Fatty acid desaturase (FADS) gene polymorphisms and insulin resistance in association with serum phospholipid polyunsaturated fatty acid composition in healthy Korean men: cross-sectional study

Background We investigated the relationship between fatty acid desaturase (FADS) gene polymorphisms and insulin resistance (IR) in association with serum phospholipid polyunsaturated fatty acid (FA) composition in healthy Korean men. Methods Healthy men (n = 576, 30 ~ 79 years old) were genotyped for rs174537 near FADS1 (FEN1-10154G>T), FADS2 (rs174575C>G, rs2727270C>T), and FADS3 (rs1000778C>T) SNPs. Dietary intake, serum phospholipid FA composition and HOMA-IR were measured. Results Fasting insulin and HOMA-IR were significantly higher in the rs174575G allele carriers than the CC homozygotes, but lower in the rs2727270T allele carriers than the CC homozygotes. The proportion of linoleic acid (18:2ω-6, LA) was higher in the minor allele carriers of FEN1-10154G>T, rs174575C>G and rs2727270C>T than the major homozygotes, respectively. On the other hand, the proportions of dihomo-γ-linolenic acid (20:3ω-6, DGLA) and arachidonic acid (20:4ω-6, AA) in serum phospholipids were significantly lower in the minor allele carriers of FEN1-10154 G>T carriers and rs2727270C>T than the major homozygotes respectively. AA was also significantly lower in the rs1000778T allele carriers than the CC homozygotes. HOMA-IR positively correlated with LA and DGLA and negatively with AA/DGLA in total subjects. Interestingly, rs174575G allele carriers showed remarkably higher HOMA-IR than the CC homozygotes when subjects had higher proportions of DLGA (≥1.412% in total serum phospholipid FA composition) (P for interaction = 0.009) or of AA (≥4.573%) (P for interaction = 0.047). Conclusion HOMA-IR is associated with FADS gene cluster as well as with FA composition in serum phospholipids. Additionally, HOMA-IR may be modulated by the interaction between rs174575C>G and the proportion of DGLA or AA in serum phospholipids.


Backgrounds
It has been reported that fatty acid (FA) composition in serum phospholipids as well as a high fat intake might influence the insulin sensitivity and the progression of obesity [1][2][3]. People with insulin resistance (IR) or metabolic syndrome (MetS) had high levels of palmitic acid (16:0, PA) and low levels of linoleic acid (18:2ω-6, LA) in serum phospholipids [4,5]. Lopes-Alvarenga et al [1] showed that serum triglyceride (TG) level in MetS subjects was positively associated with saturated FAs (SFA), but negatively with long-chain polyunsaturated fatty acids (PUFA). According to Leeson et al [5], higher proportions of docosahexaenoic acid (22:6ω-3, DHA) in erythrocyte lipids were associated with the improved endothelial function, particularly in young smokers who had some features of the IR.
FA composition in serum lipid esters is known to reflect the dietary FA composition during recent one or two months [6][7][8][9]. Serum PUFA compositions are influenced by both the FA metabolisms and the genetic variations [10][11][12]. The key enzymes involved in PUFA metabolism are the Δ5 desaturase (D5D) and Δ6 desaturase (D6D) which are encoded by the FA desaturase 1 (FADS1) and FADS2 genes located on chromosome 11 (11q12-13.1) [13][14][15][16]. This gene cluster also includes a FADS3, another desaturase gene which shares 52 and 62% sequence of the FADS1 and FADS2 genes, respectively [13], but the activity has not been fully identified yet. Among the single nucleotide polymorphisms (SNPs) in FADS gene cluster, rs174537 (FEN1, flapstructure specific endonuclease) near FADS1 shows the most significant association with FAs in a genome-wide association study. Minor T allele homozygotes of the rs174537 (FEN1-10154G>T) are associated with lower levels of plasma arachidonic acid (20:4ω-6, AA) and higher levels of LDL cholesterol and total cholesterol than the major G allele homozygotes [16].
Recently, Dupis et al. [17] and Ingelsson et al [18] reported that FADS1 (rs174550) was strongly associated with fasting glucose, fasting insulin, homeostasis model assessment of IR (HOMA-IR), HOMA-beta cell function and type 2 diabetes in Caucasian population. However, there were no studies on the relationship between FADS gene and IR in association with serum PUFA composition in healthy Koreans. Therefore, this present study aimed to investigate whether FADS gene polymorphisms are associated with IR as well as with serum phospholipid FA composition in healthy Korean men.

Study population
Study participants (men) were recruited from the Health Service Center (HSC) in the course of a routine checkup visit or by a newspaper announcement for health examination. Subjects were excluded if they have orthopedic limitations, weight loss/gain over the previous 6 months, or any diagnosis of vascular disease, diabetes mellitus, cancer (clinically or by anamnesis), renal disease, liver disease, thyroid disease, and acute or chronic inflammatory diseases. None of all subjects were taking any medications (antihypertensive, antidyslipidemic, antithrombotic and antidiabetic drugs). Finally, 576 healthy men (aged from 30 to 79 years old) were enrolled in the study. Written informed consent was obtained from all subjects, and the protocol was approved by the Institute of Review Board of Yonsei University.

Anthropometric parameters and blood collection
Body weight and height were measured unclothed and without shoes in the morning. Body mass index (BMI) was calculated as body weight in kilograms divided by height in square meters (kg/m 2 ). Blood pressure (BP) was obtained from the left arm of seated patients with an automatic blood pressure monitor (TM-2654, A&D, Tokyo, Japan) after 20 min of rest. Study subjects were interviewed for their smoking and drinking behavior at their visit. Smoking habit was categorized to 'current smoker' and 'non-smoker', and drinking habit was also categorized to 'current drinker (current alcohol consumption)' and 'non-drinker'.
After overnight fast, venous blood specimens were collected in EDTA-treated and plain tubes. The tubes were immediately placed on ice until they arrived at the analytical laboratory (within 1-3 h). Then, the blood specimens were separated into plasma (i.e. glucose and insulin) or serum (i.e. lipid profiles, apolipoproteins and fatty acid compositions), and stored at -70°C until analysis.
Genomic DNA was extracted from 5 mL whole blood using a commercially available DNA isolation kit (WIZARD R Genomic DNA purification kit, Promega, Madison, WI, USA) according to the manufacturer's protocol.

Serum Lipid profile, Apolipoprotein A1 and B
Fasting serum total cholesterol and TG were measured using commercially available kits on a Hitachi 7150 Autoanalyzer (Hitachi Ltd. Tokyo, Japan). After precipitation of serum chylomicron, low density lipoprotein (LDL) and very low density lipoprotein (VLDL) with dextran sulfate-magnesium, high density lipoprotein cholesterol (HDL-C) left in the supernatant was measured by an enzymatic method. LDL cholesterol (LDL-C) was estimated indirectly using the Friedewald formula for subjects with serum TG concentrations <400 mg/dL (4.52 mol/L). In subjects with serum TG concentrations ≥400 mg/dL (4.52 mol/L), LDL-C was measured by an enzymetic method on a Hitachi 7150 Autoanalyzer directly. Each sample was measured duplicate and their average value was used for one value. If there is big variation between the two values, the samples were remeasured. Serum apolipoprotein (Apo) A1 and B were determined by turbidometry at 340 nm using a specific anti serum (Roche, Basel, Switzerland).

Glucose, Insulin and HOMA-IR
Fasting plasma glucose was measured by a glucose oxidase method using the Glucose Analyzer (Beckman Instruments, Irvine, CA, USA). Plasma Insulin was measured by radioimmuno-assays with commercial kits from Immuno Nucleo Corporation (Stillwater, MN, USA). Insulin resistance (IR) was calculated with the homeostasis model assessment (HOMA) using the following equation: IR = {fasting insulin (μIU/ml) × fasting glucose (mmol/l)}/22.5 [22].

Fatty acid composition in serum phospholipids
Serum phospholipid fatty acid (FA) composition was analyzed using the modified method of Folch et al. [23] and Lepage et al. [24] with gas chromatography (Hewlett Packard 5890A, CA, USA). Individual FAs were calculated as a relative percentage with the elevated FAs set at 100% using Chemstation software.

The assessment of dietary intake/physical activity level
The subjects' usual diet information was obtained using both a 24-hour recall method and a semi-quantitative food frequency questionnaire (SQFFQ) of which the validity had been previously tested [25]. We used the former to carry out analyses and the latter to check if the data collected by 24-hour recall methods was representative of the usual dietary pattern. All subjects were given written and verbal instructions by a registered dietitian on completion of a 3-day (2 week days and 1 weekend) dietary record. Dietary energy values and nutrient content from a 3-day food records were calculated using the Computer Aided Nutritional analysis program (CAN-pro 2.0, Korean Nutrition Society, Seoul, Korea). Total energy expenditure (TEE) (kcal/day) was calculated from activity patterns including basal metabolic rate, physical activity for 24 hours [26], and specific dynamic action of food. Basal metabolic rate for each subject was calculated with the Harris-Benedict equation [27].

Statistical analysis
Statistical analyses were performed with SPSS version 15.0 for Windows (Statistical Package for the Social Science, SPSS Ins., Chicago, IL, USA). Hardy-Weinberg Equilibrium (HWE) and the linkage disequilibrium (LD) tests were examined using the Haploviewer 4.1 (Broad Inst, MA, USA). In order to test the difference in general characteristics and biochemical parameters (ie. lipid profiles, glucose, insulin resistance, fatty acid composition and dietary intakes), student t-test (independent ttest) between the two groups (major allele homozygotes vs minor allele carriers) and one-way analysis of variance (ANOVA) among the genotype groups followed by Bonferroni correction for multiple comparisons to reduce the rate of false positive was used. General linear model analysis (GLM, Post hoc multiple comparison tests) followed by Bonferroni correction for multiple comparisons was also performed to see the differences in IR and FA composition between genotype groups with adjustment for confounders (i.e. age, body mass index, cigarette smoking, alcohol drinking, blood pressure, triglyceride, total energy intake and expenditure etc.). The Pearson correlation analyses were used to evaluate the relationship between insulin resistance (HOMA-IR) and general or biochemical parameters including FA composition. In addition, multiple stepwise regression analysis was performed to see whether FADS SNPs remain significant influent factor on HOMA-IR together with other markers related to HOMA-IR. Frequencies were tested by chi-square test. We determined whether each variable was normally distributed before statistical testing, and logarithmic transformation was performed on skewed variables. For descriptive purposes, mean values are presented using untransformed values. Results are expressed as means ± standard deviation (S.D.) or %. A two-tailed value of P < 0.05 was considered statistically significant.
Since haplotype analysis did not provide information beyond that revealed by each SNP (data not shown), we presented only the results of individual SNPs. Particularly, we presented the results especially by the form of the homozygotes for a given allele and the carrier of the alternative allele because their general and biochemical characteristics (i.e. insulin level, HOMA-IR and FA composition) were shown similarly with those among 3 genotype groups, furthermore the subject number of the minor homozygotes of rs rs174575 was so small (n = 7).
Clinical and biochemical characteristics according to FEN1 -10154G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T Table 2 shows clinical and biochemical characteristics of study subjects according to the genotypes of the FEN1 -10154G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T, respectively. No significant genotype-associated differences were observed for age, BMI, systolic blood pressure (SBP), diastolic blood pressure (DBP), serum concentrations of TG, total cholesterol (Total-C), HDL-C, LDL-C, ApoA1, ApoB (Table 2), total energy intake, proportions of energy intake derived from carbohydrates and fat in each of 4 FADS SNPs (Additional file 2). Interestingly, fasting insulin and HOMA-IR were significantly higher in subjects having the minor G allele of FADS2 rs174575C>G than those having the major CC homozygotes. On the other hand, the levels of these two markers were significantly lower in the subjects carrying the minor T allele of FADS2 rs2727270C>T than those  Table 2 Demographic and metabolic parameters of study population according to Age ( having major CC homozygote (Table 3). These significant associations still maintained after adjustment for confounders (age, BMI, cigarette smoking, alcohol consumption, blood pressure, triglyceride, total energy expenditure and total calorie intake) which were followed by Bonferroni correction for multiple comparisons (Table 3).
Fatty acid composition in serum phospholipids according to FEN1 -10154G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T Table 3 shows the FA composition in serum phospholipids according to FEN1 -10154G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T, respectively after the adjustment for confounders (age, BMI, cigarette smoking, alcohol consumption, blood pressure, triglyceride, total energy expenditure and total calorie intake) which were followed by Bonferroni correction for multiple comparisons. The minor T allele carriers of FEN1 -10154G>T showed the higher proportion of linoleic acid (18:2ω-6, LA) (P = 0.002) and lower proportion of stearic acid (18:0, SA) (P = 0.045), dihomo-γ-linolenic acid (20:3ω-6, DGLA) (P = 0.038) and arachidonic acid (20:4ω-6, AA) (P < 0.001) in serum phospholipids than the GG homozygotes. The minor G allele carriers of FADS2 rs174575C>G had the higher proportion of LA (P = 0.001) and lower proportion of SA (P = 0.009) than the CC homozygotes. The minor T allele carriers of FADS2 rs2727270C>T showed the higher proportion of LA (P = 0.026), and lower proportions of DGLA (P = 0.035) and AA (P < 0.001) than the CC homozygtes. The minor T allele carriers of FADS3 rs1000778C>T had the lower proportion of AA (P = 0.022) than the CC homozygotes. Among FA ratios which indicate FA desaturase activity, AA/LA were lower in the subjects carrying the minor allele of FEN1 -10154G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T than the major allele homozygote, respectively. DGLA/LA were also were lower in the subjects carrying the minor allele of FEN1 -10154G>T, FADS2 rs2727270C>T and FADS3 rs1000778C>T than the major allele homozygote, respectively. Additionally, DGLA/LA were lower in the subjects carrying the minor allele of FEN1 -10154G>T and FADS2 rs2727270C>T than the major allele homozygote, respectively.

Association of FADS gene SNPs with insulin resistance according to the proportion of FA in serum phospholipids
Based on the correlation results between HOMA-IR and FA compositions mentioned above, study subjects were subdivided into two groups according to the proportion of LA, DGLA or AA in serum phospholipids: upper median (LA≥12.396%, DGLA≥1.412% and AA≥4.573%. of total FA composition in serum phospholipids) and lower median (LA < 12.396%, DGAL < 1.412% and AA < 4.573%). Then, we investigated the interaction effect of genetic variants and proportion of LA, DGLA or AA on their HOMA-IR levels ( Figure 1). Interestingly, HOMA-IR was significantly higher in minor G allele carriers of rs174575C>G than the CC homozygotes, when the proportion of DGLA or AA were higher than median value (Interaction P = 0.009 and P = 0.047, respectively). However, significant interaction effect was neither observed with other SNPs (FEN1 -10154G>T, FADS2 rs2727270C>T or FADS3 rs1000778C>T) nor with LA.
Stepwise multiple regression analysis to find the major contributors to HOMA-IR among basic parameters and FADS gene SNPs Stepwise multiple linear regression analysis was performed to determine whether FADS gene SNPs are the independent influencing factors on HOMA-IR. Based on the results of Pearson correlation coefficients (between HOMA-IR and anthropometric or basic biochemical parameters) and those of the LD test, we put HOMA-IR as a dependent variable, and body mass index, triglyceride, blood pressure and 4 SNPs (FEN1-10154 rs174537G>T, FADS2 rs174575 C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T) as independent variables in the multiple stepwise linear regression model. Among the 3 SNPs, we found that FADS2 rs174575 C>G (β-coefficient ± S.E.: 0.253 ± 0.099, p < 0.001) remained as a significant predictor of HOMA-IR together with body mass index, triglyceride and blood pressure (R 2 = 0.435, P < 0.001). In addition, when we put the 4 SNPs in the model by the forms of 'the Tested by log-transformed, Performed by general linear model analysis (GLM, Post hoc multiple comparison tests) with adjustment for age, body mass index, cigarette smoking, alcohol drinking, blood pressure, triglyceride, total energy intake and expenditure, followed by Bonferroni correction for multiple comparison.

Discussion
In this present study, we found that HOMA-IR as well as serum FA composition and FA ratios were associated with the variants of FADS SNPs. This result is meaningful because it was first reported in Korean men but already done in Caucasians [17,18]. We also confirmed that HOMA-IR was related to FA composition in serum phospholipids, particularly LA, DGLA and AA/DGLA. Interestingly, we found the interactive effect between FADS2 rs174575 and the proportion DGLA or AA in serum phospholipids on HOMA-IR: FADS2 rs174575G allele carriers showed the remarkably higher HOMA-IR levels than the major CC homozygotes, when they had a higher proportion of DGLA (≥1.412% of total FA in serum phospholipids) or AA (≥4.573%. of total FA composition in serum phospholipids) in serum phospholipids. So far, many studies have researched on the relationship of serum phospholipid FA composition or dietary fat with IR or metabolic disorder [1][2][3][4][5]. For example, Vessby and his colleague reported that people with IR or MetS had higher levels of palmitic acid and lower levels of LA in serum phospholipids [4,5]. Warensjö et al and others showed the significant positive relationship between the metabolic disorder (ie. IR and obesity) and the proportion of palmitoleic acid (16:1) and DGLA in serum phospholipids FAs [1,4,10]. They also showed the inverse association of IR with AA/DGLA which indicates the D5D activity. In our study, we found that HOMA-IR was positively associated with DGLA and negatively with AA/DGLA. However, we additionally found that HOMA-IR was positively related with LA, which is different from others. It may be related to the subject characteristics or the sources of FA. First, our study subjects were all healthy without MetS nor obesity (n), § tested by log-transformed, P for interaction was tested by multiple logistic regression analysis after adjusting for age and body mass index. HOMA-IR: homeostasis model assessment of insulin resistance. [1,4,10]. In addition, LA and α-linolenic acid (18:3ω-3, ALA) in serum phospholipids are generally known as biomarkers of long-term essential FA intake [6][7][8][9], but palmitoleic acid and DGLA are synthesized endogenously by a sequence of desaturation and elongation reactions rather than they reflect dietary intake of those FAs [1]. It may indicate that DGLA or D5D are more relevant to explain the relationship with IR or metabolic disorder.
Recently, as examined in our study, the relationship between FA composition in serum phospholipids and the variants of FADS gene cluster has been studied [13][14][15][16]. Among the SNPs in FADS gene cluster, FEN1-10154G>T near FADS1 was reported to be most significantly associated with plasma FA composition; the minor T allele homozygotes of the FEN1-10154G>T are associated with lower levels of plasma AA than the major GG homozygotes [16]. In our study, the minor allele carriers of FEN1 -10154G>T, rs2727270C>T and FADS3 rs1000778C>T showed significantly lower proportion of AA. Moreover, the proportions of DGLA, DGLA/LA and AA/DGLA (D5D activity) were lower in the minor allele carriers of FEN1 -10154G>T and FADS2 rs2727270 SNPs. In addition, HOMA-IR levels were lower in the FADS2 rs2727270T allele carriers compared with the major CC homozytoges. All of these results were still retained after the adjustment for confounders followed by Bonferroni's correction for multiple comparisons. According to Rodriguez et al., AA or long-chain ω-6 PUFA were inversely associated with fasting insulin levels which might be due to the reduced desaturation-elongation cascade under IR condition furthermore, this phenomenon may worsen as the degree of IR increases [28]. Taken all together, our results may indicate that subjects with the major homozygotes had the impaired FA metabolism which may cause the accumulation of DGLA, and in consequence increase IR.
In addition, we found that rs2727270C>T (negative direction) and particularly FADS2 rs174575C>G (positive direction) were significant influent factor on HOMA-IR together with anthropometric or basic biochemical parameters. On the other hand, we found that HOMA-IR levels were higher in the FADS2 rs174575 G allele carriers compared with the major CC homozytoges, which were different from the results shown in FADS2 rs2727270C>T. These results were still retained after the adjustment for confounders followed by Bonferroni's correction for multiple comparisons. Furthermore, the correlation between HOMA-IR and serum PUFA composition were more clearly found in major C homozygotes (data not shown). Interestingly the higher HOMA-IR in FADS2 rs174575G carriers were more remarkable when the proportion of DGLA or AA in serum phospholipids was higher than median levels (interaction P = 0.009 and P = 0.047, respectively). These results were also still retained after the adjustment for confounders followed by Bonferroni's correction for multiple comparisons. Regarding this discordant result, we can assume that FADS2 rs174575 SNP differently from other FADS SNPs may affect the FA metabolism and that rs174575 may interact with other genes, such as the nuclear receptor peroxisome proliferator-activated receptor-γ (PPAR-γ) or the adipocyte C1q and collagen-domain-containing gene (ADIPOQ). PPAR-γ is a transcriptional regulator which directly interacts with the ADIPOQ promoter, both of them are closely associated with IR, MetS and diabetes mellitus [29][30][31][32]. It is known that PUFA acts as ligands of PPAR-γ [31], thus the pharmacologic activation with PPAR-γ agonists may lead to the increased plasma adiponectin [30,32], which may affect the glucose and insulin metabolism. This part needs to be further studied.
Our study has several limitations. First, study subjects were all healthy Korean men; thus, the results may not be applicable to women, other ethnic groups, or patients with cardiometablic syndrome whose lifestyle habits or biochemical characteristics may differ from those in our subjects. For example, Warensjö et al demonstrated that serum FA composition [i.e palmitoleic acid, γ-linolenic acid (18:3ω-6, GLA), DGLA and AA] and D6D activity were different between men and women [1]. Second, this study was performed in a cross-sectional design. It is not suitable for assessing the time sequential associations because the exposure and outcomes are collected at one point in time. Third, we need to consider the involvement of other genetic backgrounds which are associated in FA metabolism, for example PPAR-γ and ADIPOQ genes as mentioned above.
Despite these limitations, the present study shows that HOMA-IR as well as serum FA composition and FA ratios were associated with the variants of FADS SNPs. Furthermore, it suggested that HOMA-IR may be modulated by the interaction between the FADS genetic variant and the proportion of FA, particularly DGLA in serum phospholipids.

Conclusion
HOMA-IR and serum FA composition or FA ratios were associated with the FADS SNPs. Interestingly, FADS2 rs174575 and the proportion DGLA or AA in serum phospholipids may interactively affect HOMA-IR even in healthy people.

Additional material
Additional file 1: Genotype and haplotype distribution of study population. It included genotype and haplotype distribution of 4 SNPs (FEN1-10154 rs174537G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 1000778C>T) in the whole study subjects. The selected 4 SNPs satisfied the HWE (P > 0.05): minor T allele frequency of FEN1-10154 rs174537G>T was 0.312, minor G allele frequency of FADS2 rs174575C>G was 0.082, minor T allele frequency of FADS2 rs2727270C>T was 0.24, and minor T allele frequency of FADS3 1000778C>T was 0.293. This file also included the haplotype distribution of rs174537-rs174575-rs2727270-rs1000778: GCCC was the most highly frequent haplotype and the nonGCCC frequency was 0.450.
Additional file 2: Macro-nutrient intake and energy expenditure of study population according to FEN1 -10154G>T, FADS2 rs174575, FADS2 rs2727270 and rsFADS3 1000778C>T. It included the information of Macro-nutrient intake and energy expenditure according to the genotypes of the FEN1 -10154G>T, FADS2 rs174575C>G, FADS2 rs2727270C>T and FADS3 rs1000778C>T, respectively. No significant genotype-associated differences were observed for total energy intake, proportions of energy intake derived from carbohydrates and fat in each of 4 FADS SNPs.