Interactions between Environmental Factors and Glutathione S-Transferase (GST) Genes with Respect to Detectable Blood Aluminum Concentrations in Jamaican Children

Aluminum (Al) is a metallic toxicant at high concentrations following natural or unnatural exposures. Dietary intake is considered as the main source of aluminum exposure in children. We used data from 366 typically developing (TD) children (ages 2–8 years) who participated as controls in an age- and sex-matched case–control study in Jamaica. We investigated additive and interactive associations among environmental factors and children’s genotypes for glutathione S-transferase (GST) genes (GSTT1, GSTM1, GSTP1), in relation to having a detectable blood aluminum concentration (BAlC) of >5.0 μg/L, using multivariable logistic regression models. Findings from interactive models revealed that the odds of having a detectable BAlC was significantly higher among children who ate string beans (p ≤ 0.01), whereas about 40% lower odds of having a detectable BAlC was observed in children with higher parental education level, (p = 0.02). A significant interaction between consumption of saltwater fish and GSTP1 in relation to having a detectable BAlC using either co-dominant or dominant genetic models (overall interaction p = 0.02 for both models) indicated that consumption of saltwater fish was associated with higher odds of having a detectable BAlC only among children with the GSTP1 Ile105Val Ile/Ile genotype using either co-dominant or dominant models [OR (95% CI) = 2.73 (1.07, 6.96), p = 0.04; and OR (95% CI) = 2.74 (1.08, 6.99), p = 0.03]. Since this is the first study from Jamaica that reports such findings, replication in other populations is warranted.


Introduction
Aluminum (Al) is one of the most plentiful elements after oxygen and silicon in the surface of the Earth and is about 8% by mass [1][2][3]. Even though Al is not required for any biological process in humans and animals, it can be a metallic toxicant at high concentrations after natural or unnatural exposure [4,5]. Exposure to Al has been linked to several lower level of GST activity compared to other combinations of genotypes, suggesting a poor aluminum detoxification ability [48]. All of these reports indicate that genetic variation can provide an explanation for differences in Al concentrations in a population.
In our previous paper, we used data from 116 age-and sex-matched pairs (ASD vs. typically developing controls (TD)) (232 children) of Jamaican children 2-8 years old. We observed that TD children with GSTP1 Ile/Ile or Val/Val genotype had a significantly higher geometric mean blood Al concentration (BAlC) than those with Ile/Val genotypes (23.75 µg/L vs. 14.57 µg/L, p < 0.03). Furthermore, none of the additive effects of food consumption were statistically significantly associated with log-transformed BAlC (all p > 0.06) [49]. In the present study, we evaluated the additive and interactive association between environmental factors and genotypes of three GST genes, as well as the possible pairwise gene-gene interactions of these genes in relation to a detectable BAlC (>5.0 µg/L) in Jamaican TD children.

Study Population
This study was conducted using data from 366 TD control children, between 2-8 years old, who were enrolled in the Epidemiological Research on Autism in Jamaica (ERAJ) studies between December 2009 and September 2017. Detailed information regarding the enrollment and assessment of TD controls has been reported earlier [50][51][52]. Relevant to the research objectives here, the age-and sex-matched TD controls (within six months of the matched ASD case) were identified from schools, churches, and well-child clinics at the University of the West Indies (UWI) and the Social Communication Questionnaire (SCQ) [53] was used to rule out developmental disorders (SCQ score of 0-6) in the TD control children. [49] We collected information about parents/guardians' sociodemographic characteristics as well as children's weekly food intake through questionnaires [49], and about 5 mL of whole blood was drawn from each child to assess exposure to the heavy metals including Al and to determine genotypes for the three GST genes. This study was approved by the Institutional Review Boards of The University of Texas Health Science Center in Houston (UTHealth), Michigan Department of Health and Human Services (MDHHS), and the University of the West Indies, Mona campus, in Kingston, Jamaica (HSC-SPH-09-0059).

Assessment of Al Exposure
In this study, we assessed BAlCs to measure Al exposure in children. BAlCs were assessed at the Trace Metals Lab at the MDHHS in Lansing, MI, USA. We have previously reported details on sample processing and storage [49,51,54]. MDHHS followed a fully authenticated protocol for analyzing Al in blood samples with a limit of detection (LoD) of 5.0 µg/L, and 37.1% (136 out of 366) of children in this study had an undetectable BAlC because it was below the LoD.

Statistical and Genetic Analysis
We conducted descriptive analyses to assess socioeconomic status (SES) characteristics, and frequencies of the GSTP1, GSTT1, and GSTM1 genotypes for the TD children. Since more than one-third (37%) of BAlCs were below the LoD, we used 5.0 µg/L as the cutoff point and converted BAlCs to a binary variable. The choice of cutoff point reflects the LoD in the ERAJ studies.
Assessment of the children's genotypes for the GSTP1 Ile105Val (rs1695) polymorphism and insertion deletion polymorphisms in GSTT1 and GSTM1 was carried out as previously described [50,55]. Choice of the genetic models that were used to test their additive and interactive associations with environmental factors was based on differences in the types of polymorphisms. Because there are 3 possible genotypes for GSTP1 rs1695 and 2 possible genotypes for GSTT1 and GSTM1 since the homozygote (I/I) and heterozygote (I/D) cannot be distinguished, only the recessive model was selected for GSTT1 and GSTM1 (D/D vs. I/I and I/D) whereas three different genetic models were specified for GSTP1 rs1695 (dominant, co-dominant, and recessive). Similarly, only the GSTP1 Ile105Val polymorphism was tested for accordance with Hardy-Weinberg equilibrium expectations using the Chi-square test.
Using logistic regression models, we assessed additive association of each independent variable including the three GST genes, sociodemographic characteristics, and consumption of different kinds of vegetables, starches, and seafoods in relation to binary BAlCs (<LoD vs. ≥LoD). Then, we evaluated the potential gene-gene interactions among the three GST genes and possible gene-environment interactions between each of the three GST genes and consumption of various types of food in relation to BAlCs. Subsequently, we developed logistic regression models that contained both additive and interactive effects of GST genes and environmental factors to evaluate the adjusted odds of having a detectable BAlC. To minimize the potential effects of multicollinearity, we only kept one of the correlated variables when the model became unstable by adding both correlated variables. Following the procedure described we used the CONTRAST statement in SAS [56] to access odds ratios and 95% confidence intervals for evaluating the interactive effects in the presence of two-way interactions. All statistical tests were evaluated at 5% level of significance using SAS 9.4 software [57].

Results
Demographic information and other characteristics are displayed in Table 1. 81.7% of the 366 TD children were male and 97.3% were Afro-Caribbean. About 25% of them were 72 months or older and 62.6% of the children were born in the Kingston parish. 11.4% of TD children were born to mothers who were at least 35 years old, and 45.5% of the children had at least one parent who attained an education level beyond high school. Moreover, 40.7% of the families owned a car, which represents high SES in Jamaica. The frequencies of null (DD) genotype for GSTM1 and GSTT1 were 26.1% and 24.6%, respectively. In addition, the frequencies of the GSTP1 genotypes were in agreement with Hardy-Weinberg equilibrium expectations (p = 0.67). a Other parishes include all 12 parishes in Jamaica, except for Kingston parish as described previously [58]. b Up to high school education included Primary/Jr. Secondary, and Secondary/High/Technical schools. c Beyond high school education included Vocational, Tertiary College, or University.
In univariable logistic regression analysis (Table 2), we found a significant inverse association between having at least one parent with education level beyond high school and a detectable BAlC in children [OR (95% CI) = 0.52 (0.34, 0.80), p < 0.01]. We also found significant associations between consumption of certain types of food and BAlCs.  .27), p = 0.01], had higher odds of having a detectable BAlC compared to those who never ate such food. We did not find any significant additive associations between BAlCs and genotypes for the three GST genes (all p > 0.08).
Unadjusted multivariable models were used to assess the two-way gene-gene interaction of GST genes in relation to BAlCs (Table 3). Using a dominant genetic model for GSTP1, there was a significant interaction between GSTP1 and GSTM1 with respect to BAlCs (overall interaction p = 0.04) indicating that among children with GSTM1 DD genotype, children with GSTP1 Ile/Val or Val/Val genotype were 68% less likely to have a detectable BAlC than those with the GSTP1 Ile/Ile genotype [OR (95% CI) = 0.32 (0.11, 0.98), p < 0.05].
Additionally, using a co-dominant model for GSTP1, although the interaction between GSTP1 and GSTM1 was marginally significant (overall interaction p = 0.06), we found that among children with GSTM1 DD genotype, the odds of having a detectable BAlC in children with the GSTP1 Val/Val genotype was 0.20 times (or 1/5 times) that of those with the Ile/Ile genotype [OR (95% CI) = 0.20 (0.05, 0.82), p = 0.03]. When we used the recessive genetic model for GSTP1 (overall interaction p = 0.10), we found that (though marginally significant) among children with GSTP1 Val/Val genotype, the odds of having a detectable BAlC in children with the GSTM1 DD genotype was 0.33 times that of those with the I/I or I/D genotype [OR (95% CI) = 0.33 (0.10, 1.07), p = 0.06]. Moreover, although the interaction between GSTM1 and GSTT1 in relation to BAlCs was not statistically significant (overall interaction p = 0.11), we found among children with DD genotype for GSTM1, the odds of having a detectable BAlC was 67% lower in children with DD genotype for GSTT1 than in those with I/I or I/D genotype for GSTT1 [OR (95% CI) = 0.33 (0.13, 0.87), p = 0.03].
In the assessment of the interactive associations of children's environmental exposures and genotypes for GST genes with respect to detectable BAlCs (Table 4), we identified a significant interaction between consumption of green banana and GSTT1 genotypes in relation to BAlCs (interaction p = 0.04). Specifically, using a recessive genetic model, among children with GSTT1 I/I or I/D genotype, the odds of having a BAlC above LoD in children who ate green banana was 0.45 times that of those who never ate such food [OR (95% CI) = 0.45 (0.25, 0.82), p = 0.01], whereas, no statistically significant associations were found between consumption of green banana and BAlCs among children with GSTT1 DD genotypes [OR (95% CI) = 1.47 (0.55, 3.90), p = 0.44]. In addition, we found a similar interactive association between child's genotypes for GSTT1 and consumption of porridge, as well as consumption of macaroni in relation to BAlCs (both overall interaction p = 0.03).    Table 3 for GSTT1, GSTM1, and GSTP1. DD, I/I, and I/D are defined for GSTT1 and GSTM1 in Table 1. Results a p values and b Overall interaction p values are described in Table 3.
Furthermore, there was a significant interaction between consumption of broad beans and GSTM1 genotypes under a recessive genetic model, in relation to BAlCs (interaction p < 0.05). Specifically, among children with GSTM1 null (DD) genotype, the odds of having a detectable BAlC in those who ate broad beans was 3.96 times that of those who never ate such food [OR (95% CI) = 3.96 (1.57, 9.97), p < 0.01], whereas, there was no significant association between consumption of broad beans and BAlCs among children with GSTM1 I/I or I/D genotypes [OR (95% CI) = 1.37 (0.82, 2.27), p = 0.23]. We also identified a significant interaction between consumption of saltwater fish and child's GSTP1 genotype in relation to a detectable BAlC using either a co-dominant or dominant genetic model (overall interaction p = 0.03, and p = 0.02, respectively). Specifically, among children with GSTP1 Ile/Ile genotypes, the odds of having detectable BAlCs in children who reported eating saltwater fish was 3.36 times that of those who never ate such seafood in both genetic models [OR (95% CI) = 3.36 (1.37, 8.24), p = <0.01 for both models]. Although the overall interaction was marginally significant when using the recessive genetic model (overall interaction p = 0.05), we have found that among children with at least one Ile allele, those who ate saltwater fish had higher odds of having detectable BAlCs than those who never ate such food [OR (95% CI) = 1. 76 (Table S1).
In the interactive multivariable models that assessed the adjusted associations of children's GST genotypes, exposure to environmental factors, and their interactions in relation to a detectable BAlC (Table 5), we identified education level of the parents and consumption of string beans as environmental factors that are additively associated with BAlCs in Jamaican TD children (all p ≤ 0.02 in all models). For example, using the codominant model for GSTP1 genotype, the odds of having detectable BAlCs in children who consumed string beans was still significantly higher than in those who never ate string beans (OR (95% CI) = 3.07 (1.07, 5.09), p < 0.01), and having at least one parent with education level beyond high school versus up to high school was associated with significantly lower odds of having a BAlC above LoD [OR (95% CI) = 0.57 (0.36, 0.91), p = 0.02]. By using three genetic models for GSTP1 genotype, we have investigated the gene-environment interaction between GSTP1 and consumption of saltwater fish in relation to BAlCs. After holding the aforementioned environmental factors constant, we found similar significant interactions between consumption of saltwater fish and GSTP1 under both co-dominant and dominant genetic models (overall interaction p = 0.02 for both models). Specifically, we found that among children with GSTP1 Ile/Ile genotype, the odds of having detectable BAlCs in children who ate saltwater fish was about 2.7 times that of those who never ate saltwater fish based on both co-dominant or dominant genetic models [OR (95% CI) = 2.73 (1.07, 6.96), p = 0.04; and OR (95% CI) = 2.74 (1.08, 6.99), p = 0.03, respectively]. Though the overall interaction is significant (p = 0.04), no statistically significant associations between saltwater fish consumption and BAlCs by children's genotypes in GSTP1 was found using the recessive genetic model for GSTP1. In addition, details about adjusted associations between children's genotypes in GSTP1 and BAlCs by saltwater fish consumption are shown in Table S2. Table 5. Adjusted associations between exposure to environmental factors and binary detectable blood Al concentrations (BAlCs) by genotypes for GSTP1 genes in typically developing children based on logistic regression models that include gene*environment interaction (N = 366).  Table 3. EF1 = Parental education level (Parental education level missing data are based on numbers reported in Table 1). EF2 = Consumption of string beans. EF3 = Consumption of saltwater fish. G1 = beyond high school education included Vocational, Tertiary College, or University. G2 = Primary/Jr. Secondary, and Secondary/High/Technical schools. a p and b Overall interaction p values are described in Table 3.

Discussion
Findings from our study suggest that the odds of having a detectable BAlC was about 3 times higher among children who ate string beans, compared to those who did not eat such beans, and 50% times lower in children with at least one parent with education level beyond high school. The association between consumption of saltwater fish and having a detectable BAlC varied by children's genotype for GSTP1 using either dominant or co-dominant genetic models (overall interaction p = 0.02 for both models), and eating saltwater fish was significantly associated with 3 times higher odds of having a detectable BAlC only among children with GSTP1 Ile/Ile genotype. Jamaica is known for its abundant and high-quality bauxite for decades. Over 20% of the surface area was covered by bauxite deposits in Jamaica, and a comparatively high level of Al was found in soils [37,59]. Since content of Al in foods varies by species and the soil pH [60], a possible explanation for our finding that consumption of string beans is associated with higher odds of having a detectable BAlC is that legumes including string beans accumulate more Al than others. Filippini et al. [26] conducted a study about dietary intake of Al by obtaining 908 food samples from Italy and measuring the Al content. Legumes were the category of food that had the highest levels of Al (7370.23 µg/kg). In addition, our finding is similar to several studies in China that reported soybeans, a member of the legume family, and bean products had a higher level of dietary Al content [34,61,62].
Our finding indicating 50% lower odds of having a detectable BAlC in children who had parents with higher education levels (at least one of the parents had education beyond high school) is consistent with several previous studies reporting that people from families with a lower level of education were exposed to more heavy metals [54,[63][64][65]. In our previous study, we also found that Jamaican children whose parents both had education levels up to high school had 1.82 times the odds of having a detectable blood arsenic concentration (>1.3 µg/L) than children who had at least one parent with an education level beyond high school (p ≤ 0.01) [54]. Jee et al. also demonstrated that a lower level of formal education contributes to significantly higher blood cadmium levels [63]. Another study in Canada reported a significant inverse relationship between education (completed high school or not) and blood mercury levels in pregnant women [65]. A possible reason for our finding is that children from families with a lower level of education tend to be exposed to more fast food that contains a high content of food additives [66]. Moreover, parents with a higher education level may reveal more health-conscious behavior in providing food for their children [67] although they may not be aware that foods such as vegetables, string beans, and lettuce may have high levels of aluminum.
The literature about the association between genetic variation in GST genes and BAlCs in TD children is very limited. Our results support the conclusion that GSTP1 Ile105Val genotype can modify the effect of consuming saltwater fish on BAlCs where only carriage of the Ile/Ile genotype was shown to confer an increased risk of having a level > 5.0 µg/L. This was observed when either a co-dominant or dominant genetic model was used, while there was no significant association between any of the three GST polymorphisms and BAlCs in the additive models. One mechanism that may help to explain this finding is that codon 105 is located in the active site of the enzyme and that GSTP1 Ile105Val has been associated with changes in substrate-specific catalytic activity [68][69][70]. Similar relationships have been reported for another heavy metal. Engström et al. found that variation in the amount of fish intake can influence the level of mercury measured in erythrocytes and that this is dependent on GSTP1 genotype. No significant association with mercury levels was found if fish consumption was low, but individuals with the Ile/Ile genotype had significantly higher mercury levels than those with either the Ile/Val or Val/Val genotype if fish was eaten at least 2.5 times a week [68,71], In addition, previous studies have demonstrated that the GST enzyme itself or glutathione reductase (GR), an enzyme that maintains the supply of the GST substrate reduced glutathione, can be potentially inhibited by heavy metal ions at toxic concentrations [72][73][74]. For example, long term low-level lead exposure in rats has shown significant inhibitory effects (up to 55% inhibition) on GST activity [75]. Cadmium was shown to play a role in the inhibition of GST that was purified from Van Lake fish gills [76]. Since saltwater fish is a source of many heavy metals including arsenic, lead, mercury, and cadmium [77,78], a joint effect of multiple heavy metal exposures through saltwater fish consumption and GST genes is possible in relation to BAlCs. More studies are needed to replicate these relationships.

Limitations
We acknowledge that this study has several limitations. First, our participants are more likely to be from the Kingston area. Hence, our findings may not be generalizable to all children in Jamaica. Second, the timing of Al exposure was not available in this study as the BAlCs we used as a biomarker are more likely to reflect recent exposure. Although we used a food frequency questionnaire that reflects the food selection in Jamaica, we cannot exclude the possibility that findings may be confounded by other unmeasured variables, such as the consumption of canned beverages or the use of Al foil that may have a strong association with BAlCs. In addition, since we categorized the frequency of food into binary variables (consumed vs. never consumed), our analysis did not account for the quantity of food intake. Furthermore, to avoid the potential multicollinearity, consumption of several food items including freshwater fish, tuna, cakes/buns, vegetables (broad beans, lettuce, cabbage, and root vegetables) that were significantly associated with BAlCs in the univariable analysis were removed from the multivariable analyses. Furthermore, since SES is associated with parental education level, we choose to use parental education level in the model to avoid any potential for multicollinearity. Therefore, we advise caution in interpretation of these findings.

Conclusions
The present work indicated that children in Jamaica may be more susceptible to Al exposures through specific environmental factors as well as variation in GST genes. Our findings from interactive multivariable logistic regression models revealed that consumption of string beans was associated with higher odds of having a detectable BAlC, whereas higher parental education level was associated with lower odds of having a detectable BAlC in TD children. Additionally, we have found that among children with the GSTP1 Ile/Ile genotype, the odds of having a detectable BAlC was higher in children who consumed saltwater fish than in those who did not eat such food under both a co-dominant and dominant genetic model for GSTP1. This finding suggests that GSTP1 rs1695 may serve as an effect modifier for the association between saltwater fish consumption and BAlCs in Jamaican children. Further research is recommended to better understand the biological explanation for these findings.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/genes13101907/s1, Table S1: Associations between children's genotypes for GST genes and binary detectable blood Al concentrations (BAlCs) by children's exposure to environmental factors based on logistic regression models that include interaction between GST genes and the main environmental exposure (N = 366); Table S2: Associations between children's genotypes for GSTP1 and binary detectable blood Al concentrations (BAlCs) by saltwater fish consumption based on logistic regression models that adjusted for parental education level and consumption of string beans (N = 366).  Informed Consent Statement: Informed consent was obtained from parents/guardians of all children involved in the study. Children's assents were also obtained if the child was 7-8 years old.

Data Availability Statement:
The data analyzed in this study are from two grants (i.e., R21 and R01). The data from R01 are or will be publicly available through the National Database for Autism Research (NDAR). Data from R21 will also be available upon request from the corresponding author.