DHA intake interacts with ELOVL2 and ELOVL5 genetic variants to influence polyunsaturated fatty acids in human milk

Endogenous synthesis of PUFAs is mediated by genes controlling fatty acid elongases 2 and 5 ( ELOVL2 and ELOVL5 ) and by exogenous DHA intake. Associations between elongases and PUFA levels probably involve genetic variants of ELOVL and changes in DHA intake, but data about their combined effect on PUFA levels are sparse. We hypothesized that each factor would directly affect PUFAs and that interactions between haplotypes and DHA intake would influence PUFAs. We explored four levels of DHA intake in pregnant Chinese Han women and 10 SNPs in the ELOVL genes to determine associations with PUFAs in breast milk. The SNP, rs3798713, and 3-SNP haplotype (rs2281591, rs12332786, and rs3798713) in ELOVL2 were associated with linoleic acid (LA) concentrations. However, carriers of the 3-SNP haplotype with higher DHA intake (second quartile, 14.58 to 43.15 mg/day) had higher concentrations of LA, arachidonic acid, EPA, and DHA compared with the interaction baseline. In ELOVL5 , five SNPs (rs2294867, rs9357760, rs2397142, rs209512, and rs12207094) correlated with PUFA changes. Compared with those who had the 5-SNP haplotype C-A-C-G-A and low DHA intake (<14.58 mg/day), carriers with other haplotypes (A-A-C-A-A or C-A-C-A-A) and high DHA intake ( ≥ 118.82 mg/day) had increased EPA levels after adjustments for age and body mass index. This study showed that maternal genetic variants in ELOVL2 and ELOVL5 were associated with PUFA levels in breast milk and that the combination of SNP haplotypes and higher DHA intake increased PUFA concentrations.


INTRODUCTION
Polyunsaturated fatty acids (PUFAs), especially long-chain PUFAs (LC-PUFAs), are essential nutrients for human health, and are associated with the development of the human brain and neurotransmitter function (1)(2)(3). In particular, the n-3 fatty acid DHA (22:6n-3), present in breast milk, has been identified in animal and human studies as being crucial for the development of the central nervous system during fetal life and early infancy (2,4). Findings from epidemiological and observational studies substantiate the association between maternal seafood consumption in pregnancy and breastfeeding and improved infant neurodevelopment (5)(6)(7)(8)(9). This may be explained partly by the resultant increase in the early supply of LC-PUFAs, especially DHA, which accumulate in the brain during the early growth spurt in children (2,10). Pregnant women who consumed high amounts of n-3 LC-PUFA in their diet showed positive effects on pregnancy outcomes, such as longer gestational duration and greater birth weight (11). Additionally, LC-PUFAs may increase infant growth and enhance the short-and long-term development of the offspring (11)(12)(13). A randomized controlled trial study showed that supplementation with n-3 LC-PUFA in the third trimester of pregnancy reduced the absolute risk of persistent wheeze or asthma and infections of the lower respiratory tract in offspring (14). The fetus is mainly supplied with LC-PUFAs by transfer from the maternal circulation via the placenta (15). Therefore, an adequate supply of LC-PUFA during pregnancy is critical for fetal life onwards.
Elongation of LC-PUFAs in the n-3 family is made possible by enzymes called elongases, which are encoded by the elongation of a very long-chain fatty acid (ELOVL) gene family on by guest, on April 28, 2019 www.jlr.org Downloaded from 4 chromosome 6. Elongases catalyze the elongation of aliphatic carbon chains leading to the formation of LC-PUFAs (16,17). Fatty acid elongase-2 and the fatty acid elongase-5, encoded by the ELOVL2 (6p24.2) and ELOVL5 (6p12.1) gene, respectively, are involved in LC-PUFA synthesis. Plasma percentages of n-3 LC-PUFA have been shown to be associated with single nucleotide polymorphisms (SNPs) in ELOVL2 (18) and ELOVL5 (19,20), and the enzyme activities are influenced by SNPs within the ELOVL gene family after supplementation of n-3 LC-PUFA (21,22).
The mechanisms underlying the observed associations between elongases and PUFAs levels are inconsistent, which probably involve genetic variants within ELOVL and changes in levels of DHA intake. Consequently, we hypothesized that ELOVL polymorphisms and DHA intake would each be directly associated with PUFA concentrations. We also hypothesized that the interactions of haplotypes within ELOVL2 and ELOVL5 with DHA consumption during pregnancy may influence PUFA concentration in the breast milk of healthy lactating Chinese Han women.

Questionnaire survey and breast milk collection
We gave each participant a face to face interview and a semi-structured food frequency questionnaire (FFQ). The FFQ was used to assess the dietary intake of enrolled subjects during their pregnancy. It included specific questions about consumption of sources containing DHA, such as freshwater fish, seafood, canned tuna, etc. To better understand the data of DHA intake, in the interview, we also asked participants the DHA supplement's brand, DHA dietary intakes were calculated as mg per day. Twenty milliliter of breast milk was collected between 9:00 a.m. and 11:00 a.m. on one day between the 22nd and 25th day after delivery. The first few drops of milk were discarded, and then the mature breast milk was collected. The samples were stored at-80 °C (24), and detected in one month averagely.

Fatty acid analysis
The levels of eight kinds of fatty acids in breast milk were determined by direct methylation

SNPs selection and genotyping
SNPs in ELOVL2 and ELOVL5 were identified using the International HapMap Project SNP database and the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/snp/). Selected SNPs of the ELOVL gene cluster (rs2281591, rs12332786, rs3798713, rs3778166, rs9468304, rs2294867, rs9357760, rs2397142, rs209512, rs12207094) have been genotyped using Sequenom Mass Array system (BO MIAO Biological Technological Company, Beijing) with validated primers and the genotyping success rate was above 96%. The genomic location of ELOVL2 and ELOVL5 gene region are on the chromosome 6(10.99-11.05 Mb and 53.28-53.34 Mb, respectively), and the selected SNPs are all intron variants. Each SNP has a minor allele frequency (MAF) above 10% in Asian population according to the SNP database of the NCBI. The milk samples were thawed at 4°C and genomic DNA was extracted from 300 µl of the cellular layer using the DNA kit (Beijing, TIANGEN) according to the manufacturer's instructions.

Statistical analysis
Normal distribution of the fatty acids was tested by the Kolmogorov-Smirnov test and distribution plots. Data were expressed as means ± SDs for normal distributed variables, as medians (25th-75th percentiles) for skewed distribution data. The skewed measurements of γ-linolenic acid (GLA, 18:3n-6), EPA (20:5n-3) and DHA concentrations were expressed as by guest, on April 28, 2019 www.jlr.org Downloaded from square roots to obtain a normal distribution. Hardy-Weinberg equilibrium was tested by Chi-square goodness of fit test for each SNP locus. Genotype association with concentration of PUFAs was tested using the online SNPstats software (https://www.snpstats.net/start.htm.).
Linear regression analysis was used to investigate the associations of ELOVL gene polymorphisms with levels of PUFAs. Statistical analysis was performed using SPSS version 16.0 (SPSS Inc., Chicago, IL, 100 USA). Haplotype analyses play an important role in genetic studies (27, 28). It is impossible to define the combination of haplotypes carried by any one individual, but all possible combinations can be computed and techniques like the EM algorithm incorporated in the haplo.stats package of R software (https://cran.r-project.org/src/contrib/Archive/haplo.stats/) can be used to assign a probability to each haplotype pair. To explore potential effects of multi-SNPs and DHA intake, the interactions between ELOVL2/5 haplotype and DHA intake on LC-PUFA levels were performed by the general linear model using R software (R version 3.5.0) adjusted for confounding factors. P0.05 (two-tailed) was considered to indicate a statistical significance.

Characteristics of the study subjects
The 422 healthy lactating mothers included in this study had an average age of 30.29±3.40 years and mainly came from middle income households (54.09%). Gestational age was 39.27±1.00 weeks. The preconception body mass index (BMI) was 20.95±3.33 kg/m 2 that was within the normal range. A total of 31.10% of subjects had a vaginal delivery, while the rest had cesareans. A total of 54.39% of the subject breastfed exclusively while the remaining 45.61% opted for mixed feeding. A total of 50.75% of the mothers had a university education  Table 1).

Effect of DHA intake during pregnancy on the concentration of PUFA in breast milk
We investigated the effect of DHA intake (including dietary and supplement DHA) during pregnancy on the concentration of eight PUFAs in breast milk, and sought to determine whether there was a dose-response relationship. The lactating mothers were classified into four subgroups depending on the quartiles of DHA intake (< 14.58, 14.58-43.15, 43. .82, and ≥118.82 mg/day) ( Table 2). However, the results showed that there was no significant difference for PUFA composition of breast milk among the four groups.

Frequency of SNPs
The distributions of genotype frequencies in the 422 subjects were in accordance with Hardy-Weinberg equilibrium (P>0.05). The selected SNPs were all in intron, and the genotyping success rate was >96%. The characteristics of the detected SNPs, including their positions on chromosome 6 and their genotypes, were summarized in Table 3. Minor allele frequencies (MAF) ranged from 10.7% ~ 46.8% of the population.

The association between ELOVL2 and ELOVL5 genotypes and PUFA concentrations
The association of three genetic models (Codominant model, Dominant model and Recessive model) for each of the 10 SNPs with breast milk PUFA was analyzed using the online SNPstats software adjusted for age, preconception BMI, and DHA intake of subjects. The best fit genetic model of the 10 SNPs was chosen based on the Akaike Information Criterion (AIC) and the Bayesian information criterion (BIC) (Figure 1). Carriers of the minor allele of rs3798713 (P=0.019) in ELOVL2 gene had lower linoleic acid (LA, 18:2n-6) concentrations of breast milk than homozygous subjects for the major allele. The subjects carrying the minor 9 allele homozygote of rs2294867 (P=0.036) within ELOVL5 had higher EPA concentrations than those who carried the major allele. Subjects who carried the minor allele of rs9357760 in  Table S1.

Interaction effect between ELOVL2 and ELOVL5 haplotypes and DHA intake on breast milk PUFA levels
The results of the ELOVL2 and ELOVL5 haplotype frequencies can be found in supplemental Table S2 and supplemental Table S3. Table 4 showed that a 3-SNP haplotype (  Table 4). Table 4 lists the significant results of the haplotype analysis for ELOVL2 and ELOVL5 (P0.05). Full details of related results can be found in supplemental Table S4 and   supplemental Table S5.

DISCUSSION
In this study, we analyzed polymorphisms in genes encoding the elongases involved in LC-PUFA synthesis in lactating mothers to disentangle their role in modifying potential nutritional advantages of LC-PUFAs in breastfeeding. Maternal genetic variants in the ELOVL gene family were associated with breast milk levels of LC-PUFA. Furthermore, DHA intake during pregnancy, including dietary intake and supplementation, were not associated with the concentrations of eight LC-PUFAs that were regulated by ELOVL2 and ELOVL5 gene variations. However, we showed an interaction effect between ELOVL2 and ELOVL5 haplotypes and DHA intake on breast milk LC-PUFA levels. We observed that rs3798713 of ELOVL2 gene influenced the LC-PUFA concentration in present study; carriers of the minor allele had lower LA concentrations than homozygotes for the major allele. Our previous study, however, found no significant association between ELOVL2 variants and LC-PUFA (19), which may be due to a small sample size. Higher EPA levels, in early studies, have been shown in individuals containing minor alleles of rs953413 (29), rs2236212 (18) and rs3798719 (20) with another study showing no effect of the rs3734398 (21) within ELOVL2 gene, while lower DHA levels have been observed in individuals carrying rs953413 (29), rs2236212 (18) and rs3734398 (21) with another study showing no effect of the rs3798719 (20). Inconsistencies between our study and other studies may be because the subjects in our study were all Chinese Han lactating women, and population heterogeneity among studies such as variations in SNP frequencies across the major population groups may also lead to differences in study outcomes.
Regarding the ELOVL5 gene, we found that subjects carrying the minor allele homozygote of rs2294867 had higher EPA concentrations than those who carried the major allele. GLA, DGLA, ARA, and DTA concentrations were significantly affected by rs9357760 in the ELOVL5 gene, and the subjects who carried the minor allele had higher concentrations than those who were homozygous for the major allele. Carriers of the minor allele of rs2397142 had higher DTA levels compared to major allele homozygotes. The subjects homozygous for the minor allele of rs209512 had lower GLA and DGLA concentrations than those carrying the major allele. A significant association was also observed between the minor allele of rs12207094 and a higher level of GLA. In Spanish population-based birth cohorts, a trend was observed for minor allele of rs12207094 and higher EPA levels. Additionally, minor allele by guest, on April 28, 2019 www.jlr.org Downloaded from of rs17544159, rs9395855, and rs12207094 were associated with a high EPA/ARA ratio. This study also found that children of mothers carrying the rs17544159-C allele or carrying the rs12207094-T allele had higher cognition scores compared to children of mothers homozygous for the major allele (20). Our results mainly suggest that minor allele of specific SNPs within ELOVL5 could be prior related to n-6 PUFA levels, probably due to increased transcription or to a more enzyme activity of ELOVL5.
In the second half of pregnancy, DHA accumulates rapidly in neural cortex tissues (30) and retinal membrane synapses (31). DHA comes from the diet or adipose stores, and is synthesized from precursor fatty acids circulating in the maternal bloodstream for uptake by the placenta and fetus. It is not clear how much DHA is released from adipose stores or is synthesized, but a remarkable association of maternal DHA intake during pregnancy with maternal circulating DHA shows that diet may be a major source of DHA for the developing fetus (32-34). Thus, we examined the effect of maternal DHA intake during pregnancy on breast milk LC-PUFAs. However, we did not find any difference. It may be the lower intake of DHA (<200mg/d) (35) among most subjects (86.49%) in our study that affected generalizability of the findings to a wider population.
Haplotypes effectively capture joint marker correlations and evolutionary history; a progressive knowledge of haplotype structures holds great promise for the use of haplotype information to understand genetic factors (36). Our results suggested an association of a haplotype (A-G-G) of ELOVL2 with lower LA, EPA, and DHA levels, but not ARA levels.
Analysis of the impact of DHA intake on these FA levels, however, showed no differences.
Interestingly, there was a significant interaction effect between this haplotype and the second In the present study, carriers with haplotype (A-A-C-A-A) of 5-SNPs in ELOVL5 who consumed high DHA (Q 4 ) had a higher level of DGLA, ALA, and EPA concentrations compared with those who had the haplotype (C-A-C-G-A) with low DHA intakes (Q 1 ) adjusted for age and BMI. Interaction between the haplotype (C-A-C-A-A) and DHA intake (Q 2 and Q 4 , but not Q 3 ) increased EPA concentrations. The reason may be due to the lower frequency of the haplotype in Q 3 group compared to Q 2 and Q 4 group resulting in lower power.
It has been known that DHA supplementation results in an increase in EPA concentrations (38, 39). Metabolically, increases in EPA concentration may be due to increased EPA biosynthesis or decreased EPA degradation. Increased EPA in our study is assumed to result from higher DHA intake via the slowed elongation of EPA. However, there is no significant effect of high DHA intake or haplotype of ELOVL5 on EPA level in our study until high DHA intake interacting with haplotype. This suggests a role for specific ELOVL haplotypes that maybe affect EPA elongation when DHA intake is high.
We speculate that DHA intake for mothers during pregnancy may modulate the ELOVL2 and Limitations of our study included the reliance on estimates of n-3 LC-PUFA consumption from food frequency questionnaires and recorded DHA intake, rather than using controlled doses of DHA. In addition, all quartiles of DHA intake in our study were relatively low level, which may be due to the fact that participants came from inland areas and consumed less DHA-rich seafood in their daily diet. It also explained the result that there were no differences in breast milk PUFA levels among the lactating mothers with different quartiles of DHA intake. Therefore, low DHA intake in the study population limits the generalizability of our results, and further studies included DHA supplemented group are needed in the future, which would be more helpful for us to explore the interaction between DHA intake and ELOVL genes. Most genes contributing to quantitative phenotypes confer modest effects, requiring large sample sizes for detection with high power. Our finding of significant gene and diet interactions as a determinant of the breast milk PUFA concentration in 422 subjects must therefore be regarded with caution. Furthermore, we anticipate that the increase in dietary variability would tend to blur the potential effects of ELOVL genes. Studies carry the risk of a false-positive finding, which can arise by chance, for genotyping errors or failure to correct for multiple testing across the number of SNPs or phenotypes tested. However, our genotyping success rate was high and the genotypes met Hardy-Weinberg equilibrium.
In conclusion, the results of our study showed that ELOVL2 and ELOVL5 genetic variants associated with alterations in PUFA levels in Han Chinese lactating mothers. Additionally, we observed interactions between DHA intake during pregnancy and specific haplotypes within