Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Tissue-specific impact of FADS cluster variants on FADS1 and FADS2 gene expression

  • Lindsay M. Reynolds,

    Roles Conceptualization, Data curation, Investigation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Epidemiology & Prevention, Wake Forest School of Medicine, Winston-Salem, North Carolina, United States of America

  • Timothy D. Howard,

    Roles Conceptualization, Writing – original draft

    Affiliation Department of Biochemistry, Wake Forest School of Medicine, Winston-Salem, North Carolina, United States of America

  • Ingo Ruczinski,

    Roles Conceptualization, Investigation, Writing – review & editing

    Affiliation Division of Allergy and Clinical Immunology, Department of Medicine, Johns Hopkins University, Baltimore, Maryland, United States of America

  • Kanika Kanchan,

    Roles Visualization

    Affiliation Division of Allergy and Clinical Immunology, Department of Medicine, Johns Hopkins University, Baltimore, Maryland, United States of America

  • Michael C. Seeds,

    Roles Writing – review & editing

    Affiliation Department of Internal Medicine/Section on Molecular Medicine, Wake Forest School of Medicine, Winston-Salem, North Carolina, United States of America

  • Rasika A. Mathias,

    Roles Conceptualization, Investigation, Visualization, Writing – review & editing

    Affiliation Division of Allergy and Clinical Immunology, Department of Medicine, Johns Hopkins University, Baltimore, Maryland, United States of America

  • Floyd H. Chilton

    Roles Conceptualization, Investigation, Supervision, Writing – original draft, Writing – review & editing

    schilton@wakehealth.edu

    Affiliation Department of Physiology and Pharmacology, Wake Forest School of Medicine, Winston-Salem, North Carolina, United States of America

Tissue-specific impact of FADS cluster variants on FADS1 and FADS2 gene expression

  • Lindsay M. Reynolds, 
  • Timothy D. Howard, 
  • Ingo Ruczinski, 
  • Kanika Kanchan, 
  • Michael C. Seeds, 
  • Rasika A. Mathias, 
  • Floyd H. Chilton
PLOS
x

Abstract

Omega-6 (n-6) and omega-3 (n-3) long (≥ 20 carbon) chain polyunsaturated fatty acids (LC-PUFAs) play a critical role in human health and disease. Biosynthesis of LC-PUFAs from dietary 18 carbon PUFAs in tissues such as the liver is highly associated with genetic variation within the fatty acid desaturase (FADS) gene cluster, containing FADS1 and FADS2 that encode the rate-limiting desaturation enzymes in the LC-PUFA biosynthesis pathway. However, the molecular mechanisms by which FADS genetic variants affect LC-PUFA biosynthesis, and in which tissues, are unclear. The current study examined associations between common single nucleotide polymorphisms (SNPs) within the FADS gene cluster and FADS1 and FADS2 gene expression in 44 different human tissues (sample sizes ranging 70–361) from the Genotype-Tissue Expression (GTEx) Project. FADS1 and FADS2 expression were detected in all 44 tissues. Significant cis-eQTLs (within 1 megabase of each gene, False Discovery Rate, FDR<0.05, as defined by GTEx) were identified in 12 tissues for FADS1 gene expression and 23 tissues for FADS2 gene expression. Six tissues had significant (FDR< 0.05) eQTLs associated with both FADS1 and FADS2 (including artery, esophagus, heart, muscle, nerve, and thyroid). Interestingly, the identified eQTLs were consistently found to be associated in opposite directions for FADS1 and FADS2 expression. Taken together, findings from this study suggest common SNPs within the FADS gene cluster impact the transcription of FADS1 and FADS2 in numerous tissues and raise important questions about how the inverse expression of these two genes impact intermediate molecular (such a LC-PUFA and LC-PUFA-containing glycerolipid levels) and ultimately clinical phenotypes associated with inflammatory diseases and brain health.

Introduction

Landmark studies by Burr and Burr almost 100 years ago demonstrated that dietary eighteen carbon (18C-) polyunsaturated fatty acids (PUFAs), linoleic acid (LA, 18:2n-6) and alpha-linolenic (ALA, 18:3n-3), are essential to health [1, 2]. Later it was discovered that these 18C-PUFAs are substrates that enter omega-6 (n-6) and omega-3 (n-3) long chain (LC-) PUFA biosynthetic pathways resulting in the formation of biologically-active LC-PUFAs. In two parallel and competing pathways, desaturase and elongase enzymes convert n-6 or n-3 PUFAs into several LC-PUFAs. In the n-6 arm of the pathway, arachidonic acid (ARA, 20:4n-6) is synthesized from LA, utilizing 2 desaturation and 1 elongation enzymatic steps encoded by FADS2, FADS1 and ELOVL5, respectively [35]. The n-3 LC-PUFA, eicosapentaenoic acid (EPA, 20:5n-3) is synthesized from ALA employing the same three enzymatic steps. Twenty-two carbon n-6 and n-3 LC-PUFAs, adrenic acid (ADA, 22:4n-6) and docosapentaenoic acid (DPA, 22:5n-3) can then be formed utilizing an additional elongation step (encoded by ELOVL2), and docosahexaenoic acid (DHA, 22:6n-3) synthesized from DPA with another desaturation step (encoded by FADS2) or three additional biosynthetic steps (2 elongation, 1 desaturation and 1 β-oxidation) [69]. Once formed, LC-PUFAs are transported to cells and tissues in circulation as free fatty acids bound to albumin or esterified to complex lipids such as phospholipids (PL), cholesterol esters, and triglycerides in lipoprotein particles [10].

Until recently, it was assumed that the metabolic capacity of the LC-PUFA biosynthetic pathway was fairly uniform for all humans. This premise was supported by stable isotope, metabolic studies carried out in European ancestry populations, which indicated that small proportions of ingested dietary 18-carbon PUFAs are converted into LC-PUFAs [11]. The desaturase enzymes within the pathway, encoded by the three genes (FADS1, FADS2, and FADS3) in the FADS gene cluster region (chr11:61,540,615–61,664,170, hg19) have long been recognized as the rate-limiting steps in the conversion of 18C-PUFAs to LC-PUFAs [3]. Studies over the past decade have challenged the concept that biosynthesis of LC-PUFAs is uniform among individuals and populations by demonstrating that common genetic and epigenetic variation in and near FADS1 and FADS2 are strongly associated with the levels of LC-PUFAs found in circulation, blood cells, and tissues [4, 1221]. Moreover, the frequency of these variants and thus the capacity to synthesize LC-PUFAs can differ dramatically across diverse human populations [15, 16]. In addition to LC-PUFA levels themselves, genetic variation in the FADS cluster is associated with numerous human phenotypes, including inflammatory [22] and cardiovascular disorders [21, 23], blood lipid levels including low-density lipoprotein (LDL) and triglyceride levels [17, 2427], coronary artery disease [23, 28, 29], insulin resistance [30], perinatal depression [31], atopic diseases [3234], attention deficit disorder/hyperactivity, intelligence and memory in children [35, 36]. Taken together, these findings support a relationship between genetic variations in the FADS gene cluster and several human diseases that may involve tissue-specific levels of LC-PUFAs.

It has also been generally presumed that the liver is the primary organ responsible for the biosynthesis of LC-PUFAs that are released into circulation and eventually incorporated into tissues. However, other cells and tissues are known to have their own endogenous capacity to synthesize LC-PUFAs from dietary 18C-PUFAs [3739]. While associations between genetic variants in FADS1 and FADS2 and LC-PUFA levels, as well as desaturase activities and certain phenotypes have been established, little is known about the range of tissues most capable of LC-PUFA biosynthesis and the molecular mechanisms by which these FADS polymorphisms exert their effects. Expression quantitative trait loci (eQTL) mapping is a powerful approach to examine specific genetic variation(s) associations with gene expression levels, but eQTLs are typically examined in only a few accessible tissues. This creates challenges in understanding how genetic variants may influence gene expression levels and in this case LC-PUFA biosynthesis across a wide range of tissues. The Genotype-Tissue Expression (GTEx) Project was designed to address many of these limitations by providing a platform to examine the relationships between genetic variation and gene expression across a wide range of human tissues [40]. The current study has leveraged the publicly available GTEx data to better understand which tissues have the capacity to express FADS cluster genes and to better understand the potential for tissue-specific influences of genetic variation in the FADS cluster on FADS gene expression and PUFA desaturation, by examining the associations between FADS cluster single nucleotide polymorphisms (SNPs) and FADS1 and FADS2 expression in 44 tissues.

Methods

FADS cluster gene expression and eQTL analysis in GTEx

The GTEx Project [40] data portal was used to investigate FADS1 and FADS2 gene expression across human tissues and genetic influences on FADS1 and FADS2 expression across 44 tissues (data source: GTEx data version V6p). As described previously, GTEx data includes gene expression measured in various tissues collected during autopsies within 24 hours of death for 572 individuals (34.4% female; 84.3% White, 13.7% African American, 1% Asian, 1% not reported). GTEx version V6p eQTL analysis utilized gene expression measured using RNA-sequencing (Illumina TrueSeq) for samples with an RNA Integrity Number (RIN, as measured by Agilent Bioanalyzer) of 6.0 or higher. GTEx performed the following quality control measures for all results accessed through the data portal. Gene expression levels passed the quality control thresholds of >0.1 reads per kilobase of transcript per million (RPKM) in at least 10 individuals and ≥6 reads in at least 10 individuals. Expression values were quantile normalized to the average empirical distribution across samples. For each gene, expression values were inverse quantile normalized to a standard normal distribution across samples. Outliers were identified and excluded using a correlation-based statistic similar to methods previously described [41]. GTEx genotyped single nucleotide polymorphisms (SNPs) using whole genome sequencing (HiSeq X; first batch on HiSeq 2000), whole exome sequencing (Agilent or ICE target capture, HiSeq 2000), or microarray (Illumina OMNI 5M Array or 2.5M SNP Array, or Illumina Human Exome SNP Array) from blood samples, and imputed using 1000 Genomes Project Phase I, version 3 for 449 donors. A call rate threshold of 95% and a minor allele frequency (MAF) > = 1% (a tissue specific cutoff, as sample sets vary by tissue) was used.

Publicly available eQTL results and trait expression levels were downloaded from the GTEx Project data portal [40]. Briefly, GTEx generated the association results using linear regression, associating normalized log-transformed gene expression levels with SNPs. A threshold of ≥ 70 samples per tissue was used for eQTL analysis, resulting in testing of 44 tissues for eQTL analysis. Association analyses were adjusted for sex, genotyping platform, and the top three genotyping-based principal components to adjust for potential population stratification by race/ethnicity background. The effect size of the eQTLs was defined as the slope of the linear regression, comparing the alternative allele (ALT) to the reference allele (REF) in the human genome reference GRCh37/hg19. GTEx used a false discovery rate (FDR) < 0.05 to correct for multiple hypothesis testing [42]. This paper includes the publicly available GTEx eQTL results from 7,161 SNPs located within one megabase of FADS1 and FADS2 transcription start sites.

In silico functional analysis

RegulomeDB [43] version 1.1 was used to prioritize the FADS cluster eQTLs identified by GTEx data, based on predicted function. RegulomeDB in silico prediction utilizes known and predicted DNA regulatory elements from public datasets including the NCBI Gene Expression Omnibus (GEO), the ENCODE project [44], and published literature. Regulatory regions predictions are based on previously identified overlapping features characterized to regulate gene expression, including DNAase hypersensitive sites, transcription factor binding sites, and promoter/enhancer regions.

Results

FADS1 and FADS2 gene expression

FADS1 and FADS2 mRNA expression were detected (at an expression threshold of >0.1 RPKM in at least 10 individuals and ≥6 reads in at least 10 individuals) in all 44 human tissues found in GTEx (Fig 1). Median gene expression levels across tissues were highest in the adrenal gland and brain tissues for both FADS1 and FADS2. Overall, FADS2 was expressed at higher median levels than FADS1, with FADS1 ranging from 0.89 RPKM in whole blood to 16.2 RPKM in adrenal gland tissue, and FADS2 ranging from 0.89 RPKM in skeletal muscle to 112.7 RPKM in adrenal gland tissue.

thumbnail
Fig 1. FADS1 and FADS2 expression across 44 human tissues.

Boxplots of FADS1 (top) and FADS2 (bottom) mRNA expression (log10RPKM—y axis) are shown across tissues (x-axis) from GTEx; outliers are shown as circles; median expression is represented as the center line.

https://doi.org/10.1371/journal.pone.0194610.g001

FADS1 and FADS2 eQTL analysis

Overall, 27 of the 44 tissues from GTEx contained significant cis-eQTLs for either FADS1 or FADS2 after adjusting for multiple testing (FDR<0.05), as depicted in Fig 2. For FADS1, 167 SNPs were identified as significant (FDR<0.05) cis-eQTLs, which were found in at least one of 12 tissues, including artery (tibial), brain (cerebellar hemisphere, cerebellum, putamen basal ganglia), esophagus (Mucosa), heart (left ventricle), muscle (skeletal), nerve (tibial), pancreas, stomach, testis, and thyroid tissues. For FADS2, 233 SNPs were identified as significant (FDR<0.05) cis-eQTLs, which were identified in at least one of 23 tissues, including adipose (subcutaneous, visceral omentum), artery (aorta, tibial), breast (mammary tissue), transformed fibroblasts, colon (sigmoid, transverse), esophagus (gastroesophageal junction, mucosa, muscularis), heart (atrial appendage, left ventricle), lung, muscle (skeletal), nerve (tibial), pancreas, small intestine (terminal ileum), spleen, stomach, testis, thyroid, and whole blood. The 233 SNPs found to be FADS2 eQTLs included 146 SNPs that were also FADS1 eQTLs, of which 73 SNPs were significantly (FDR<0.05, p≤ 8.30x10-5) associated with both expression of FADS1 and FADS2 in same tissue (in at least one of the six following tissues: artery (tibial), esophagus (mucosa), heart (left ventricle), muscle (skeletal), nerve (tibial), and thyroid, see S1 Table). Surprisingly, all of these SNPs were associated with FADS1 and FADS2 expression in opposite directions as depicted by the blue and red colors in Fig 2.

thumbnail
Fig 2. FADS1 and FADS2 cis-eQTLs.

Significant eQTLS (FDR<0.05) associated with gene expression of FADS1 (A) or FADS2 (B) identified in 27 human tissues from GTEx. The significance of the detected association is represented as the size of the circle (larger the circle, more significant the result), and the color represents the direction of effect of the alternate (alt) allele (red for increased transcription with alt allele, blue for decreased transcription). FADS1 and FADS2 gene transcription start sites (TSS) and transcription end sites (TES) illustrate the location of FADS1 and FADS2 genes relative to the cis-eQTLs.

https://doi.org/10.1371/journal.pone.0194610.g002

In silico predicted functional effects of FADS1 and FADS2 eQTLs

Identification of predicted regulatory features for gene expression (from ENCODE[44] and RegulomeDB [43]) such as promoters, enhancers, insulators, DNase hypersensitive regions, and transcription factor binding sites (TFBS) overlapping the identified eQTLs provided additional evidence supporting the potential functional effects of ~30 of the 73 eQTLs for both FADS1 and FADS2. (S2 Table and S1 Fig). High linkage disequilibrium (LD) was observed between most of these potentially functional variants (r2≥0.8), as shown in S1 Fig. Five of the predicted functional eQTLs were not in high LD with other eQTLs (rs61896141, rs968567, rs61897793, rs174575, and rs174627). Six of the potential functional SNPs were previously identified as eQTLs for FADS2 in monocytes (rs174534, rs174547, rs174549, rs174548, rs174583, and rs174577) [45]. All six of these SNPs overlap transcription factor binding sites, three of which also lie in DNase hypersensitive regions (rs174534, rs174547, and rs174583). These six SNPs are in moderate to high LD (ranging from r2 = 0.58 between rs174548 and rs174534 to r2 = 0.97 between rs174548 and rs174549). Only one of these SNPs (rs174548) was previously annotated as a FADS1 eQTL (in liver cells [46]) in RegulomeDB.

Rs174548 (chr 11:61,571,348, hg19) was chosen as a representative candidate functional eQTL for FADS1 and FADS2 for a number of reasons, including that it was among the most significant eQTLs for FADS1 and FADS2 in GTEx. Additionally, rs174548 had a RegulomeDB score of 1f, indicating that it was likely to affect protein binding and was previously linked to expression of genes, including FADS1 (in liver [46]), FADS2 (in monocytes [45]), and FADS3 (in liver [46]). Rs174548 also overlaps a binding site for RNA polymerase II subunit A (POLR2A in B cell line) and has been previously demonstrated to be associated with numerous traits, including plasma levels of dihomogamma-linolenic acid, DGLA, 20:3n-6, p = 7 x 10−31 [47], blood levels of ARA (p = 1 x 10−84) [48], triglycerides (p = 5 x 10−14) [49], high density lipoprotein (HDL) cholesterol levels (p = 1 x 10−12) [50], and ratios of ARA/DGLA (p = 2 x 10−361) in blood [48].

Similar to other potentially functional SNPs in high LD (such as rs174560, r2 = 0.99), rs174548 was associated with FADS1 and FADS2 expression with opposite effect directions in numerous tissues. Fig 3 illustrates the effect size and 95% confidence intervals for the rs174548 eQTL (alternative allele G, reference allele C) across tissues for FADS1 and FADS2. The strongest associations between rs174548 and FADS1 were identified in pancreas (beta = -0.71, p = 4.5x10-15) and brain tissue (beta = -0.62, p = 6.1x10-10), with lower FADS1 expression associated with increased copies of the minor allele (G, minor allele frequency = 0.30 in HAPMAP CEU). Notably, unlike brain tissue, FADS1 was not highly expressed in pancreas tissue (Fig 1). Increased copies of the rs174548 G allele was also nominally (p<0.05) associated with lower FADS1 expression in tissues thought to be particularly relevant to LC-PUFA biosynthesis, including adipose (subcutaneous, beta = -0.22, p = 2.2x10-4) and liver (beta = -0.36, p = 3.3x10-3). In contrast, the rs174548 G allele was nominally associated with higher FADS2 expression in adipose tissue (subcutaneous, beta = 0.15, p = 7x10-3), but was not significantly associated with FADS2 expression in brain or liver. The rs174548 G allele was also associated with increasing FADS2 expression in other tissues, including artery, breast, colon, esophagus, heart, lung, muscle, nerve, ovary, skin (sun exposed), stomach, testis, and thyroid tissues. Of note, the only tissue with similar effect directions observed for rs174548 associations with FADS1 and FADS2 was whole blood, where the G allele was associated with higher expression of both FADS1 and FADS2; however, neither FADS1 nor FADS2 were highly expressed in whole blood (Fig 1). Also, although the association between rs174548 and FADS1 was nominally significant in tissues such as whole blood and subcutaneous adipose tissue, the associations did not reach the multiple testing correction threshold (FDR<0.05, p≤ 8.30x10-5).

thumbnail
Fig 3. rs174548 genotype associations with FADS1 and FADS2 across tissues.

Nominal associations (effect size and 95% confidence interval) between rs174548 genotypes (G, alt allele vs C, ref allele) with FADS1 (A) and FADS2 (B) are shown across GTEx tissues. The effect size of the eQTLs was defined as the slope of the linear regression, comparing the alt allele (G) to the reference allele (C).

https://doi.org/10.1371/journal.pone.0194610.g003

Associations between rs174548 and both FADS1 and FADS2 were significant at the multiple testing correction threshold (FDR<0.05, p≤ 8.30x10-5) in five tissues. Specifically, the G allele of rs174548 was associated with lower expression of FADS1 and higher expression of FADS2 in all five tissues. Tissues included artery (tibial), heart (left ventricle), muscle (skeletal), nerve (tibial), and thyroid (Fig 4 and S3 Table).

thumbnail
Fig 4. FADS1 and FADS2 expression by rs174548 genotype.

FADS1 (A) FADS2 (B) rank normalized gene expression (Y-axis) by rs174548 genotype (G: alt allele, C: ref allele) in the five tissues with significant (FDR<0.05) associations between rs174548 and FADS1 and FADS2 in the same tissues.

https://doi.org/10.1371/journal.pone.0194610.g004

Discussion

Candidate gene-based studies and GWAS have revealed that SNPs throughout the FADS cluster are associated with LC-PUFA levels, 18C-PUFA to LC-PUFA conversion efficiencies, as well as numerous disease-related molecular phenotypes and the risk of human diseases [5, 21, 22, 2830, 4749, 51, 52]. However, determining causative variants, molecular mechanisms, and relevant tissues contributing to these associations has been difficult, due in large part to extensive LD throughout the FADS cluster and the capacity of numerous tissues to synthesize LC-PUFAs. The liver has long been assumed to be the primary tissue that synthesizes LC-PUFAs found in circulation. We and others have demonstrated that in the liver, several genetic and epigenetic variants within and near the FADS cluster are associated with LC-PUFA levels [53, 54]. Wang and colleagues demonstrated that six SNPs within the FADS cluster (FADS1, FADS2, and FADS3) were consistently and significantly associated with FADS1 gene expression in liver, but not with expression of FADS2 or FADS3, with rs174556 exerting the highest correlation[50]. Pan and colleagues systematically tested SNPs in putative regulatory regions in the FADS cluster using a luciferase reporter assay and found a SNP in FADS1 (rs174557) in an Alu element that they postulated affects FADS1 enhancer activity in a liver cell line (HepG2) by altering competitive transcription factor binding between the suppressive PATZ1 and the activating complex of SP1/SREBP1c [55]. These studies suggest that FADS1 gene expression plays a causal role in the previously observed associations between FADS cluster SNPs and LC-PUFA biosynthesis. While these studies are clearly important initial steps in determining putative regulatory SNPs that impact FADS1 expression, they are limited by the fact that they have not examined these eQTLs across a wide range of tissues.

To better clarify the relationship between genetic polymorphisms identified by GWAS and disease etiology, GTEx has investigated the relationships between genetic polymorphisms and gene expression in several organs and tissues. The current study utilized publicly available results generated by GTEx to examine the relationship between genetic variation within or near the FADS cluster and FADS1 and FADS2 gene expression in 44 tissues. There were several interesting and surprising observations from this analysis. As expected, the levels of FADS1 and FADS2 expression, and the association of FADS cluster SNPs, varied greatly among different organs and tissues. While all tissues examined demonstrated FADS1 and FADS2 expression, significant cis-eQTLs for FADS1 or FADS2 were detected in 27 tissues, with several cis-eQTLs common to both FADS1 and FADS2 expression within the same tissue. The most unexpected result from this study was the observation that in tissues with common eQTLs for both FADS1 and FADS2, the variant associations with expression of FADS2 gene expression were in the opposite effect directions as the associations with FADS1 gene expression.

Rs174548 was chosen as a candidate functional eQTL for FADS1 and FADS2 for many reasons, including its predicted functional score from RegulomeDB, and its previous identification as a FADS1 eQTL in liver [46] and a FADS2 eQTL in monocytes [45]. Additionally, it has strong associations with LC-PUFA biosynthetic capacity and other important molecular phenotypes [26, 4749], and is among the most significant eQTLs for FADS1 and FADS2 in GTEx. Rs174548 is also in high LD with many nearby variants predicted to have functional effects on transcription factor binding (by RegulomeDB), including one of the most significant SNPs (rs174556, r2 ~ 0.80) previously associated with FADS1 in liver [50], located ~10 kb away. In GTEx, rs174556 was also an eQTL for both FADS1 and FADS2 expression in multiple tissues (but only FADS1 in liver). The other previously identified FADS1 eQTL in liver (rs174557) that was reported to alter the activity of a SREBP1c inducible FADS1 enhancer [55] was not among the variants captured in GTEx. However, rs174560, which was located on the same 646-bp luciferase construct as rs174557 [55], was among the 73 cis-eQTLs identified for both FADS1 and FADS2 in GTEx in the same tissues, and was in almost perfect LD (r2>0.99) with the candidate functional eQTL chosen here, rs174548.

In tissues thought to be particularly relevant to LC-PUFA biosynthesis such as the liver and brain, the minor allele of rs174548 (G) was associated with lower expression of FADS1, but was not associated with FADS2. The minor allele of rs174548 in subcutaneous adipose tissue was also associated with lower expression of FADS1 but higher expression of FADS2 (with nominal significance, not after multiple testing corrections). A similar pattern was observed in all five tissues (including tibial artery, heart left ventricle, skeletal muscle, tibial nerve, and thyroid tissues) that had significant associations after genome-wide multiple testing corrections between rs174548 and both FADS1 and FADS2, where the minor allele of rs174548 was associated with lower expression of FADS1 and higher expression of FADS2. The only exception to this pattern was the observation in whole blood, which did not reach genome-wide significance, where the minor allele of rs174548 was nominally associated with increased FADS1 and FADS2 expression. However, it is important to point out that whole blood had very low levels of FADS1 and FADS2 expression.

The inverse gene expression patterns of FADS1 and FADS2 with cis-eQTLs is intriguing as these two genes are situated in a “head-to-head” orientation, with intergenic regulatory regions (containing putative promoters and an enhancer) that may affect both genes. This regulatory landscape is relatively common in the genome, and it has been reported that approximately 10% of protein coding transcripts are encoded in this orientation [5659]. Although expression of head-to-head genes is usually positively correlated, genome-wide analysis has shown that negatively correlated expression also occurs [58, 59]. For FADS1 and FADS2, inverse expression may be due to steric hindrance caused by the binding of transcription factors targeting expression of one of the two genes. Given the importance of these genes in sequential steps in this PUFA biosynthetic pathway, it is also conceivable that the genes may alternate between positive and negative correlation, depending on the availability of their substrate. Additionally, allele-specific methylation analysis has revealed that SNPs in the FADS cluster of human liver tissue may interact with the DNA methylation in the common regulatory region between FADS1 and FADS2, and such a mechanism could also impact gene expression [53, 60]. Future investigation will be necessary to better delineate the effects of genetic variation in the FADS gene cluster on the regulatory mechanisms that impact the expression of these two genes.

How does the inverse eQTL pattern for FADS1 and FADS2 affect our understanding of LC-PUFA biosynthesis and related phenotypes? Studies that combine GWAS and metabolomics (in serum or whole blood) have proven very powerful for identifying genes that participate in the synthesis of LC-PUFA and LC-PUFA complex lipids [61]. Certainly, when examining levels of circulating and liver 18C-PUFAs and LC-PUFAs, and precursor-product ratios of the FADS1 and FADS2 reactions, most studies point to the importance of polymorphisms near FADS1. For example, Geiger and colleagues showed that phospholipid metabolites with four double bonds (i.e., ARA) were associated with SNPs in FADS1 [6163] and this effect was observed for all major phospholipid species [61]. Additionally, the association with SNPs in FADS1 increased up to 14-fold when examining ratios of phospholipids that contained FADS1 precursors (DGLA) and products (ARA). The authors point out that this effect is so strong that “if the molecular function of FADS1 had not been already known, the association between the SNP and the different glycerophospholipid concentrations would have allowed one to deduce its enzymatic activity.” Similarly, rs174548 (along with numerous other SNPs in high LD) is strongly associated with levels of ARA (p = 1 x 10−84), but the effect size is increased 4-fold when measuring the FADS1 product to precursor ratio of ARA/DGLA (p = 2 x 10−361) in human blood [48]. A recent study indicated that FADS1 polymorphisms play a key role in hepatic total fat content by modulating levels of LC-PUFA-containing phospholipids and FADS1 expression [50].

Numerous tissues have their own endogenous capacity to synthesize LC-PUFAs, and this study shows that many of these same tissues have eQTLs inversely correlated for FADS1 and FADS2. A major question that arises from this paper is how FADS cluster SNPs impact levels of LC-PUFAs or FADS1 and FADS2 precursor-product ratios in these tissues. We have recently examined prostate cancer tissues from patients undergoing radical prostatectomy and demonstrated that prostate cancer tissues have great capacity to synthesize LC-PUFAs from 18-C PUFAs[39]. The presence of the G allele at FADS SNP rs174537 was significantly associated with higher levels of both n-6 and n-3 LC-PUFAs. Importantly, rs174537 genotypes had much greater effects on the ratio of ARA to DGLA than the effects on ARA alone or precursor-product ratios of FADS2 activity, suggesting that FADS1 activity, expression, or both, are related to genotypic variation in the FADS cluster. However, few studies have examined the relationship between FADS cluster variants and LC-PUFA biosynthesis in other tissues that have endogenous capacities to synthesize LC-PUFAs. It remains unclear how alterations in FADS1 and FADS2 gene expression in difficult to ascertain tissues, such as those identified in this study including the artery, heart, muscle, nerve, and thyroid tissues, may influence LC-PUFA biosynthesis. More studies are necessary to understand the impact of FADS gene cluster eQTLs in inverse directions for FADS1 and FADS2 on LC-PUFA levels in these tissues.

Interpretation of the findings from this study is limited by the combined analysis of both sexes and Caucasian and African American populations. However, alleviating concerns that the FADS1 and FADS2 eQTL results from GTEx were driven by population stratification, all analyses included an adjustment for sex and the top three principal components to adjust for genetic ancestry. Additionally, rs174548 and other potentially functional eQTLs are common variants in both Caucasian and African American populations (MAF in a European population = 0.31 and MAF in an African population = 0.18, from 1000 genomes). Future studies with larger sample sizes will be necessary to confirm the opposite effects of increasing minor allele copies of rs174548 on FADS1 and FADS2 expression across sex- and ethnic-specific strata.

It is important to note that results of this study are limited to the associations between FADS cluster variants and gene expression of FADS1 and FADS2. It remains unclear how the opposing relationships between FADS variants and FADS1 and FADS2 gene expression affect protein expression or LC-PUFA biosynthesis, as changes in gene expression may not translate to changes in enzyme activity. A major strength of the current study is use of large-scale published data to explore the relationship between genetics and FADS1 and FADS2 gene expression in many tissues that are difficult to ascertain. However, because the study leverages gene expression measured in tissues collected over a 24-hour window following death, it is possible that the measured FADS1 and FADS2 expression levels may not accurately reflect expression levels in vivo.

In conclusion, our study for the first time examines the impact of genetic variants in and near the FADS gene cluster on FADS1 and FADS2 expression in 44 tissues. The analysis revealed that 27 tissues contained significant cis-eQTLs for FADS1 or FADS2, with several tissues containing cis-eQTLs for both. For those tissues containing both, FADS1 expression and FADS2 expression were associated with eQTLs in opposite directions. Future studies will be necessary to determine the molecular mechanisms leading to opposite effects of eQTLs on FADS1 and FADS2, and their ultimate impact on LC-PUFA levels and related molecular and clinical phenotypes.

Supporting information

S1 Fig. Regulatory and genomic features surrounding eQTLs for FADS1 and FADS2.

Linkage disequilibrium (r2) is shown (top panel) for 32 eQTLs for both FADS1 and FADS2 which had evidence supporting potential functional effects on gene expression (from ENCODE and RegulomeDB). The bottom panel shows the genomic location of eQTLs on chromosome 11, as well as nearby genes, histone marks indicative of regulatory elements (H3K27ac), transcription factor binding sites, DNase clusters, and other predicted regulatory features (from ChromHMM) such as promoters (red), enhancers (orange), and insulators (blue). eQTLs with high prediction scores for functional effects from RegulomeDB (score = 1) are also indicated, as well as variants previously identified as genome-wide significant loci for GWAS investigated traits.

https://doi.org/10.1371/journal.pone.0194610.s001

(PDF)

S1 Table. 73 SNPs were significantly (FDR<0.05, p≤ 8.30x10-5) associated with both expression of FADS1 and FADS2 in same tissue from GTEx.

https://doi.org/10.1371/journal.pone.0194610.s002

(XLSX)

S2 Table. Evidence for function for 73 eQTLs significantly (FDR<0.05) associated with both FADS1 and FADS2 in same tissues from GTEx.

https://doi.org/10.1371/journal.pone.0194610.s003

(XLSX)

S3 Table. rs174848 (G vs C) associations (FDR<0.05) with both FADS1 and FADS2 in the same tissues from GTEx.

https://doi.org/10.1371/journal.pone.0194610.s004

(XLSX)

Acknowledgments

The data used for the analyses described in this manuscript were obtained from the GTEx Portal on 12/02/16. The GTEx Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health, and by NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS.

References

  1. 1. Burr GO, Burr MM. On the nature and role of the fatty acids essential in nutrition. J Biol Chem. 1930;86:587–621.
  2. 2. Burr GO, Burr MM. A new deficiency disease produced by the rigid exclusion of fat from the diet. J Biol Chem. 1929;82:345–67.
  3. 3. Sprecher H. Biochemistry of essential fatty acids. Prog Lipid Res. 1981;20:13–22. pmid:7342077
  4. 4. Glaser C, Lattka E, Rzehak P, Steer C, Koletzko B. Genetic variation in polyunsaturated fatty acid metabolism and its potential relevance for human development and health. Matern Child Nutr. 2011;7 Suppl 2:27–40.
  5. 5. Chilton FH, Murphy RC, Wilson BA, Sergeant S, Ainsworth H, Seeds MC, et al. Diet-gene interactions and PUFA metabolism: A potential contributor to health disparities and human diseases. Nutrients. 2014;6(5):1993–2022. pmid:24853887
  6. 6. Poisson JP, Dupuy RP, Sarda P, Descomps B, Narce M, Rieu D, et al. Evidence that liver microsomes of human neonates desaturate essential fatty acids. Biochim Biophys Acta. 1993;1167(2):109–13. pmid:8466936
  7. 7. Horrobin DF. Fatty acid metabolism in health and disease: the role of delta-6-desaturase. Am J Clin Nutr. 1993;57(5 Suppl):732S–6S. pmid:8386433
  8. 8. el Boustani S, Causse JE, Descomps B, Monnier L, Mendy F, Crastes de PA. Direct in vivo characterization of delta 5 desaturase activity in humans by deuterium labeling: effect of insulin. Metabolism. 1989;38(4):315–21. pmid:2498608
  9. 9. Park HG, Park WJ, Kothapalli KS, Brenna JT. The fatty acid desaturase 2 (FADS2) gene product catalyzes Delta4 desaturation to yield n-3 docosahexaenoic acid and n-6 docosapentaenoic acid in human cells. FASEB J. 2015;29:3911–9. pmid:26065859
  10. 10. Spector AA. Plasma free fatty acid and lipoproteins as sources of polyunsaturated fatty acid for the brain. J Mol Neurosci. 2001;16(2–3):159–65. pmid:11478370
  11. 11. Pawlosky RJ, Hibbeln JR, Novotny JA, Salem N Jr. Physiological compartmental analysis of alpha-linolenic acid metabolism in adult humans. J Lipid Res. 2001;42(8):1257–65. pmid:11483627
  12. 12. Tanaka T, Shen J, Abecasis GR, Kisialiou A, Ordovas JM, Guralnik JM, et al. Genome-wide association study of plasma polyunsaturated fatty acids in the InCHIANTI Study. PLoS Genet. 2009;5(1):e1000338 pmid:19148276
  13. 13. Schaeffer L, Gohlke H, Muller M, Heid IM, Palmer LJ, Kompauer I, et al. 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. 2006;15(11):1745–56. pmid:16670158
  14. 14. Mathias RA, Fu WQ, Akey JM, Ainsworth HC, Torgerson DG, Ruczinski I, et al. Adaptive evolution of the FADS gene cluster within Africa. PLoS One. 2012;7(9):e44926. pmid:23028684
  15. 15. Mathias RA, Sergeant S, Ruczinski I, Torgerson DG, Hugenschmidt CE, Kubala M, et al. The impact of FADS genetic variants on omega 6 polyunsaturated fatty acid metabolism in African Americans. BMC Genet. 2011;12:50. pmid:21599946
  16. 16. Sergeant S, Hugenschmidt CE, Rudock ME, Ziegler JT, Ivester P, Ainsworth HC, et al. Differences in arachidonic acid levels and fatty acid desaturase (FADS) gene variants in African Americans and European Americans with diabetes or the metabolic syndrome. Brit J Nutr. 2012;107(4):547–55. pmid:21733300
  17. 17. Hong SH, Kwak JH, Paik JK, Chae JS, Lee JH. Association of polymorphisms in FADS gene with age-related changes in serum phospholipid polyunsaturated fatty acids and oxidative stress markers in middle-aged nonobese men. Clin Interv Aging. 2013;8:585–96. pmid:23818766
  18. 18. Lattka E, Koletzko B, Zeilinger S, Hibbeln JR, Klopp N, Ring SM, et al. Umbilical cord PUFA are determined by maternal and child fatty acid desaturase (FADS) genetic variants in the Avon Longitudinal Study of Parents and Children (ALSPAC). Br J Nutr. 2013;109(7):1196–210. pmid:22877655
  19. 19. Koletzko B, Lattka E, Zeilinger S, Illig T, Steer C. Genetic variants of the fatty acid desaturase gene cluster predict amounts of red blood cell docosahexaenoic and other polyunsaturated fatty acids in pregnant women: findings from the Avon Longitudinal Study of Parents and Children. Am J Clin Nutr. 2011;93(1):211–9. pmid:21106917
  20. 20. Mozaffarian D, Kabagambe EK, Johnson CO, Lemaitre RN, Manichaikul A, Sun Q, et al. Genetic loci associated with circulating phospholipid trans fatty acids: a meta-analysis of genome-wide association studies from the CHARGE Consortium. Am J Clin Nutr. 2015;101(2):398–406. pmid:25646338
  21. 21. O’Neill CM, Minihane AM. The impact of fatty acid desaturase genotype on fatty acid status and cardiovascular health in adults. Proc Nutr Soc. 2017;76(1):64–75. pmid:27527582
  22. 22. Mathias RA, Pani V, Chilton FH. Genetic variants in the FADS gene: Implications for dietary recommendations for fatty acid intake. Curr Nutr Rep. 2014;3(2):139–48. pmid:24977108
  23. 23. Hu Y, Li H, Lu L, Manichaikul A, Zhu J, Chen YD, et al. Genome-wide meta-analyses identify novel loci associated with n-3 and n-6 polyunsaturated fatty acid levels in Chinese and European-ancestry populations. Hum Mol Genet. 2016;25(6):1215–24. pmid:26744325
  24. 24. Buckley MT, Racimo F, Allentoft ME, Jensen MK, Jonsson A, Huang H, et al. Selection in europeans on fatty acid desaturases associated with dietary changes. Mol Biol Evol. 2017;34(6):1307–18. pmid:28333262
  25. 25. Standl M, Lattka E, Stach B, Koletzko S, Bauer CP, von BA, et al. FADS1 FADS2 gene cluster, PUFA intake and blood lipids in children: results from the GINIplus and LISAplus studies. PLoS One. 2012;7(5):e37780. pmid:22629455
  26. 26. Lettre G, Palmer CD, Young T, Ejebe KG, Allayee H, Benjamin EJ, et al. Genome-wide association study of coronary heart disease and its risk factors in 8,090 African Americans: the NHLBI CARe Project. PLoS Genet. 2011;7(2):e1001300. pmid:21347282
  27. 27. Kathiresan S, Willer CJ, Peloso GM, Demissie S, Musunuru K, Schadt EE, et al. Common variants at 30 loci contribute to polygenic dyslipidemia. Nat Genet. 2009;41(1):56–65. pmid:19060906
  28. 28. Del Gobbo LC, Imamura F, Aslibekyan S, Marklund M, Virtanen JK, Wennberg M, et al. Omega-3 polyunsaturated fatty acid biomarkers and coronary heart disease: Pooling project of 19 cohort studies. JAMA Intern Med. 2016;176(8):1155–66. pmid:27357102
  29. 29. Li SW, Lin K, Ma P, Zhang ZL, Zhou YD, Lu SY, et al. FADS gene polymorphisms confer the risk of coronary artery disease in a Chinese Han population through the altered desaturase activities: based on high-resolution melting analysis. PloS One. 2013;8(1):e55869 pmid:23383292
  30. 30. Dupuis J, Langenberg C, Prokopenko I, Saxena R, Soranzo N, Jackson AU, et al. New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet. 2010;42(2):105–16. pmid:20081858
  31. 31. Xie L, Innis SM. Association of fatty acid desaturase gene polymorphisms with blood lipid essential fatty acids and perinatal depression among Canadian women: a pilot study. J Nutrigenet Nutrigenomics. 2009;2(4–5):243–50. pmid:20395685
  32. 32. Standl M, Sausenthaler S, Lattka E, Koletzko S, Bauer CP, Wichmann HE, et al. FADS gene variants modulate the effect of dietary fatty acid intake on allergic diseases in children. Clinical and experimental allergy: J Brit Soc Allergy Clin Immunol. 2011;41(12):1757–66. pmid:21793953
  33. 33. Lattka E, Illig T, Heinrich J, Koletzko B. FADS Gene Cluster Polymorphisms: Important Modulators of Fatty Acid Levels and Their Impact on Atopic Diseases. J Nutrigenet Nutrigenomics. 2009;2(3):119–28. pmid:19776639
  34. 34. Standl M, Sausenthaler S, Lattka E, Koletzko S, Bauer CP, Wichmann HE, et al. FADS gene cluster modulates the effect of breastfeeding on asthma. Results from the GINIplus and LISAplus studies. Allergy. 2012;67(1):83–90. pmid:21933193
  35. 35. Lattka E, Klopp N, Demmelmair H, Klingler M, Heinrich J, Koletzko B. Genetic variations in polyunsaturated fatty acid metabolism—implications for child health? Ann Nutr Metab. 2012;60 Suppl 3:8–17. pmid:22614814
  36. 36. Morales E, Bustamante M, Gonzalez JR, Guxens M, Torrent M, Mendez M, et al. Genetic variants of the FADS gene cluster and ELOVL gene family, colostrums LC-PUFA levels, breastfeeding, and child cognition. PloS One. 2011;6(2):e17181. pmid:21383846
  37. 37. Bezard J, Blond JP, Bernard A, Clouet P. The metabolism and availability of essential fatty acids in animal and human tissues. Reprod Nutr Dev. 1994;34(6):539–68. pmid:7840871
  38. 38. Youdim KA, Martin A, Joseph JA. Essential fatty acids and the brain: possible health implications. Int J Dev Neurosci. 2000;18(4–5):383–99. pmid:10817922
  39. 39. Cui T, Hester AG, Seeds MC, Rahbar E, Howard TD, Sergeant S, et al. Impact of genetic and epigenetic variations within the FADS cluster on the composition and metabolism of polyunsaturated fatty acids in prostate cancer. Prostate. 2016;76(13):1182–91. pmid:27197070
  40. 40. Lonsdale J T J.; Slavatore M.; Phillips R.; Lo E.; Shad S.; Hasz R.; Walters G.; Garcia F.; Young N. The Genotype-Tissue Expression (GTEx) project. Nat Genet. 2013;45(6):580–5. pmid:23715323
  41. 41. Wright FA, Sullivan PF, Brooks AI, Zou F, Sun W, Xia K, et al. Heritability and genomics of gene expression in peripheral blood. Nature Genetics. 2014;46(5):430–7. pmid:24728292
  42. 42. Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A. 2003;100(16):9440–5. pmid:12883005
  43. 43. Boyle AP, Hong EL, Hariharan M, Cheng Y, Schaub MA, Kasowski M, et al. Annotation of functional variation in personal genomes using RegulomeDB. Genome Res. 2012;22(9):1790–7. pmid:22955989
  44. 44. Rosenbloom KR, Sloan CA, Malladi VS, Dreszer TR, Learned K, Kirkup VM, et al. ENCODE data in the UCSC Genome Browser: year 5 update. Nucleic Acids Res. 2013;41(Database issue):D56–63. pmid:23193274
  45. 45. Zeller T, Wild P, Szymczak S, Rotival M, Schillert A, Castagne R, et al. Genetics and beyond—the transcriptome of human monocytes and disease susceptibility. PLoS One. 2010;5(5):e10693. pmid:20502693
  46. 46. Schadt EE, Molony C, Chudin E, Hao K, Yang X, Lum PY, et al. Mapping the genetic architecture of gene expression in human liver. PLoS Biol. 2008;6(5):e107. pmid:18462017
  47. 47. Dorajoo R, Sun Y, Han Y, Ke T, Burger A, Chang X, et al. A genome-wide association study of n-3 and n-6 plasma fatty acids in a Singaporean Chinese population. Genes Nutr. 2015;10(6):53. pmid:26584805
  48. 48. Shin SY, Fauman EB, Petersen AK, Krumsiek J, Santos R, Huang J, et al. An atlas of genetic influences on human blood metabolites. Nat Genet. 2014;46(6):543–50. pmid:24816252
  49. 49. Waterworth DM, Ricketts SL, Song K, Chen L, Zhao JH, Ripatti S, et al. Genetic variants influencing circulating lipid levels and risk of coronary artery disease. Arterioscler Thromb Vasc Biol. 2010;30(11):2264–76. pmid:20864672
  50. 50. Wang L, Athinarayanan S, Jiang G, Chalasani N, Zhang M, Liu W. Fatty acid desaturase 1 gene polymorphisms control human hepatic lipid composition. Hepatology. 2015;61(1):119–28. pmid:25123259
  51. 51. Kwak JH, Paik JK, Kim OY, Jang Y, Lee SH, Ordovas JM, et al. FADS gene polymorphisms in Koreans: association with omega6 polyunsaturated fatty acids in serum phospholipids, lipid peroxides, and coronary artery disease. Atherosclerosis. 2011;214(1):94–100. pmid:21040914
  52. 52. Lattka E, Illig T, Heinrich J, Koletzko B. Do FADS genotypes enhance our knowledge about fatty acid related phenotypes? Clin Nutr. 2010;29(3):277–87. pmid:19948371
  53. 53. Howard TD, Mathias RA, Seeds MC, Herrington DM, Hixson JE, Shimmin LC, et al. DNA methylation in an enhancer region of the FADS cluster is associated with fads activity in human liver. PLoS One. 2014;9(5):e97510. pmid:24842322
  54. 54. Hoile SP, Irvine NA, Kelsall CJ, Sibbons C, Feunteun A, Collister A, et al. Maternal fat intake in rats alters 20:4n-6 and 22:6n-3 status and the epigenetic regulation of FADS2 in offspring liver. J Nutr Biochem. 2013;24(7):1213–20. pmid:23107313
  55. 55. Pan G, Ameur A, Enroth S, Bysani M, Nord H, Cavalli M, et al. PATZ1 down-regulates FADS1 by binding to rs174557 and is opposed by SP1/SREBP1c. Nucleic Acids Res. 2017;45(5):2408–22. pmid:27932482
  56. 56. Chen Y, Li Y, Wei J, Li YY. Transcriptional regulation and spatial interactions of head-to-head genes. BMC Genomics. 2014;15:519. pmid:24962804
  57. 57. Chen YQ, Yu H, Li YX, Li YY. Sorting out inherent features of head-to-head gene pairs by evolutionary conservation. BMC Bioinformatics. 2010;11 Suppl 11:S16.
  58. 58. Li YY, Yu H, Guo ZM, Guo TQ, Tu K, Li YX. Systematic analysis of head-to-head gene organization: evolutionary conservation and potential biological relevance. PLoS Comput Biol. 2006;2(7):e74. pmid:16839196
  59. 59. Trinklein ND, Aldred SF, Hartman SJ, Schroeder DI, Otillar RP, Myers RM. An abundance of bidirectional promoters in the human genome. Genome Res. 2004;14(1):62–6. pmid:14707170
  60. 60. Rahbar E, Ainsworth HC, Howard TD, Hawkins GA, Ruczinski I, Mathias R, et al. Uncovering the DNA methylation landscape in key regulatory regions within the FADS cluster. PLoS One. 2017;12(9):e0180903. pmid:28957329
  61. 61. Gieger C, Geistlinger L, Altmaier E, Hrabe de AM, Kronenberg F, Meitinger T, et al. Genetics meets metabolomics: a genome-wide association study of metabolite profiles in human serum. PLoS Genet. 2008;4(11):e1000282 pmid:19043545
  62. 62. Illig T, Gieger C, Zhai G, Romisch-Margl W, Wang-Sattler R, Prehn C, et al. A genome-wide perspective of genetic variation in human metabolism. Nat Genet. 2010;42(2):137–41. pmid:20037589
  63. 63. Draisma HHM, Pool R, Kobl M, Jansen R, Petersen AK, Vaarhorst AAM, et al. Genome-wide association study identifies novel genetic variants contributing to variation in blood metabolite levels. Nat Commun. 2015;6:7208. pmid:26068415