We measured the desaturation index, the ratio of plasma 18:1 / 18:0 and 16:1 /16:0, in 39 FCHL probands and 70 normolipidemic control individuals and in 400 family members from 18 extended pedigrees for whom plasma samples were available. The results and relevant clinical characteristics for these study samples are listed in . The mean 16:1/16:0 and 18:1/18:0 desaturation indices in FCHL probands was significantly higher than that of normolipidemic spouse control individuals (0.134 ± 0.060 vs. 0.101 ± 0.030 for 16:1/16:0, P<0.003, and 3.95 ± 1.08 vs. 2.93 ± 0.882 for 18:1/18:0, P≤0.0001, in probands (n=39) and spouses (n=70), respectively). A similar increase in desaturation indices was also observed in hyperlipidemic relatives when compared to spouse controls ().
Clinical Characteristics of the 18 FCHL families
As expected, there was strong correlation between the 18:1/18:0 and 16:1/16:0 values for the 400 FCHL family members (Pearson correlation = 0.53, p=2.2×10−16
), therefore for all subsequent analyses we used the 18:1/18:0 molar ratio to define the desaturation index. To assess whether the desaturation index is quantitatively correlated with total cholesterol, triglycerides, HDL-C, as well as other metabolic parameters, we conducted regression analyses in order to identify covariates that are significantly correlated with the desaturation index in these pedigrees. Both triglyceride (β = 0.34, P=2.0 × 10−9
) and to a lesser extent, HDLC (β = −0.19, P=0.0005) levels were significant predictors of the desaturation index. Although the desaturation index has been associated with measures of body adiposity in other studies 14–18
, BMI was not a significant predictor of the desaturation index in these FCHL families (β = −0.002, P=0.97). This finding could be due to the strong correlation of the desaturation index with plasma triglycerides in this population such that it obscures the relationship of SCD1 activity with BMI. We tested this hypothesis by testing the triglyceride- and HDLC-adjusted desaturation index residuals for correlation with BMI, however BMI was still not a significant predictor of the desaturation index (β =−0.001, P=0.79) in these families.
In order to estimate the heritability of the desaturation index, we used variance components analysis in all 400 individuals that were measured for the desaturation index. Unadjusted heritability was estimated to be 0.47 (P= 1.0 × 10−9). After adjustment for sex (β=0.18,P=0.05), age (β=−0.004, P=0.2), triglyceride (β=0.34, P= 2.0×10−9) and HDL-C (β = −0.19, P= 0.0005) levels, the estimated heritability (h2) of the desaturation index increased slightly to h2=0.48 (P=2.2 × 10−11), indicating that the desaturation index is a highly heritable trait in this FCHL cohort even after adjustment for covariates.
Given the evidence that genetic factors contribute to the desaturation index, we performed quantitative nonparametric sibpair analysis to identify regions of the genome that may harbor genes that contribute to variation in the desaturation index. We performed the analysis using both age- and sex- adjusted and unadjusted desaturation index values. No evidence of genome-wide significant linkage of the desaturation index was obtained in this nonparametric genome scan. We did, however, obtain suggestive evidence for linkage to chromosome 3p26.1-3p13 (z=2.7, P=0.003), 7p22.2-p15.3 (z=2.1, P=0.02), and 20p11.21-20q13.32 (z=1.7, P=0.04). All regions exhibiting a z-score >1.64 (P≤0.05) are shown in and . Two of these regions, 3p26.1-3p13 and 20p11.21-20q13.32 are located near genes which have been previously associated with either free fatty acid or triglyceride levels in FCHL cohorts 19,20
. No evidence for linkage was obtained for the region containing the SCD1 gene on chromosome 10 (). Additional adjustment of the desaturation index for triglyceride and BMI levels did not significantly alter the results (). Linkage of each region to 16:1/16:0 was consistent with the linkage results of 18:1/18:0, although the overall linkage scores were lower for 16:1/16:0.
Chromosomal Locations of Desaturation Index QTL in 18 FCHL Pedigrees
Figure 1 Non-parametric z-scores > 1.64 (p<0.05) obtained for age- and sex- adjusted (solid black line), age- , sex-, tg-, and hdlc- adjusted (open circles) 18:1/18:0 desaturation index values and for 16:1/16:0 (dashed lines) in 18 extended FCHL (more ...)
The strongest evidence for linkage was located on chromosome 3 near the PPARγ gene (z=2.7, P=0.003). The empirical p-value for this linkage region was P=0.01, indicating that there is a 1% chance that the linkage of the desaturation index to this region is a false positive. We identified tagging SNPs in the 148.5 kb genomic region containing the PPARγ gene (+/− 1kb) using HAPMAP data from 30 CEPH trios (). Seven tagging SNPs and two coding SNPs (rs1801282, also known as the Pro12Ala polymorphism and rs3856806, also known as C161T), which had previously been studied in our case-control set, were genotyped in our FCHL families 19
. These nine SNPs span a total of 141 kb. Pairwise linkage disequilibrium (LD) was similar between FCHL probands and normolipidemic spouses for al SNPs tested (Figure SI). The D’ statistic suggests that there is strong LD among all 9 of the PPARg SNPs in both probands and spouses, whereas the r2
statistics suggest that there is little LD in the region. Since the D’ statistics are not dependent on allele frequencies, the measures of LD indicated by the D’ statistic may be inflated given the small size of our case-control cohort.
We observed only modest association of the desaturation index with four SNPs (rs2972164, rs10510418, rs2938395, rs709157) out of the nine that were genotyped (Table SI, P<0.05), although these results do not remain significant after correction for multiple testing. We tested for haplotype association using the four moderately associated SNPs (P≤0.05 before correction for multiple testing). We observed no significant association of the four-SNP haplotype; however, haplotypes comprised of three SNPs (rs2972164, rs10510418, and rs2938395) defined both a risk and a protective haplotype for the desaturation index trait (). The most significant association was observed for haplotypes composed of two of these SNPs rs2972164 and rs2938395. Under a dominant model, we observed marginal evidence of association of the T-T haplotype of rs2972164 and rs2938395 with the desaturation index (, P=0.02), and under an additive model, the C-C haplotype of rs2972164 and rs2938395 was also marginally associated with the desaturation index (P=0.01). The effect of each haplotype can be interpreted by evaluating the sign of the FBAT statistic. For the T-T haplotype, the sign of the FBAT statistic was positive, suggesting that the effect of the haplotype is to raise the desaturation index value. We have therefore designated this haplotype as the risk haplotype. Similarly, the negative FBAT statistic obtained for the C-C haplotype indicates that the effect of the haplotype is to decrease the desaturation index value, and it is therefore designated as a protective haplotype. Although the association of PPARγ with the desaturation index is only weakly significant, linkage analysis of families which contribute to the association signal revealed an increase in the linkage signal to 3.15, suggesting that subsetting families on the basis of PPARg haplotypes strengthens the linkage signal at the chromosome 3 locus. These data suggest that PPARγ haplotypes may contribute to the variation in desaturation indices in FCHL families; however confirmation of these findings in an independent population is warranted.
Haplotype Association of PPARγ and HNF4α with desaturation index in FCHL families.
We also observed suggestive evidence of linkage of the desaturation index to chromosome 20p11.21-20q13.32 (p= 0.01), a region previously linked to triglycerides and HDL-C in Finnish FCHL and low-HDL families respectively 21,22
. The chromosome 20 linkage peak is located directly over the HNF4α gene which has been recently associated with elevated serum lipids in Finnish and Mexican FCHL families 20
. We genotyped five SNPs (rs2144908, rs6031558, rs2425640, rs745975, rs3212198) that were associated individually or as haplotypes with triglycerides in the Finnish FCHL cohort and with glucose levels in the Mexican FCHL cohort, in our extended FCHL pedigrees (Table SII). Pairwise LD patterns for the five SNPs tested did not differ from those obtained by Weisglas-Volkov et al. (Figure S1) 20
. We observed only marginal association of rs3212198 with the desaturation index in our FCHL families (P=0.03), which does not remain significant after correction for multiple testing. In the study by Weissglas-Volkov et al., a risk haplotype comprised of rs6031558, rs745975, and rs3212198 was significantly associated with serum triglycerides in Finnish and Mexican FCHL families. We tested for association of this haplotype with the desaturation index in our FCHL cohort. We observed quantitative association of the G-G-A haplotype of rs6031558 - rs745975 - rs3212198 with the desaturation index (P=0.01, ) in our FCHL families, although most of the association signal seems to come from rs6031558 - rs745975 (P=0.002). The G-G-A haplotype is the same risk haplotype that was found to be associated with triglycerides in Finnish and Mexican FCHL families20
, supporting the hypothesis that specific HNF4α haplotypes contribute to the development of the FCHL phenotype. Subset analysis of the HNF4a associated families revealed that these families provide no evidence for linkage to the peak marker D20S481, however, the linkage signal in the non-associated families remains and even increases to 1.9. These data suggest that although HNF4a haplotypes are associated with the desaturation index, this association does not account for the linkage of the desaturation index to this chromosome 20 region.