with 2 mutations, and 3 with 3 mutations. The remaining 381 patterns had a median of 3 NNRTI resistance mutations (data not shown).
K103N was the most common NNRTI resistance mutation, occurring alone in 18.2% of sequences and in 12 of the remaining patterns. The top 30 patterns included mutations at 13 of the 15 NNRTI resistance positions. Mutations at positions 227 and 236 were uncommon, occurring in 62 and 4 sequences, respectively. Four of the patterns—K103R, V106I, K101R, and K238R—are polymorphisms that do not occur with increased frequency in persons receiving NNRTIs and do not confer phenotypic resistance (Table (Table3).3). Three of the top 30 patterns included mutations (K101P, M230L, and K238T) that have only recently been recognized as causing NNRTI resistance (W. Huang, N. T. Parkin, Y. S. Lie, T. Wrin, R. Haubrich, S. Deeks, N. Hellmann, C. J. Petropoulos, and J. M. Whitcomb, abstract from the 40th Interscience Conference on Antimicrobial Agents and Chemotherapy 2000, Antivir. Ther. 5(Suppl. 3):24-25, 2000; C. J. Petropoulos, C. Chappey, and N. T. Parkin, Abstr. 43rd Intersci. Conf. Antimicrob. Agents Chemother., abstr. H-451, 2003). There was no significant change in the NNRTI resistance patterns between the first and second halves of the study (data not shown).
With the exception of atazanavir and tenofovir, multiple drug susceptibility results were available for most patterns of drug resistance mutations (Table (Table4).4). The mean absolute deviation in resistance (n-fold) from the median susceptibility for each pattern ranged from 0.05 log10 (1.1-fold) for stavudine and didanosine to 0.32 log10 for delavirdine (2.1-fold), suggesting that results were consistent within most patterns of drug-resistance mutations.
In this sample of >6,000 sequences from nearly 4,200 persons, the top 30 PI, NRTI, and NNRTI mutation patterns accounted for 55, 46, and 66%, respectively, of all sequences with drug resistance mutations. A much larger number of mutation patterns, however, were required to account for the remaining mutant sequences. Studies that examine the clinical significance of drug resistance mutations should focus on the commonly occurring patterns of mutations that we identified in this analysis, because they affect the largest numbers of patients and form the foundation for many of the more complex patterns.
Although about one-third of patients had more than one sequence, clustering of patterns within patients did not explain the observed patterns, because nearly identical patterns were observed in analyses that considered only one sequence per patient (Tables (Tables11 to to3,3, footnotes). All the HIV-1 isolates in this study were from northern California; however, we have previously shown with heavily treated persons that the patterns of drug resistance mutations in northern California are similar to those in subtype B sequences from other parts of the world (3). Although each of the known protease and RT drug resistance mutations occurs in non-B-subtype isolates (R. Kantor, D. Katzenstein, R. Camacho, P. R. Harrigan, A. Tanuri, D. Pillay, A.-M. Vandamme, P. Phanuphak, W. Sugiura, V. Soriano, L. Morris, Z. Grossman, L. F. Brigido, J. Schapiro, and R. W. Shafer, Abstr. 10th Conf. Retrovir. Opportun. Infect., abstr. 623, 2003), preliminary data suggest that the patterns of these mutations may be different in other subtypes (1, 2; Z. Grossman, E. Paxinos, D. Auerbuch, S. Maayan, N. Parkin, D. Engelhard, M. Lorber, E. Kedem, F. Mileguir, N. Vardinon, Z. Bentwich, C. Petropoulos, and J. M. Schapiro, abstract from the 11th International HIV Drug Resistance Workshop 2002, Antivir. Ther. 7:S39, 2002).