FADS1 FADS2 variants modify the association between fish intake and the docosahexaenoic acid proportions in human milk1–4

Carolina Molto´-Puigmartı´, Jogchum Plat, Ronald P Mensink, Andre´ Mu¨ller, Euge`ne Jansen, Maurice P Zeegers, and Carel Thijs Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 ABSTRACT phospholipids (3–5). Both fatty acids derive mainly from ma- Background: The encoding D5- and D6-desaturases (FADS1 ternal stores. Therefore, maternal (prolonged) supplementation FADS2 gene cluster) were reported to be associated with n23 with preformed AA and DHA (3, 4, 6) and habitual dietary intake (omega-3) and n26 (omega-6) proportions in human of foods rich in these 2 fatty acids will influence maternal stores plasma, tissues, and milk. Docosahexaenoic acid (DHA) can be and, in turn, the proportions of milk AA and DHA. Red meat, supplied especially by dietary fish or fish oil and synthesized from chicken, and eggs are the main dietary sources of AA, whereas a-linolenic acid through a pathway involving these desaturases. fish (especially oily fish) is the main dietary source of DHA (7). Objective: We evaluated whether FADS gene variants modify the The maternal AA and DHA status can also depend on the effect of maternal fish and fish-oil intake on plasma and milk DHA endogenous synthesis of these 2 fatty acids from the essential proportions. fatty acids, linoleic acid (LA) and a-linolenic acid (ALA), re- Design: FADS1 rs174561, FADS2 rs174575, and intergenic spectively. This conversion is catalyzed by the D5- and rs3834458 single nucleotide polymorphisms were genotyped in D6-desaturases (D5D and D6D), as well as elongases, through 309 women from the KOALA Birth Cohort Study in the Nether- the pathway shown in Figure 1. Both desaturases are expressed lands. Plasma was collected at 36 wk of pregnancy, and milk was in a majority of human tissues, with highest concentrations collected at 1 mo postpartum. Fish and fish-oil intake was assessed shown in liver but also in adipose tissue (8), brain, heart, and by using a food-frequency questionnaire at 34 wk of pregnancy and lung, whereas minor amounts of the desaturases are shown in updated for the week of milk collection. Gene-diet interactions were the placenta, skeletal muscle, kidney, pancreas, and pregnant tested by linear regression analysis. uterus (9, 10). A study in rats (11) indicated that both desa- Results: DHA proportions were lower in women homozygous for turases were also present in the mammary gland, which suggests the minor allele than in women who were homozygous for the major the possibility that the human mammary gland participates in allele (DHA proportions in plasma phospholipids: P , 0.01; DHA the synthesis of LC-PUFAs for their subsequent incorporation proportions in milk: P , 0.05). Fish intake ranged from 0 to 2.5 into human milk. portions of fatty fish/wk, and 12 women took fish-oil supplements during pregnancy. DHA proportions in plasma phospholipids in- 1 creased with increasing fish and fish-oil intake, irrespective of the From the Department of Nutrition and Food Science, Faculty of Phar- genotype. DHA proportions in milk increased only with fish and macy, University of Barcelona, Barcelona, Spain (CM-P); the Departments of Human Biology (JP and RPM), Complex Genetics, Cluster of Genetics, fish-oil intake in the major-allele carriers. and Cell Biology (MPZ), and the Department of Epidemiology (CT), School Conclusion: Lower proportions of DHA in milk from women who of Public Health and Primary Care (CAPHRI), and the School for Nutrition, were homozygous for the minor allele could not be compensated for Toxicology and Metabolism (NUTRIM), Maastricht University, Maastricht, by increasing fish and fish-oil intake, possibly because of limited Netherlands; the Institute of Food Chemistry and Biochemistry, Department incorporation into milk. Am J Clin Nutr 2010;91:1368–76. of Food Chemistry, University of Hamburg, Hamburg, Germany (AM); the Laboratory for Health Protection Research, National Institute for Public Health and the Environment, Bilthoven, Netherlands (EJ); and the Unit of Genetic Epidemiology, Department of Public Health and Epidemiology, INTRODUCTION University of Birmingham, Birmingham, United Kingdom (MPZ). 2 The funders had no role in the study design, data collection and analysis, The supply of n23 (omega-3) and n26 (omega-6) long-chain decision to publish, or preparation of the manuscript. polyunsaturated fatty acids (LC-PUFAs) during gestation to the 3 Collection of breast milk within the KOALA Birth Cohort Study was fetus and postpartum to the newborn is very important because supported by a grant from the Netherlands Organisation for Health Research of the involvement of these LC-PUFAs in many regulatory and Development and the program of Innovative Prevention Research (pre- functions (1) and because LC-PUFA (AA; vention program 1; 210-00-090). CM-P was supported by a research fellow- 20:4n26) and docosahexaenoic acid (DHA; 22:6n23) are ship from the Spanish Ministry of Education and Science. 4 Address correspondence to C Thijs, Department of Epidemiology, constituents of the neural tissue, favoring the cognitive and vi- Maastricht University, PO Box 616, NL-6200 MD Maastricht, Netherlands. sual development of children (2). E-mail: [email protected]. Human milk contains AA and DHA in proportions that sig- Received October 7, 2009. Accepted for publication February 23, 2010. nificantly correlate with those in plasma or erythrocyte-membrane First published online March 24, 2010; doi: 10.3945/ajcn.2009.28789.

1368 Am J Clin Nutr 2010;91:1368–76. Printed in USA. Ó 2010 American Society for Nutrition FADS1 FADS2 GENE VARIANTS, FISH INTAKE, AND DHA 1369 studies of Xie and Innis (16) and Tanaka et al (20), which re- ported the subjects homozygous for the minor allele for the SNPs rs154575 and rs174537, respectively, to have significantly lower proportions of DHA in milk or erythrocyte-membrane phospholipids. Therefore, in the present study we wanted to confirm pre- viously reported associations between FADS SNPs and fatty acid proportions, in particular those of DHA, in plasma phospholi- pids and human milk. The final aim was to investigate whether lower DHA proportions in women homozygous for the minor allele, if observed, could be compensated for by diet and, if so, to what extent. Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021

SUBJECTS AND METHODS

Study design This study is part of the KOALA Birth Cohort Study, a pro- spective birth cohort study in the south of the Netherlands de- scribed previously in detail (21). KOALA is an acronym for Kind, Ouders en gezondheid: Aandacht voor Leefstijl en Aanleg (Child, Parents and Health: Lifestyle and Genetic Constitution). Briefly, participants consisted of pregnant women who were recruited between 2000 and 2002 at 34 wk of gestation. At 36 wk of pregnancy, they were visited at home for collection of venous blood. In addition, from October 2002 participants were asked to consent to breast-milk sampling at 1 mo postpartum. For the present study, we used a group of 309 women, for which we had plasma and breast-milk samples taken between December 2002 and September 2003. Besides these samples, we also moni- tored participants’ dietary habits since they had filled in a food- frequency questionnaire (FFQ) at 34 wk of gestation and a short questionnaire about the intake of fatty fish and fish-oil supple- ments during the week previous to breast milk sampling (22). The study was approved by the Ethics Committee of the Maastricht University/University Hospital Maastricht, and all participants gave written informed consent. All clinical investigations were conducted according to the principles expressed in the Decla- FIGURE 1. Metabolic pathway of fatty acid synthesis and ration of Helsinki. interconversion. Broken lines represent routes not known to be present in humans. D6D, D6-desaturase; D5D, D5-desaturase; D4D, D4-desaturase; E, elongase; b-ox, b-oxidation. Plasma sampling and measurement of fatty acids in plasma The genes that encode for the D5D and D6D are FADS1 phospholipids (GeneID 3992; www.ncbi.nlm.nih.gov/gene) and FADS2 (GeneID Blood was collected into a 10-mL EDTA-coated tube (VT- 9415), respectively, both located on the desaturase gene cluster 100STK; TERUMO, Leuven, Belgium). Plasma was separated on 11q12-q13.1 with a head-to-head orientation from the buffy coat and red blood cells after centrifugation of the (12, 13). There is evidence about the existence of a third desa- EDTA-coated tube at 3000 rpm (1500 · gmax) for 10 min at room turase gene, FADS3 (GeneID 3995), but its implication for the temperature with no brake. The buffy coat was prepared by metabolism of fatty acids is still not known (12). recentrifuging the EDTA-coated tube at 3000 rpm (1580 · gmax) Several single nucleotide polymorphisms (SNPs) in the FADS for 5 min at 4°C with no brake and separating it from red blood genes were reported in humans, and some studies showed cells with a sterile Pasteur capillary pipette of glass. EDTA- associations between FADS SNPs and fatty acids in serum or plasma and buffy coat vials were stored at 280°C until use. plasma phospholipids (14–18), plasma (19, 20), erythrocyte Analyses of fatty acids in plasma phospholipids were per- membranes (15, 17, 18, 20), erythrocyte-membrane phospholi- formed according to the method described in Mamalakis et al pids (16), and adipose tissue (19). In addition, a recent study by (23). Briefly, plasma lipids were extracted with a mixture of Xie and Innis (16) showed that the fatty acid composition of chloroform:methanol (1:1; vol:vol). Different lipid classes were human milk is also related to these FADS SNPs. However, most separated by solid-phase extraction with an amino propyl solid- of the studies did not find a significant association between phase column (Bond-Elut NH2 200 mg; Varian Inc, Middelburg, DHA proportions and FADS SNPs. The only exceptions are the Netherlands). Phospholipid fatty acid methyl esters (FAMEs) 1370 MOLTO´ -PUIGMARTI´ ET AL were prepared with a mixture of 100 lL toluene and 0.5 mL tables, we estimated that each portion of fatty fishes and blue BF3/MeOH (60 min at 100°C in a heating block) and extracted mussels per week contributed equivalent amounts of weekly into hexane. FAMEs were separated on a 100 · 0.25-mm inside n23 LC-PUFA intake, and that each portion of other fishes diameter WCOT (Wall Coated Open Tubular) fused silica cap- contributed one-third of that equivalent (29). From the standard illary column coated with 0.25 lm CP-Select CB (Varian Inc) dosage of fish-oil supplements (which provide ’300 mg n23 with a GC-3900 gas chromatograph (Varian Inc) equipped with LC-PUFAs/1-g capsule), we estimated that they contributed an a CP 8400 autoinjector. A baseline separation of .50 FAMEs equivalent of one portion of fatty fish per week. Therefore, we was accomplished by using mixed FAME standards (Sigma, St estimated the total dietary n23 LC-PUFA intake from seafood Louis, MO), and fatty acids were expressed as the proportion of and supplements as equivalents of portions of fatty fish per week the total fatty acids present in the chromatogram. (FFEq; sum of portions of herring, salmon, mackerel, and blue mussels, plus one-third of total portions of other fishes, plus one equivalent for fish-oil supplement use). Breast-milk sampling and measurement of fatty acids Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 The methods of breast-milk collection and fatty acid analyses were described in detail previously (24). Briefly, mothers were Selection of SNPs instructed to collect their milk in the morning before breast- The selection of SNPs was based on 3 criteria: 1) SNPs should feeding their child from the second breast after finishing the first have been already included in previous studies (14–19, 30) re- and to keep the sample in the refrigerator (at ’4°C) until it was garding essential fatty acid metabolism in humans (6 SNPs in collected on the same day by one of the researchers. the FADS1 gene, 14 SNPs in the FADS2 gene, and 5 SNPs in Lipids were extracted from milk with a mixture of chloroform: intergenic regions fulfilled this requirement); 2) among these methanol (2:1; vol:vol) containing 0.001% butylated hydroxy- SNPs, only those for which an association with the essential toluene, and the lipid fraction was stored at –80°C at the Bio- fatty acid or LC-PUFA concentrations in tissues had been de- bank Maastricht (Maastricht, Netherlands) until use. FAMEs scribed were considered as candidates; and 3) the minor allele were prepared from the lipid fraction and analyzed by using gas frequency had to be 18% so that the expected number of chromatography with flame ionization detection and Ag+-HPLC subjects homozygous for the rare allele in our study population with diode array detection as previously described (24). All was 10. measurements were done in duplicate and reported as averages. According to these criteria, 3 SNPs from the FADS1 FADS2 A total of 60 fatty acids were identified, from 4:0 to 22:6n23, gene cluster were selected: SNP rs174561 in the FADS1 gene, including also some branched chain fatty acids and trans fatty which is thought to have a direct influence on the transcriptional acids. For data analyses, we ended up with a list of 36 fatty level of D5D (14); SNP rs3834458, which is located in the in- acids. Fatty acid concentrations were expressed as the propor- tergenic region between the FADS1 and FADS2 genes and was tion of total milk [weight percentage (wt%) in mg/100 mg]. suggested to play an important role in metabolism because of its The total fat content of human milk was estimated by calcu- location near potential regulatory regions (14, 19, 31); and SNP lating the creamatocrit of all samples following the method rs174575 in the FADS2 gene, which has no known function. developed by Lucas et al (25). SNP genotyping FFQ DNAwas extracted from the buffy coat with the QIAamp DNA The FFQ was included in a self-administered questionnaire in Blood Mini Kit (QIAGEN, Hilden, Germany). SNPs were the 34th wk of pregnancy. It was based on an existing validated genotyped with TaqMan SNP Genotyping assays (Applied Bio- FFQ (26) and consisted of ’160 food items, for which the systems, Foster City, CA) with real-time polymerase chain re- frequency of consumption in the previous month and portion action according to the instructions of the manufacturer. size had to be estimated by the participants. For estimation of the intake of n23 LC-PUFAs, the sources considered were fish and seafood because they contain much higher concentrations of Statistical analyses these fatty acids than other sources (eg, eggs and poultry) (27, Correlations between the proportions of a fatty acid in plasma 28). Of note, n23 LC-PUFA–fortified products were not com- phospholipids and human milk were computed by using Pear- mercially available in the Netherlands at the time of the study. son’s correlation coefficient. Items on seafood intake included different types of the fatty Deviations from Hardy-Weinberg proportions for the geno- fish most frequently eaten in the Netherlands with cold and types of each SNP were tested by using chi-square tests. Genetic warm meals (ie, herring, salmon, mackerel, and blue mussels as data were analyzed for each SNP separately and categorized as separate items), and other (nonfatty and canned) fish types. In homozygous for the major allele, heterozygous, and homozygous addition, participants were asked about the use of fish-oil sup- for the minor allele by using ANOVA to compare mean fatty acid plements during pregnancy. On the day of breast-milk sampling, proportions. An additive model was assumed so that the increase a short questionnaire was taken on the intake of fatty fish and or decrease of fatty acid proportions was tested for each in- fish-oil supplements during the week previous to the milk crement of one minor allele (ie, homozygous for the major allele: sampling. coded 1; heterozygous: coded 2; and homozygous for the minor On the basis of the content of n23 LC-PUFAs [eicosa- allele: coded 3; tested as an interval variable in the ANOVA). A pentaenoic acid (EPA; 20:5n23), docosapentaenoic acid (DPA; Bayesian statistical model for reconstructing uncertain hap- 22:5n23), and DHA] in seafood in the Dutch food-composition lotypes was applied with the PHASE program (version 2.1) (32). FADS1 FADS2 GENE VARIANTS, FISH INTAKE, AND DHA 1371 We used linear regression as a formal statistical test of gene- compared with the carriers of the major allele) but did not reach diet interaction. The proportion of DHA in plasma phospholipids the same level of significance. The proportions of DHA, the end or breast milk was used as the dependent variable, and fish and product of the n23 family, were significantly lower in the plasma fish-oil intake (equivalents), genotypes (2 dummy variables for phospholipids for the minor-allele carriers, whereas osbond acid heterozygous and homozygous for the minor allele, respectively, (22:5n26), its n26 counterpart, could not be quantified (because it with the homozygous for the major allele as the reference group), was below the limit of sensitivity). Substantially analogous results and interaction terms (equivalents · dummy variables) were used were shown in human milk except for GLA and ALA, the pro- as independent variables. The t statistics of the interaction terms portions of which did not significantly vary according to genotype. in the linear regression analysis were used as significance tests The results generally agreed with an additive model [an in- for interactions. The same was done for the sum of EPA and DPA crease or decrease of fatty acid proportions with an increasing (EPA+DPA) as dependent variables. We considered P , 0.05 as number of minor alleles (ie, proportionate differences between statistically significant. DHA and EPA+DPA were log10- carriers homozygous for the major allele, heterozygous, and transformed to normalize their distribution. All data were analyzed homozygous for the minor allele)]. The pattern of decreased or Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 with SPSS 15.0 statistical software (SPSS, Chicago, IL). increased fatty acid proportions was similar for the 3 SNPs, indicating that none of the SNPs were uniquely related to pre- RESULTS cursors or conversion products of D5D or D6D. A total of 309 pregnant women with a mean (6SD) age at 6 delivery of 33.2 3.9 y were included in the study. The fatty Fish-oil intake and n23 LC-PUFA status acid composition of plasma phospholipids and human milk and  the creamatocrit for milk samples from all participants were In the current study, 12 women (3.9%) consumed 2 portions measured. The SNPs rs174561, rs174575, and rs3834458 on the of fatty fish/wk, whereas 117 women (37.9%) reported having FADS1 FADS2 gene cluster were genotyped with a success rate no fatty fish intake at all. The remaining 180 women (58.2%) of 100%. The genotype distribution for each SNP did not deviate had intakes in between these amounts. Also, 12 women (3.9%) from Hardy-Weinberg equilibrium (Hardy-Weinberg equilib- used a fish-oil supplement during pregnancy, and all of them had , rium P values . 0.05; Table 1). Minor allele frequencies ranged eaten 2 portions of fatty fish/wk during pregnancy. Regarding between 25.24% and 30.10% of the study population (Table 1). the FFEq calculated from dietary intake of fish and fish-oil , 6 The total fat content of milk (measured as creamatocrit) did not supplements, 249 women (80.6%) consumed 2 FFEq (mean 6 significantly vary according to genotype. SD intake: 1.19 0.89). FFEq intake was strongly related with DHA proportions in plasma phospholipids (regression coefficient 6 SE: 0.508 6 SNP associations with fatty acids 0.042; P , 10210); ie, the DHA proportions increased 0.508 wt With the use of the genotyping data, we analyzed the asso- % with each portion per week of FFEq. Similarly, FFEq was ciation of the 3 FADS SNPs with the fatty acid composition of strongly related with EPA+DPA proportions in plasma phos- plasma phospholipids and human milk (Table 2). pholipids (regression coefficient 6 SE: 0.184 6 0.023; P , For the n26 family, subjects homozygous for the minor allele 10210). FFEq intake was also strongly related with DHA pro- (for any of the 3 SNPs) had significantly higher proportions of portions in human milk (regression coefficient 6 SE: 0.102 6 LA, 20:2n26 (eicosadienoic acid), and 20:3n26 (dihomo-c- 0.013; P , 10210) and with EPA+DPA proportions in human linolenic acid, DGLA) and lower proportions of 18:3n26(c- milk (regression coefficient 6 SE: 0.053 6 0.007; P , 10210). linolenic acid, GLA), 20:4n–6 (AA), and 22:4n26 (adrenic In the week before breast-milk sampling, 107 women con- acid) in their plasma phospholipids. For the n23 family, the sumed fatty fish, 4 took fish-oil supplements, and 3 consumed results for ALA (18:3n23), EPA (20:5n23), and DPA fatty fish and took fish-oil supplements. When the FFEq intake (22:5n23) in plasma phospholipids were fairly consistent with was adapted by using this information, this did not essentially their n26 counterparts with respect to the direction of the dif- modify the prediction of the proportions of DHA and EPA+DPA ferences by genotype (increased ALA and decreased EPA and in human milk (always P , 10210), and therefore, FFEq intake DPA proportions in subjects homozygous for the minor allele during pregnancy was used as estimate of intake of sources of

TABLE 1 Characteristics of the 3 studied single nucleotide polymorphisms (SNPs)1 SNP Gene M/m alleles Genotype MAF HWE2

% rs174561 FADS1 T/C TT TC CC 166 (53.7)3 125 (40.5) 18 (5.8) 26.05 0.380 rs174575 FADS2 C/G CC CG GG 168 (54.4) 126 (40.8) 15 (4.9) 25.24 0.157 rs3834458 Intergenic T/del TT Tdel deldel 144 (46.6) 144 (46.6) 21 (6.8) 30.10 0.058 1 M/m, major and minor alleles; MAF, minor allele frequency; HWE, Hardy-Weinberg equilibrium; del, deletion. 2 P values of deviation from HWE among all subjects were tested by chi-square tests. 3 Number of subjects for each genotype; percentage in parentheses (all such values). 1372 MOLTO´ -PUIGMARTI´ ET AL

TABLE 2 Fatty acid composition (percentage of total fatty acids identified) of plasma phospholipids and human milk in women classified according to their genotype and Pearson’s correlation coefficients between plasma phospholipids and human milk fatty acids1 Plasma phospholipid proportions Human milk proportions

SNP Fatty acid MM Mm mm P MM Mm mm P Correlation r rs174561 16:1n27 0.73 6 0.202 0.76 6 0.22 0.70 6 0.18 0.335 2.49 6 0.80 2.43 6 0.75 2.37 6 0.76 0.730 0.3413 rs174575 0.73 6 0.20 0.77 6 0.23 0.67 6 0.17 0.067 2.54 6 0.82 2.39 6 0.70 2.08 6 0.62 0.0474 rs3834458 0.73 6 0.21 0.77 6 0.22 0.69 6 0.18 0.140 2.51 6 0.83 2.42 6 0.72 2.29 6 0.73 0.386 rs174561 cis 18:1n27 1.13 6 0.17 1.19 6 0.17 1.18 6 0.16 0.0114 ———— rs174575 1.16 6 0.18 1.17 6 0.17 1.17 6 0.15 0.776 — — — — rs3834458 1.14 6 0.18 1.18 6 0.17 1.18 6 0.15 0.063 — — — — rs174561 cis 20:1n29 0.17 6 0.04 0.19 6 0.05 0.22 6 0.04 ,0.0014 0.36 6 0.21 0.36 6 0.21 0.37 6 0.24 0.982 20.070 4 rs174575 0.18 6 0.04 0.19 6 0.05 0.22 6 0.03 0.001 0.35 6 0.21 0.38 6 0.22 0.35 6 0.24 0.548 Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 rs3834458 0.17 6 0.04 0.19 6 0.05 0.22 6 0.04 ,0.0014 0.36 6 0.20 0.36 6 0.22 0.35 6 0.23 0.932 rs174561 cis 22:1n–9 0.19 6 0.08 0.20 6 0.07 0.26 6 0.10 0.0014 ———— rs174575 0.19 6 0.08 0.20 6 0.08 0.23 6 0.10 0.098 — — — — rs3834458 0.19 6 0.08 0.20 6 0.07 0.24 6 0.11 0.0074 ———— rs174561 18:2n26 (LA) 20.33 6 2.31 21.60 6 2.03 23.37 6 2.16 ,0.0014 13.62 6 3.47 13.75 6 3.31 15.32 6 4.17 0.140 0.3773 rs174575 20.59 6 2.34 21.34 6 2.23 23.08 6 1.66 ,0.0014 13.69 6 3.40 13.66 6 3.41 15.60 6 4.15 0.109 rs3834458 20.25 6 2.22 21.43 6 2.16 23.40 6 2.06 ,0.0014 13.50 6 3.41 13.80 6 3.38 15.46 6 3.97 0.0394 rs174561 20:2n26 0.42 6 0.07 0.47 6 0.07 0.55 6 0.05 ,0.0014 ———— rs174575 0.44 6 0.07 0.46 6 0.08 0.53 6 0.06 ,0.0014 ———— rs3834458 0.42 6 0.06 0.47 6 0.07 0.55 6 0.06 ,0.0014 ———— rs174561 18:3n26 (GLA) 0.05 6 0.03 0.04 6 0.03 0.04 6 0.02 0.0034 0.14 6 0.05 0.14 6 0.05 0.13 6 0.05 0.329 0.1573 rs174575 0.05 6 0.03 0.05 6 0.03 0.04 6 0.03 0.128 0.14 6 0.05 0.14 6 0.05 0.16 6 0.04 0.328 rs3834458 0.05 6 0.03 0.04 6 0.02 0.04 6 0.02 0.0014 0.14 6 0.04 0.14 6 0.05 0.14 6 0.05 0.920 rs174561 20:3n26 (DGLA) 3.26 6 0.47 3.53 6 0.54 3.81 6 0.58 ,0.0014 0.45 6 0.10 0.49 6 0.12 0.55 6 0.15 ,0.0014 0.4373 rs174575 3.26 6 0.45 3.53 6 0.57 3.95 6 0.52 ,0.0014 0.46 6 0.10 0.49 6 0.11 0.57 6 0.14 0.0014 rs3834458 3.26 6 0.47 3.49 6 0.55 3.80 6 0.55 ,0.0014 0.46 6 0.10 0.48 6 0.12 0.53 6 0.15 0.0084 rs174561 20:4n26 (AA) 7.94 6 1.01 7.01 6 0.87 5.75 6 0.85 ,0.0014 0.54 6 0.09 0.49 6 0.08 0.39 6 0.06 ,0.0014 0.4973 rs174575 7.82 6 1.13 7.10 6 0.87 6.00 6 0.90 ,0.0014 0.54 6 0.09 0.49 6 0.08 0.41 6 0.06 ,0.0014 rs3834458 8.05 6 0.98 7.06 6 0.88 5.86 6 0.85 ,0.0014 0.55 6 0.09 0.49 6 0.08 0.40 6 0.06 ,0.0014 rs174561 22:4n26 0.45 6 0.09 0.43 6 0.09 0.35 6 0.09 ,0.0014 0.14 6 0.03 0.13 6 0.03 0.11 6 0.03 ,0.0014 0.4043 rs174575 0.45 6 0.09 0.43 6 0.09 0.38 6 0.07 0.0254 0.14 6 0.03 0.13 6 0.03 0.10 6 0.03 ,0.0014 rs3834458 0.46 6 0.09 0.43 6 0.09 0.36 6 0.09 ,0.0014 0.14 6 0.03 0.13 6 0.03 0.11 6 0.03 ,0.0014 rs174561 22:5n26 — — — — 0.07 6 0.02 0.07 6 0.02 0.06 6 0.03 0.457 rs174575 — — — — 0.07 6 0.02 0.07 6 0.02 0.07 6 0.03 0.816 rs3834458 — — — — 0.07 6 0.02 0.07 6 0.02 0.06 6 0.02 0.206 rs174561 18:3n23 (ALA) 0.27 6 0.09 0.28 6 0.08 0.30 6 0.08 0.230 0.98 6 0.35 1.00 6 0.38 0.96 6 0.33 0.898 0.2143 rs174575 0.27 6 0.09 0.28 6 0.08 0.28 6 0.08 0.876 0.99 6 0.39 0.98 6 0.33 0.98 6 0.35 0.961 rs3834458 0.27 6 0.09 0.28 6 0.08 0.29 6 0.08 0.210 0.97 6 0.35 1.01 6 0.38 0.96 6 0.32 0.638 rs174561 18:4n23 — — — — 0.04 6 0.05 0.05 6 0.08 0.04 6 0.08 0.381 rs174575 — — — — 0.04 6 0.06 0.05 6 0.08 0.05 6 0.06 0.814 rs3834458 — — — — 0.04 6 0.06 0.04 6 0.08 0.05 6 0.08 0.682 rs174561 20:4n23 — — — — 0.07 6 0.02 0.08 6 0.03 0.07 6 0.02 0.155 rs174575 — — — — 0.07 6 0.02 0.08 6 0.03 0.07 6 0.02 0.434 rs3834458 — — — — 0.07 6 0.02 0.08 6 0.03 0.07 6 0.02 0.0404 rs174561 20:5n23 (EPA) 0.47 6 0.32 0.43 6 0.35 0.41 6 0.33 0.580 0.11 6 0.06 0.10 6 0.06 0.06 6 0.02 0.0134 0.2893 rs174575 0.45 6 0.27 0.46 6 0.41 0.31 6 0.21 0.269 0.10 6 0.06 0.10 6 0.07 0.06 6 0.01 0.0484 rs3834458 0.46 6 0.27 0.44 6 0.39 0.39 6 0.31 0.617 0.11 6 0.06 0.10 6 0.07 0.06 6 0.02 0.0134 rs174561 22:5n23 (DPA) 0.56 6 0.11 0.55 6 0.12 0.53 6 0.13 0.702 0.24 6 0.07 0.22 6 0.06 0.18 6 0.03 ,0.0014 0.3233 rs174575 0.56 6 0.12 0.55 6 0.12 0.49 6 0.09 0.076 0.24 6 0.07 0.22 6 0.06 0.18 6 0.03 0.0014 rs3834458 0.56 6 0.11 0.55 6 0.12 0.53 6 0.12 0.398 0.24 6 0.07 0.22 6 0.06 0.18 6 0.03 ,0.0014 rs174561 22:6n23 (DHA) 3.93 6 0.78 3.72 6 0.80 3.29 6 0.73 0.0024 0.44 6 0.23 0.41 6 0.22 0.30 6 0.07 0.0444 0.4813 rs174575 3.94 6 0.81 3.70 6 0.76 3.24 6 0.64 0.0014 0.43 6 0.22 0.42 6 0.22 0.29 6 0.07 0.064 rs3834458 3.95 6 0.79 3.74 6 0.79 3.26 6 0.68 ,0.0014 0.43 6 0.23 0.42 6 0.22 0.30 6 0.07 0.0294 1 The numbers of women who were homozygous for the major allele (MM), heterozygous (Mm), and homozygous for the minor allele (mm) were as follows: 166, 125, and 18, respectively, for single nucleotide polymorphism (SNP) rs174561; 168, 126, and 15, respectively, for SNP rs174575; and 144, 144, and 21, respectively, for SNP rs3834458. LA, linoleic acid; DGLA, dihomo-c-linolenic acid; GLA, c-linolenic acid; AA, arachidonic acid; ALA, a-linolenic acid; EPA, ; DPA, docosapentaenoic acid; DHA, docosahexaenoic acid. Differences between plasma phospholipids and human milk fatty acid proportions between genotypes were assessed by using ANOVA assuming an additive model (FADS SNP genotypes coded 1, 2, and 3 for homozygous for the major allele, heterozygous, and homozygous for the minor allele, respectively). 2 Mean 6 SD (all such values). 3 Significant at 0.01 (2-tailed). 4 Significant difference (P , 0.05). FADS1 FADS2 GENE VARIANTS, FISH INTAKE, AND DHA 1373 n23 LC-PUFAs during pregnancy and lactation in further In human milk, by contrast, significant relations between FFEq analyses. The total fat content of breast milk was not associated intake and DHA and EPA+DPA proportions were shown in the with FFEq intake. carriers of the major allele but not in the women homozygous for the minor allele; ie, proportions of these fatty acids in milk did not increase with higher FFEq intakes in the minor allele homo- Gene-diet interaction for fish intake and FADS SNPs zygotes (Figure 2, B and D; flat regression lines with slope near zero). In the test for the statistical interaction, P values for As expected, there were no significant differences in FFEq differences between slopes of different genotypes of the SNP intake between women with different genotypes. DHA and EPA rs174575 in human milk were P = 0.077 for DHA and P = 0.019 +DPA proportions in plasma phospholipids increased with FFEq for EPA+DPA (regression coefficient 6 SE for interaction = intake in all 3 genotypes for rs174575 (Figure 2, A and C). 20.081 6 0.046 for DHA and 20.084 6 0.019 for EPA+DPA). Although DHA proportions (Figure 2A) were lower in the Results for the other 2 SNPs were essentially the same (P values women homozygous for the minor allele over the whole range of for gene-diet interaction: 0.090 for DHA and 0.036 for EPA FFEq intake, the DHA proportions increased with higher FFEq Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 +DPA for SNP rs174561; 0.071 for DHA; and 0.021 for EPA to a similar extent for the 3 genotypes (the slopes of the re- +DPA for SNP rs3834458). gression lines were similar); this was also true for EPA+DPA proportions (Figure 2B). The differences of slopes between the minor genotypes compared with major genotypes were tested by assessing the interaction with FFEq intake in linear regression Haplotype association with DHA and gene-diet interaction (with 10log transformation of DHA and EPA+DPA to derive Analyzing the data at the allelic level, we showed 7 haplotypes a normal distribution); P values for this test of gene-diet in- for combinations of the SNPs rs174561, rs174575, and teraction were .0.05, indicating no significant differences be- rs3834458. Ordered according to the number of alleles in each tween the slopes. Results for the other 2 SNPs were essentially group and considering A, B, and C to represent the major alleles the same (data not shown), with gene-diet interaction P . 0.05. for the SNPs rs174561, rs174575 and rs3834458, respectively,

FIGURE 2. Docosahexaenoic acid (DHA) and the sum of eicosapentaenoic acid and docosapentaenoic acid (EPA+DPA) as a function of fatty fish or fish- oil intake for only one single nucleotide polymorphism. The percentages of DHA and EPA+DPA in plasma phospholipids (A and C) and in human milk (B and D) according to the weekly intake of equivalents of fatty fish during pregnancy and lactation for women classified by their genotype for the FADS2 single nucleotide polymorphism rs174575. Regression lines for subjects homozygous for the major allele (n = 168; blue diamonds), heterozygous (n = 126; red hyphens), and homozygous for the minor allele (n = 15; green crosses) are represented by blue, red, and green lines, respectively. Regression formulas for subjects homozygous for the major allele (blue), heterozygous (red), and homozygous for the minor allele (green) are as follows: A: (blue) y = 3.303 + 0.521x, (red) y =3.143+0.479x, and (green) y =2.755+0.441x;B:(blue)y = 0.302 + 0.106x, (red) y = 0.294 + 0.107x, and (green) y =0.277+0.014x; C: (blue) y =0.793+ 0.183x, (red) y =0.792+0.183x, and (green) y =0.600+0.190x;andD:(blue)y =0.274+0.057x, (red) y = 0.259 + 0.053x, and (green) y =0.25132 0.011x. 1374 MOLTO´ -PUIGMARTI´ ET AL and a, b, and c to represent the minor alleles for the SNPs that each SNP conveys most of the information contained in the rs174561, rs174575 and rs3834458, respectively, the haplotypes other SNPs. In general, previous studies point to a lower activity were: ABC (n = 428), abc (n = 128), aBc (n = 27), abC (n = 26), or transcription of the desaturases to explain the observed in- Abc (n = 4), ABc (n = 3) and aBC (n = 2). These totals add up to creases in the proportions of the substrates of the mentioned 618 (2 alleles for each SNP · 309 subjects). The alleles of the enzymes and the subsequent decreases in the proportions of their markers were in some linkage disequilibrium (D . 0.10, R2 . products. Our results follow this same pattern, with increased 0.19), which may explain why the abc haplotype was far more proportions of substrates (LA and DGLA), decreased pro- common (128 of 618; 20.7% observed) than expected from portions of products (GLA, AA, EPA), and changes in other random combination of the minor allele frequencies (2.0% ex- fatty acid proportions in the expected direction (decreased pro- pected; this is the product of minor allele frequencies shown in portions of 22:4n26, DPA, and DHA). By the time that our Table 1). To assess whether the results were different at the study was finished, an article (20) reported that the most sig- haplotype level than at the genotype level (as presented in Table nificant SNP for AA was rs174537 on the FADS1 FADS2 gene 2), we combined the 2 most common haplotypes across the cluster. However, this marker was not included in our list of Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 maternal and paternal to create the following 3 candidate SNPs because none of the previously published works haplogenotype groups: subjects homozygous for all major al- genotyped it. leles on all 3 SNPs (ABC/ABC, n = 140 subjects), subjects One of the most important implications for the development of homozygous for all minor alleles (abc/abc, n = 12), and subjects newborns is that the percentages of AA and DHA in plasma heterozygotes for all SNPs (ABC/abc and vice versa, n = 97), phospholipids and milk of the mothers homozygous for the minor leaving out the group with other combinations (n = 60). allele were lower compared with the percentages for carriers of Results for the haplogenotypes were very similar compared the major allele. However, we expected that higher fish and fish- with results for the single SNPs, with the homozygous minor oil intake could permit those women homozygous for the minor alleles (abc/abc) having lower proportions of products of the allele to compensate for their lower desaturase activity/expres- desaturases than of substrates compared with the homozygous sion (because of incorporation of the preformed fatty acids in- wild types (ABC/ABC), and again, the results for the hetero- cluding DHA in fatty fish and fish-oil supplements) and to reach zygotes were in between (see supplemental Table 1 under DHA proportions not far below the DHA proportions observed “Supplemental data” in the online issue). for the carriers of the major allele, both in their plasma and milk. The gene-diet interaction was also very similar for the hap- Indeed, in plasma phospholipids, DHA proportions increased logenotypes compared with the individual SNPs genotypes (see with fish and fish-oil intake for all of the 3 genotypes and to supplemental Figure 1 and Table 2 under “Supplemental data” in a similar extent (parallel slope in Figure 2A), and this was also the online issue). Contrasting the homozygous haplogenotype apparent for EPA+DPA proportions (Figure 2C). By contrast, groups ABC/ABC and abc/abc, their slopes of increasing DHA proportions of DHA and EPA+DPA in milk only increased with proportions in human milk with increasing FFEq diverged fish and fish-oil intake in the major-allele carriers, whereas no similarly as shown in Figure 2B, and the test for interaction gave increase was seen in the mothers homozygous for the minor a P value of 0.081. For the proportions of EPA+DPA in human allele, leading to an augmentation of the genetic differences at milk, this divergence of slopes between haplogenotype groups higher FFEq intakes. The finding of a different response to the ABC/ABC and abc/abc was similar to the divergence shown in intake of n23 LC-PUFA depending on the FADS1 FADS2 ge- Figure 2D, and was significant (P = 0.022). For DHA and EPA notype can be described as a gene-diet interaction, and may +DPA in plasma phospholipids, the slopes were parallel for the indicate that mothers homozygous for the minor allele take no haplogenotype groups, similar to the results for the single SNPs advantage of fish-oil intake or at least have less advantage than (Figure 2, A and C), and not significantly different (P . 0.35). the major-allele carriers. We found it surprising that this response differed between plasma phospholipids and human milk and that the response was DISCUSSION similar for DHA and EPA+DPA proportions. We speculate that, In the present study we provided evidence from a group of 309 in the minor allele homozygous mothers, there is an alteration in women that LC-PUFAs proportions in their plasma phospholi- the incorporation of n23 LC-PUFAs into milk, and that the pids at the 34th wk of pregnancy and in their milk at 1 mo FADS gene cluster markers could be linked somehow to fatty postpartum depended on their genotypes for 3 selected SNPs on acid transportation (for instance, related to common genetic reg- the FADS1 FADS2 gene cluster. In particular, we observed that ulation). In a recent genome-wide association study, Tanaka et al DHA proportions in plasma phospholipids and milk were de- (20) showed EPA and DHA in plasma to be associated with the creased in the women homozygous for the minor allele com- genotype for SNP rs2277324 in the gene SLC26A10 (GeneID pared with the proportions in carriers of the major allele, and, 65012) on chromosome 12, a gene encoding for a solute carrier for the first time to our knowledge, showed that a higher fish (or family member. Other possible genetic candidates are those that fish-oil) intake compensated for the lower DHA proportions in code for fatty acid binding (FABP) and fatty acid the women’s plasma phospholipids but not in their milk. transport proteins (FATP), but all known variants are located on Our results of the fatty acid composition of plasma phos- different chromosomes than the FADS gene cluster (FABP5, pholipids and human milk by FADS gene variants agree well expressed in the mammary gland, for instance, is located on with those of previous studies (14–16, 18, 19). Also in agree- chromosome 8). Interestingly, an FABP5 homolog, FABP5 like ment with other studies (14–17), the haplogenotype analyses 7, is located at chromosome 11q12.1, close to FADS (11q12.2), confirmed the results of the single markers. This agrees with the but this is considered a nonfunctional pseudogene. Finally, a fact that the SNPs were highly linked with each other and shows polymorphism in FABP2 was shown to be related to D6D FADS1 FADS2 GENE VARIANTS, FISH INTAKE, AND DHA 1375 activity in obese children (33), but FABP2 is not known to be REFERENCES expressed in the mammary gland. 1. Gottrand F. Long-chain polyunsaturated fatty acids influence the im- If incorporation of preformed n23 LC-PUFAs into milk is mune system of infants. J Nutr 2008;138:1807S–1812S. 2. Lauritzen L, Hansen HS, Jorgensen MH, Michaelsen KF. The essenti- truly limited in mothers with certain FADS variants, this could ality of long chain n23 fatty acids in relation to development and imply that it is not helpful for these mothers to increase their fish function of the brain and retina. Prog Lipid Res 2001;40:1–94. or fish-oil intake to raise the content of n23 LC-PUFAs (in 3. Weseler AR, Dirix CEH, Bruins MJ, Hornstra G. Dietary arachidonic particular DHA) in their milk, at least up to the studied upper acid dose-dependently increases the arachidonic acid concentration in human milk. J Nutr 2008;138:2190–7. range of 3 portions of fatty fish/wk. In any case, these results 4. Bergmann RL, Haschke-Becher E, Klassen-Wigger P, et al. Supple- should be replicated in other studies, preferably by using stable mentation with 200 mg/day docosahexaenoic acid from mid-pregnancy isotopes to compute conversion rates and to calculate fluxes. In through lactation improves the docosahexaenoic acid status of mothers addition, our results are based on a limited number of subjects with a habitually low fish intake and of their infants. Ann Nutr Metab 2008;52:157–66. with the minor allele genotype and with a maximum intake of 3 5. Arterburn LM, Hall EB, Oken H. Distribution, interconversion, and dose equivalents of fatty fish/wk. Future supplementation studies 2 response of n 3 fatty acids in humans. Am J Clin Nutr 2006;83(suppl): Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 could evaluate whether higher amounts of fatty fish intake or 1467S–76S. n23 LC-PUFA supplementation than currently recommended 6. van Goor SA, Dijck-Brouwer DAJ, Hadders-Algra M, et al. Human milk arachidonic acid and docosahexaenoic acid contents increase following can overcome the limited incorporation into the milk of mothers supplementation during pregnancy and lactation. Prostaglandins Leukot homozygous for the minor allele. Essent Fatty Acids 2009;80:65–9. The majority of the population of the Netherlands and other 7. Muskiet FAJ, van Goor SA, Kuipers RS, et al. Long-chain poly- Western European countries do not reach the recommended fatty unsaturated fatty acids in maternal and infant nutrition. Prostaglandins fish intake of 2 times/wk. For those with low fish intake (0 or Leukot Essent Fatty Acids 2006;75:135–44. 8. Sjo¨gren P, Sierra-Johnson J, Gertow K, et al. Fatty acid desaturases in 1 time/wk), milk DHA proportions do not greatly differ according human adipose tissue: relationships between gene expression, desatu- to genotype (Figure 2B). However, at the amount of 2 portions of ration indexes and insulin resistance. Diabetologia 2008;51:328–35. fatty fish/wk, DHA proportions already differed in 0.2 wt% 9. Cho HP, Nakamura M, Clarke SD. Cloning, expression, and fatty acid regulation of the human delta-5 desaturase. J Biol Chem 1999;274: between subjects homozygous for the minor allele and subjects 37335–9. with the other 2 genotypes (Figure 2B; mean ’0.3 wt% for 10. Cho HP, Nakamura MT, Clarke SD. Cloning, expression, and nutritional minor-allele homozygous subjects and ’0.5 wt% for major- regulation of the mammalian delta-6 desaturase. J Biol Chem 1999;274: allele carriers). This means that children at risk of an insufficient 471–7. 11. Rodriguez-Cruz M, Tovar AR, Palacios-Gonzalez B, del Prado M, DHA supply are those whose mothers have low fish and fish-oil Torres N. Synthesis of long-chain polyunsaturated fatty acids in lactat- intake and the children from mothers homozygous for the minor ing mammary gland: role of Delta-5 and Delta-6 desaturases, SREBP-1, allele, regardless of fish intake or supplementation at the rec- PPAR-alpha, and PGC-1. J Lipid Res 2006;47:553–60. ommended amounts. In addition, the child’s own FADS genotype 12. Marquardt A, Sto¨hr H, White K, Weber BHF. cDNA cloning, genomic structure, and chromosomal localization of three members of the human may be an important factor determining his or her LC-PUFA family. Genomics 2000;66:175–83. status. Caspi et al (30) suggested an association between chil- 13. Nakamura MT, Nara TY. Structure, function, and dietary regulation of dren’s genotypes for the SNP rs174575 on FADS2 and their AA delta 6, delta 5, and delta 9 desaturases. Annu Rev Nutr 2004;24: and DHA status. This observation and the findings of the present 345–76. 14. Schaeffer L, Gohlke H, Muller M, et al. Common genetic variants of the study raise new questions: eg, What would the DHA status of FADS1 FADS2 gene cluster and their reconstructed haplotypes are as- a breastfed child be if mother and child shared the minor allele sociated with the fatty acid composition in phospholipids. Hum Mol homozygous genotype? Genet 2006;15:1745–56. In conclusion, we confirmed the previously reported associ- 15. Malerba G, Schaeffer L, Xumerle L, et al. SNPs of the FADS gene ations between FADS SNPs and fatty acid profiles in human milk cluster are associated with polyunsaturated fatty acids in a cohort of patients with cardiovascular disease. Lipids 2008;43:289–99. within a different population and by using a larger dataset. To 16. Xie L, Innis SM. Genetic variants of the FADS1 FADS2 gene cluster are our knowledge, the major novelty in this study is the gene-diet associated with altered (n26) and (n23) essential fatty acids in plasma interaction, which was shown to differ between plasma phos- and erythrocyte phospholipids in women during pregnancy and in breast pholipids and human milk for DHA and EPA+DPA proportions. milk during lactation. J Nutr 2008;138:2222–8. 17. Martinelli N, Girelli D, Malerba G, et al. FADS genotypes and desa- If confirmed by supplementation studies, a challenge is to dis- turase activity estimated by the ratio of arachidonic acid to linoleic acid entangle the possible mechanisms of incorporation of n23 LC- are associated with inflammation and coronary artery disease. Am J Clin PUFAs in human milk and its genetic regulation. Also, further Nutr 2008;88:941–9. investigation is warranted on whether these findings may have 18. Rzehak P, Heinrich J, Klopp N, et al. Evidence for an association be- tween genetic variants of the fatty acid desaturase 1 fatty acid desaturase any functional consequences for the child. If so, a mother’s diet 2 (FADS1 FADS2) gene cluster and the fatty acid composition of and the combination of FADS genotypes of the mother and child erythrocyte membranes. Br J Nutr 2009;101:20–6. would be crucial for the child’s development and health. 19. Baylin A, Ruiz-Narvaez E, Kraft PCH. a-Linolenic acid, D6-desaturase gene polymorphism, and the risk of nonfatal myocardial infarction. Am We thank Maurice Konings for technical assistance in the genotyping and J Clin Nutr 2007;85:554–60. Hans Cremers for technical assistance in the fatty acid analyses. 20. Tanaka T, Shen J, Abecasis GR, et al. Genome-wide association study of The authors’ responsibilities were as follows—CM-P, JP, RPM, and CT: plasma polyunsaturated fatty acids in the InCHIANTI Study. PLoS conception and design of the study; CM-P: genotyping; AM and EJ: fatty acid Genet 2009;5:e1000338. 21. Kummeling I, Thijs C, Penders J, et al. Etiology of atopy in infancy: analyses; CM-P, JP, RPM, MPZ, and CT: analysis and interpretation of the the KOALA Birth Cohort Study. Pediatr Allergy Immunol 2005;16: data; MPZ: assistance in genetic data analysis; CM-P and CT: writing of 679–84. the manuscript draft; and CM-P, JP, RPM, MPZ, and CT: critical review 22. Snijders BE, Damoiseaux JG, Penders J, et al. Cytokines and soluble of the manuscript. None of the authors had a personal or financial conflict CD14 in breast milk in relation with atopic manifestations in mother and of interest. infant (KOALA Study). Clin Exp Allergy 2006;36:1609–15. 1376 MOLTO´ -PUIGMARTI´ ET AL

23. Mamalakis G, Kiriakakis M, Tsibinos G, et al. Lack of an association of docosahexaenoic acid: impact on public health. Int J Food Sci Nutr depression with n23 polyunsaturated fatty acids in adipose tissue and serum 2008;59:491–501. phospholipids in healthy adults. Pharmacol Biochem Behav 2008;89:6–10. 29. NEVO-tabel: Nederlands Voedingsstoffenbestand 2001. [Dutch food 24. Rist L, Mueller A, Barthel C, et al. Influence of organic diet on the composition table.] Stichting Nederlands Voedingsstoffenbestand 2001. amount of conjugated linoleic acids in breast milk of lactating women in The Hague, Netherlands: Voedingscentrum, 2001. the Netherlands. Br J Nutr 2007;97:735–43. 30. Caspi A, Williams B, Kim-Cohen J, et al. Moderation of breastfeeding 25. Lucas A, Gibbs JA, Lyster RL, Baum JD. Creamatocrit: simple clinical effects on the IQ by genetic variation in fatty acid metabolism. Proc Natl technique for estimating fat concentration and energy value of human Acad Sci USA 2007;104:18860–5. milk. BMJ 1978;1:1018–20. 31. Truong H, DiBello JR, Ruiz-Narvaez E, Kraft P, Campos H, Baylin A. 26. Grootenhuis PA, Westenbrink S, Sie CMTL, De Neeling JN, Kok FJ, Does genetic variation in the D6-desaturase promoter modify the asso- Bouter LM. A semiquantitative food frequency questionnaire for use in ciation between a-linolenic acid and the prevalence of metabolic syn- epidemiologic research among the elderly: validation by comparison drome? Am J Clin Nutr 2009;89:920–5. with dietary history. J Clin Epidemiol 1995;48:859–68. 32. Stephens M, Smith NJ, Donelly P. A new statistical method for haplo- 27. Hibbeln JR, Nieminen LR, Blasbalg TL, Riggs JA, Lands WE. Healthy type reconstruction from population data. Am J Hum Genet 2001;68: intakes of n23 and n26 fatty acids: estimations considering worldwide 978–89. diversity. Am J Clin Nutr 2006;83(suppl):1483S–93S. 33. Okada T, Sato-Furuhashi N, Iwata F, Mugishima H. The interaction 28. Bourre JM, Paquotte PM. Contributions (in 2005) of marine and fresh between intestinal fatty acid-binding 2 polymorphism and Downloaded from https://academic.oup.com/ajcn/article/91/5/1368/4597347 by guest on 28 September 2021 water products (finfish and shellfish, seafood, wild and farmed) to the delta-6 desaturase activity in obese children. Am J Clin Nutr 2008;87: French dietary intakes of vitamins D and B12, selenium, iodine and 1066–7.