Structural forms of the human amylase locus and their relationships to SNPs, haplotypes and obesity

Abstract

Hundreds of genes reside in structurally complex, poorly understood regions of the human genome. One such region contains the three amylase genes (AMY2B, AMY2A and AMY1) responsible for digesting starch into sugar. Copy number of AMY1 is reported to be the largest genomic influence on obesity, although genome-wide association studies for obesity have found this locus unremarkable. Using whole-genome sequence analysis, droplet digital PCR and genome mapping, we identified eight common structural haplotypes of the amylase locus that suggest its mutational history. We found that the AMY1 copy number in an individual's genome is generally even (rather than odd) and partially correlates with nearby SNPs, which do not associate with body mass index (BMI). We measured amylase gene copy number in 1,000 obese or lean Estonians and in 2 other cohorts totaling ∼3,500 individuals. We had 99% power to detect the lower bound of the reported effects on BMI, yet found no association.

Figure 1. The copy number distributions of the amylase genes and the structural haplotypes of the amylase locus

Whole-genome sequence data from European (CEU), Yoruba (YRI), Chinese (CHB), and Japanese (JPT) population samples from the 1000 Genomes Project were analyzed for AMY1, AMY2A, and AMY2B copy number using Genome STRiP. (a) The populations have similar distributions of AMY2A and AMY2B, but a different distribution for AMY1 (KS test, P value = 3 × 10⁻⁶). For AMY1, even copy numbers greatly outnumber odd copy numbers. (b) In the YRI and CEU population samples, AMY2A and AMY1 share parity in 98% of individuals: when AMY2A copy number is odd, so is AMY1 copy number. JPT and CHB samples had very few odd copy numbers and were excluded from the figure. (c) A preliminary map of eight common structural forms of the amylase locus in Europeans, derived from the above analysis and with structural features ordered using earlier data from cosmid mapping and (d) genome mapping experiments to analyze the haplotypes (AH2, AH3, AH4, AH2B2, AH4B2). Additional structural forms found only in African genomes are described in Supplementary Fig. 4.

Figure 2. The relationship of the amylase structural haplotypes to SNPs and SNP haplotypes

(a) Displayed are the SNP haplotypes flanking the structural alleles of the amylase locus in the Europeans (CEU+GBR+TSI+IBS+FIN) of the 1000 Genomes Project. The amylase alleles are represented by the colored leaves, though the locus actually resides within the center of the plot. The colored columns are the SNP alleles, and the gray is the invariant surrounding region. The branchpoints mark where the SNP haplotypes diverge due to mutation or recombination. Note that the AMY1-odd structures (brown) share multiple SNP haplotype backgrounds, while other amylase structures (blue, green) segregate on distinct branches. Also note that specific SNP haplotypes (branches) appear to associate with greater or lesser average AMY1 copy number than others do. (b) The relationship of nearby SNPs to AMY1 copy number is consistent across two European-ancestry cohorts.

Figure 3. The association analysis of AMY1 copy number to obesity or BMI in three cohorts

Out of a cohort of 51,535 Estonians, those in the tails of the BMI distribution (500 individuals with BMI < 22 | 500 individuals with BMI > 33) were measured for copy number of all three amylase genes and genotyped for SNPs in obesity-related genes. (a) Measurements of AMY1 copy number in the Estonian cohort. (b) Obese and lean individuals show indistinguishable distributions of AMY1 copy number (P > 0.05). Statistical tests were performed on raw measurements as well as AMY2A-informed AMY1 copy number (Methods). (c) Measurements of AMY1 copy number and (d) association are shown for the GoT2D cohort controls. (e) Measurements of AMY1 copy number and (f) association are shown for the InCHIANTI cohort. Points are the mean BMI for each AMY1 copy number. Error bars are the 95% confidence intervals.