Identification of 6 novel loci associated with amino acid levels using single-variant and gene-based tests. T. M. Teslovich1, J. R. Perry2, J. R. Huyghe1, A. U. Jackson1, A. Stančáková3, H. M. Stringham1, P. S. Chines4, J. M. Romm5, H. Ling5, I. McMullen5, R. Ingersoll5, E. W. Pugh5, K. F. Doheny5, J. Kuusisto3, F. S. Collins4, K. L. Mohlke6, M. Laakso3, M. Boehnke1 1) Center for Statistical Genetics, Department of Biostatistics, University of Michigan, Ann Arbor, MI; 2) Genetics of Complex Traits, Exeter Medical School, University of Exeter, Exeter, UK; Wellcome Trust Centre for Human Genetics, UK; and Department of Twin Research and Genetic Epidemiology, King's College London, London, UK; 3) Department of Medicine, University of Eastern Finland and Kuopio University Hospital, Kuopio, Finland; 4) Genome Technology Branch, National Human Genome Research Institute, Bethesda, Maryland, USA; 5) Center for Inherited Disease Research, Johns Hopkins University, Baltimore, Maryland, USA; 6) Department of Genetics, University of North Carolina, Chapel Hill, North Carolina, USA.

   Amino acid levels (AA) are correlated with several disease states, including Alzheimer disease and type 2 diabetes. While candidate gene and GWA studies have implicated common variants underlying variability in AA, the impact of low-frequency (LF; MAF .5% - 5%) and rare (MAF < .5%) variants on AA is largely unknown. To determine the effect of LF (nonsynonymous) variants on AA, we genotyped ~242,000 SNPs on the Illumina HumanExome Beadchip in 8,850 Finnish men from the population-based METSIM study. 60,517 variants were non-monomorphic and passed QC filters. The levels of nine amino acids (Ala, Gln, Gly, His, Ile, Leu, Phe, Tyr, Val) were measured by nuclear magnetic resonance in all subjects; after log-normalization, we generated inverse normalized residuals adjusting for age, age2, and BMI, and tested for association using a linear mixed model assuming an additive genetic effect. We identified 17 associations at genome-wide significance (p<5x10-8) at 12 unique loci. The 3 novel loci are all associated with Gly level. Intergenic variant rs4841132 (MAF=17%, pGly=2x10-23), proximal to PPP1R3B, was included on the array due to its associations with fasting insulin, fasting glucose, and free cholesterol in medium HDL cholesterol particles. Synonymous variant rs16954698 (MAF=12%, pGly=8x10-13) creates a new initiation codon for PKD1L2; however, an adjacent gene, glycine cleavage system H (GCSH), is a strong biological candidate, and rs16954698 may be tagging an untyped causal variant in or near GCSH. Finally, 2 independent LF, nonsynonymous variants in glycine dehydrogenase (GLDC) are strongly associated with Gly (Q996H, MAF=.9%, p=4x10-58; V735L, MAF=4%, p=3x10-17). We performed gene-level tests of association using SKAT and the variable threshold (VT) test to aggregate nonsynonymous and essential splice site variants (SKAT MAF < 1%). SKAT and VT analyses each identified 3 genes associated with AA at genome-wide significance (p<2x10-6). While three of the associations are driven by single variants with large effect, we identified 3 gene-level associations due to aggregate evidence across multiple variants: ALDH1L1 associated with Gly, SKAT p=9x10-8, 9 variants, minimum single-marker p (MSMP)=1x10-5; BCAT2 associated with Val, VT p=2x10-6, 3 variants, MSMP=5x10-5; and HAL associated with His, VT p=1x10-7, 7 variants, MSMP=8x10-5. Taken together, our results suggest that low-frequency and rare variants may have a substantial impact on amino acid levels.

You may contact the first author (during and after the meeting) at