|Home | About | Journals | Submit | Contact Us | Français|
Although the Mexican population has a high predisposition to dyslipidemias and premature coronary artery disease, this population is underinvestigated for the genetic factors conferring the high susceptibility.
First, we investigated apolipoprotein B (apoB) levels in Mexican extended families with familial combined hyperlipidemia (FCHL) using a two-step testing strategy. In the screening step, we screened 5,721 SNPs for linkage signals with apoB. In the test step, we analyzed the 130 SNPs residing in regions of suggestive linkage signals for association with apoB. We identified significant associations with two SNPs, rs1424032 (P=6.07×10−6) and rs1349411 (P=2.72×10−4), that surpassed the significance level for the number of tests performed in the test step (P<3.84×10−4). Second, these SNPs were tested for replication in Mexican hyperlipidemic cases-control samples. The same risk alleles as in the FCHL families were significantly associated (P<0.05) with apoB in the case-control samples. The rs1349411 resides near the apoB mRNA editing enzyme (APOBEC1) involved in the processing of APOB mRNA in the small intestine. The rs1424032 resides in a highly conserved non-coding region predicted to function as a regulatory element.
We identified two novel variants, rs1349411 and rs1424032, for serum apoB levels in Mexicans.
Familial combined hyperlipidemia (FCHL) is the most common genetic dyslipidemia affecting 1%–6% of the general population and 10%–20% of subjects with premature coronary artery disease (CAD).1 FCHL was originally described by premature CAD and elevated levels of serum total cholesterol (TC), triglycerides (TG),or both.1 The lipoprotein profile in FCHL patients is characterized by increased levels of small dense LDL (sdLDL) and very large low density lipoprotein (VLDL).2 As apoB is the major protein on VLDL and LDL particles, elevated levels of apoB is considered the hallmark of FCHL.3
In the past 30 years several linkage and candidate-gene studies have been performed to identify genes for FCHL. Three chromosomal regions with significant evidence of linkage have been replicated in several FCHL study samples originating from different populations, the 1q21-234, 11p5, and 16q6 regions. Fine-mapping of the 1q21-23 region resulted in the characterization of the gene encoding upstream transcription factor 1 (USF1), the most established susceptibility gene for FCHL.7 The associations with the lipoprotein lipase (LPL) gene and apolipoprotein A1/C3/A4/A5 gene cluster have also been replicated in several candidate-gene studies for FCHL.8 However, the total number of variants and their relative contributions to the susceptibly of FCHL are largely unknown yet.
Several epidemiological studies have demonstrated that the Mexican population has an increased predisposition to combined hyperlipidemia and premature CAD.9, 10 However, little is known about the genetic factors that may contribute to the increased susceptibility in this population, as Mexicans have been underinvestigated in genetic studies. To date no genome-wide linkage or association studies for serum lipids have been performed in Mexicans.
Previous genome-wide association studies (GWAS) identified nearly 40 loci for serum lipid levels11 with about half of these genes already known to carry mutations with larger effects in dyslipidemic families.12 To date, all GWAS for lipids have examined the concentrations of high and low-density lipoprotein (HDL and LDL) cholesterol and TGs in study samples that were not ascertained for dyslipidaemia.11, 13 Furthermore, the majority of individuals in these studies were White Caucasians. As the types of loci identified in a GWAS reflect the characteristics of the samples utilized in the study, it is likely that additional variants remain to be identified in dyslipidemic study samples, in ethnicities other than Caucasians, or with alternative specific lipid traits. In the present study, we identified two common variants that associate with to apoB concentrations in Mexicans using genome-wide linkage followed by family-based association and replication of the association signals in an unrelated Mexican case-control study sample.
For complete description of the Methods, please also see the online supplementary material available at http://atvb.ahajournals.org
The study design was approved by the ethics committees of the participating centers and all subjects provided a written informed consent. Clinical characteristics of the study samples are shown in supplementary table 1.
A total of 52 extended Mexican FCHL families were included in this study (n=567). These families were recruited at the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (INCMNSZ ) in Mexico City, as previously described.14 Measurements of fasting lipid levels were performed with commercially available standardized methods.14 None of the family members were using lipid lowering medication when the blood sample was drawn.
A total of 1,446 Mexican hypertriglyceridemic cases and controls were recruited at the INCMNSZ in Mexico City. The inclusion criteria were fasting serum TGs >2.3 mmol/L (200 mg/dL) for the cases and <1.7 mmol/L (150 mg/dL) for the controls. Exclusion criteria were type 2 diabetes (T2DM), morbid obesity (BMI > 40 kg/m2), and the use of lipid lowering medication for the controls. Measurements of fasting lipid levels were performed with commercially available standardized methods.14
A total of 567 Mexican family members were genotyped for 6,090 SNPs using the Illumina infinium Human-Linkage-12 BeadChip (Illumina Inc.). Genotyping quality control procedures are described in the Supplementary Methods.
Genotyping of the SNPs rs960345, rs1024461, rs2725769, rs11607526, rs1349411 and rs1424032 in Mexican dyslipidemic cases and controls was performed by the pyrosequencing technique (Biotage, Uppsala, Sweden). The genotyping call rate was >97% for all SNPs, and all SNPs were in HWE in the controls (P>0.05).
In the screening step, the genome-wide set of SNPs was analyzed for linkage. To be able to compare the Mexican data with previous Caucasian data for FCHL, we performed the two-point parametric linkage analysis of the apoB status applying the same linkage approach and parameters as in the previous linkage analyses for FCHL in Caucasians,4, 15 using the ‘Location-Score’ option in the Mendel software16 and allowing for genetic heterogeneity, as described in detail in the supplementary methods.
Since we consider this screening step as exploratory, we also performed multipoint nonparametric linkage as a second analytic approach. This approach makes fewer assumptions about the underling trait model and enables us to capture linkage evidence from multiple SNPs. Nonparametric multipoint linkage analyses for qualitative and quantitative apoB was performed using the variance-component methods implemented in SOLAR,17 as described in detail in the supplementary methods. Residual kurtosis was within the normal range (supplementary figure 2A), and therefore should not influence the quantitative variance component analysis that is known to be sensitive to kurtosis in the trait distributions.
In the test step, we tested a total of 130 SNPs for association with apoB levels. Association analysis was performed utilizing the quantitative transmission disequilibrium test (QTDT) implemented in the genetic analysis package SOLAR13, 17 using an additive model with age and sex as covariates. QTDT is robust against spurious associations due to population stratification and admixture, and was therefore utilized in the admixed Mexican population.13, 17
Association analyses of the SNPs rs960345, rs1024461, rs2725769, rs11607526, rs1349411 and rs1424032 with apoB levels were performed by multivariate linear regression for the additive model in R (R2.8.0) with adjustment for affection status, age, sex and BMI. Cases on lipid lowering therapy at the time of the blood drawing (n=88) were excluded. We considered P-values ≤0.05 as significant replication. We also analyzed the SNPs rs1349411 and rs1424032 for the hypertriglyceridemia case/control status by logistic regression in R.
To identify common genetic variants increasing risk of dyslipidemia in Mexicans, we studied 52 extended Mexican FCHL families for apoB levels, the most consistent abnormality in both young and adult FCHL patients. In these extended families, the heritability estimate for apoB was 0.63 (SE=0.07; P=1.6×10−16). To reduce multiple comparisons while maintaining our chances to detect regions of large effect, we utilized genome-wide linkage as a screening step. In the screening step, 5,721 SNPs, evenly distributed in the genome, were screened for linkage signals. In the test step, we tested for association only those SNPs within regions (±1cM) of suggestive linkage signals (lod score ≥2).
To date no genome-wide linkage or association studies for serum lipids have been performed in Mexicans and thus no prior data were available about the informativeness of the SNPs in Mexicans. The markers of the HumanLinkage-12 array were originally selected using the HapMap Project PhaseII (HapMapII) data for maximum coverage (i.e. tagSNPs) and heterozygosity (i.e. high MAF). Since the Mexican population was not represented by the three HapMapII populations, we first examined whether these markers are common tagSNPs in the Mexican population as well. We observed that the average MAF in Mexican founders (33%, SD=0.11) is similar to the CEU founders of European descent (36%, SD=0.11) (supplementary fig.1), and that the correlation between the allele frequencies is high (r=0.7; P=2×10−16). Furthermore, recently the HapMap PhaseIII was released with genotype data for ~1.4 million SNPs in Mexican-Americans. Although most of these data were selected for nonredundant SNPs, we calculated that >50% of the SNPs on the HumanLinkage-12 array also tag two or more SNPs in the Mexican population (r2>0.8) (supplementary methods). Taken together, these data suggest that the genotyped SNPs are highly prevalent tagSNPs in the Mexican population as well.
To identify suggestive regions (lod>2.0)18 for association analyses, we performed two-point parametric and multipoint nonparametric linkage analyses. The parametric two-point analyses yielded 16 regions with lod scores >2 (fig 1A). The nonparametric multipoint analyses yielded 4 regions with lod scores >2 (fig 1B). Supplementary table 2 provides the results of these suggestive loci with comparison to previous Caucasian data for FCHL. We selected non-redundant SNPs (r2=0.8) present on the Ilumina HumanLinkage-12 array that were located within ±1cM interval of these suggestive regions, resulting in a total of 130 SNPs that were analyzed in the test step.
The 130 SNPs selected from regions of suggestive linkage were tested with continuous apoB levels in family-based association analyses using QTDT (supplementary table 2). Three SNPs were significant after Bonferroni correction for 130 independent tests (Bonferroni significant level P=3.84×10−4), rs1424032 (P=6.07×10−6) and rs30882 (P=5.82×10−5) on 16q21 and rs1349411 on 12p13.31 (P=2.7×10−4) (fig. 1C and supplementary table 2). As the SNPs rs1424032 and rs30882 were in high LD (r2=0.78), only the SNPs rs1349411 and rs1424032 were investigated in further analyses.
To ensure that the association signals of rs1424032 and rs1349411 were not confounded by the linkage signal (allele-sharing) we also performed a test of association given linkage (supplementary methods). We obtained a P-value of 8.8×10−6 for rs1424032 and 8.4×10−4 for rs1349411 in these conditional analyses, confirming the association signals of rs1424032 and rs1349411 in the presence of linkage. Furthermore, supplementary table 2 also provides association results for all SNPs using the empirical variance option −e of the family based association test (FBAT)19 software that tests the null hypothesis of no association in the presence of linkage (supplementary methods). Although the FBAT −e test is less powerful than QTDT as it utilizes only the offspring with heterozygote parents, these association results are not confounded by linkage.
In addition, we examined whether the risk alleles of rs1349411 and rs1424032 can account for the linkage signal for apoB at 12p13.31 and 16q21, respectively. Using subset linkage analyses, as described previously14 and in the supplementary methods, we observed that the associated risk alleles of rs1349411 and rs1424032 each account for a substantial fraction of the linkage signals (42% for both regions). However, these data suggest that additional variants in these regions also influence apoB levels.
To extend our investigation to unrelated and non-FCHL study samples, we examined rs1349411 and rs1424032 in 1,446 Mexican hypertriglyceridemic cases and controls for association with apoB levels. This sample size provides sufficient power (> 80%) to detect effects as little as 0.12 standard deviation (SD) change per each copy of the rare allele (0.6% variance explained) with common SNPs such as rs1349411 and rs1424032 (MAF ≥ 26%). Based on the within-family regression coefficient (β-coefficient) of QTDT, the estimated standardized effect-sizes of rs1349411 and rs1424032 were −0.42 and −0.43, respectively, representing the proportion of SD change in apoB levels (after adjustment for stratification, relatedness, age and sex) per each copy of the rare allele. Hence, the replication sample size should provide adequate power to detect an effect even if the FCHL family sample may have overestimated the effect of these SNPs as “discovery” samples, especially if small, tend to do.
Additionally, our strongest linkage signal (lod>4) was obtained on the chromosomal region 4q. This chromosomal region and a region on chromosome 11p have been implicated in several FCHL studies.5, 20 Therefore, we also further investigated all nominally significant SNPs (P≤0.05 using QTDT) on chromosomes 4 (rs960345, rs1024461 and rs2725769), and 11 (rs11607526) (supplementary table 2). ApoB levels were normally distributed (supplementary figure 2B) and adjusted for age, sex, and BMI as well as for the affection status in order to correct for the sampling variable. Consistent with the family-based results, both the SNPs rs1349411 and rs1424032 were also associated with apoB levels for the additive model and the same risk allele in this unrelated Mexican sample (P=0.016 and P=0.046, respectively) (table 1). As a secondary analysis we tested the SNPs rs1349411 and rs1424032 for the hypertriglyceridemia case/control status by logistic regression analyses. No significant results (P>0.05) were obtained in these analyses. None of the SNPs on chromosomes 4 and 11 were replicated for apoB levels in the case/control study sample (P>0.05, for the same risk allele; supplementary table 3).
Taken together, we observed significant associations between apoB levels and the SNPs rs1349411 and rs1424032 for the same risk allele, consistently throughout the family-based and unrelated study samples. These analyses included a total of 1,998 Mexican subjects.
Population admixture may only confound allelic association if both the trait distribution and the allele frequency differ between ancestries. Family-based association has the advantage that it is robust to spurious associations due to population admixture. We used QTDT for association analysis in the Mexican FCHL families, because it is robust against spurious associations due to population stratification and admixture. However, to minimize the possibility of spurious associations due to population admixture in the unrelated study sample, we evaluated whether apoB distribution and/or the effect of the two significant SNPs on apoB differs with individual ancestry estimates (i.e. IA proportions) in 1,049 Mexican case-control subjects for which we had ancestry informative markers (AIMs) available (supplementary methods). The SNPs on chromosomes 4q and 11p were not significant in the case/control study sample and therefore, they were not examined for spurious associations due to admixture.
In agreement with previous estimates,21 we obtained an average Amerindian/European IA proportion of 0.50 (SE=0.005) (supplementary fig. 3). ApoB levels were not significantly associated with ancestry (P=0.2) (supplementary fig. 4). Similarly, we did not observe a significant interaction between the genotype and IA on apoB levels with either SNP rs1349411 or rs1424032 (P>0.25). Taken together, these data suggest that admixed ancestry should not influence our association analyses in the unrelated study sample.
SNP rs1424032 is located on chromosome 16q21, and its flanking region (±500kb) does not contain any known genes. We calculated pairwise LD in 50 Mexican-American founders of the HapMapIII data (supplementary methods) and observed a 70-kb region of LD containing 16 SNPs in r2 ≥0.7 with rs1424032 (fig. 2). This region is enriched for non-coding conserved elements as predicted by the phastCons program (UCSC Genome Browser)(fig. 2). We calculated by extracting 7,677 similar genomic regions (i.e. 70 kb of non-coding sequence) using the Galaxy tool (http://galaxy.psu.edu/) that the numbers of conserved elements and base-pairs in rs1424032 region fall in the 95th and 90th percentiles, respectively (supplementary methods), making it unlikely that this high degree of regional conservation would be observed by chance alone. Additionally, 10 regions from this LD block were computationally predicted to function as regulatory elements (VISTA Enhancer Browser) (fig. 2).
The SNP rs1349411 is located on chromosome 12p13.31. The region surrounding rs1349411 (±500kb) includes the apoB mRNA editing enzyme (APOBEC1) gene, which resides 400kb downstream from rs1349411. APOBEC1 is an excellent candidate gene for apoB levels and lipoprotein metabolism, as it is involved in the processing of APOB mRNA in the small intestine. As rs1349411 was not included as part of HapMapIII data, we could not determine its region of LD. However, we observed that LD extends between the surrounding region (±50kb) of rs1349411 and APOBEC1 in Mexicans (r2>0.5), but not in Europeans (r2<0.1) (fig. 3 and supplementary table. 4), making it possible that rs1349411 or SNPs in LD with rs1349411 may affect the APOBEC1 gene in the Mexican population.
We identified two variants that associate with apoB concentrations in Mexicans. Our genetic analysis showed association, significant after adjusting for multiple testing, with rs1424032 located on 16q21 and rs1349411 on 12p13.31 in a total of 1,998 individuals from Mexican dyslipidemic families and a case/control study sample. The 16q21 locus is a highly conserved non-coding region, and the 12p13.31 locus includes the APOBEC1 gene, which is an excellent candidate gene for serum apoB levels, as it is involved in the editing of APOB mRNA in the small intestine.
As our sample size was not sufficiently powered for genome-wide association we focused our association analyses on regions of the genome that were supported by linkage analysis using highly prevalent tagSNPs, for which we have greatest statistical power. We are also encouraged by the number of coincident location of the linkage signals in the Mexican FCHL families with previous positive signals in Caucasian FCHL families (supplementary table 2).6, 8, 20 Importantly, both SNPs rs1424032 and rs1349411 are located in regions that provided genome-wide evidence of linkage (lod > 3.3) in previous studies of FCHL.6, 20
Mexicans are an admixed population, descended from a recent mix of Amerindian and European ancestry with a small proportion of African ancestry.21 To screen for true association signals we used family-based association method that is robust to population stratification and admixture.13 We also examined whether apoB levels and/or the associations of rs1424032 and rs1349411 were influenced by global ancestry in the unrelated case/control sample using European/Amerindian informative markers. To reduce sources of heterogeneity both the dyslipidemic families and case/control study sample were recruited by the same dyslipidemia clinic. We recognize that differences in ascertainment criteria may potentially cause heterogeneity. However, we feel that additional significance and conformation were obtained for the SNPs rs1424032 and rs1349411 by utilizing a variety of resources that includes families with multiple affected individuals and population based case/control individuals.
We studied apoB as our primary phenotype in individuals at increased risk to develop CAD. The results from several prospective epidemiological studies have demonstrated that apoB levels are superior to LDL-C and TC levels in predicting the risk of CAD.22 This is because apoB is present in all atherogenic lipoprotein particles with each particle containing exactly one molecule of apoB. Thus, the measurement of apoB represents the total number of atherogenic particles,22 yet apoB levels are not routinely measured and investigated in genetic studies for lipids and/or CAD.
ApoB exists in two forms: apoB-100 and apoB-48. ApoB-100 is synthesized in the liver and is present in VLDL, IDL and LDL particles. ApoB-48 is produced in the small intestine from apoB-100 by RNA editing and is necessary for the assembly of chylomicrons for the absorption of dietary fats.23 The SNP rs1349411 is an excellent candidate for apoB levels and lipoprotein metabolism as it resides near the APOBEC1 gene that is necessary for the production of apoB-48 from apoB-100.23 In previous reports, mice homozygous for targeted deletion of the Apobec1 gene were viable and fertile with no phenotype other than unfavorable changes in lipoprotein metabolism, such as elevated LDL fraction and low HDL-C levels.24 However, mice express Apobec1 in both liver and small intestine, while in humans, APOBEC1 is expressed exclusively in the small intestine.23 As apoB-48 concentrations increase after a fat-rich meal relative to that of apo B-100,23 we intend to compare the postprandial amount of apoB-48 relative to apoB-100 between rs1349411genotype groups in future studies. Although extensive resequencing is warranted to identify all putative regional causal variants, such a diet study could provide a direct conformation for the molecular mechanism of this locus.
The SNP rs1424032 resulted in a P-value of 6 × 10−6 in the FCHL families. This SNP is located in a region that has been consistently replicated in numerous linkage studies for FCHL and HDL-C.6, 8 Although there are no known genes, RNAs or spliced expressed sequence tags near rs1424032, the LD interval of rs1424032 contains many conserved sequences predicted to function as regulatory elements. Recent findings suggest that functional effects may be mediated by remote regulatory elements.25,26 Hence, the susceptibility gene could lie beyond the interval of the association. As rs1424032 is located in the most replicated region for FCHL, it is tempting to speculate that it may potentially influence underling susceptibility gene(s) in this region. The next-by flanking genes, cadherin 8 and 11 which are 1 and 2 Mb away, have not been implicated in lipid metabolism previously. This potential long-range regulatory effect warrants investigation in future studies.
To conclude, it is important to identify genes contributing to the increased susceptibility to hyperlipidemia and CAD in Mexicans, as this population is underinvestigated for the genetic factors conferring the high susceptibility. We identified two loci that are significantly associated with a clinically important atherosclerotic lipid phenotype in the Mexican population.
We thank the Mexican individuals who participated in this study. We also thank Milla Lupsakko, Elina Nikkola, and Salvador Ramirez for laboratory technical assistance.
Sources of Funding
This research was supported by the NIH grants HL-095056, HL-082762 and HL-28481. D.W.-V. is supported by NHGRI grant T32 HG02536, and A.H.-V. by the AHA grant 072523Y
No conflicts to disclose