Positive Selection Detection in 40,000 HumanImmunodeficiency Virus (HIV) Type 1 Sequences Automatically IdentifiesDrug Resistance and Positive Fitness Mutations in HIV Proteaseand ReverseTranscriptase

Abstract

Drug resistance is a major problem in the treatment of AIDS, due to the very high mutation rate of human immunodeficiency virus (HIV) and subsequent rapid development of resistance to new drugs. Identification of mutations associated with drug resistance is critical for both individualized treatment selection and new drug design. We have performed an automated mutation analysis of HIV Type 1 (HIV-1) protease and reverse transcriptase (RT) from approximately 40,000 AIDS patient plasma samples sequenced by Specialty Laboratories Inc. from 1999 to mid-2002. This data set provides a nearly complete mutagenesis of HIV protease and enables the calculation of statistically significant K a /K s values for each individual amino acid mutation in protease and RT. Positive selection (i.e., a K a /K s ratio of >1, indicating increased reproductive fitness) detected 19 of 23 known drug-resistant mutation positions in protease and 20 of 34 such positions in RT. We also discovered 163 new amino acid mutations in HIV protease and RT that are strong candidates for drug resistance or fitness. Our results match available independent data on protease mutations associated with specific drug treatments and mutations with positive reproductive fitness, with high statistical significance (the P values for the observed matches to occur by random chance are 10 ؊5.2 and 10 ؊16.6 , respectively). Our mutation analysis provides a valuable resource for AIDS research and will be available to academic researchers upon publication at http://www.bioinformatics.ucla.edu/HIV. Our data indicate that positive selection mapping is an analysis that can yield powerful insights from high-throughput sequencing of rapidly mutating pathogens.The emergence of drug-resistant mutants of human immunodeficiency virus type 1 (HIV-1) protease and reverse transcriptase (RT) genes is an ongoing problem in the fight against AIDS. HIV's mutation rate is high, approximately 4 ϫ 10 Ϫ5 mutations per base per replication cycle, or about one mutation every three generations, yielding at least 10 14 mutations per day worldwide, based on available replication rates and AIDS population estimates (2,12,15). This can lead to the rapid development of resistance to new drug treatments. Researchers and clinicians have made enormous efforts to identify drug-resistant mutations in HIV protease and reverse transcriptase (RT), the molecular targets of the 18 antiretroviral drugs currently approved by the Food and Drug Administration. The discovery of a new drug-resistant mutation typically requires a combination of clinical data (e.g., AIDS patients displaying drug resistance), sequencing, and basic science (e.g., obtaining viral samples and performing phenotypic assays). Fast, automatic methods for identifying drug-resistant mutations could be of great value for researchers studying HIV and other pathogens.Fortunately, the rapid evolution of HIV itself may provide a powerful tool for gaining understanding of its function in general and drug resistance in particular. HIV's high mutati...

Evidence of a novel RNA secondary structurein the coding region of HIV-1 pol gene

“…The first indication of this novel secondary structure came from our previous studies of HIV mutation covariance in clinical HIV samples (Chen et al 2004;Wang and Lee 2007), which identified the unusual pattern of synonymous mutation covariance shown in Figure 4. The clear ''diagonal'' pattern in the covariance matrix, the remarkable conservation of Watson-Crick base-pairing by compensatory mutations, and the obvious sequence complementarity of these regions immediately indicated an RNA secondary structure.…”

“…We computed a measure of synonymous variability from a data set of about 20,000 HIV-1 subtype B sequences (Chen et al 2004). Following the previous analysis of Gog et al (2007), synonymous variability was calculated for each codon position as the normalized mean pairwise distance (MPD) (raw data shown in Supplemental Fig.…”

“…Since synonymous mutations do not alter the amino acid sequence, covariance between synonymous mutations cannot be due to amino acid selection pressure. We therefore analyzed covariance between all possible pairs of synonymous mutations in pol, as an independent test of possible base pairings, using the same data set of 20,000 HIV-1 subtype B sequences (Chen et al 2004). …”

“…This data set contained just subtype B sequences, mostly from patients under antiretroviral drug treatment (Chen et al 2004;Pan et al 2007). These sequences cover 450 codons, including the whole protease (99 codons) and the first 351 codons of the reverse transcriptase.…”

“…These sequences cover 450 codons, including the whole protease (99 codons) and the first 351 codons of the reverse transcriptase. Multiple sequence alignments and mutation detection were performed as previously described (Chen et al 2004). To ensure sufficient diversity of the data and minimize potential bias from phylogenetic effects, we excluded sequences with less than 2% nucleotide diversity.…”