Prenatal antioxidant-enriched and pro-oxidant-contained food, IL4 and IL13 pathway genes, and cord blood IgE

Prenatal oxidative balance might influence cord blood IgE (cIgE) levels. We aimed to explore if certain prenatal dietary sources of antioxidants and pro-oxidants are associated with cIgE elevation and if they interact with IL4 and IL13 pathway genes. A structured questionnaire was completed during the third trimester of pregnancy for 1107 full-term newborns. Surveyed antioxidant-enriched food included fish, shellfish, and fruit, whereas surveyed pro-oxidant-contained food included fried fish sticks and canned fish. Cord blood was collected for measuring cIgE levels and genotyping IL13 rs1800925, rs20541, rs848, IL4 rs2243250, and STAT6 rs324011. Fairly lean fish consumption showed protection against cIgE elevation (odds ratio [OR] 0.66; 95% CI 0.49–0.90) in the whole sample, while daily fruit (OR 0.46; 95% CI 0.27–0.79) and ≥ monthly canned fish (OR 2.81; 95% CI 1.24–6.36) exhibited associations only in genetically susceptible babies. A prenatal food protective index, comprising any fairly lean fish, daily fruit, and the absence of any canned fish, exerted dose–response protection against cIgE elevation in babies carrying the IL13 rs20541 GA or AA genotype (P for trend < 0.0001; P for interaction = 0.004). We concluded that prenatal antioxidant-enriched and pro-oxidant-contained food consumption may influence cIgE, especially in genetically susceptible babies.


Differential associations of fairly lean fish with cIgE elevation by the intake of canned fish.
To clarify if there was any association between surveyed antioxidant-enriched food and surveyed pro-oxidant-contained food, we investigated dietary habits. Prenatal intake of any canned fish was associated with more frequent prenatal intake of any fairly lean fish (see Supplementary Table S1 online). Further, any prenatal fairly lean fish was protectively associated with cIgE elevation (OR, 0.63 [95% CI 0.40-0.98]) only in participants without any canned fish. This association disappeared in the presence of any canned fish (see Supplementary Table S2 online).
No evidence for sex-food interaction. We also tested the models for neonatal sex-by-prenatal food interactions. Prenatal foods were distributed evenly between female and male babies, and there was no significant interaction between sex and prenatal food for cIgE elevation (P for interactions were all > 0.05) (see Supplementary Tables S3 and S4 online).

Gene-food interactions.
To explore gene-food interactions, we stratified the total study sample by genetic polymorphisms of IL4-IL13 pathway genes ( Table 3) and revealed that daily prenatal fruit intake significantly reduced the risk of cIgE elevation in term babies carrying the IL4 rs2243250 TC or CC genotype (P = 0.005; P after FDR = 0.039) but not in those carrying the IL4 rs2243250 TT genotype, with a significant interaction between daily fruit intake and IL4 rs2243250 (P for interaction = 0.02). Moreover, we found borderline significant interactions between any canned fish and IL13 rs1800925 and between any canned fish and IL13 rs20541 (P www.nature.com/scientificreports/ for interaction = 0.049 and 0.0497, respectively). To clarify these relationships, we further examined the canned fish of different frequencies (Table 4), and demonstrated that canned fish significantly enhanced the risk of cIgE elevation in a dose-dependent manner for those carrying the IL13 rs1800925 CT or TT genotype (P for trend = 0.01), with a significant interaction between canned fish and IL13 rs1800925 (P for interaction = 0.02), but not for those carrying IL13 rs20541 GA or AA genotype.

Prenatal food protective index (PFPI).
To explore whether there is a potentially beneficial dietary pattern, we compiled prenatal intake of any fairly lean fish, daily fruit, and absence of any canned fish into a composite index, namely the prenatal food protective index (PFPI), and participants were divided into four subgroups (i.e., none, one, two, or three protective factors). The corresponding proportion in each food loading category was 60 (5.4%) for none, 340 (30.7%) for one, 509 (46.0%) for two, and 198 (17.9%) for three. We observed a significant dose-response beneficial association of the PFPI with reduced risk for cIgE elevation in the whole study sample (P for trend = 0.002) ( Table 5). After stratification by genetic polymorphisms, PFPI significantly reduced the risk of cIgE elevation in a dose-response manner in term babies carrying the IL13 rs20541 GA or AA genotype (P for trend = < 0.0001) or in those carrying the IL13 rs848 CA or AA genotype (P for trend = 0.0002), with significant gene-food matrix interactions (P for interaction = 0.004 and 0.03, respectively) ( Table 5). Among them, having all three PFPI was associated with 77% (GA or AA of IL13 rs20541) and 73% (CA or AA of IL13 rs848) reductions in the risks of cIgE elevation. These interactions are also illustrated as proportions in Fig. 1. For babies having all three PFPI, the proportion ± standard error of cIgE elevation among those carrying IL13 rs20541 GA or AA genotype (11 ± 3%) was significantly lower than that (22 ± 4%) of babies carrying IL13 rs20541 GG genotype (P = 0.04; P for interaction = 0.004).

Discussion
In this study, we evaluated certain prenatal foods with antioxidants or pro-oxidants and demonstrated their associations with cIgE elevation and their interactions with IL4-IL13 pathway genes. Prenatal intake of fairly lean fish was the only food associated with reduced risk for cIgE elevation in the whole study sample. Regarding foods with significant gene-food interactions, prenatal intake of daily fruit in term babies carrying the risk alleles of IL4 rs2243250 reduced the risk of cIgE elevation. Prenatal intake of canned fish enhanced the risk of cIgE elevation in a dose-response manner in term babies carrying the risk allele of IL13 rs1800925. Moreover, a summary score PFPI exerted beneficial dose-response associations with reduced risk for cIgE elevation not only in the whole study sample but also in the strata of babies carrying the risk allele of IL13 rs20541 or IL13 rs848, with significant gene-food interactions. Among these strata, the potentially beneficial prenatal dietary pattern considering all three indexes resulted in a > 70% reduction in risks. cIgE elevation increases the risks of allergic diseases in childhood 14 . Despite plenty of studies on the relations between prenatal food and childhood allergies 9 , few focused on cIgE as an early outcome. We demonstrated a reduced risk of cIgE elevation by prenatal intake of fairly lean fish, e.g., flounder and shark, in a dose-response manner in the whole study sample. Flounder is commonly found in the coastal waters of East Asian countries 19 . www.nature.com/scientificreports/ Table 3. Associations of prenatal surveyed antioxidant-enriched and pro-oxidant-contained food consumption with cord blood total IgE elevation modified by genetic polymorphisms. aOR, adjusted odds ratio; 95% CI, 95% confidence interval. # P for interaction. *P < 0.05, **P < 0.01, † P < 0.05 after FDR correction. a Adjusted for neonatal sex, gestational age, birth body weight, maternal age, parental education, family history of atopic diseases, environmental tobacco smoke, and perception of mildewy odour.  www.nature.com/scientificreports/ Flounder muscle contains peptides with high antioxidant activity 20 , and is considered as one of the best choices during pregnancy due to its low mercury level 21 . Moreover, flounder possesses antioxidant defensive responses to mercury. A recent study shows that activities of antioxidants in the flounder muscle are enhanced by mercury, even if its exposure dose is low 22 . Shark muscle is also a natural source of antioxidants 23 . However, because of its high mercury level, the Dietary Guideline of Taiwan cautions pregnant women against eating more than 35 g of shark per week 24 , and the Dietary Guideline for Americans recommends avoidance of shark during pregnancy 21 . Fatty fish are a rich source of antioxidants too 8 . However, our study did not demonstrate any association between prenatal fatty fish and cIgE elevation, which is consistent with a prenatal fatty fish intervention study 25 . Similarly, there are conflicting reports about its associations with childhood allergic diseases [9][10][11] . This might be due to the counteracting effect of histamine 26 . Fatty fish is rich in histidine, which is easily transformed to histamine when not adequately cryopreserved 27 .
Prenatal fruit intake has been associated with decreased risk of atopic dermatitis and asthma 9 , while its influence on cIgE remains largely unexplored. Fruits are an important source of dietary antioxidants 28 , including folic acid, vitamin C, and vitamin E, prenatal supplementation of which appear to decrease the risks of atopic dermatitis 29 , allergic sensitization 9 , and infantile wheezing or eczema 9 , respectively. Vitamin E has also been shown to suppress IL-4 secretion in human peripheral blood T cells in a dose-dependent manner 30 , supporting our result of a significant interaction between prenatal fruit intake and IL4 rs2243250 to reduce cIgE levels.
The influences of prenatal canned fish on cord blood biomarkers remain largely uninvestigated. Canned fish is high in histamine 13 . Histamine not only triggers oxidative stress 31 but also promote chemotaxis of eosinophils and neutrophils 26 . In addition, histamine has been shown to enhance the production of Th2 cytokines, such as IL-13 32 , and results in elevated serum IgE levels 26 , supporting our finding on the significant interactions between prenatal intake of canned fish and IL13 rs1800925 in a dose-response manner to increase cIgE levels.
The most ideal prenatal dietary pattern to protect fetuses against cIgE elevation has not been elucidated. Fairly lean fish and fruit are rich sources of antioxidants. One study showed a beneficial combined effect of high prenatal consumption of fish plus fruit 8 , but their influences on cIgE remain unexplored. Antioxidant vitamin C, which is rich in fruit, has a unique ability to catalyze the degradation of histamine 33 , which is rich in canned fish 13 . Our study further revealed that the benefit of fairly lean fish might be counteracted by canned fish. Because of these potential combined effects and mutually counteracted influences, we investigated and established the summary score PFPI, and revealed its dose-response protection against cIgE elevation not only in the whole study sample but also in term babies carrying the risk alleles of IL13 rs20541 or IL13 rs848, with significant gene-food matrix interactions. The potentially most beneficial prenatal dietary pattern demonstrated in this study was that including all three indexes. Table 5. Dose-response associations of prenatal food protective index with cord blood total IgE elevation modified by genetic polymorphisms. PFPI, prenatal food protective index; aOR, adjusted odds ratio; 95% CI, 95% confidence interval. # PFPI includes any fairly lean fish, daily fruit, and absence of any canned fish. a Adjusted for neonatal sex, gestational age, birth body weight, maternal age, parental education, family history of atopic diseases, environmental tobacco smoke, and perception of mildewy odour. *P < 0.05, **P < 0.01, † P < .05 after FDR correction, ‡ P < .01 after FDR correction.  Relations between prenatal food protective index (i.e., any fairly lean fish, daily fruit, and absence of any canned fish) and cord blood total IgE elevation in (a) total study sample, (b) newborns with IL13 rs20541 GG genotype or GA or AA genotype, and (c) newborns with IL13 rs848 CC genotype or CA or AA genotype. Data represent proportions ± standard error of the proportion. P for interactions are the P-values between the prenatal food protective index and IL13 rs20541 (b) and between the prenatal food protective index and IL13 rs848 (c). www.nature.com/scientificreports/ This study has some limitations. We did not use semi-quantitative food frequency questionnaires 6 to collect information on prenatal nutrient intake, such as vitamin C, vitamin E, and folic acid. Instead, data on real foods were collected and analyzed, enabling us to suggest a potentially beneficial dietary combination. Besides, we did not include foods extensively. Instead, we incorporated several foods only, based on our hypotheses. Since we did not collect information on maternal body size, maternal asthma history, and maternal asthma status (exacerbated or stable) during pregnancy, we were unable to investigate if these maternal characteristics modified the associations of prenatal foods with cIgE elevation and their interactions with other factors (such as sex and genes). Furthermore, we did not measure prenatal biomarkers of oxidant or antioxidant system, which, based on our exploratory results, warrant further investigation to evaluate their relations to cIgE.

PFPI #,a
In conclusion, we revealed possibly beneficial (fairly lean fish and fruit) and unfavorable (canned fish) prenatal food for cIgE levels either in the total study sample or in term babies carrying the risk alleles of certain IL4-IL13 pathway genes, with significant gene-food interactions. The potentially most advantageous dietary pattern in this study, which included all three indexes, decreased the risk of cIgE elevation not only in the total study sample but also in the genetic strata, with the largest effect (77% reduction in risk) in term babies carrying the risk allele of IL13 rs20541. Implementation of this prenatal dietary pattern may be helpful for early prevention of allergies.

Methods
The GEIBCS cohort. GEIBCS is a birth cohort study of 1107 Han Chinese term newborns enrolled from March 2008 to April 2011 in Taiwan. Its study design and baseline characteristics have been described in detail elsewhere 17,18 . Briefly, gestational age was first estimated by maternal report of the first day of the last menstrual period and then determined by a series of periodic gestational ultrasonography. Structured questionnaires 18 were completed during the third trimester of pregnancy by pregnant women who underwent prenatal care at Min-Sheng General Hospital, and their fetuses were prospectively followed. Full-term birth was defined as the delivery of a baby at or after 37 weeks' gestation. Written informed consent was obtained from the mothers of all participants. This study was approved by the Institutional Review Board of Min-Sheng General Hospital (approval number: MSEIRB0980109). All methods were performed in accordance with the relevant guidelines and regulations.

Prenatal surveyed antioxidant-enriched and pro-oxidant-contained food consumption.
We collected data on certain prenatal food intake by using our structured questionnaires during the third trimester of pregnancy, which included questions for selected food, but not for total daily food. Prenatal food of interest was divided into two categories: surveyed antioxidant-enriched food vs. surveyed pro-oxidant-contained food. Surveyed antioxidant-enriched food included fish 8 , shellfish (shrimp or crab), and fruit (no specific types) 28 . To contrast and emphasize the potentially inverse effect of certain processing methods on fish 12 , we included fried fish sticks 12 and canned fish 13 (tuna, salmon, or sardines) as surveyed pro-oxidant-contained food. Fish were further divided into four subcategories: 10, 11 .
The frequency of prenatal food intake was grouped into five categories: never, rarely (i.e., less than once a month), monthly, weekly, and daily. More than never was defined as any. The fish intake was not energy-adjusted because we did not attempt to estimate the effects of individual dietary components and there has been considerable debate about which of the strategies of energy adjustment is most appropriate 34 .
PFPI was composed of prenatal intake of any fairly lean fish, daily fruit, and absence of any canned fish and was not energy-adjusted. PFPI was calculated by counting the accomplishment of each component of this index. Babies with any one, any two, and all of these 3 protective factors got one point, two points, and three points of PFPI, respectively.
Outcome measures. Serum cIgE levels were measured using a CAP FEIA system (Pharmacia, Uppsala, Sweden), and levels ≥ 0.5 IU/mL were defined as elevated 16 .
DNA collection and genotyping. The rationale for choosing three IL13 SNPs (rs1800925, rs20541, and rs848), IL4 rs2243250, and STAT6 rs324011 was described in our previous studies 17,18 . Genotyping of the five SNPs was performed using TaqMan probe assays (TaqMan ® SNP Genotyping Assays, Applied Biosystems, Foster City, CA, USA) on a 7900 Fast Real-Time PCR System (Applied Biosystems). Genotyping of each sample was automatically attributed using SDS 1.3 software for allelic discrimination. www.nature.com/scientificreports/ Statistical analysis. Statistical analyses were performed with SAS software version 9.4 and the SAS/Genetics module (SAS Institute Inc., Cary, NC, USA). The distributions of prenatal surveyed antioxidant-enriched food and pro-oxidant-contained food by cIgE elevation were analyzed by chi-square test. Further, we performed univariate logistic regression to investigate the associations of these prenatal foods with cIgE elevation. We also adjusted for known predictors of atopy (i.e., gestational age, birth body weight, maternal age, and family history of atopic diseases) and correlates of cIgE elevation identified in our previous study 18 (i.e., neonatal sex, parental education, environmental tobacco smoke, and mildewy odour). The relations of these covariates to prenatal foods and cIgE were illustrated in a directed acyclic graph (see Supplementary Fig. S1 online). Multiple tests were corrected by means of false discovery rate (FDR). Deviations of genotypes from Hardy-Weinberg equilibrium were assessed using the χ 2 test. Based on the findings of our previous study 17,18 , the dominant genetic models were implemented. To further investigate if genetic polymorphisms modified the influences of prenatal surveyed antioxidantenriched food and pro-oxidant-contained food on cIgE elevation, firstly we stratified the association analyses by genotypes of each genetic polymorphism to explore whether there was any stratum-specific main effect. Significance levels of the associations in each stratification strategy were corrected for multiple testing by means of FDR. A two-tailed P value < 0.05 after FDR correction was considered significant. After that, the significance level of statistical interaction was determined using likelihood ratio tests that compared the model with an interaction term versus the model without it.
Ethical approval. This study was approved by the Institutional Review Board of Min-Sheng General Hospital (approval number: MSEIRB0980109). Written informed consent was obtained from the mothers of all participants.