We hypothesized that a common SNP in the 3' untranslated region of the upstream transcription factor 1 (USF1), rs3737787, may affect lipid traits by influencing gene expression levels, and we investigated this possibility utilizing the Mexican population, which has a high predisposition to dyslipidemia. We first associated rs3737787 genotypes in Mexican Familial Combined Hyperlipidemia (FCHL) case/control fat biopsies, with global expression patterns. To identify sets of co-expressed genes co-regulated by similar factors such as transcription factors, genetic variants, or environmental effects, we utilized weighted gene co-expression network analysis (WGCNA). Through WGCNA in the Mexican FCHL fat biopsies we identified two significant Triglyceride (TG)-associated co-expression modules. One of these modules was also associated with FCHL, the other FCHL component traits, and rs3737787 genotypes. This USF1-regulated FCHL-associated (URFA) module was enriched for genes involved in lipid metabolic processes. Using systems genetics procedures we identified 18 causal candidate genes in the URFA module. The FCHL causal candidate gene fatty acid desaturase 3 (FADS3) was associated with TGs in a recent Caucasian genome-wide significant association study and we replicated this association in Mexican FCHL families. Based on a USF1-regulated FCHL-associated co-expression module and SNP rs3737787, we identify a set of causal candidate genes for FCHL-related traits. We then provide evidence from two independent datasets supporting FADS3 as a causal gene for FCHL and elevated TGs in Mexicans.
By integrating a genetic polymorphism with genome-wide gene expression levels, we were able to attribute function to a genetic polymorphism in the USF1 gene. The USF1 gene has previously been associated with a common dyslipidemia, FCHL. FCHL is characterized by elevated levels of total cholesterol, triglycerides, or both. We demonstrate that this genetic polymorphism in USF1 contributes to FCHL disease risk by modulating the expression of a group of genes functionally related to lipid metabolism, and that this modulation is mediated by USF1. One of the genes whose expression is modulated by USF1 is FADS3, which was also implicated in a recent genome-wide association study for lipid traits. We demonstrated that a genetic polymorphism from the FADS3 region, which was associated with triglycerides in a GWAS study of Caucasians, was also associated with triglycerides in Mexican FCHL families. Our analysis provides novel insight into the gene expression profile contributing to FCHL disease risk, and identifies FADS3 as a new gene for FCHL in Mexicans.
Citation: Plaisier CL, Horvath S, Huertas-Vazquez A, Cruz-Bautista I, Herrera MF, Tusie-Luna T, et al. (2009) A Systems Genetics Approach Implicates USF1, FADS3, and Other Causal Candidate Genes for Familial Combined Hyperlipidemia. PLoS Genet 5(9): e1000642. https://doi.org/10.1371/journal.pgen.1000642
Editor: John D. Storey, Princeton University, United States of America
Received: April 9, 2009; Accepted: August 12, 2009; Published: September 11, 2009
Copyright: © 2009 Plaisier et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by NIH grants HL-28481 and HL082762; American Heart Association grant 0725232Y; National Human Genome Research Institute grant T32 HG02536. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Familial combined hyperlipidemia (FCHL) is a common atherogenic dyslipidemia conferring nearly two-fold greater risk for coronary heart disease . FCHL is characterized by familial segregation of elevated fasting plasma triglycerides (TGs), total cholesterol (TC), or both ,. Another common characteristic of FCHL is elevated levels of fasting plasma apolipoprotein B (ApoB) . In Mexico 12.6% of the general population have combined hyperlipidemia, suggesting that FCHL is a common dyslipidemia in the Mexican population .
We previously identified an association within the region of chromosome 1q21-q23 consistently linked to FCHL – with the associated linkage disequilibrium (LD) bin containing variants in upstream transcription factor 1 (USF1), and the adjacent gene, F11 receptor (F11R) . The association with this LD bin containing USF1 has been replicated for FCHL, and its component traits TC, TGs and low density lipoprotein cholesterol (LDL-C) in different populations –, including Mexicans . It has become evident that the SNP rs3737787 residing in the 3′ UTR of USF1 captures the disease-associated signal, although its relationship to the causal polymorphism contributing to the etiology of FCHL is unknown. Previous studies involving direct sequencing, extensive genotyping and gene expression analyses of the USF1 region have not identified any SNPs in the rs3737787 LD bin altering the coding sequence or the expression of USF1 itself in fat  or lymphoblasts . It has, however, been demonstrated that genes known to be regulated by USF1 were differentially expressed between rs3737787 genotype groups in Finnish fat biopsies . The direct targets of USF1 were previously identified using chromatin immunoprecipitation and high-resolution promoter microarrays (ChIP-Chip) . Indirect USF1 target genes remain unknown.
We hypothesized that rs3737787 affects the function of USF1, and that this effect could be identified as the differential expression of genes directly or indirectly regulated by USF1. To test this hypothesis we first identified the direct and indirect targets of USF1 by transiently over-expressing USF1 in vitro and assaying for differential expression using gene expression microarrays. Next we showed that there is overlap between the genes regulated by USF1 and genes correlated with rs3737787 genotypes in 70 Mexican FCHL case/control fat biopsies. Adipose tissue is the major storage depot for TGs in the body, and has been suggested to play an important role in the hypertriglyceridemia component of FCHL –. It is also worth noting that TGs are the major focus of our study because the most associated FCHL-component trait with USF1 has been TGs ,,. Liver would also be a highly relevant tissue for gene expression studies of dyslipidemia due to its key role in the regulation of lipid metabolism. However, adipose tissue biopsies represent significantly less severe complications to the research subjects than liver biopsies. Adipose tissue has thus become the most studied tissue in FCHL because of its relevance to TG metabolism and the relatively minimal invasiveness of collecting subcutaneous fat biopsies , , , –.
Recent systems genetic strategies that characterize interactions between genotype data and co-expression modules have successfully been applied to complex diseases –. Here we integrate weighted gene co-expression network analysis (WGCNA) with the SNP rs3737787 genotypes to identify an FCHL-related module and develop a systems genetic gene-screening strategy for causal candidate genes of FCHL related traits. We were able to identify two TG related co-expression modules, which are clusters of genes likely to be co-regulated by similar factors (i.e. transcription factors, genetic variants, or environmental effects). We tested the co-expression modules for association with FCHL, FCHL component traits, and rs3737787 genotypes. Several authors have suggested that genetic markers can be used as causal anchors for dissecting the causal relationships between traits. Since randomization is the most convincing method for establishing causal relationships between two correlated traits, it is natural to make use of genetically randomized genotypes (implied by Mendel's laws) to derive causality tests that are less susceptible to confounding by hidden variables –. Here we utilized the network edge orienting (NEO) software  coupled with the SNP rs3737787 genotypes to screen for causal candidate genes of FCHL and FCHL component traits. We followed-up a causal candidate gene for both FCHL and TGs, fatty acid desaturase 3 (FADS3), residing in the FADS1-2-3 locus, which was recently implicated for TGs in a meta-analysis of genome-wide association studies for lipids in Caucasians . In an independent Mexican FCHL dataset we replicate the association for TGs with the previously associated genome-wide SNP. To summarize, we identify two co-expression modules associated with TGs, the key component trait of FCHL. We then demonstrate that one of these co-expression modules is also associated with the SNP rs3737787. Our genetic marker based causal screening analysis implicates 18 FCHL causal candidate genes inside this module. Multiple lines of evidence validate the role of FADS3 as a causal candidate gene for FCHL.
Relating rs3737787 Correlated Gene Expression to USF1
We hypothesized that if the SNP rs3737787 tagged LD bin alters the function of the transcription factor USF1, the effect could be assayed on the downstream expression of USF1 target genes, or the subsequent targets of those genes. We transiently over-expressed USF1 in the HEK293T cell line and assayed for differential expression using gene expression microarrays to detect both direct and indirect targets of USF1. Over-expression of USF1 was validated by western blot (Figure S1). In our over-expression analysis, we identified 2,897 genes that were directly or indirectly regulated by USF1 over-expression in vitro (Table S1). We validated the differential expression of nine genes (AGT, AK5, CNR1, GALR2, IGFBP7, KISSR1, PLXDC1, PLXNC1 and PROK2) using quantitative real-time PCR (Figure S2).
Next we utilized an independent gene expression dataset of 70 Mexican FCHL case/control fat biopsies whose clinical characteristics are given in Table 1. We observed 972 genes (gene expression profiles) significantly correlated with rs3737787 genotypes using an additive model (Table S2). The rs3737787 correlated genes had significant overlap both with i) the set of USF1 regulated genes identified in our USF1 over-expression experiment (n = 277; p-value = 3.0×10−5; fold-enrichment = 1.22) and ii) the previously published genes identified by ChIP-Chip  which are directly regulated by USF1 (n = 117; p-value = 0.0051; fold-enrichment = 1.23). Furthermore, we also observed significant overlap between the rs3737787 correlated genes and the 2,189 genes differentially expressed between FCHL cases and normolipidemic controls (n = 245; p-value = 0.0030; fold-enrichment = 1.16) (Table S3), supporting a link from rs3737787 to FCHL etiology. Taken together, the overlap between rs3737787 correlated genes and genes regulated by USF1 suggest that the effect of rs3737787 on FCHL is mediated through the transcription factor USF1.
Weighted Gene Co-Expression Network Analysis
The gene expression variation between individuals in the Mexican FCHL fat biopsies is due to a combination of genetic and environmental factors. Ascertainment for disease status enriches the cases with factors leading to hyperlipidemia, and contrasting the cases with the controls will allow these factors to be more easily identified. We utilized weighted gene co-expression networking analysis (WGCNA) to identify gene co-expression modules summarizing the main trends in variation for the Mexican FCHL case/control fat biopsies. The WGCNA method clustered the 14,942 gene expression probes on the Mexican FCHL case/control microarrays into 28 gene co-expression modules (Table S4). Each of the modules was labeled with a unique color as an identifier. We then characterized each module for enrichment of functionally-related genes (Table S5). To summarize the gene expression profiles of the highly correlated genes inside a given module, we use the first principal component, which is referred to as the module eigengene (ME). We tested the amount of trait variance explained by each ME (Table S6), which provides insight into how much influence each module has on a particular trait. We also tested each ME for correlation with the quantitative FCHL component traits: TC, TG and ApoB (Figure 1). Given the high level of correlation between the FCHL component traits (correlation TC & TG = 0.57; correlation TC & ApoB = 0.90; correlation TG & ApoB = 0.59), we approximate the total number of independent tests to be 2. Therefore, a Bonferroni correction would have to account for a total of 56 multiple comparisons (28 modules×2 independent tests). Thus using Bonferroni correction we would require a p-value less than or equal to 8.9×10−4 in order to maintain an experiment-wide type I error rate of 0.05. This highlights a major statistical advantage of our module based analyses over conventional differential expression analyses which have to account for tens of thousands of multiple comparisons.
The rows are labeled by the ME color, and in parentheses is the number of genes in the module. The columns are labeled by SNP or clinical trait. The quantitative clinical traits, all except for FCHL which is qualitative, were corrected for significant covariates (age and/or sex) and standardized before use in analyses. The correlation coefficients are shown for each cell, and in parentheses is the p-value for the significance of the correlation. Cells are colorized based on the strength and sign of the correlation according to the scale on the right hand side of the figure. AThe tan module has been renamed to the USF1-regulated FCHL-associated (URFA) module.
TG-Associated Module Eigengenes
We observed two modules (or more precisely MEs) significantly associated with TGs (tan ME correlation = 0.43, p-value = 2×10−4, Bonferroni-corrected p-value = 0.011; midnightblue ME correlation = 0.52, p-value = 5×10−6, Bonferroni-corrected p-value = 0.00028), the key component trait of FCHL. Interestingly, the associations of the tan and midnightblue ME with TGs were experiment-wide significant as they were less than the most conservative multiple comparison correction (Bonferroni-corrected p-value for 56 tests, p-value≤8.9×10−4). Using a multivariate regression model, we observed that the two TG associated MEs explained 30% of TG trait variance, which is relatively high considering this is a complex trait. Both the tan and midnightblue MEs were also marginally associated with qualitative FCHL disease status (tan ME correlation = 0.33, p-value = 0.006; midnightblue ME correlation = 0.36, p-value = 0.002), although these associations did not beat multiple testing correction (Figure 1). The association pattern of the tan ME with the FCHL component traits was characteristic of FCHL: elevated plasma levels of TC, TG and ApoB (TC correlation = 0.28, p-value = 0.02; TG correlation = 0.43, p-value = 2×10−4; ApoB correlation = 0.32, p-value = 0.009) (Figure 1). The midnightblue ME had a similar association pattern with the FCHL component traits (TG correlation = 0.52, p-value = 5×10−6; ApoB correlation = 0.31, p-value = 0.01), although it was not associated with TC (correlation = 0.17, p-value = 0.20) (Figure 1). We observed significant overlap (p-value≤2.7×10−10) between the FCHL differentially expressed genes and the tan and midnightblue module genes (tan module overlap p-value = 2.7×10−10; midnightblue module overlap p-value = 4.6×10−31) (Table 2). Given that FCHL is known to be a complex disease, it is not surprising that FCHL status is associated with multiple modules. Both of these modules are likely to be regulated by different genetic or environmental factors which contribute to the etiology of FCHL.
rs3737787-Associated Module Eigengenes
Integrating the co-expression modules with genetic data provides insight into the function and regulation of each module. The turquoise and tan MEs were associated with the rs3737787 genotypes (turquoise ME correlation = 0.28, p-value = 0.02; tan ME correlation = 0.33, p-value = 0.006) (Figure 1). We observed a significant overlap between the rs3737787 correlated genes and both the turquoise and tan module genes (Table 3). The tan module was the only module to be associated with FCHL, the FCHL component trait signature, and rs3737787 genotypes. To emphasize the biological implications of the tan module, we renamed it as the USF1-regulated FCHL-associated (URFA) module. The turquoise module was not significantly associated with FCHL, suggesting that while the expression of turquoise module genes depends on rs3737787 genotypes, the gene functions of the turquoise module are less likely to be related to FCHL (Figure 1, Table S5). These results demonstrate that rs3737787 genotypes predict the expression of many genes, some of which predispose to FCHL.
We should point out that our sample size of 70 individuals provides 80% power to detect a significant association (p-value≤0.05) with correlation coefficient = 0.33, and only limited power to rule out associations with a correlation coefficient<0.33. To have 80% power to detect a significant association (p-value≤0.05) when the true correlation coefficient = 0.2 requires 194 individuals, and when the true correlation coefficient = 0.1 requires 783 individuals. Therefore our sample size does not permit us to rule out associations of smaller magnitude (correlation coefficient<0.33). However, our sample size of 70 individuals provides enough power to detect significant moderate correlations (correlation coefficient≥0.33) between eigengenes, SNPs, and clinical traits.
USF1-Regulated FCHL-Associated (URFA) Module Characterization
The URFA module contains 504 co-expressed genes that are enriched for the gene ontology (GO) Biological Process categories of Cellular Lipid Metabolic Processes and Lipid Metabolic Processes (p-values = 1.0×10−5 and 9.3×10−6; Benjamini-Hochberg multiple comparison corrected p-values = 0.022 and 0.039) (Table S5). The URFA ME accounts for 10% of the variation of FCHL, 6% of TC, 17% of TG, and 9% of ApoB (Table S6). The fact that the URFA ME was associated with rs3737787 reflects the fact that most module genes are at least partially regulated by this SNP. To evaluate whether the module causally affects FCHL component traits, we utilized the Network Edge Orienting (NEO) R software package  which takes SNP genotypes as input into structural equation models that compute causal edge orienting scores (referred to as LEO.NB scores). Since we are only considering a single SNP (rs3737787) we computed LEO.NB.SingleMarker scores for the causal orientation of a ME→trait. The LEO.NB.SingleMarker score is a relative model fitting index for the causal model (rs3737787→ME→trait) relative to alternative causal models, and the larger the value for the LEO.NB.SingleMarker score the stronger the evidence that this causal orientation is correct. We required that the our causal model fit at least two times better than the next best alternative model, which equates to a LEO.NB.SingleMarker score of 0.3  (please see the Materials and Methods for a detailed description). We found sufficient evidence to infer a causal relationship between the URFA ME and fasting plasma TGs levels (LEO.NB.SingleMarker score = 0.31), the key component trait of FCHL. The LEO.NB.SingleMarker score for the URFA ME to FCHL was 0.25. We then used the NEO software to prioritize genes inside the URFA module by calculating the LEO.NB.SingleMarker scores evaluating the causal model (rs3737787→gene expression→trait). We identified 18 causal candidate genes for FCHL, and 171 causal candidate genes for fasting plasma TGs levels (Table S7). We observed 13 causal candidate genes for both FCHL and TG. None of the URFA module genes showed evidence for causally affecting TC or ApoB. Since our interest was in FCHL disease status, we characterized the 18 causal candidate genes for FCHL disease status as potential candidate genes for genetic association studies in Mexican FCHL families (Table 4).
Replication of Association to the FADS1-2-3 Gene Locus
Fatty acid desaturase 3 (FADS3) was 1 of the 18 genes causally linked to FCHL. Interestingly, variation from the FADS1-2-3 genomic region was previously associated with TGs in a recent meta-analysis of GWAS in Caucasians . We chose to follow-up the SNP rs174547 residing in the FADS1-2-3 locus which was significantly associated with TGs at the genome-wide level in this previous meta-analysis . The same study demonstrated that the SNP rs102275, in complete LD with rs174547 in Caucasians, predicted the expression of FADS1 and to a lesser extent FADS3 in human liver . We hypothesized that because FADS3 expression was associated with FCHL, any variation affecting the expression of FADS3 could be associated with FCHL or an FCHL component trait, especially TGs. Therefore we genotyped both rs174547 and rs102275 in the Mexican FCHL case/control fat biopsies (n = 70). We observed a large allele frequency discrepancy between the Mexican and Caucasian populations for both rs174547 and rs102275, which was also present in the HapMap Phase 3 data (rs174547: Caucasian T allele = 0.66, Mexican T allele = 0.34; rs102275: Caucasian T allele = 0.65, Mexican T allele = 0.30). The LD between rs174547 and rs102275 was reduced in the Mexican samples compared to the Caucasian population (Mexican r2 = 0.90; Caucasian r2 = 1.0), indicating that the SNPs are no longer completely redundant in the Mexican population. Therefore, we report the results for both SNPs (Table 5). In the 70 Mexican individuals undergoing fat biopsies, the major alleles of both SNPs were associated with hypertriglyceridemia as defined by TGs dichotomized by the 90th age-sex specific population percentiles (rs174547 p-value = 0.024, rs102275 p-value = 0.012). Despite the switch in allele frequencies, the association was observed with the same allele (C) as in Caucasians and thus, the direction of the association is identical to the previous genome-wide association study . We then proceeded to test these two SNPs for replication in the Mexican families with FCHL and its component traits (Table 6). The major alleles of both SNPs were associated with hypertriglyceridemia as defined by TGs dichotomized by the 90th age-sex specific population percentiles (rs174547 p-value = 0.0071, rs102275 p-value = 0.0034), and again the direction of the association was consistent with the previous genome-wide association study . The association in the Mexican FCHL families with rs102275 was Bonferroni corrected significant (rs102275 corrected p-value = 0.034). These results replicate the findings for the FADS1-2-3 locus in the Mexican population.
Predicting the Expression of FADS1 and FADS3 in Adipose Tissue
Of the three fatty acid desaturase genes in the FADS1-2-3 locus, we observed that only the expression of FADS3 was associated with FCHL disease status in adipose tissue (204257_at correlation = 0.31, p-value = 0.0084; 216080_s_at correlation = 0.31, p-value = 0.0092). This result indicates that increasing expression of FADS3 is associated with increased risk for FCHL. We then tested whether the two SNPs rs174547 and rs102275 predicted the expression of FADS1 and/or FADS3 in the Mexican FCHL case/control fat biopsies. We observed a marginally significant prediction of FADS1 expression (rs174547 correlation = 0.26, p-value = 0.027; rs107225 correlation = 0.20, p-value = 0.089), and little to no evidence for prediction of FADS3 expression (rs174547 correlation = −0.15, p-value = 0.22; rs102275 correlation = −0.14, p-value = 0.25). These results were consistent with the results from Caucasian human liver samples where rs102275 predicted the expression of FADS1 and to a lesser extent FADS3 expression.
By integrating both genetic and transcriptome profile data we were able to attribute a function to the FCHL associated, rs3737787-tagged LD bin. We first provide evidence that the SNP rs3737787 has an effect on global expression profiles, and further evidence suggesting that this effect is mediated by USF1. We then identified two TG-associated co-expression modules one of which (the URFA module) was also associated with the FCHL, the other FCHL component traits, and rs3737787 genotypes. The URFA module was enriched for the GO Biological Process categories Lipid Metabolic Process and Cellular Lipid Metabolic process. The URFA module also contained eighteen genes causally linked to FCHL. One of the causal genes for FCHL, FADS3, coincides with a previous genome-wide significant association signal for TGs . We demonstrate that the genome-wide associated SNP also has an effect on TGs in the Mexican FCHL families. In summary, we demonstrate the function of the rs3737787 LD bin, construct and characterize a co-expression network for Mexican FCHL case/control fat biopsies, and demonstrate that the loci identified as causal can harbor variants associated with FCHL or FCHL component traits.
There are three SNPs from the rs3737787 tagged LD bin residing within the USF1 gene. Although our current study does not allow us to pinpoint which of these three USF1 SNPs in LD is the actual causative SNP, we were able to demonstrate that an allele-dependent effect of the rs3737787 tagged LD bin on downstream genes is mediated by USF1 and not the adjacent gene F11R. To identify the causal SNP in USF1, additional functional studies are warranted that investigate allele-dependent effects of each of the three SNPs on USF1 isoform and protein levels.
Gene co-expression networking methods have been successfully applied in a variety of different settings –. We utilized weighted gene co-expression network analysis (WGCNA) for the following reasons. First, it is assumed that co-expression is a biological process, and as such a co-expression module is a set of genes that are likely to be co-regulated by similar factors (e.g. shared transcription factors, genetic variants or environmental effects). Second, modules (and corresponding module eigengenes) represent a biologically motivated data reduction method which greatly alleviates the multiple comparison problem inherent in genomic data analysis. Third, we provide annotation tables for module membership (Table S4) and URFA module causal candidate genes measures (Table S7), which provides a resource for other investigators. Our unbiased module detection analysis identified a module that was associated with rs3737787 genotypes, fasting plasma TG levels, FCHL disease status, and contained genes that are causal drivers of TG levels. It is difficult to ascertain if the URFA module can be identified in other tissues. Nevertheless, we consider that the identification of the URFA module in fat does further implicate a biological role of fat tissue in FCHL, although we cannot establish whether these FCHL modules are tissue-specific.
Our approach provides insight to how the SNP rs3737787 confers increased risk for FCHL, by demonstrating that it regulates the URFA module eigengene which in turn contributes to increased TG levels, a key component trait of FCHL. The URFA module can be considered rs3737787-genotype dependent, while the other TG associated midnightblue module was not dependent on rs3737787 genotypes. The variation regulating other modules remains to be discovered.
We conducted a literature search for links between the 18 FCHL causal candidate genes from the URFA module to lipid related phenotypes and/or atherogenic processes. Three of the FCHL causal candidate genes were directly related to lipid phenotypes (ABCC6, AKT2, HSD11B1), and two others were likely to be related to lipid phenotypes (FADS3, PNPLA4) via homology. We also identified genes which were related to atherogenic processes such as inflammation (CARD8, ICSBP1, STX6) and reactive oxygen species (GCLM). Among the 18 genes there are also putative genes and genes with little known function. The function of these novel candidate genes can be studied in vivo or in vitro, and will be useful in understanding how the URFA module contributes to the etiology of FCHL. Importantly, some of the genes causally linked to FCHL (ABCC6, AKT2, FADS3, GCLM, HSD11B1) have already been associated with FCHL related traits in humans (Table 3 and Table S8). Studies in mice have also demonstrated that genetic manipulation of Abcc6, Akt2 and Hsd11b1 affect FCHL component traits or related phenotypes –. Together these data support a causal association between the 18 causal candidate genes from the URFA module and FCHL. Although we present strong statistical and biological evidence in support of our findings, we want to emphasize that additional biological validation studies are warranted.
Identifying the associated gene(s) from the genetic association evidence for the FADS1-2-3 locus in humans is complicated by the fact that multiple LD blocks have been associated with several different but related lipid traits , –, as well as coronary artery disease  and nonfatal myocardial infarction . It has been established that both FADS1 and FADS2 function as a fatty acid desaturase , in the pathway generating polyunsaturated fatty acids from the essential Omega-3 and Omega-6 fatty acids. The function of FADS3 is unknown, although it shares 52% sequence identity with FADS1 and 62% with FADS2 , suggesting that FADS3 may also function as a fatty acid desaturase. Our findings demonstrate that FADS3 expression is regulated by rs3737787 genotypes in adipose tissue, and that this differential expression causes increased risk for FCHL. Neither FADS1 nor FADS2 were regulated by rs3737787 genotypes, nor were the expression of these two genes associated with FCHL. Future studies to identify the function of FADS3 will be required to elucidate the mechanisms by which differential regulation of FADS3 in adipose tissue predisposes to FCHL.
Our systems genetic analysis integrated SNP genotypes with gene expression levels to unravel the complex disorder that is familial combined hyperlipidemia. With this study we have identified new intriguing causal candidate genes for genetic studies with FCHL, and we provide causal candidate genes for FCHL related component traits (Table S7). These data may form the starting point for future studies that further elucidate the etiology of FCHL. The fact that we were able to account for 30% of fasting plasma TG trait variance is very promising, and that it required the expression of hundreds of genes to observe this demonstrates that FCHL is a complex and heterogeneous disorder. The URFA module is regulated by the SNP rs3737787 in USF1, suggesting that variation in the USF1 gene acts as a master regulator of genes contributing to the FCHL phenotype. Our analysis of gene expression and SNP genotype data from few but carefully selected individuals demonstrates that a well designed systems genetic study can lead to insights on complex disease traits.
Materials and Methods
Transient transfection and over-expression of USF1 were conducted in HEK293T cells, which express the USF1 protein (Figure S1). The USF1 coding region was PCR amplified from cDNA (forward-primer = ATGAAGGGGCAGCAGAAAACAG, reverse primer = TTAGTTGCTGTCATTCTTGATG) and cloned into pcDNA4/HisMax TOPO TA mammalian expression vector (Invitrogen, Cat. No. K864-20). Plasmid for transient transfection was prepared using the Endo Free Plasmid Kit Mega (Qiagen, Cat. No. 12363). Over-expression of USF1 was validated by western blot using Anti-HisG-HRP (Invitrogen, Cat No. R941-25) and Rabbit Anti-USF1 (Santa-Cruz, Cat. No. sc-8983) (Figure S1). Total RNA was collected using the RNeasy Plus Mini Kit (Cat No. 74134) 48 hours after transient transfection of HEK293T cells. Microarray hybridization has been described previously . After applying the quality control pipeline, a total of 16,570 probesets passed and were utilized in the analyses. Differential expression between USF1 and empty vector control transfected cells was analyzed using a Student's T-test (Table S1). The differential expression of nine genes (AGT, AK5, CNR1, GALR2, IGFBP7, KISSR1, PLXDC1, PLXNC1 and PROK2) was successfully validated using quantitative real-time PCR (Figure S2 and Table S9). The transient transfection and over-expression of USF1 microarray data can be accessed in MIAME compliant format from NCBI GEO database (GSE17300).
Mexican FCHL Case/Control Gene Expression Microarrays
We collected 70 Mexican FCHL case/control fat biopsies from umbilical subcutaneous adipose tissue under local anesthesia. Each participant provided written informed consent. The study design was approved by the ethics committees of the INCMNSZ and UCLA. Clinical characteristics are summarized in Table 1. Differential expression for a subset of these microarrays has been validated by qRT-PCR previously ,. We have previously described in detail the procedures for RNA extraction ,, and microarray hybridization . After applying the microarray quality control pipeline, a total of 14,942 probesets passed and were utilized in analyses. The microarray data can be accessed in MIAME compliant format from the NCBI Gene Expression Omnibus (GEO) database (GSE17170).
Microarray Quality Control Pipeline
A rigorous quality control pipeline was developed to prepare the gene expression microarray data for analysis. First, the CEL files were imported into R 2.8.0 using the justRMA function of the Affymetrix library from Bioconductor, which applies background subtraction and quantile normalization. An alternate CDF file (U133Plus2msk.cdf.gz) was used to exclude mis-targeted and nonspecific probes from the microarrays , and probesets with less than 7 remaining probes were also excluded. Probesets with more than 50% absent calls were excluded, as calculated using the panp package from Bioconductor. Finally, the ComBat library in R was used to correct for batch effect .
Mexican Microarray Analyses
Only residuals of the clinical traits corrected for significant covariates, such as age and sex, were used in analyses. There was minimal relatedness between individuals, but to avoid confounding due to relatedness all analyses with the Mexican FCHL cases/controls were corrected for kinship. The lmekin function from the kinship package in R was utilized for the analyses of differential gene expression. Differential expression of genes based on the FCHL status was assayed by regressing FCHL status with the covariates age, sex and kinship on the expression of each gene. Differential expression of genes based on the additive model for rs3737787 genotypes was assayed by regressing the rare allele counts with the covariates age, sex, FCHL status and kinship on the expression of each gene. Significance of the overlap of independent gene lists was calculated using Fisher's exact test. We report two-tailed p-values and fold change of enrichment to demonstrate whether there are more or less overlapping genes than expected. Functional enrichment analysis was conducted using the DAVID software package . The Benjamini-Hochberg (B–H) multiple testing corrected p-values are reported. Power calculations were conducted utilizing pwr.r.test function from the pwr package in R.
Weighted Gene Co-Expression Network Analysis (WGCNA)
For the 70 Mexican FCHL case/control fat biopsies, a weighted gene co-expression network was constructed using the blockwiseModules function from the WGCNA package in R . We utilized residuals of gene expression corrected for age, sex and kinship in the WGCNA. The blockwiseModules function allows the entire dataset of 14,942 genes to be utilized in the construction of the weighted gene co-expression network, by splitting the genes into sets of less than 2,000 genes . For each set of genes the pair-wise correlation matrix is computed, and using the power function the correlations are weighted to a power of β . An advantage of WGCNA is the fact that the results are highly robust with respect to the choice of the parameter β . WGCNA is designed to identify modules of densely interconnected genes by searching for genes with similar patterns of connectivity with other genes, which can be summarized as the topological overlap between genes . Topological overlap is then calculated from the weighted co-expression matrix, and turned into a dissimilarity measure (1 - topological overlap) for average linkage hierarchical clustering. The dynamic tree-cutting algorithm  is then used to identify modules of co-expressed genes. After all blocks have been processed, a gene is reassigned to another module if it is found to have higher connectivity to this other module, and modules whose eigengenes are highly correlated are merged . Genes that are not assigned to a module are assigned to the grey module. The maximum block size was set at 2,000 genes; a power (β) was set to 6 as this was found to be optimal (and also happens to be the default setting); minimum module size was set at 100; the minimum height for merging modules was set at 0.2; and the maximum height at which the tree could be cut was set to 1. All the other parameters were left at default settings.
To infer causal flow between genetic markers, gene expression traits and clinical traits, we utilized the single.marker.analysis function from the Network Edge Orienting (NEO) package in R . We also utilized residuals of gene expression corrected for age, sex and kinship in causality testing. There are five different models for causal flow that are tested in the single.marker.analysis function. The model of interest for these studies infers a causal flow anchored by the SNP rs3737787 which affects the expression of a gene or gene co-expression module that in turn alters a clinical trait (M→A→B). The fit of this model was assessed using the local structural equation model (SEM) based, edge orienting, next best single marker (LEO.NB.SingleMarker) score, which is the log10 probability of this model divided by the log10 probability of the next best fitting alternative model . We chose to use the same threshold as utilized by Presson et al., where the LEO.NB.SingleMarker score was required to be positive and at least twice as probable as the next best alternative model, a LEO.NB.SingleMarker score of 0.3≈log10(2) . Thus any LEO.NB.SingleMarker score greater than or equal to 0.3 was considered sufficient to infer causal flow.
Mexican FCHL Families
A total of 872 individuals from 74 Mexican FCHL families were collected at Instituto Nacional de Ciencias Medicas y Nutricion, Salvador Zubiran, as described previously ,. The 90th age-sex specific Mexican population percentiles for TGs and TC were used to determine the affection status, as described previously ,. Each participant provided written informed consent. The study design was approved by the ethics committees of the INCMNSZ and UCLA.
The rs3737787, rs174547 and rs102275 SNPs were genotyped using the pyrosequencing technique (Biotage). Primers for PCR were designed using the Primer3 software , and labeled detection primers using SNP Primer Design software (Biotage). Sequencing was performed using an ABI 377XL automated DNA sequencer (Applied Biosystems). All SNPs had at least 90% genotype call rate in each study sample, and when necessary individuals were sequenced to achieve at least a 90% genotype call rate for each SNP.
Genetic Association Testing
All genetic association testing was conducted using qualitative phenotypes, where the continuous FCHL related phenotypes are dichotomized by the 90th age-sex specific percentiles of TGs and TC. Association for the Mexican FCHL fat biopsy case/control study sample was calculated by regressing genotypes in an additive model to the dichotomized trait of interest using logistic regression with the glm function in R. We utilized the Family Based Association Testing (FBAT) software package to test for association in the Mexican FCHL families . Because the FADS1-2-3 locus resides in a linked region, the -e option of FBAT was used. This option utilizes a null model of no association in the presence of linkage. FBAT implements a generalization of the transmission disequilibrium test (TDT)  in the extended pedigrees. Thus, FBAT is not sensitive to population stratification or admixture –. The FBAT –e option also corrects for cryptic relatedness. Due to these features FBAT -e provides a robust test for association in these extended Mexican families.
Western blot demonstrating HEK293T endogenous USF1 expression and USF1 over-expression at the protein level. Protein collected from HEK293T cells transfected with USF1 is labeled USF1 overexpression, protein collected from HEK293T cells transfected with an empty vector is labeled Empty Vector, and protein collected from HEK293T cells which were not tranfected is labeled HEK293T. Protein amounts were quantitated using BCA assay and 10 µg of protein was loaded for each lane. (A) Probed with Anti-HisG-HRP antibody which should detect only the over expressed USF1 protein with a polyhistidine tag. (B) Probed with Anti-USF1 antibody, which will detect over expressed as well as endogenous USF1 and demonstrates the relative amount of over-expression of USF1. This also demonstrates that the HEK293T cell line does express USF1 protein at a detectable level by western blot.
(1.00 MB EPS)
qRT-PCR validation of genes identified as differentially expressed due to USF1 over-expression on microarrays. The validation genes were: AGT = angeotensin, AK5 = adenylate kinase 5, CNR1 = canaboid receptor 1, GALR2 = galanin receptor 2, IGFBP7 = insulin-like growth factor binding protein 7, KISS1R = KISS1 receptor, PLXDC1 = plexin domain containing 1, PLXNC1 = plexin C1, and PROK2 = prokineticin 2. The endogenous housekeeping controls were: B2M = beta-2-microglobulin, and HPRT = hypoxanthine phosphoribosyltransferase. Student's t-test p-values were calculated between USF1 over-expression and Empty Vector control, and a single asterisk (*) indicates a p-value≤0.05, a double asterisk (**) indicates a p-value≤0.005, and a triple asterisk (***) indicates a p-value≤0.0005.
(0.99 MB EPS)
Genes differentially expressed by USF1 over-expression in HEK293T cells (p-value≤0.05).
(1.80 MB PDF)
Genes correlated to FCHL-associated rs3737787 genotypes (additive model) in Mexican FCHL case/control fat biopsies (p-value≤0.05).
(0.64 MB PDF)
Genes differentially expressed between FCHL cases and normolipidemic controls in Mexican FCHL case/control fat biopsies (p-value≤0.05).
(1.42 MB PDF)
WGCNA module membership for all probes from the Mexican FCHL case/control fat biopsies.
(11.40 MB ZIP)
Functionally enriched annotation terms for the set of genes comprising each co-expression module (Benjamini-Hochberg corrected p-value≤0.05).
(0.11 MB PDF)
Trait variance explained by WGCNA co-expression module eigengenes (Adjusted R2).
(0.02 MB PDF)
Causality testing results for URFA module genes terminating in FCHL, TC, TG, and ApoB.
(0.38 MB PDF)
Previous association evidence with lipid or atherogenic traits for genes causally linked to FCHL from the URFA module.
(0.07 MB PDF)
We thank the individuals who participated in this study. We thank F. M. Ranchos for fat biopsy collection, and E. Nikkola, M. Lupsakko, L. Riba, and S. Ramírez-Jiménez for laboratory technical assistance. Also we would like to thank the UCLA Microarray Core Facility.
Conceived and designed the experiments: CLP TTL CAS PP. Performed the experiments: CLP AHV PP. Analyzed the data: CLP PP. Contributed reagents/materials/analysis tools: CLP SH AHV ICB MFH TTL CAS PP. Wrote the paper: CLP SH TTL CAS PP.
- 1. Hopkins PN, Heiss G, Ellison RC, Province MA, Pankow JS, et al. (2003) Coronary artery disease risk in familial combined hyperlipidemia and familial hypertriglyceridemia: a case-control comparison from the National Heart, Lung, and Blood Institute Family Heart Study. Circulation 108: 519–523.PN HopkinsG. HeissRC EllisonMA ProvinceJS Pankow2003Coronary artery disease risk in familial combined hyperlipidemia and familial hypertriglyceridemia: a case-control comparison from the National Heart, Lung, and Blood Institute Family Heart Study.Circulation108519523
- 2. Goldstein JL, Schrott HG, Hazzard WR, Bierman EL, Motulsky AG (1973) Hyperlipidemia in coronary heart disease. II. Genetic analysis of lipid levels in 176 families and delineation of a new inherited disorder, combined hyperlipidemia. J Clin Invest 52: 1544–1568.JL GoldsteinHG SchrottWR HazzardEL BiermanAG Motulsky1973Hyperlipidemia in coronary heart disease. II. Genetic analysis of lipid levels in 176 families and delineation of a new inherited disorder, combined hyperlipidemia.J Clin Invest5215441568
- 3. Nikkila EA, Aro A (1973) Family study of serum lipids and lipoproteins in coronary heart-disease. Lancet 1: 954–959.EA NikkilaA. Aro1973Family study of serum lipids and lipoproteins in coronary heart-disease.Lancet1954959
- 4. Aguilar-Salinas CA, Olaiz G, Valles V, Torres JM, Gomez Perez FJ, et al. (2001) High prevalence of low HDL cholesterol concentrations and mixed hyperlipidemia in a Mexican nationwide survey. J Lipid Res 42: 1298–1307.CA Aguilar-SalinasG. OlaizV. VallesJM TorresFJ Gomez Perez2001High prevalence of low HDL cholesterol concentrations and mixed hyperlipidemia in a Mexican nationwide survey.J Lipid Res4212981307
- 5. Huertas-Vazquez A, del Rincon JP, Canizales-Quinteros S, Riba L, Vega-Hernandez G, et al. (2004) Contribution of chromosome 1q21-q23 to familial combined hyperlipidemia in Mexican families. Ann Hum Genet 68: 419–427.A. Huertas-VazquezJP del RinconS. Canizales-QuinterosL. RibaG. Vega-Hernandez2004Contribution of chromosome 1q21-q23 to familial combined hyperlipidemia in Mexican families.Ann Hum Genet68419427
- 6. Coon H, Myers RH, Borecki IB, Arnett DK, Hunt SC, et al. (2000) Replication of linkage of familial combined hyperlipidemia to chromosome 1q with additional heterogeneous effect of apolipoprotein A-I/C-III/A-IV locus. The NHLBI Family Heart Study. Arterioscler Thromb Vasc Biol 20: 2275–2280.H. CoonRH MyersIB BoreckiDK ArnettSC Hunt2000Replication of linkage of familial combined hyperlipidemia to chromosome 1q with additional heterogeneous effect of apolipoprotein A-I/C-III/A-IV locus. The NHLBI Family Heart Study.Arterioscler Thromb Vasc Biol2022752280
- 7. Allayee H, Krass KL, Pajukanta P, Cantor RM, van der Kallen CJ, et al. (2002) Locus for elevated apolipoprotein B levels on chromosome 1p31 in families with familial combined hyperlipidemia. Circ Res 90: 926–931.H. AllayeeKL KrassP. PajukantaRM CantorCJ van der Kallen2002Locus for elevated apolipoprotein B levels on chromosome 1p31 in families with familial combined hyperlipidemia.Circ Res90926931
- 8. Pajukanta P, Nuotio I, Terwilliger JD, Porkka KV, Ylitalo K, et al. (1998) Linkage of familial combined hyperlipidaemia to chromosome 1q21-q23. Nat Genet 18: 369–373.P. PajukantaI. NuotioJD TerwilligerKV PorkkaK. Ylitalo1998Linkage of familial combined hyperlipidaemia to chromosome 1q21-q23.Nat Genet18369373
- 9. Pei W, Baron H, Muller-Myhsok B, Knoblauch H, Al-Yahyaee SA, et al. (2000) Support for linkage of familial combined hyperlipidemia to chromosome 1q21-q23 in Chinese and German families. Clin Genet 57: 29–34.W. PeiH. BaronB. Muller-MyhsokH. KnoblauchSA Al-Yahyaee2000Support for linkage of familial combined hyperlipidemia to chromosome 1q21-q23 in Chinese and German families.Clin Genet572934
- 10. Pajukanta P, Lilja HE, Sinsheimer JS, Cantor RM, Lusis AJ, et al. (2004) Familial combined hyperlipidemia is associated with upstream transcription factor 1 (USF1). Nat Genet 36: 371–376.P. PajukantaHE LiljaJS SinsheimerRM CantorAJ Lusis2004Familial combined hyperlipidemia is associated with upstream transcription factor 1 (USF1).Nat Genet36371376
- 11. van der Vleuten GM, Isaacs A, Hijmans A, van Duijn CM, Stalenhoef AF, et al. (2007) The involvement of upstream stimulatory factor 1 in Dutch patients with familial combined hyperlipidemia. J Lipid Res 48: 193–200.GM van der VleutenA. IsaacsA. HijmansCM van DuijnAF Stalenhoef2007The involvement of upstream stimulatory factor 1 in Dutch patients with familial combined hyperlipidemia.J Lipid Res48193200
- 12. Putt W, Palmen J, Nicaud V, Tregouet DA, Tahri-Daizadeh N, et al. (2004) Variation in USF1 shows haplotype effects, gene : gene and gene : environment associations with glucose and lipid parameters in the European Atherosclerosis Research Study II. Hum Mol Genet 13: 1587–1597.W. PuttJ. PalmenV. NicaudDA TregouetN. Tahri-Daizadeh2004Variation in USF1 shows haplotype effects, gene : gene and gene : environment associations with glucose and lipid parameters in the European Atherosclerosis Research Study II.Hum Mol Genet1315871597
- 13. Coon H, Xin Y, Hopkins PN, Cawthon RM, Hasstedt SJ, et al. (2005) Upstream stimulatory factor 1 associated with familial combined hyperlipidemia, LDL cholesterol, and triglycerides. Hum Genet 117: 444–451.H. CoonY. XinPN HopkinsRM CawthonSJ Hasstedt2005Upstream stimulatory factor 1 associated with familial combined hyperlipidemia, LDL cholesterol, and triglycerides.Hum Genet117444451
- 14. Huertas-Vazquez A, Aguilar-Salinas C, Lusis AJ, Cantor RM, Canizales-Quinteros S, et al. (2005) Familial combined hyperlipidemia in Mexicans: association with upstream transcription factor 1 and linkage on chromosome 16q24.1. Arterioscler Thromb Vasc Biol 25: 1985–1991.A. Huertas-VazquezC. Aguilar-SalinasAJ LusisRM CantorS. Canizales-Quinteros2005Familial combined hyperlipidemia in Mexicans: association with upstream transcription factor 1 and linkage on chromosome 16q24.1.Arterioscler Thromb Vasc Biol2519851991
- 15. Naukkarinen J, Gentile M, Soro-Paavonen A, Saarela J, Koistinen HA, et al. (2005) USF1 and dyslipidemias: converging evidence for a functional intronic variant. Hum Mol Genet 14: 2595–2605.J. NaukkarinenM. GentileA. Soro-PaavonenJ. SaarelaHA Koistinen2005USF1 and dyslipidemias: converging evidence for a functional intronic variant.Hum Mol Genet1425952605
- 16. Odom DT, Dowell RD, Jacobsen ES, Nekludova L, Rolfe PA, et al. (2006) Core transcriptional regulatory circuitry in human hepatocytes. Mol Syst Biol 2: 2006 0017.DT OdomRD DowellES JacobsenL. NekludovaPA Rolfe2006Core transcriptional regulatory circuitry in human hepatocytes.Mol Syst Biol22006 0017
- 17. de Graaf J, Veerkamp MJ, Stalenhoef AF (2002) Metabolic pathogenesis of familial combined hyperlipidaemia with emphasis on insulin resistance, adipose tissue metabolism and free fatty acids. J R Soc Med 95: Suppl 4246–53.J. de GraafMJ VeerkampAF Stalenhoef2002Metabolic pathogenesis of familial combined hyperlipidaemia with emphasis on insulin resistance, adipose tissue metabolism and free fatty acids.J R Soc Med95Suppl 424653
- 18. Eurlings PM, Van Der Kallen CJ, Geurts JM, Kouwenberg P, Boeckx WD, et al. (2002) Identification of differentially expressed genes in subcutaneous adipose tissue from subjects with familial combined hyperlipidemia. J Lipid Res 43: 930–935.PM EurlingsCJ Van Der KallenJM GeurtsP. KouwenbergWD Boeckx2002Identification of differentially expressed genes in subcutaneous adipose tissue from subjects with familial combined hyperlipidemia.J Lipid Res43930935
- 19. Arner P (1997) Is familial combined hyperlipidaemia a genetic disorder of adipose tissue? Curr Opin Lipidol 8: 89–94.P. Arner1997Is familial combined hyperlipidaemia a genetic disorder of adipose tissue?Curr Opin Lipidol88994
- 20. Lee JC, Weissglas-Volkov D, Kyttala M, Sinsheimer JS, Jokiaho A, et al. (2007) USF1 Contributes to High Serum Lipid Levels in Dutch FCHL Families and U.S. Whites With Coronary Artery Disease. Arterioscler Thromb Vasc Biol 27: 2222–2227.JC LeeD. Weissglas-VolkovM. KyttalaJS SinsheimerA. Jokiaho2007USF1 Contributes to High Serum Lipid Levels in Dutch FCHL Families and U.S. Whites With Coronary Artery Disease.Arterioscler Thromb Vasc Biol2722222227
- 21. Reynisdottir S, Angelin B, Langin D, Lithell H, Eriksson M, et al. (1997) Adipose tissue lipoprotein lipase and hormone-sensitive lipase. Contrasting findings in familial combined hyperlipidemia and insulin resistance syndrome. Arterioscler Thromb Vasc Biol 17: 2287–2292.S. ReynisdottirB. AngelinD. LanginH. LithellM. Eriksson1997Adipose tissue lipoprotein lipase and hormone-sensitive lipase. Contrasting findings in familial combined hyperlipidemia and insulin resistance syndrome.Arterioscler Thromb Vasc Biol1722872292
- 22. Plaisier CL, Kyttala M, Weissglas-Volkov D, Sinsheimer JS, Huertas-Vazquez A, et al. (2009) Galanin preproprotein is associated with elevated plasma triglycerides. Arterioscler Thromb Vasc Biol 29: 147–152.CL PlaisierM. KyttalaD. Weissglas-VolkovJS SinsheimerA. Huertas-Vazquez2009Galanin preproprotein is associated with elevated plasma triglycerides.Arterioscler Thromb Vasc Biol29147152
- 23. Meex SJ, van der Kallen CJ, van Greevenbroek MM, Eurlings PM, El Hasnaoui M, et al. (2005) Up-regulation of CD36/FAT in preadipocytes in familial combined hyperlipidemia. Faseb J 19: 2063–2065.SJ MeexCJ van der KallenMM van GreevenbroekPM EurlingsM. El Hasnaoui2005Up-regulation of CD36/FAT in preadipocytes in familial combined hyperlipidemia.Faseb J1920632065
- 24. Horra A, Salazar J, Ferre R, Vallve JC, Guardiola M, et al. (2009) Prox-1 and FOXC2 gene expression in adipose tissue: A potential contributory role of the lymphatic system to familial combined hyperlipidaemia. Atherosclerosis 2009 March 11 (Epub ahead of print). A. HorraJ. SalazarR. FerreJC VallveM. Guardiola2009Prox-1 and FOXC2 gene expression in adipose tissue: A potential contributory role of the lymphatic system to familial combined hyperlipidaemia.Atherosclerosis 2009 March 11 (Epub ahead of print)
- 25. Huertas-Vazquez A, Plaisier C, Weissglas-Volkov D, Sinsheimer J, Canizales-Quinteros S, et al. (2008) TCF7L2 is associated with high serum triacylglycerol and differentially expressed in adipose tissue in families with familial combined hyperlipidaemia. Diabetologia 51: 62–69.A. Huertas-VazquezC. PlaisierD. Weissglas-VolkovJ. SinsheimerS. Canizales-Quinteros2008TCF7L2 is associated with high serum triacylglycerol and differentially expressed in adipose tissue in families with familial combined hyperlipidaemia.Diabetologia516269
- 26. Ylitalo K, Nuotio I, Viikari J, Auwerx J, Vidal H, et al. (2002) C3, hormone-sensitive lipase, and peroxisome proliferator-activated receptor gamma expression in adipose tissue of familial combined hyperlipidemia patients. Metabolism 51: 664–670.K. YlitaloI. NuotioJ. ViikariJ. AuwerxH. Vidal2002C3, hormone-sensitive lipase, and peroxisome proliferator-activated receptor gamma expression in adipose tissue of familial combined hyperlipidemia patients.Metabolism51664670
- 27. Ghazalpour A, Doss S, Zhang B, Wang S, Plaisier C, et al. (2006) Integrating genetic and network analysis to characterize genes related to mouse weight. PLoS Genet 2: e130.A. GhazalpourS. DossB. ZhangS. WangC. Plaisier2006Integrating genetic and network analysis to characterize genes related to mouse weight.PLoS Genet2e130
- 28. Chen Y, Zhu J, Lum PY, Yang X, Pinto S, et al. (2008) Variations in DNA elucidate molecular networks that cause disease. Nature 452: 429–435.Y. ChenJ. ZhuPY LumX. YangS. Pinto2008Variations in DNA elucidate molecular networks that cause disease.Nature452429435
- 29. Ferrara CT, Wang P, Neto EC, Stevens RD, Bain JR, et al. (2008) Genetic networks of liver metabolism revealed by integration of metabolic and transcriptional profiling. PLoS Genet 4: e1000034.CT FerraraP. WangEC NetoRD StevensJR Bain2008Genetic networks of liver metabolism revealed by integration of metabolic and transcriptional profiling.PLoS Genet4e1000034
- 30. Fuller TF, Ghazalpour A, Aten JE, Drake TA, Lusis AJ, et al. (2007) Weighted gene coexpression network analysis strategies applied to mouse weight. Mamm Genome 18: 463–472.TF FullerA. GhazalpourJE AtenTA DrakeAJ Lusis2007Weighted gene coexpression network analysis strategies applied to mouse weight.Mamm Genome18463472
- 31. Presson AP, Sobel EM, Papp JC, Suarez CJ, Whistler T, et al. (2008) Integrated weighted gene co-expression network analysis with an application to chronic fatigue syndrome. BMC Syst Biol 2: 95.AP PressonEM SobelJC PappCJ SuarezT. Whistler2008Integrated weighted gene co-expression network analysis with an application to chronic fatigue syndrome.BMC Syst Biol295
- 32. Chaibub Neto E, Ferrara CT, Attie AD, Yandell BS (2008) Inferring causal phenotype networks from segregating populations. Genetics 179: 1089–1100.E. Chaibub NetoCT FerraraAD AttieBS Yandell2008Inferring causal phenotype networks from segregating populations.Genetics17910891100
- 33. Schadt EE, Lamb J, Yang X, Zhu J, Edwards S, et al. (2005) An integrative genomics approach to infer causal associations between gene expression and disease. Nat Genet 37: 710–717.EE SchadtJ. LambX. YangJ. ZhuS. Edwards2005An integrative genomics approach to infer causal associations between gene expression and disease.Nat Genet37710717
- 34. Schafer J, Strimmer K (2005) An empirical Bayes approach to inferring large-scale gene association networks. Bioinformatics 21: 754–764.J. SchaferK. Strimmer2005An empirical Bayes approach to inferring large-scale gene association networks.Bioinformatics21754764
- 35. Smith GD, Ebrahim S (2004) Mendelian randomization: prospects, potentials, and limitations. Int J Epidemiol 33: 30–42.GD SmithS. Ebrahim2004Mendelian randomization: prospects, potentials, and limitations.Int J Epidemiol333042
- 36. Little J, Khoury MJ (2003) Mendelian randomisation: a new spin or real progress? Lancet 362: 930–931.J. LittleMJ Khoury2003Mendelian randomisation: a new spin or real progress?Lancet362930931
- 37. Aten JE, Fuller TF, Lusis AJ, Horvath S (2008) Using genetic markers to orient the edges in quantitative trait networks: the NEO software. BMC Syst Biol 2: 34.JE AtenTF FullerAJ LusisS. Horvath2008Using genetic markers to orient the edges in quantitative trait networks: the NEO software.BMC Syst Biol234
- 38. Kathiresan S, Willer CJ, Peloso GM, Demissie S, Musunuru K, et al. (2009) Common variants at 30 loci contribute to polygenic dyslipidemia. Nat Genet 41: 56–65.S. KathiresanCJ WillerGM PelosoS. DemissieK. Musunuru2009Common variants at 30 loci contribute to polygenic dyslipidemia.Nat Genet415665
- 39. Grieve IC, Dickens NJ, Pravenec M, Kren V, Hubner N, et al. (2008) Genome-wide co-expression analysis in multiple tissues. PLoS ONE 3: e4033.IC GrieveNJ DickensM. PravenecV. KrenN. Hubner2008Genome-wide co-expression analysis in multiple tissues.PLoS ONE3e4033
- 40. Ala U, Piro RM, Grassi E, Damasco C, Silengo L, et al. (2008) Prediction of human disease genes by human-mouse conserved coexpression analysis. PLoS Comput Biol 4: e1000043.U. AlaRM PiroE. GrassiC. DamascoL. Silengo2008Prediction of human disease genes by human-mouse conserved coexpression analysis.PLoS Comput Biol4e1000043
- 41. Keller MP, Choi Y, Wang P, Davis DB, Rabaglia ME, et al. (2008) A gene expression network model of type 2 diabetes links cell cycle regulation in islets with diabetes susceptibility. Genome Res 18: 706–716.MP KellerY. ChoiP. WangDB DavisME Rabaglia2008A gene expression network model of type 2 diabetes links cell cycle regulation in islets with diabetes susceptibility.Genome Res18706716
- 42. Saxena V, Orgill D, Kohane I (2007) A set of genes previously implicated in the hypoxia response might be an important modulator in the rat ear tissue response to mechanical stretch. BMC Genomics 8: 430.V. SaxenaD. OrgillI. Kohane2007A set of genes previously implicated in the hypoxia response might be an important modulator in the rat ear tissue response to mechanical stretch.BMC Genomics8430
- 43. Wei H, Persson S, Mehta T, Srinivasasainagendra V, Chen L, et al. (2006) Transcriptional coordination of the metabolic network in Arabidopsis. Plant Physiol 142: 762–774.H. WeiS. PerssonT. MehtaV. SrinivasasainagendraL. Chen2006Transcriptional coordination of the metabolic network in Arabidopsis.Plant Physiol142762774
- 44. Li H, Chen H, Bao L, Manly KF, Chesler EJ, et al. (2006) Integrative genetic analysis of transcription modules: towards filling the gap between genetic loci and inherited traits. Hum Mol Genet 15: 481–492.H. LiH. ChenL. BaoKF ManlyEJ Chesler2006Integrative genetic analysis of transcription modules: towards filling the gap between genetic loci and inherited traits.Hum Mol Genet15481492
- 45. Stuart JM, Segal E, Koller D, Kim SK (2003) A gene-coexpression network for global discovery of conserved genetic modules. Science 302: 249–255.JM StuartE. SegalD. KollerSK Kim2003A gene-coexpression network for global discovery of conserved genetic modules.Science302249255
- 46. Zhou X, Kao MC, Wong WH (2002) Transitive functional annotation by shortest-path analysis of gene expression data. Proc Natl Acad Sci U S A 99: 12783–12788.X. ZhouMC KaoWH Wong2002Transitive functional annotation by shortest-path analysis of gene expression data.Proc Natl Acad Sci U S A991278312788
- 47. Gorgels TG, Hu X, Scheffer GL, van der Wal AC, Toonstra J, et al. (2005) Disruption of Abcc6 in the mouse: novel insight in the pathogenesis of pseudoxanthoma elasticum. Hum Mol Genet 14: 1763–1773.TG GorgelsX. HuGL SchefferAC van der WalJ. Toonstra2005Disruption of Abcc6 in the mouse: novel insight in the pathogenesis of pseudoxanthoma elasticum.Hum Mol Genet1417631773
- 48. Garofalo RS, Orena SJ, Rafidi K, Torchia AJ, Stock JL, et al. (2003) Severe diabetes, age-dependent loss of adipose tissue, and mild growth deficiency in mice lacking Akt2/PKB beta. J Clin Invest 112: 197–208.RS GarofaloSJ OrenaK. RafidiAJ TorchiaJL Stock2003Severe diabetes, age-dependent loss of adipose tissue, and mild growth deficiency in mice lacking Akt2/PKB beta.J Clin Invest112197208
- 49. Masuzaki H, Paterson J, Shinyama H, Morton NM, Mullins JJ, et al. (2001) A transgenic model of visceral obesity and the metabolic syndrome. Science 294: 2166–2170.H. MasuzakiJ. PatersonH. ShinyamaNM MortonJJ Mullins2001A transgenic model of visceral obesity and the metabolic syndrome.Science29421662170
- 50. Sabatti C, Service SK, Hartikainen AL, Pouta A, Ripatti S, et al. (2009) Genome-wide association analysis of metabolic traits in a birth cohort from a founder population. Nat Genet 41: 35–46.C. SabattiSK ServiceAL HartikainenA. PoutaS. Ripatti2009Genome-wide association analysis of metabolic traits in a birth cohort from a founder population.Nat Genet413546
- 51. Aulchenko YS, Ripatti S, Lindqvist I, Boomsma D, Heid IM, et al. (2009) Loci influencing lipid levels and coronary heart disease risk in 16 European population cohorts. Nat Genet 41: 47–55.YS AulchenkoS. RipattiI. LindqvistD. BoomsmaIM Heid2009Loci influencing lipid levels and coronary heart disease risk in 16 European population cohorts.Nat Genet414755
- 52. Malerba G, Schaeffer L, Xumerle L, Klopp N, Trabetti E, et al. (2008) SNPs of the FADS gene cluster are associated with polyunsaturated fatty acids in a cohort of patients with cardiovascular disease. Lipids 43: 289–299.G. MalerbaL. SchaefferL. XumerleN. KloppE. Trabetti2008SNPs of the FADS gene cluster are associated with polyunsaturated fatty acids in a cohort of patients with cardiovascular disease.Lipids43289299
- 53. Tanaka T, Shen J, Abecasis GR, Kisialiou A, Ordovas JM, et al. (2009) Genome-wide association study of plasma polyunsaturated fatty acids in the InCHIANTI Study. PLoS Genet 5: e1000338.T. TanakaJ. ShenGR AbecasisA. KisialiouJM Ordovas2009Genome-wide association study of plasma polyunsaturated fatty acids in the InCHIANTI Study.PLoS Genet5e1000338
- 54. Schaeffer L, Gohlke H, Muller M, Heid IM, Palmer LJ, et al. (2006) Common genetic variants of the FADS1 FADS2 gene cluster and their reconstructed haplotypes are associated with the fatty acid composition in phospholipids. Hum Mol Genet 15: 1745–1756.L. SchaefferH. GohlkeM. MullerIM HeidLJ Palmer2006Common genetic variants of the FADS1 FADS2 gene cluster and their reconstructed haplotypes are associated with the fatty acid composition in phospholipids.Hum Mol Genet1517451756
- 55. Martinelli N, Girelli D, Malerba G, Guarini P, Illig T, et al. (2008) FADS genotypes and desaturase activity estimated by the ratio of arachidonic acid to linoleic acid are associated with inflammation and coronary artery disease. Am J Clin Nutr 88: 941–949.N. MartinelliD. GirelliG. MalerbaP. GuariniT. Illig2008FADS genotypes and desaturase activity estimated by the ratio of arachidonic acid to linoleic acid are associated with inflammation and coronary artery disease.Am J Clin Nutr88941949
- 56. Rzehak P, Heinrich J, Klopp N, Schaeffer L, Hoff S, et al. (2009) Evidence for an association between genetic variants of the fatty acid desaturase 1 fatty acid desaturase 2 ( FADS1 FADS2) gene cluster and the fatty acid composition of erythrocyte membranes. Br J Nutr 101: 20–26.P. RzehakJ. HeinrichN. KloppL. SchaefferS. Hoff2009Evidence for an association between genetic variants of the fatty acid desaturase 1 fatty acid desaturase 2 ( FADS1 FADS2) gene cluster and the fatty acid composition of erythrocyte membranes.Br J Nutr1012026
- 57. Baylin A, Ruiz-Narvaez E, Kraft P, Campos H (2007) alpha-Linolenic acid, Delta6-desaturase gene polymorphism, and the risk of nonfatal myocardial infarction. Am J Clin Nutr 85: 554–560.A. BaylinE. Ruiz-NarvaezP. KraftH. Campos2007alpha-Linolenic acid, Delta6-desaturase gene polymorphism, and the risk of nonfatal myocardial infarction.Am J Clin Nutr85554560
- 58. Cho HP, Nakamura M, Clarke SD (1999) Cloning, expression, and fatty acid regulation of the human delta-5 desaturase. J Biol Chem 274: 37335–37339.HP ChoM. NakamuraSD Clarke1999Cloning, expression, and fatty acid regulation of the human delta-5 desaturase.J Biol Chem2743733537339
- 59. Cho HP, Nakamura MT, Clarke SD (1999) Cloning, expression, and nutritional regulation of the mammalian Delta-6 desaturase. J Biol Chem 274: 471–477.HP ChoMT NakamuraSD Clarke1999Cloning, expression, and nutritional regulation of the mammalian Delta-6 desaturase.J Biol Chem274471477
- 60. Marquardt A, Stohr H, White K, Weber BH (2000) cDNA cloning, genomic structure, and chromosomal localization of three members of the human fatty acid desaturase family. Genomics 66: 175–183.A. MarquardtH. StohrK. WhiteBH Weber2000cDNA cloning, genomic structure, and chromosomal localization of three members of the human fatty acid desaturase family.Genomics66175183
- 61. Zhang J, Finney RP, Clifford RJ, Derr LK, Buetow KH (2005) Detecting false expression signals in high-density oligonucleotide arrays by an in silico approach. Genomics 85: 297–308.J. ZhangRP FinneyRJ CliffordLK DerrKH Buetow2005Detecting false expression signals in high-density oligonucleotide arrays by an in silico approach.Genomics85297308
- 62. Johnson WE, Li C, Rabinovic A (2007) Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8: 118–127.WE JohnsonC. LiA. Rabinovic2007Adjusting batch effects in microarray expression data using empirical Bayes methods.Biostatistics8118127
- 63. Sherman BT, Huang da W, Tan Q, Guo Y, Bour S, et al. (2007) DAVID Knowledgebase: a gene-centered database integrating heterogeneous gene annotation resources to facilitate high-throughput gene functional analysis. BMC Bioinformatics 8: 426.BT ShermanW. Huang daQ. TanY. GuoS. Bour2007DAVID Knowledgebase: a gene-centered database integrating heterogeneous gene annotation resources to facilitate high-throughput gene functional analysis.BMC Bioinformatics8426
- 64. Langfelder P, Horvath S (2008) WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9: 559.P. LangfelderS. Horvath2008WGCNA: an R package for weighted correlation network analysis.BMC Bioinformatics9559
- 65. Zhang B, Horvath S (2005) A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4: Article17.B. ZhangS. Horvath2005A general framework for weighted gene co-expression network analysis.Stat Appl Genet Mol Biol4Article17
- 66. Langfelder P, Zhang B, Horvath S (2008) Defining clusters from a hierarchical cluster tree: the Dynamic Tree Cut package for R. Bioinformatics 24: 719–720.P. LangfelderB. ZhangS. Horvath2008Defining clusters from a hierarchical cluster tree: the Dynamic Tree Cut package for R.Bioinformatics24719720
- 67. Rozen S, Skaletsky H (2000) Primer3 on the WWW for general users and for biologist programmers. Methods Mol Biol 132: 365–386.S. RozenH. Skaletsky2000Primer3 on the WWW for general users and for biologist programmers.Methods Mol Biol132365386
- 68. Horvath S, Xu X, Laird NM (2001) The family based association test method: strategies for studying general genotype–phenotype associations. Eur J Hum Genet 9: 301–306.S. HorvathX. XuNM Laird2001The family based association test method: strategies for studying general genotype–phenotype associations.Eur J Hum Genet9301306
- 69. Spielman RS, McGinnis RE, Ewens WJ (1993) Transmission test for linkage disequilibrium: the insulin gene region and insulin-dependent diabetes mellitus (IDDM). Am J Hum Genet 52: 506–516.RS SpielmanRE McGinnisWJ Ewens1993Transmission test for linkage disequilibrium: the insulin gene region and insulin-dependent diabetes mellitus (IDDM).Am J Hum Genet52506516
- 70. Laird NM, Lange C (2006) Family-based designs in the age of large-scale gene-association studies. Nat Rev Genet 7: 385–394.NM LairdC. Lange2006Family-based designs in the age of large-scale gene-association studies.Nat Rev Genet7385394