Genome-wide Association Study of Plasma Efavirenz Pharmacokinetics in ACTG Protocols
B Grady1, M Ritchie2, E Acosta3, G Morse4, E Daar5, R Gulick6, G Robbins7, P McLaren8, David Haas1, and ACTG
1Vanderbilt Univ Sch of Med, Nashville, TN, US; 2Pennsylvania State Univ, Univ Park, US; 3Univ of Alabama at Birmingham Sch of Med, US; 4Univ at Buffalo, State Univ of New York, US; 5Los Angeles Biomed Res Inst at Harbor-UCLA Med Ctr, Torrance, CA, US; 6Weill Cornell Med Coll, New York, NY, US; 7Harvard Univ, Boston, MA, US; and 8Harvard Univ, Broad Inst of MIT and Harvard, Cambridge, MA, US
Background: Inter-individual variability in plasma efavirenz (EFV) pharmacokinetics (PK) is due, at least in part, to human genetic variants. Prior candidate gene studies found associations between CYP2B6 (chromosome 19) variants (especially 516G>T [rs3745274] and 983C>T [rs28399499]) and increased plasma EFV exposure. We used a genome-wide approach to identify novel variants associated with EFV PK.
Methods: ART-naïve ACTG studies A5202, A5095, and ACTG 384 included EFV-containing arms, with plasma sampling for PK. Log-transformed trough EFV concentrations (Cmin) were previously estimated by population PK modeling. Genotyping utilized Illumina HumanHap 650Y or 1MDuo platforms (~500,000 polymorphisms shared across platforms) plus 51 CYP2B6 and 21 CYP2A6 variants genotyped by MassARRAY iPLEX Gold. Genetic consent was obtained under ACTG protocol A5128. Associations were identified by step-wise linear regression, and included the top 10 principal component vectors generated with smartPCA (part of EIGENSTRAT) to adjust for genetic ancestry. Genome-wide significance was P<5x10–8.
Results: The present analyses include 856 participants from ACTG studies. In a first analysis, CYP2B6 516G>T was associated with EFV Cmin at P = 8.5x10–41 (Figure, Panel A). After adjusting for CYP2B6 516G>T, CYP2B6 983C>T was associated at P = 9.9x10–11 (Panel B). After adjusting for both CYP2B6 516G>T and 983C>T, an intronic CYP2B6 variant (rs4803419) was associated at P = 4.4x10–15 (Panel C). This variant was previously shown to be in linkage disequilibrium with a hepatocyte nuclear factor 4 (HNF4) binding site polymorphism, but not previously associated with EFV PK. After adjusting for all 3 variants, 2 non-coding variants on chromosomes 2 and 8 were associated at P <5x10–8, while 3 variants on chromosome 18 approached significance (P <5x10–7) (Panel D). The latter 5 variants did not clearly involve drug metabolism or transport genes.
Conclusions: Among ART-naïve patients, genome-wide significant associations were identified between EFV Cmin and several CYP2B6 variants. After adjusting for CYP2B6 variants, there were possible associations with novel variants elsewhere in the genome. The CYP2B6 associations appear robust; non-CYP2B6 associations must be replicated in independent cohorts.