A common variant in ARHGEF10 alters delta-6 desaturase activity and influence susceptibility to hypertriglyceridemia

Authors: Juan de Toro-Martín1,2, Frédéric Guénard1,2, Iwona Rudkowska3, Simone Lemieux1,2, Patrick Couture1,3 and Marie-Claude Vohl1,2*

1 Institute of Nutrition and Functional Foods (INAF), Laval University, Quebec City, QC, G1V 0A6, Canada. 2 School of Nutrition, Laval University, Quebec City, QC, G1V 0A6, Canada. 3 Endocrinology and Nephrology unit, Centre de recherche du CHU de Québec – Laval University, Quebec City, QC, G1V 4G2, Canada.

Running title: A novel D6D-associated SNP in ARHGEF10

* Corresponding author: Marie-Claude Vohl, Ph.D. Institute of Nutrition and Functional Foods (INAF) Laval University 2440 Hochelaga Blvd Quebec, QC Canada, G1V 0A6 [email protected] +1 (418) 656-2131 (4676) Abstract

Background. Numbers of single nucleotide polymorphisms (SNPs) associated with fatty acid desaturase activities have been previously identified within the FADS1-FADS2 cluster, which encodes delta-5 (D5D) and delta-6 (D6D) desaturases, respectively.

Objective. We aimed at further characterizing the genetic variability associated with D5D and D6D activities on a genome-wide scale.

Methods. We conducted a genome-wide association study of D5D and D6D activities in a cohort of 141 individuals from the greater Quebec City metropolitan area using the Illumina HumanOmni5-Quad BeadChip. Estimates of D5D and D6D activities were computed using product-to-precursor fatty acid ratios, arachidonic acid (AA)/dihomo- gamma-linolenic acid (DGLA) for D5D, and DGLA/linoleic acid (LA) for D6D. Levels of fatty acids were measured by gas chromatography in plasma phospholipids.

Results. We identified 24 previously reported SNPs associated with fatty acid levels and desaturase activities as significantly associated with D5D activity within the FADS1- FADS2 gene cluster (lead SNP rs174566/A>G). Furthermore, we identified 5 novel loci potentially associated with D5D activity at 1, 6, 4, 8 and 19. A novel SNP associated with D6D activity and mapped to the ARHGEF10 locus (rs2280885/A>G) was identified, with carriers of the rare allele showing a significant increase in D6D activity and plasma triglyceride levels. After multiple testing correction by permutation, only rs174566 and rs2280885 remained significantly associated to D5D and D6D activity estimates, respectively.

Conclusions. These results confirm previous genetic associations within the FADS1- FADS2 gene cluster with D5D activity. A novel genetic variation associated with higher D6D activity within the ARHGEF10 gene is potentially altering plasma triglyceride levels. Introduction Polyunsaturated fatty acids (PUFA) have been extensively reported as relevant factors involved in maintaining several physiological functions. Among others, relative amounts of omega-3 (n-3) and omega-6 (n-6) PUFA, specially low n-3/n-6 ratios, have been associated with metabolic disorders, such as obesity, metabolic syndrome or inflammation 1–3, as well as with an adequate lipid profile control 4. Although PUFA levels in the organism mainly depends on dietary intake, an alternative and not negligible source of PUFA comes from the conversion of the essential n-3 alpha-linolenic (ALA) and n-6 linoleic (LA) fatty acids 5.

The process of de novo synthesis of long-chain PUFA is governed by enzymes belonging to two different classes, called elongases and desaturases, encoded by from ELOVL (ELOVL2 and ELOVL5) and FADS families (FADS1 and FADS2), respectively 6. The process of elongation and desaturation ends up giving way to the principal forms of n-3, eicosapentaenoic (EPA) and docosahexaenoic acid (DHA), and n-6 long-chain PUFA, dihomo-gamma-linolenic (DGLA) and arachidonic acid (AA) 7. Specifically, desaturation steps are catalyzed by two enzymes, called delta-6 (D6D) and delta-5 (D5D) desaturases, encoded by FADS1 and FADS2, respectively.

Previous studies have revealed that single nucleotide polymorphisms (SNPs) located within or near the FADS1-FADS2 gene cluster on 11 may alter desaturation rates of both D6D and D5D, as well as the associated endogenous levels of n-3 and n-6 PUFA 8–10. Therefore, the aim of this study was to replicate these findings and to reveal novel loci associated to desaturase activities in the Fatty Acid Sensor (FAS) study population 11.

Furthermore, in view of the pronounced inter-individual variability in the plasma triglyceride (TG) response to a n-3 PUFA supplementation 12 and the consistent association between genetic variants altering desaturase activities and plasma lipid levels 13,14, we also analyzed the impact of desaturase-associated SNPs on plasma lipid profile. Methods

Study population The cohort used in this work comprised a total of 141 subjects from the FAS study. In this previous study, 254 subjects from the Quebec City metropolitan area were recruited (registered at ClinicalTrials.gov as NCT01343342) in order to identify determinants of the plasma TG response to an n-3 PUFA (EPA and DHA) supplementation. Trial details and participant selection criteria are extensively described in 12. Briefly, participants of the FAS study were metabolically healthy with a BMI between 25 and 40 kg/m 2 and not taking any medication to treat lipid disorders or n-3 PUFA supplements for at least 6 months prior the intervention study. The experimental protocol was approved by the ethics committees of Centre Hospitalier Universitaire de Québec Research Center and Laval University and was conducted in accordance with the Declaration of Helsinki. From the 254 subjects, 210 completed the intervention protocol and were classified as responders (having a reduction in plasma TG levels > 0.01mM) and non-responders (having no reduction in plasma TG levels). From these 210 participants, genotypes of those exhibiting the most extreme TG response after the n-3 PUFA supplementation were selected for the present study. Extreme TG responders were selected according to a threshold based on the technical error of measurement (TEM) 15. Briefly, participants classified as responders and showing changes in plasma TG levels greater than 0.22 mM (1.5 TEM) after the n-3 PUFA supplementation were considered as extreme TG responders (n = 81). Non-responders were always defined as those showing no reduction in plasma TG levels (n = 60).

Determination of biochemical parameters and desaturase activities Anthropometric and biochemical parameters – plasma TG, total cholesterol (TC), HDL- cholesterol (HDL-C) and LDL-cholesterol (LDL-C) – of study participants were performed after a two-week run-in period, and before the n-3 PUFA supplementation, and they are described in detail elsewhere 10,16. Fatty acid extraction was performed in blood samples collected after 12h overnight fast prior the n-3 PUFA supplementation. Plasma was separated by centrifugation and frozen for subsequent analyses. Plasma lipids were extracted with chloroform:methanol (2:1, v/v), according to a modified Folch method, as previously described 16. Total phospholipids were separated by thin layer chromatography using a combination of acetic acid and isopropyl ether. Fatty acids of isolated phospholipids were then methylated and capillary gas chromatography was used to obtain n-3 (ALA, EPA and DHA) and n-6 (AL, DGLA and AA) fatty acid profiles. Capillary gas chromatography was performed using a temperature gradient on a HP5890 gas chromatograph (Hewlett Packard, Toronto, Canada) equipped with a HP- 88 capillary column (100m x 0.25 mm inner diameter x 0.20µm film thickness; Agilent Technologies, Palo Alto, CA) coupled with a flame ionization detector. Helium was used as carrier gas (split ratio 1:80). The identification of n-3 and n-6 PUFA was performed according to their retention time, using the following standard mixtures as a basis for comparison: the FAME 37 mix (Supelco Inc., Bellefonte, PA) and the GLC-411 FA mix (NuChek Prep Inc, Elysian, MN), as well as the methylated n-6 docosapentaenoic acid (n-6 DPA; Larodan AB, Malmö, Sweden) and n-3 DPA (Supelco Inc., Bellefonte, PA). Fatty acid levels were finally expressed as a percentage of total fatty acids in plasma phospholipids. Estimates of desaturase activities were calculated as n-6 PUFA product- to-precursor ratios as follows: D5D activity was determined as the ratio AA/DGLA, and D6D activity as the ratio DGLA/LA, as previously described 9.

Genotyping Genomic DNA was extracted form whole blood using the GenElute Gel Extraction Kit (Sigma-Aldrich Co., St. Louis, MO) and quantified with an Agilent 2100 Bioanalyser (Agilent Technologies, Palo Alto, CA). Samples were genotyped using the Illumina HumanOmni-5-Quad Bead-Chip (Illumina, San Diego, CA) at the McGill University and Génome Québec Innovation Centre (Montreal, Canada), according to the manufacturer’s instructions. Each HumanOmni-5-Quad BeadChip contained 4,301,331 markers. Sample quality was assessed by signal intensity using R 17 and overall call rate using PLINK 1.9 18. The mean call rate across all samples was 99.84%. None of the 141 samples analyzed were excluded due to low signal intensity or low overall call rate (<90%). Genome-wide association study The genome-wide association study (GWAS) SNP inclusion criteria were a minor allele frequency >1%, a call rate >95%, and a Hardy-Weinberg equilibrium (HWE) P>1×10-7. Allele frequencies and HWE tests were calculated using PLINK 1.9. A total of 1,632,526 SNPs were excluded, leaving 2,668,805 SNPs for statistical analyses. For each SNP, linear associations under an additive model of inheritance were tested in PLINK 1.9 with estimates of D5D and D6D activities used as continuous variables. A potential population structure was corrected by principal component analysis (PCA) and eigenvectors were included as co-variables in the model of linear association. Genomic control inflation factors (λ) were obtained for both association tests to account for false positives. The genomic inflation observed in the association tests for estimates of D5D (λ = 1.26) and D6D (λ = 1.06) activities was corrected, yielding a final genomic control value of 1.09 and 1.03, for D5D and D6D, respectively. Quantile-quantile plots after correction of population structure are shown in Supplementary Figure 1.

After quality control and exclusion of SNPs, the calculated threshold for statistical significance was P<1.87×10-8 (0.05/2,668,805 SNPs). Correction for multiple testing was further performed by permutation (1000 permutations) and empirical P values (Pemp) were computed. The actual power to detect significant associations of medium (f=0.25) and large (f=0.4) effect sizes at P<1.87×10-8 with 141 individuals is 80% (Supplementary Table 1). Power calculations were performed using Quanto 1.2.4 19. Annotation of SNPs was performed using the Variant Annotation Integrator (VAI) tool at the UCSC genome browser (https://genome.ucsc.edu/cgi-bin/hgVai). The subset of lead SNPs was determined by the linkage disequilibrium (LD) clumping approach in PLINK 1.9 (SNPs located less than 250 kb away from an index variant and LD r2>0.5).

Transcriptomic profiling For the present study, we used gene expression levels obtained before the n-3 PUFA supplementation for the first 29 subjects who completed the FAS study. As described in detail in 11, the expression levels of 48,803 mRNA transcripts representing 37,804 genes, were assessed by the Human-6 v3 Expression BeadChips (Illumina) at the McGill University and Génome Québec Innovation Center (Montreal, Canada). The microarray analysis was performed using FlexArray software and the Lumi algorithm was used to normalize Illumina microarray data via the Robust Multiarray Average (RAM) algorithm, as previously reported 11.

Post-hoc association analysis D5D- and D6D-associated SNPs were tested for further associations with fatty acid (LA, ALA, DGLA, AA, DHA and EPA) and plasma lipid levels (TG, TC, HDL-C and LDL-C) in the entire population of 141 participants using the analysis of variance (general linear model, type III sum of squares) with age, sex and BMI included in the model as covariates. An expression quantitative trait loci (eQTL) analysis was also performed with D5D- and D6D-associated SNPs in the sub-population of 29 subjects with the same statistical procedure and parameters as for fatty acid and plasma lipid associations. Pearson correlations were performed among gene expression levels, estimates of desaturase activities, fatty acid and plasma lipid levels. Post-hoc association analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC, USA) and statistical significance defined as False Discovery Rate (FDR) P≤0.05.

Results

Linoleic and arachidonic acid levels prevailed over other fatty acids in study participants Anthropometric parameters and plasma lipid profile of study participants (68 men and 72 women) are summarized in Table 1. PUFA profiles, expressed as a percentage of total fatty acids, revealed that n-6 fatty acids LA (mean 19.4 ± SEM 2.2) and AA (11.2 ± 1.8) were the most abundant as compared to other fatty acids (ANOVA, P<0.001). Regarding desaturase activities, D5D estimates (3.54 ± 1.16) were significantly higher (Student's t-test, P<0.001) than those of D6D (0.18 ± 0.05) (Table 1). Table 1. Characteristics of study participants Number of subjects (men/women) 141 (68/73) Age 31.6 ± 8.8 Anthropometric profile Weight (kg) 83.4 ± 14.2 Height (cm) 171.1 ± 9.4 Body mass index (kg/m2) 28.4 ± 3.8 Lipid profile Triglycerides (mmol/L) 1.32 ± 0.68 Total-cholesterol (mmol/L) 4.90 ± 0.97 HDL-cholesterol (mmol/L) 1.42 ± 0.38 LDL-cholesterol (mmol/L) 2.88 ± 0.88 n3-PUFA (%)* -Linolenic acid 0.19 ± 0.16 Eicosapentaenoic acid 1.14 ± 0.52 휶Docosahexaenoic acid 3.52 ± 0.80 n6-PUFA (%)* Linoleic acid 19.37 ± 2.17 Dihomo- -linolenic acid 3.35 ± 0.74 Arachidonic acid 11.17 ± 1.82 Desaturase휸 activities** D5D 3.54 ± 1.16 D6D 0.18 ± 0.05

Values are reported as mean ± SD. *Polyunsaturated fatty acid (PUFA) levels are expressed as a percentage (%) of total PUFA in plasma phospholipids. **Estimates of desaturase activities are calculated as the product-to-precursor ratios: D6D=%Dihomo-gamma-linolenic acid/%Linoleic acid; D5D= %Arachidonic acid/%Dihomo-gamma-linolenic acid.

GWAS revealed novel loci associated to D6D and D5D activities On one hand, a novel SNP significantly associated with estimates of D6D activity was identified at , within the ARHGEF10 locus, as shown in Figure 1a. Concretely, this new D6D-associated SNP, rs2280885 (P=8.96x10-9; =0.0430.01), was located within intron 4 of the ARHGEF10 gene (Figure 1b). This association remained significant after correction for multiple testing by permutation analysis

(Pemp=0.05). Carriers of the rare allele showed significantly higher estimates of D6D activity than carriers of the common allele (Table 2). On the other hand, a total of 32 SNPs distributed across nine loci were identified as significantly associated with estimates of D5D activity. Most of D5D-associated SNPs were identified at chromosome 11 (Figure 1c). Concretely, 16 SNPs where mapped to the FADS1-FADS2 gene cluster, 7 to MYRF and 1 to FEN1 locus, both located near the FADS1-FADS2 region. Figure 1. D6D and D5D GWAS results Manhattan plots showing GWAS significant associations (red line = 1.87x10 -8) and suggestive associations (blue line = 1x10-5) with estimates of D6D (a) and D5D (c) activities. The figure shows 1 SNP significantly associated with D6D (a) and 32 SNPs significantly associated with D5D activity (c). Relevant desaturase-associated loci are pointed out with arrows. Panels b and d show regional association plots of D6D- and D5D-associated SNPs within ARHGEF10 (b) and FADS1-FADS2 (d) regions, respectively. SNPs are plotted by genetic position against association (-logP) and denoted by diamonds in a white-to- red scale, representing the extent of linkage disequilibrium (LD r2) with the lead SNP (big diamond). Blue lines in the background represents genetic recombination rates estimated using 1000 Genomes Project CEU panel (Pilot 1) and genes within the region are shown at the bottom of the picture (GRCh37/hg19). Regional plots were prepared using SNAP v2.1. After clumping, rs174566 was identified as the lead SNP (P=1.21x10-12; =- 0.9040.12), capturing all D5D-associated SNPs within this chromosomal region (Figure 1d). Furthermore, we identified five novel loci associated with D5D activity at genome-wide significance level: an intronic SNP at chromosome 4 within the LINC01378 locus, a SNP upstream the UBXN6 gene at chromosome 19, and six intergenic SNPs at chromosomes 1, 6 and 8 (Table 2). Carriers of the rare allele of lead D5D-associated SNPs from these new five loci showed a significant increase in estimates of D5D activity (Table 2). By contrast, the lead SNP of the FADS1-FADS2 region, rs174566, showed the opposite association, with carriers of the rare allele showing a significant decrease in D5D activity. Among D5D-associated SNPs, only rs174566 (FADS1-FADS2) remained significant after correction for multiple testing by

permutation analysis (Pemp=0.01) (Table 2).

Table 2. Estimates of D6D and D5D activities by genotype of GWAS-associated lead SNPs

CHR COMMON HMZ HTZ RARE HMZ MAF GWAS P β (SE) Pemp D6D rs2280885 (A>G) chr8:1830025 0.16 ± 0.00 0.21 ± 0.01b 0.23 ± 0.02b 0.13 8.9x10-9 0.043 (0.01) 0.05

D5D rs12093443 (A>G) chr1:69721651 6.84 ± 0.50a 3.45 ± 0.08b 0.01 9.4x10-9 2.996 (0.49) 1.00 rs11932518 (A>G) chr4:118373185 3.46 ± 0.08a 7.65 ± 0.56b 0.01 9.5x10-10 3.697 (0.56) 0.91 rs9451947 (A>G) chr6:92869939 3.42 ± 0.08a 4.85 ± 0.34b 9.71 ± 0.96c 0.04 1.5x10-8 1.755 (0.29) 1.00 rs58174579 (A>G) chr8:58814827 3.39 ± 0.08a 5.36 ± 0.35b 9.68 ± 0.92c 0.03 1.1x10-9 2.787 (0.43) 0.92 rs174566 (A>G) chr11:61592112 4.09 ± 0.12c 3.20 ± 0.13b 2.48 ± 0.27a 0.32 1.2x10-12 -0.904 (0.12) 0.01 rs113832155 (A>G) chr19:4458281 6.95 ± 0.49a 3.45 ± 0.08b 0.01 5.9x10-9 3.061 (0.49) 0.99 Values are adjusted least square means (LS-means ± SEM) derived from analysis of variance models in the entire study population (n = 141). P values are adjusted for age, sex, and BMI. Superscripts stand for significant differences between genotype means following post hoc LS-means pairwise comparisons (P≤0.01). D6D and D5D stand for estimates of delta-6 and delta-5 desaturase activities, respectively. SNP: single nucleotide polymorphism (common allele>rare allele). CHR: chromosomal localization of SNPs (GRCh37/hg19). HMZ: homozygote. HTZ: heterozygote. MAF: minor allele frequency. GWAS P: genome-wide association study P value. β (SE): beta coefficient (standard error). Pemp: empirical P values derived from permutation tests. Desaturase-associated SNPs were associated with n-6 but not with n-3 PUFA levels According to the increase in D6D activity exhibited by rs2280885 rare allele carriers, a significant association was found with LA levels (FDR-P=3.5x10 -2), but only heterozygotes showed a significant decrease as compared to common homozygotes (Figure 2). A significant and stronger association was found with DGLA levels (FDR- P=1.0x10-4), with heterozygotes and rare homozygotes showing a significant elevation of DGLA levels, as compared to common homozygotes (Figure 2). On the other hand, the lead SNP within the FADS1-FADS2 gene cluster, rs174566, showed a significant association with DGLA (FDR-P=4.9x10-5), AA (FDR-P=4.7x10-7) and LA levels (FDR- P=3.6x10-3). As expected, rs174566 was associated with higher levels of the D5D precursor, DGLA, and lower levels of its product, AA (Figure 2). No significant associations were found between rs2280885 or rs174566 with any of the n-3 PUFA levels analyzed (ALA, DHA and EPA; data not shown).

25 5 15 *** * 20 4 *** ** ** *** ** ) l ) )

a 10 l l t *** a a o t t

15 t 3

o o t t

% (

% % ( ( A

L A A 10 2 G L A 5 D

5 1

0 0 0 AA AG GG AA AG GG AA AG GG AA AG GG AA AG GG AA AG GG rs2280885 rs174566 rs2280885 rs174566 rs2280885 rs174566

Figure 2. Association of desaturase-associated lead SNPs with n-6 PUFA levels. Values are adjusted least square means (LS-means ± SEM) derived from analysis of variance models in the entire study population (n = 141). P values are adjusted for age, sex, and BMI. Asterisks stand for significant differences of heterozygotes and rare homozygotes as compared to common homozygotes following post hoc LS-means pairwise comparisons. *P≤0.05, **P≤0.01, ***P≤0.001. LA: linoleic acid. DGLA: dihomo-gamma-linolenic acid. AA: arachidonic acid. rs2280885 and D6D activity were associated with plasma triglyceride levels The post-hoc association test between desaturase-associated SNPs and plasma lipid levels reported a significant association between the D6D-associated SNP, rs2280885 (Figure 3a), and plasma TG levels (P=1.9x10-4; FDR-P=5.3x10-3), with heterozygotes and rare homozygotes exhibiting significantly higher levels than common homozygotes (Figure 3b). Moreover, a significant and positive correlation between D6D activity estimates and plasma TG levels was also found (r=0.54, P<0.0001) (Figure 3c). As expected, both LA and DGLA levels also exhibited a significant correlation with plasma TG levels, but in opposite directions, i.e. LA showed a negative relationship (r=-0.30, P=0.0003) and DGLA exhibited a positive correlation (r=0.43, P<0.0001). Although rs147566 also showed an association with plasma TG levels (P=0.01) in the same direction as rs2280885, this association did not remain significant after FDR correction for multiple comparisons (FDR-P=0.35). No further significant associations were found between rs2280885 or rs147566 and the other lipid traits analyzed (TC, LDL-C and HDL-C; data not shown).

ARHGEF10 gene expression positively correlated with estimates of D6D activity To determine whether rs2280885 was altering plasma TG levels by modulating ARHGEF10 gene expression, an exploratory eQTL analysis was carried out. As shown in Figure 3d, rs2280885 did not show a significant association with ARHGEF10 expression levels, although a slight increase was shown for rare homozygotes. Nevertheless, a positive and significant correlation was found between ARHGEF10 gene expression and estimates of D6D activity (r=0.46, P=0.02) (Figure 3e), as well as a trend for positive correlation with TG levels (r=0.32, P=0.10) (Figure 3f). A negative and significant correlation was also observed between ARHGEF10 gene expression and LA levels (r=-0.36, P=0.04), as well as a trend for a positive correlation with DGLA levels (r=0.36, P=0.06). a b c 0.3 3 3 P<0.0001 ) )

s r=0.54 L ** L

** / / e l l t o o a

** m m m

i 2 2 t

0.2 m m ( ( s

*** e s s

e e y t i d d i i v r r i t e e c 0.1 c 1 c 1 a y y

l l D g g i i r r 6 T T D 0.0 0 0 AA GA GG AA GA GG 0.0 0.1 0.2 0.3 0.4 D6D activity d rs2280885 e rs2280885 f ) ) ) s s s l l l

e 8000 e 8000 P=0.02 e 8000 P=0.10 v ns v v e e e

l l l r=0.32

r=0.46 n n n o o o i 6000 ns i 7000 i 7000 s s s 0 0 0 s s s 1 1 1 e e e F F F r r r E E E p p p x x x

G 4000 G 6000 G 6000 e e e

H H H e e e R R R n n n A A A e e e

g 2000 g 5000 g 5000

e e e v v v i i i t t t a a a l l l

e 0 e 4000 e 4000 r r r ( AA GA GG ( 0.0 0.1 0.2 0.3 0.4 ( 0 1 2 3 4 5 rs2280885 D6D activity Triglycerides (mmol/L) Figure 3. rs2280885 and ARHGEF10 associations with D6D activity and plasma triglyceride levels. Panels a, b and c represent association results in the entire population (n = 141). Panels d, e and f represent association results in the transcriptomic sub-population (n = 29). Black bars are adjusted least square means (LS-means ± SEM) derived from analysis of variance models. P values are adjusted for age, sex, and BMI. Asterisks stand for significant differences of heterozygotes and rare homozygotes as compared to common homozygotes following post hoc LS-means pairwise comparisons. **P≤0.01, ***P≤0.001, ns: not significant. All r values are Pearson correlation coefficients. D6D activity stands for estimates of delta-6 fatty acid desaturase activity (D6D activity = %dihomo-gamma-linolenic/%linoleic acid).

Discussion Several studies have previously tested the impact of genetic variations on desaturase activities and PUFA levels. Most of these studies have focused on polymorphisms located within target genes potentially relevant to PUFA metabolism, such as those encoding for D5D and D6D fatty acid desaturases 20–22. In the present work, we aimed to perform a GWAS in order to replicate previous desaturase-associated SNPs, as well as to discover new loci associated to desaturase activities across the entire genome. We successfully replicated previous results pointing to the FADS1-FADS2 cluster as an especially relevant locus in PUFA metabolism. Concretely, SNPs identified within the FADS1-FADS2 cluster or in high LD with them were previously reported as associated with desaturase activities and/or PUFA levels. Among others, Merino et al. 8 identified nine polymorphisms within this locus as significantly associated with aggregate desaturase activity, i.e. desaturase activity estimated as n-3 (EPA/ALA) or n-6 (AA/LA), with rs174547 showing the strongest association among Caucasian and Asian populations. Carriers of the rs174547 minor allele among the Caucasian population, also identified by us and in high LD with the lead SNP rs174566 (Supplementary Figure 2), showed a significantly lower desaturation activity, as we report in the present study. Guan et al. 23 also identified rs174547 as significantly associated with PUFA n-6 levels, concretely with increased DGLA and reduced AA, i.e. with lower D5D activity. Similarly, Tanaka et al. 24 found the same relationship between D5D activity and rs174537, a SNP in high LD with rs174547. Herein, these results were mirrored for both AA and DGLA levels by the lead SNP rs174566, reinforcing the role of genetic variants within the FADS1-FADS2 locus in PUFA metabolism. Furthermore, five novel loci not previously reported to be associated with desaturase activity or PUFA levels showed significant associations with D5D activity at genome-wide significance. Three of them were located within regions without known genes at chromosomes 1, 6 and 8, and the other two were mapped to LINC01378 and UBXN6 loci, at chromosomes 4 and 19, respectively. None of these novel loci showed significant associations with any of the plasma lipid levels analyzed, and they have not been previously reported in GWAS. It should be noted, however, that two SNPs near LINC01378 (rs11098403 and rs35553880) have been previously associated with schizophrenia 25,26, a pathology closely related with PUFA levels 27,28. It is worth highlighting that only the lead SNP rs174566 located within the FADS1-FADS2 cluster remained significant after correction for multiple testing by permutation analysis. Although most of β coefficients showed by D5D-associated SNPs are relatively high (several exceeding three standard deviations), their computed effect sizes should be considered as small, given their low MAF (<5%) and the modest sample size (Supplementary Figure 3). Thus, while the present study is adequately powered to detect medium and large size effects, it would be, however, slightly underpowered to detect such weak associations. Given that lack of power could also lead to unreliable results in terms of true positives 29, we decided to perform an additional adjustment for multiple testing by permutation. In light of permutation results, we acknowledge the need for further studies (with larger sample sizes), in order to validate these associations for such rare variants. The most intriguing result of this study was the novel and strong association found between the intronic SNP rs2280885 within the ARHGEF10 gene and estimates of D6D activity, with carriers of the rare allele having significantly higher D6D activity. This association has not been reported previously with PUFA levels or desaturase activities and it remained significant after permutation testing. ARHGEF10 belongs to the Rho guanine nucleotide-exchange factors and it is involved in RhoA activation 30. ARHGEF10 is expressed in many tissues and involved in nerve myelination, cerebral atherosclerosis and neuropsychiatric disorders 31,32. Moreover, previous works have identified several SNPs within ARHGEF10 gene as significantly associated with increased susceptibility to ischemic stroke in Japanese 30 and Chinese populations 33,34. Concretely, two associated intronic SNPs, rs2280887 and rs4480162, in total LD with the functional variant rs4376531 were identified in the Japanese population 30. On the other hand, the missense SNP rs9657362 identified in the Chinese population 33 was in high LD with our D6D-associated SNP rs2280885 (Supplementary Figure 2). Another recent study investigating the role of Rho-associated kinases (ROCKs) in a Caucasian population 35 found three other SNPs associated with increased risk of ischemic stroke within the ARHGEF10 locus (rs7007884, rs17683288, rs4876268) and not in LD with rs2280885 (Supplementary Figure 2), suggesting a significant relevance of the entire locus in the increased risk of ischemic stroke, which would be worthy for further research.

Particularly noteworthy was also the strong association found between the D6D- associated SNP rs2280885 and plasma TG levels that mirrored the association found with estimates of D6D activity. Indeed, this relationship was further supported by a positive and significant correlation between estimates of D6D activity and plasma TG levels. Paradoxically, although we could expect that increasing desaturase activities would lead to increasing EPA and DHA levels and, accordingly, a decrease in plasma TG levels, as previously reported 36, we found the opposite relationship. This singular situation has been already addressed and it has been proposed that D6D, the rate- limiting enzyme of desaturation 37,38, despite having a specific preference for n-3 ALA 39, converts considerably more n-6 LA into DGLA, due to a much higher dietary intake of LA as compared to ALA, which leads to a greater and unbalanced LA/ALA ratio. Concretely, absolute levels of LA exhibited by the participants of the present study were around a hundred times higher than those of ALA (227.3  3.1 vs 2.2  0.13 µg/ml), which would explain the deleterious consequences of increased D6D activity in terms of maintaining adequate plasma TG levels. Accordingly, a recent study carried out in patients with type 2 diabetes 40 found positive correlations between levels of the n-6 product of D6D desaturation, DGLA, and obesity-related parameters, such as body mass index, waist circumference, leptin levels and, more importantly, plasma TG levels, which was attributed to excessive intake of n-6 fatty acids. Moreover, baseline DGLA levels have been identified in a panel of PUFA as the best predictive marker of the risk of developing metabolic syndrome in subjects with obesity, also showing the strongest decrease among PUFA after a dietary intervention 41. In this sense, we previously found that plasma TG levels were positively associated with the observed decrease in D6D activity after a n-3 PUFA supplementation, where a significant decrease of DGLA levels was also observed 7.

The genetic background also plays an important role in the different TG response after n-3 PUFA supplementation, and we previously reported several gene variants significantly associated to this inter-individual variability 12. Herein, the significant increase of D6D activity shown by rs2280885 rare allele carriers was accompanied by a strong relationship with plasma TG levels. Curiously, despite rs2280885 rare allele carriers did not show a significant increase of ARHGEF10 gene expression, the positive correlation found between ARHGEF10 gene expression and estimates of D6D activity suggests a role of this gene in the increased D6D activity. The exploratory eQTL study also showed a trend for correlation between ARHGEF10 gene expression and plasma TG levels. These results then point to a potential role of ARHGEF10 in the positive correlation found between D6D activity and plasma TG levels. Moreover, together with previous evidences linking ARHGEF10 and ischemic stroke, the present findings joining increased ARHGEF10 gene expression and D6D activity suggest a potential physiopathological implication of this gene in ischemic stroke through changes in plasma TG levels 42,43.

However, the mechanism linking ARHGEF10 with D6D activity and TG levels remains elusive. Interestingly, a recent finding pointing to the involvement of the RhoA/ROCK pathway in the transcriptional regulation of FADS1 and FADS2 44 suggests that alterations of this pathway, eventually induced by genetic variations within the ARHGEF10 locus, could be involved in the transcriptional regulation of fatty acid desaturases. Although ARHGEF10 expression levels were not significantly different between rs2280885 genotypes in the present study, a similar gene dosage-effect as that observed for rs4376531 30 may be taking place herein. In this regard, whether rs2280885 is involved in the regulation of ARHGEF10 expression through a mechanism akin to that shown by rs4376531, which increases the transcriptional activity of ARHGEF10 by altering the binding affinity to the transcription factor Sp1 30, may be worth to be further studied. On the other hand, given that ARHGEF10 is transcribed in at least two splice variants 45,46, we cannot discard a role of rs2280885 in a mechanism involving alternative splicing of ARHGEF10 via the missense SNP rs9657362 (Leu370Phe) 33, in high LD with rs2280885.

On the other hand, it is well established that TG levels are an independent risk factor of atherosclerosis 47,48 and that oral administration of n-3 PUFA reduces plasma TG levels 49,50. In this regard, the mechanisms by which PUFA would regulate plasma TG levels seem circumscribed to n-3 PUFAs (mainly to EPA and DHA) and comprise a decreased TG release into peripheral circulation, increased fatty acid oxidation, and increased TG clearance via the interaction of n-3 PUFAs with PPAR and PPAR receptors 49,51. As already discussed, increasing D6D activity would move the balance towards greater n-6 PUFA desaturation rates to the detriment of n-3 PUFA, with the consequent reduction in the TG-lowering effect of n-3 PUFA. Thus, in order to elucidate the actual relevance of ARHGEF10 polymorphisms on PUFA and TG metabolism, the association found between rs2280885 and estimates of D6D activity has to be replicated in an independent cohort, and the specific role of ARHGEF10 in this relationship further investigated by functional analyses.

In conclusion, results emerging from this study confirm the relevance of the genetic variance within the FADS1-FADS2 gene cluster in D5D desaturation rates and suggest a potential association of five novel loci associated with D5D activity. In addition, a novel polymorphism strongly associated with higher D6D activity estimates and potentially involved in altered plasma TG levels was identified within the ARHGEF10 gene, pointing to this gene as potentially involved in altered PUFA profile and dyslipidemia.

Acknowledgments The authors acknowledge the contribution of Catherine Raymond, Ann-Marie Paradis, Catherine Ouellette, Véronique Garneau, Élisabeth Thifault and Annie Bouchard- Mercier for laboratory work and technical assistance. This work was supported by an operating grant from the Canadian Institutes of Health Research (CIHR) (MOP110975). JTM has a postdoctoral position at the Institute of Nutrition and Functional Foods (Laval University). IR holds a Junior 1 Research Scholar from the Fonds de Recherche du Québec – Santé (FRQ-S). MCV is Canada Research Chair in Genomics Applied to Nutrition and Health.

Authors contributions JTM performed statistical analysis, interpreted data and drafted the manuscript. FG contributed to the interpretation of the data and critical revision of the manuscript. IR, SL and MCV conceived and designed research. PC was responsible for the medical follow- up. All authors read and approved the final manuscript.

Disclosure The authors declare no conflict of interest. References 1. Simopoulos AP. An increase in the Omega-6/Omega-3 fatty acid ratio increases the risk for obesity. Nutrients. 2016;8(3):1-17. doi:10.3390/nu8030128. 2. Kröger J, Schulze MB. Recent insights into the relation of Δ5 desaturase and Δ6 desaturase activity to the development of type 2 diabetes. Curr Opin Lipidol. 2012;23(1):4-10. doi:10.1097/MOL.0b013e32834d2dc5. 3. de Oliveira Otto MC, Wu JHY, Baylin A, et al. Circulating and dietary omega-3 and omega-6 polyunsaturated fatty acids and incidence of CVD in the Multi-Ethnic Study of Atherosclerosis. J Am Heart Assoc. 2013;2(6):e000506. doi:10.1161/JAHA.113.000506. 4. Merino DM, Ma DWL, Mutch DM. Genetic variation in lipid desaturases and its impact on the development of human disease. Lipids Health Dis. 2010;9:63. doi:10.1186/1476-511X-9-63. 5. Patterson E, Wall R, Fitzgerald GF, Ross RP, Stanton C. Health Implications of High Dietary Omega-6 Polyunsaturated Fatty Acids. J Nutr Metab. 2012;2012:1- 16. doi:10.1155/2012/539426. 6. Bond LM, Miyazaki M, O’Neill LM, Ding F, Ntambi JM. Chapter 6 – Fatty Acid Desaturation and Elongation in Mammals. In: Biochemistry of Lipids, Lipoproteins and Membranes. ; 2016:185-208. doi:10.1016/B978-0-444-63438-2.00006-7. 7. Cormier H, Rudkowska I, Lemieux S, Couture P, Julien P, Vohl M-C. Effects of FADS and ELOVL polymorphisms on indexes of desaturase and elongase activities: results from a pre-post fish oil supplementation. Genes Nutr. 2014;9(6):437. doi:10.1007/s12263-014-0437-z. 8. Merino DM, Johnston H, Clarke S, et al. Polymorphisms in FADS1 and FADS2 alter desaturase activity in young Caucasian and Asian adults. Mol Genet Metab. 2011;103(2):171-178. doi:10.1016/j.ymgme.2011.02.012. 9. Bokor S, Dumont J, Spinneker A, et al. Single nucleotide polymorphisms in the FADS gene cluster are associated with delta-5 and delta-6 desaturase activities estimated by serum fatty acid ratios. J Lipid Res. 2010;51(8):2325-2333. doi:10.1194/jlr.M006205. 10. Cormier H, Rudkowska I, Thifault E, Lemieux S, Couture P, Vohl MC. Polymorphisms in Fatty Acid Desaturase (FADS) gene cluster: Effects on glycemic controls following an omega-3 Polyunsaturated Fatty Acids (PUFA) supplementation. Genes. 2013;4(3):485-498. doi:10.3390/genes4030485. 11. Rudkowska I, Paradis A-M, Thifault E, et al. Differences in metabolomic and transcriptomic profiles between responders and non-responders to an n-3 polyunsaturated fatty acids (PUFAs) supplementation. Genes Nutr. 2013;8(4):411-423. doi:10.1007/s12263-012-0328-0. 12. Rudkowska I, Guenard F, Julien P, et al. Genome-wide association study of the plasma triglyceride response to an n-3 polyunsaturated fatty acid supplementation. J Lipid Res. 2014;55(7):1245-1253. doi:10.1194/jlr.M045898. 13. Kathiresan S, Willer CJC, Peloso GMG, et al. Common variants at 30 loci contribute to polygenic dyslipidemia. Nat Genet. 2008;41(1):56-65. doi:10.1038/ng.291.Common. 14. Aulchenko YS, Ripatti S, Lindqvist I, Boomsma D, Iris M. Loci influencing lipid levels and coronary heart disease risk in 16 European population cohorts. Nat Genet. 2009;41(1):47-55. doi:10.1038/ng.269.Loci. 15. de Lannoy L, Clarke J, Stotz PJ, Ross R, Senn S, Meyer T. Effects of intensity and amount of exercise on measures of insulin and glucose: Analysis of inter- individual variability. Barengo NC, ed. PLoS One. 2017;12(5):e0177095. doi:10.1371/journal.pone.0177095. 16. Tremblay BL, Cormier H, Rudkowska I, Lemieux S, Couture P, Vohl M-C. Association between polymorphisms in phospholipase A2 genes and the plasma triglyceride response to an n-3 PUFA supplementation: a clinical trial. Lipids Health Dis. 2015;14(1):12. doi:10.1186/s12944-015-0009-2. 17. Team RDC. R: A Language and Environment for Statistical Computing. 2008. 18. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second- generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4(1):7. doi:10.1186/s13742-015-0047-8. 19. Gauderman WJ, Morrison J. QUANTO 1.2 : A computer program for power and sample size calculations for genetic-epidemiology studies. http://hydra.usc.edu/gxe. 2006. http://ci.nii.ac.jp/naid/10030808189/. Accessed August 19, 2017. 20. Zietemann V, Kröger J, Enzenbach C, et al. Genetic variation of the FADS1 FADS2 gene cluster and n-6 PUFA composition in erythrocyte membranes in the European Prospective Investigation into Cancer and Nutrition-Potsdam study. Br J Nutr. 2010;104(12):1748-1759. doi:10.1017/S0007114510002916. 21. Zhu J, Sun Q, Zong G, et al. Interaction between a common variant in FADS1 and erythrocyte polyunsaturated fatty acids on lipid profile in Chinese Hans. J Lipid Res. 2013;54(5):1477-1483. doi:10.1194/jlr.P027516. 22. Al-Hilal M, Alsaleh A, Maniou Z, et al. Genetic variation at the FADS1-FADS2 gene locus influences delta-5 desaturase activity and LC-PUFA proportions after fish oil supplement. J Lipid Res. 2013;54(2):542-551. doi:10.1194/jlr.P032276. 23. Guan W, Steffen BT, Lemaitre RN, et al. Genome-wide association study of plasma N6 polyunsaturated fatty acids within the cohorts for heart and aging research in genomic epidemiology consortium. Circ Cardiovasc Genet. 2014;7(3):321-331. doi:10.1161/CIRCGENETICS.113.000208. 24. Tanaka T, Shen J, Abecasis GR, et al. Genome-wide association study of plasma polyunsaturated fatty acids in the InCHIANTI Study. PLoS Genet. 2009;5(1):e1000338. doi:10.1371/journal.pgen.1000338. 25. Goes FS, McGrath J, Avramopoulos D, et al. Genome-wide association study of schizophrenia in Ashkenazi Jews. Am J Med Genet Part B Neuropsychiatr Genet. 2015;168(8):649-659. doi:10.1002/ajmg.b.32349. 26. Lencz T, Guha S, Liu C, et al. Genome-wide association study implicates NDST3 in schizophrenia and bipolar disorder. Nat Commun. 2013;4:2739. doi:10.1038/ncomms3739. 27. Solberg DK, Bentsen H, Refsum H, Andreassen OA. Association between serum lipids and membrane fatty acids and clinical characteristics in patients with schizophrenia. Acta Psychiatr Scand. 2015;132(4):293-300. doi:10.1111/acps.12388. 28. van der Kemp WJM, Klomp DWJ, Kahn RS, Luijten PR, Hulshoff Pol HE. A meta- analysis of the polyunsaturated fatty acid composition of erythrocyte membranes in schizophrenia. Schizophr Res. 2012;141(2):153-161. doi:10.1016/j.schres.2012.08.014. 29. Button KS, Ioannidis JPA, Mokrysz C, et al. Power failure: why small sample size undermines the reliability of neuroscience. Nat Rev Neurosci. 2013;14(5):365- 376. doi:10.1038/nrn3475. 30. Matsushita T, Ashikawa K, Yonemoto K, et al. Functional SNP of ARHGEF10 confers risk of atherothrombotic stroke. Hum Mol Genet. 2010;19(6):1137-1146. doi:10.1093/hmg/ddp582. 31. Tabarés-Seisdedos R, Rubenstein J. Chromosome 8p as a potential hub for developmental neuropsychiatric disorders: implications for schizophrenia, autism and cancer. Mol Psychiatry. 2009;14:563-589. doi:10.1038/mp.2009.2. 32. Verhoeven K, De Jonghe P, Van de Putte T, et al. Slowed conduction and thin myelination of peripheral nerves associated with mutant Rho guanine-nucleotide exchange factor 10. Am J Hum Genet. 2003;73(4):926-932. doi:10.1086/378159. 33. Li H, Yu S, Wang R, et al. ARHGEF10 gene polymorphism is closely associated with the risk of ischemic stroke in Northern Han Chinese population. Neurol Res. 2017;39(2):158-164. doi:10.1080/01616412.2016.1263175. 34. Yin Y-Y, Zhang B, Zhou M-K, et al. The functional SNP rs4376531 in the ARHGEF gene is a risk factor for the atherothrombotic stroke in Han Chinese. Neurol India. 2011;59(3):408-412. doi:10.4103/0028-3886.82755. 35. Zee RYL, Wang Q-M, Chasman DI, Ridker PM, Liao JK. Gene variations of ROCKs and risk of ischaemic stroke: the Women’s Genome Health Study. Clin Sci. 2014;126(12):829-835. doi:10.1042/CS20130652. 36. Kastelein JJP, Maki KC, Susekov A, et al. Omega-3 free fatty acids for the treatment of severe hypertriglyceridemia: The EpanoVa for Lowering Very high triglyceridEs (EVOLVE) trial. J Clin Lipidol. 2014;8(1):94-106. doi:10.1016/j.jacl.2013.10.003. 37. Portolesi R, Powell BC, Gibson R a. Competition between 24:5n-3 and ALA for Delta 6 desaturase may limit the accumulation of DHA in HepG2 membranes. J Lipid Res. 2007;48(7):1592-1598. doi:10.1194/jlr.M700081-JLR200. 38. Sprecher H. The roles of anabolic and catabolic reactions in the synthesis and recycling of polyunsaturated fatty acids. Prostaglandins Leukot Essent Fatty Acids. 2002;67(2-3):79-83. doi:10.1054/plef.402. 39. D’andrea S, Guillou H, Jan S, et al. The same rat Delta6-desaturase not only acts on 18- but also on 24-carbon fatty acids in very-long-chain polyunsaturated fatty acid biosynthesis. Biochem J. 2002;364(1):49-55. 40. Yamashita K, Higa M, Kunishita R, et al. Dihomo-gamma-linolenic acid levels and obesity in patients with type 2 diabetes. Diabetol Int. 2015;6(3):206-211. doi:10.1007/s13340-014-0187-6. 41. Ni Y, Zhao L, Yu H, et al. Circulating Unsaturated Fatty Acids Delineate the Metabolic Status of Obese Individuals. EBioMedicine. 2015;2(10):1513-1522. doi:10.1016/j.ebiom.2015.09.004. 42. Lee JS, Chang P-Y, Zhang Y, Kizer JR, Best LG, Howard B V. Triglyceride and HDL-C Dyslipidemia and Risks of Coronary Heart Disease and Ischemic Stroke by Glycemic Dysregulation Status: The Strong Heart Study. Diabetes Care. 2017:dc161958. 43. Freiberg JJ, Tybjærg-Hansen A, Jensen JS, Nordestgaard BG. Nonfasting Triglycerides and Risk of Ischemic Stroke in the General Population. JAMA. 2008;300(18):2142. doi:10.1001/jama.2008.621. 44. Ishihara N, Suzuki S, Tanaka S, et al. Atorvastatin increases Fads1, Fads2 and Elovl5 gene expression via the geranylgeranyl pyrophosphate-dependent Rho kinase pathway in 3T3-L1 cells. Mol Med Rep. 2017. doi:doi: 10.3892/mmr.2017.7141. 45. Mohl M, Winkler S, Wieland T, Lutz S. Gef10—the third member of a Rho-specific guanine nucleotide exchange factor subfamily with unusual architecture. Naunyn Schmiedebergs Arch Pharmacol. 2006;373(5):333-341. doi:10.1007/s00210-006-0083-0. 46. Verhoeven K, De Jonghe P, Van de Putte T, et al. Slowed Conduction and Thin Myelination of Peripheral Nerves Associated with Mutant Rho Guanine-Nucleotide Exchange Factor 10. Vol 73.; 2003. doi:10.1086/378159. 47. Cullen P. Evidence that triglycerides are an independent coronary heart disease risk factor. Am J Cardiol. 2000;86(9):943-949. doi:10.1016/S0002- 9149(00)01127-9. 48. Talayero BG, Sacks FM. The Role of Triglycerides in Atherosclerosis. Curr Cardiol Rep. 2011;13(6):544-552. doi:10.1007/s11886-011-0220-3. 49. Calder PC. Mechanisms of Action of (n-3) Fatty Acids. J Nutr. 2012;142(3):592S- 599S. doi:10.3945/jn.111.155259. 50. Casula M, Soranna D, Catapano AL, Corrao G. Long-term effect of high dose omega-3 fatty acid supplementation for secondary prevention of cardiovascular outcomes: A meta-analysis of randomized, double blind, placebo controlled trials. Atheroscler Suppl. 2013;14(2):243-251. doi:10.1016/S1567-5688(13)70005-9. 51. Shearer GC, Savinova O V, Harris WS. Fish oil -- how does it reduce plasma triglycerides? Biochim Biophys Acta. 2012;1821(5):843-851. doi:10.1016/j.bbalip.2011.10.011.