Interaction of Blood Manganese Concentrations with GSTT1 in Relation to Autism Spectrum Disorder in Jamaican Children

Abstract

Using data from 266 age- and sex-matched pairs of Jamaican children with autism spectrum disorder (ASD) and typically developing (TD) controls (2–8 years), we investigated whether glutathione S-transferase theta 1 (GSTT1) modifies the association between blood manganese concentrations (BMC) and ASD. After adjusting conditional logistic regression models for socioeconomic status and the interaction between GSTT1 and GSTP1(glutathione S-transferase pi 1), using a recessive genetic model for GSTT1 and either a co-dominant or dominant model for GSTP1, the interaction between GSTT1 and BMC was significant (P = 0.02, P = 0.01, respectively). Compared to controls, ASD cases with GSTT1-DD genotype had 4.33 and 4.34 times higher odds of BMC > 12 vs. ≤ 8.3μg/L, respectively. Replication in other populations is warranted.

Introduction

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder characterized by impaired social interaction and communication, and repetitive, stereotyped behavior (American Psychiatric Association 2013). Although the median global prevalence of ASD has been estimated to be 0.62% (Elsabbagh et al. 2012), the prevalence of ASD is unknown in many low- and middle-income countries (LMICs) such as Jamaica. The etiology of ASD is not well understood but there is growing evidence that the effect of individual factors, either genetic or environmental is usually insufficient to explain the ASD phenotype (Hertz-Picciotto et al. 2018). The general agreement among scientists is that complex interplay among genes, and between genes and environmental factors contributes to the etiology of ASD (Hallmayer et al. 2011; Kim et al. 2009; Schieve et al. 2018) during crucial stages in brain development (Landrigan et al. 2012).

Manganese (Mn) is a naturally occurring trace element essential for human health and brain growth and development (ATSDR 2012b; EPA 2007; Guan et al. 2014; Lucchini et al. 2017). Neurotoxic effects of excess exposure to Mn are well documented (ATSDR 2012b; Bhang et al. 2013; Bouchard et al. 2011; Claus Henn et al. 2010; Ericson et al. 2007; Khan et al. 2011; Khan et al. 2012; Kim et al. 2009; Lucchini et al. 2017; J. A. Menezes-Filho et al. 2011; Roels et al. 2012; Takser et al. 2003; Wasserman et al. 2011; Wright et al. 2006; S. Zoni et al. 2007). Specifically, children are more vulnerable than adults to both insufficient and increased exposures to Mn (Claus Henn et al. 2010; Janka 2019; Silvia Zoni and Lucchini 2013). While insufficient amounts of Mn can lead to poor neurodevelopmental outcomes (Janka 2019), exposure to higher levels of Mn in children has been associated with adverse neurological effects including increased behavioral problems (Carvalho et al. 2014; Hong et al. 2014; Khan et al. 2011; Khan et al. 2012; José A Menezes-Filho et al. 2014; Rahman et al. 2017), reduced verbal and full-scale intelligence quotient (IQ) scores (Bouchard et al. 2011; Chung et al. 2015; Haynes et al. 2015; Lin et al. 2013; J. A. Menezes-Filho et al. 2011; Wasserman et al. 2011), attention deficit (Hong et al. 2014; Takser et al. 2003), and lower academic achievement (Bhang et al. 2013; Khan et al. 2012). Several studies examined the relationship between Mn exposure and ASD, as measured by air distribution (Rossignol et al. 2014; Windham et al. 2006) and levels in tooth enamel (Abdullah et al. 2012), hair (Abdullah et al. 2012; Adams et al. 2006; Al-Ayadhi 2005; Blaurock-Busch et al. 2011; De Palma et al. 2012), urine (Blaurock-Busch et al. 2011), red blood cells (Jory and McGinnis 2008), and whole blood (Rahbar et al. 2014), though with conflicting findings from additive models. At least part of the reason for the inconsistent findings of the previous studies could be that exposure to Mn may interact with other environmental and genetic factors in relation to ASD.

Some studies have identified an association between increased blood Mn concentrations (BMC) and increased oxidative stress (Erikson et al. 2004; Fernsebner et al. 2014; Wu et al. 2010), that could be a possible mechanism for Mn neurotoxicity (Lucchini et al. 2017; Neal and Guilarte 2013). On the other hand, Mn is an essential cofactor for metalloenzyme superoxide dismutase, which protects cells against oxidative stress (Hope et al. 2006; Koh et al. 2014; Rucker et al. 2010). Cells are also protected from oxidative stress by intracellular antioxidants like glutathione along with antioxidant enzymes like Mn superoxide dismutase (MnSOD) (Tamai et al. 2011). In addition, serum MnSOD level is often used as a biomarker for oxidative stress (Tamai et al. 2011). Chronic Mn accumulation in the mitochondria appears to affect MnSOD activity, resulting in increased oxidative stress in the mitochondria (Koh et al. 2014). Glutathione S-transferase (GST) genes, including glutathione S-transferase pi 1 (GSTP1), glutathione S-transferase mu 1 (GSTM1), and glutathione S-transferase theta 1 (GSTT1), encode a family of phase II enzymes that have a critical role in protecting cells from oxidative stress by detoxification of exogenous chemicals and heavy metals (Garrecht and Austin 2011), as well as endogenous metabolites that are associated with oxidative stress (Hayes and Strange 2000; Zhang et al. 2011; Zhou et al. 2014). GST genes are highly polymorphic (Hayes and Strange 2000), and null alleles of some of the genes in this family including GSTT1 and GSTM1 have been identified that can lead to the lack of functional enzyme and decreased detoxification capacity that has been implicated in increased susceptibility to oxidative stress-related chronic diseases such as cancer (Cai et al. 2014; Ritambhara et al. 2019), and cardiovascular disease (Tang et al. 2012; Vojtková et al. 2013; Yuan et al. 2014; Yuan et al. 2013). Polymorphisms in the GST genes have also been associated with neurological (Ercegovac et al. 2015) and neurodevelopmental disorders such as ASD (Frustaci et al. 2012; James et al. 2006; Schmidt et al. 2011). For example, a null allele of the GSTT1 gene (DD genotype) has been associated with increased susceptibility to brain tumors (Diedrich et al. 2006; Ercegovac et al. 2015), neurodegenerative diseases (Ercegovac et al. 2015; Landi 2000), epilepsy (Ercegovac et al. 2015), and attention deficit hyperactivity disorder (Lee et al. 2016).

Since 2009, our research team at the University of Texas Health Science Center at Houston (UTHealth) has collaborated with faculty at the University of the West Indies (UWI) in Jamaica to investigate the role of six heavy metals including Mn, and their potential interaction with the GST genes (GSTP1, GSTT1, and GSTM1) in relation to ASD in Jamaica. Based on data from 109 pairs of sex- and age-matched ASD cases and typically developing (TD) controls from our Jamaican Autism Study, we have previously reported a lack of an additive effect of BMC in ASD (Rahbar et al. 2014). However, we found a significant interaction between GSTP1 and BMC indicating that among children with the Ile/Ile genotype for the GSTP1 Ile105Val polymorphism, the odds of having BMC ≥ 12 μg/L in ASD cases was 4 times that in the TD control group under a co-dominant genetic model (P = 0.03) (Rahbar et al. 2015a). In addition, we have demonstrated that the interaction between Mn and GSTP1 in relation to ASD remained significant with a similar magnitude of associations after adjusting for the mixture of four other metals (Pb, Hg, As, and Cd) based on an estimated mixture score from the Weighted Quantile Sum (WQS) model (Rahbar et al. 2018). We have also previously reported a significant interaction between GSTP1 and GSTT1 in relation to ASD. Specifically, for children who were heterozygous for the GSTP1 Ile105Val polymorphism under the co-dominant genetic model, the odds of ASD for those with the GSTT1 null genotype (DD) was three times that of those with either the GSTT1 I/I or I/D genotype (P = 0.03) (Rahbar et al. 2015b). These gene-environment (GSTP1*Mn) and gene-gene (GSTP1*GSTT1) interactions that we observed in relation to ASD in separate interactive models, raise a question whether GST family genes may have a more complex relationship with the association between BMC and ASD. The aim of this study is to investigate the interactive associations of BMC with GSTT1 genotypes in relation to ASD in Jamaica. In addition, we investigate whether this relationship would be impacted after accounting for the interaction between GSTT1 and GSTP1 in relation to ASD. Furthermore, since it has been shown in previous studies that socioeconomic characteristics including levels of family income and parental education may impact the associations between Mn exposure and neurodevelopment (Chiu et al. 2017; de Water et al. 2019; Mora et al. 2015), we also investigate the potential modifying effect of parental education and socioeconomic status (SES) in the association between BMC and ASD status. We also examine whether including these variables as covariates would impact the interactive associations of BMC with GSTT1 genotypes in relation to ASD.

Methods

Study design and populations

The Epidemiological Research on Autism in Jamaica (ERAJ) is an age- and sex-matched case-control study (children age 2–8 years) that began enrollment in December 2009, investigating the potential role of environmental exposures and three GST genes (GSTM1, GSTT1, GSTP1), individually or interactively, in relation to ASD. Detailed information regarding the recruitment and assessment of ASD cases and TD controls has been reported earlier (Rahbar et al. 2014; Rahbar et al. 2018; Rahbar et al. 2015a, b). In brief, after provision of written consent and child assent when applicable, we administered an SES questionnaire at the time of enrollment to assess demographic characteristics, education level of the parents, and the ownership of a car as a measure of the family’s SES in Jamaica. In addition, at the end of the interview and other assessments that included administration of ADOS-2 (Lord et al. 2015) and ADI-R (Rutter et al. 2003b) to suspected ASD cases from the Jamaican Autism Database to ascertain ASD status and administration of SCQ (Rutter et al. 2003a) to rule out developmental disorders in the TD control children, 4–5 mL of whole blood was collected from each child for assessment of exposure to the metals including Mn and to determine the genotypes for each of the GST genes. For the current study, we used data from 266 1:1 matched pairs of children 2–8 years of age [one ASD case and one age-(± 6 months) and sex-matched TD control] who were born in Jamaica and enrolled in the ERAJ study between December 2009 and September 2017.

Methods for assessment of the Mn exposure

Assays for BMC were performed at the Analytical Chemistry Lab at Michigan Department of Health and Human Services (MDHHS) in Lansing, Michigan, USA, that is certified by the Centers for Disease Control and Prevention (CDC). Venous whole blood samples were diluted and analyzed using a PerkinElmer Elan DRCII inductively-coupled plasma mass spectrometer (PerkinElmer, Waltham, MA) for trace metal analyses. All BMC in our samples are above the limit of detection of 1μg/L. The methods for analysis of BMC in blood by MDHHS have been described previously (Rahbar et al. 2014; Rahbar et al. 2018).

Genetic assessment

Whole blood was processed by the CARIGEN lab at UWI and then shipped to the UTHealth School of Public Health (UTSPH) Human Genetics Center (HGC) Laboratory in Houston, Texas. Genomic DNA was isolated from buffy coat using the Gentra PUREGENE Blood Kit (Qiagen, N.V., Venlo, The Netherlands). Regions of the GSTM1 and GSTT1 genes were amplified in two independent TaqMan Copy Number Assay reactions; GSTM1 Assay ID: Hs02575461_cn and GSTT1 Assay ID: Hs00010004_cn (www.thermofisher.com). Each quantitative PCR reaction contained 10 ng of genomic DNA, TaqMan Genotyping Master Mix, TaqMan Copy Number Assay, and TaqMan Copy Number Reference Assay (RNaseP) in a 5μL reaction in accordance with instructions provided by the manufacturer. Thermal cycling conditions were 2 min at 50°C, 10 min at 95°C, followed by 40 cycles of 15 sec at 95°C and 1 min at 60°C. The real-time QuantStudio and 12K Flex Software v1.2.2 were used to assay each sample in quadruplicate. Samples were normalized to RNaseP and averaged to obtain a single ΔCt value (FAM dye Ct – VIC dye Ct) which was imported into CopyCaller Software v2.1 (www.thermofisher.com) to determine the copy number of each genomic DNA target. GSTM1 and GSTT1 homozygous deletions were coded as DD and the presence of an insertion was coded as I*. Methods for genetic analysis of the GSTP1 Ile105Val polymorphism (rs1695) using the TaqMan Drug Metabolism SNP genotyping assay C_3217198_20 have been described in detail previously (Rahbar et al. 2018; Rahbar et al. 2015a, b).

Statistical analysis

In order to account for the possible correlation between data from ASD cases and TD controls due to the matched study design, we used conditional logistic regression (CLR) for all the analyses. Specifically, we used univariable CLR to compare the distributions of sociodemographic and socioeconomic characteristics, as well as the GSTP1, GSTT1, and GSTM1 genotypes between ASD cases and TD controls. Evidence from previous studies showed a nonlinear association between BMC and concurrent mental development scores (Claus Henn et al. 2010), and a direct association of cord blood Mn concentrations above the 75^th percentile with adverse cognitive, language, and overall quotients of the Comprehensive Developmental Inventory for Infants and Toddlers (Lin et al. 2013). Thus, for this analysis we categorized the BMC into four quartiles (i.e., BMC ≤ 8.3 μg/L, 8.3 < BMC ≤ 9.9 μg/L, 9.9 < BMC ≤ 12 μg/L, and BMC > 12 μg/L) and used the lowest 25% of BMC as the referent category. In order to minimize any potential effect of multicollinearity due to high association between the maternal and paternal education levels, we created a binary variable indicating whether both parents had education up to high school or at least one of the parents obtained education beyond high school. As mentioned earlier, based on evidence suggesting that socioeconomic characteristics may impact the associations between Mn exposure and neurodevelopment (Chiu et al. 2017; de Water et al. 2019; Mora et al. 2015), we assessed the potential modifying effect of parental education, as well as SES measured by ownership of a car by the family in the association between BMC (in 4 quartiles) and ASD status using multivariable CLR interactive models.

We examined a GSTT1 insertion-deletion polymorphism for which a homozygous deletion or a null genotype indicates the activities or functionality of these genes are reduced or interrupted completely. Since the genotyping assay does not distinguish between a normal homozygote (I/I) and a heterozygote (I/D) for GSTT1, we considered only a recessive model using a binary variable: I/* (i.e., I/I or I/D) and DD (null). For the GSTP1 Ile105Val polymorphism, there are three common genotypes (Ile/Val, Ile/Ile, Val/Val) and the replacement of adenine by guanine at nucleotide 562 results in the change of an amino acid from isoleucine to valine at codon 105 of the GSTP1 protein. We analyzed the GSTP1 gene using two different genetic models: co-dominant (Ile/Ile, Ile/Val, and Val/Val) and dominant (Ile/Ile vs Val/*). In addition, we tested whether the GSTP1 polymorphism in the TD control group met Hardy-Weinberg equilibrium expectations using the Chi-square test.

We used multivariable CLR models with ASD status as the dependent variable, and BMC divided into 4 quartiles and GSTT1 genotypes as independent variables to assess the interactive associations of BMC with GSTT1 genotype in relation to ASD. In addition, considering our previous study finding of a significant interaction between GSTT1 and GSTP1 genotypes in relation to ASD (Rahbar et al. 2015b), we used other multivariable interactive CLR models to account for this gene-gene interaction (i.e., GSTT1*GSTP1) using both co-dominant and dominant genetic models for GSTP1, while assessing the possible modifying effect of GSTT1 genotype on the association of BMC and ASD. Furthermore, we assessed whether including parental SES and level of education as covariates would impact the interactive associations of BMC with GSTT1 genotypes in relation to ASD. Covariates with Wald test p-value of less than 0.05 were kept in the final adjusted model. Additional details about all these models are provided in Tables 2, 3A, and 3B. To calculate the matched odds ratios (MORs) and 95% confidence intervals (CIs) for the association between categorized BMC and ASD among children with various GSTP1 genotypes, we used the CONTRAST statement in SAS statistical software (Kleinbaum and Klein 2010). All statistical tests were performed at the 0.05 level of significance without making any adjustments for multiple comparisons and were conducted using SAS 9.4 statistical software (SAS Institute 2013).

Table 2.

Interaction between blood manganese concentrations (BMC) and GSTT1 genotypes in relation to ASD in Jamaican children based on CLR model^*. (N = 258 matched pairs)

BMC quartiles	GSTT1 a Genotype (s)	Unadjusted*				Adjusted b
		Matched OR (MOR)	95% CI for MOR	P-Value c	P-Value for GSTT1*Mn Interaction	Matched OR (MOR)	95% CI for MOR	P-Value c	P-Value for GSTT1*Mn Interaction
2 vs. 1	I*	1.30	0.74, 2.29	0.36	0.02	1.28	0.72, 2.29	0.40	0.01
3 vs. 1		1.15	0.67, 1.97	0.62		1.17	0.67, 2.04	0.59
4 vs. 1		0.91	0.47, 1.75	0.77		0.95	0.48, 1.86	0.88
2 vs. 1	DD	1.45	0.54, 3.90	0.46		1.59	0.57, 4.41	0.38
3 vs. 1		0.52	0.20, 1.38	0.19		0.58	0.21, 1.58	0.29
4 vs. 1		3.32	1.08, 10.17	0.04		4.35	1.36, 13.92	0.01

^*CLR unadjusted Model 1: logit P(ASD=1) = β1(BMC quartile 2) + β2(BMC quartile 3) + β3(BMC quartile 4) + β4(GSTT1 DD) + β5(BMC quartile 2*GSTT1 DD) + β6(BMC quartile 3*GSTT1 DD) + β7(BMC quartile 4 *GSTT1 DD)

^aGSTT1 was missing for 3 ASD cases and 4 TD controls.

^bIn addition to the variables in the unadjusted model we adjusted for socioeconomic status.

^cP-Values are based on the Wald’s test in conditional logistic regression models.

Table 3A.

Interaction between blood manganese concentrations (BMC) and GSTT1 genotypes in relation to ASD in Jamaican children based on CLR model that accounts for gene-gene interaction between GSTT1 and GSTP1 (co-dominant model) ^*. (N = 257 matched pairs)

BMC quartiles	GSTT1 a Genotype (s)	Unadjusted*				Adjusted b
		Matched OR (MOR)	95% CI for MOR	P-Value c	P-Value for GSTT1*Mn Interaction	Matched OR (MOR)	95% CI for MOR	P-Value c	P-Value for GSTT1*Mn Interaction
2 vs. 1	I*	1.30	0.73, 2.29	0.37	0.03	1.27	0.71, 2.28	0.42	0.02
3 vs. 1		1.13	0.66, 1.95	0.66		1.15	0.65, 2.03	0.62
4 vs. 1		0.90	0.47, 1.74	0.76		0.95	0.48, 1.87	0.88
2 vs. 1	DD	1.42	0.52, 3.89	0.50		1.55	0.54, 4.42	0.41
3 vs. 1		0.56	0.21, 1.50	0.25		0.64	0.23, 1.76	0.39
4 vs. 1		3.27	1.06, 10.14	0.04		4.33	1.34, 14.04	0.02

^*CLR unadjusted Model 2: logit P(ASD=1) = β1(BMC quartile 2) + β2(BMC quartile 3) + β3(BMC quartile 4) + β4(GSTT1 DD) + β5(GSTP1 Ile/Val) + β6(GSTP1 Val/Val) + β7(BMC quartile 2 *GSTT1 DD) + β8(BMC quartile 3*GSTT1 DD)+ β9(BMC quartile 4*GSTT1 DD) + β10(GSTT1 DD*GSTP1 Ile/Val) + β11(GSTT1 DD*GSTP1 Val/Val) (under a co-dominant model for GSTP1)

^aGSTT1 missing for 3 ASD cases and 4 TD controls, and GSTP1 genotype was missing for 3 ASD cases.

^bIn addition to the variables in the unadjusted model we adjusted for socioeconomic status.

^cP-Values are based on the Wald’s test in conditional logistic regression models.

Table 3B.

Interaction between blood manganese concentrations (BMC) and GSTT1 genotypes in relation to ASD in Jamaican children based on CLR model that accounts for gene-gene interaction between GSTT1 and GSTP1 (dominant model) ^*. (N = 257 matched pairs)

BMC quartiles	GSTT1 a Genotype (s)	Unadjusted*				Adjusted b
		Matched OR (MOR)	95% CI for MOR	P-Value c	P-Value for GSTT1*Mn Interaction	Matched OR (MOR)	95% CI for MOR	P-Value c	P-Value for GSTT1*Mn Interaction
2 vs. 1	I*	1.30	0.74, 2.30	0.36	0.02	1.27	0.71, 2.28	0.42	0.01
3 vs. 1		1.15	0.67, 1.98	0.62		1.17	0.66, 2.05	0.59
4 vs. 1		0.91	0.47, 1.75	0.78		0.96	0.48, 1.88	0.90
2 vs. 1	DD	1.48	0.55, 4.02	0.44		1.65	0.59, 4.62	0.34
3 vs. 1		0.53	0.20, 1.39	0.20		0.59	0.22, 1.60	0.30
4 vs. 1		3.30	1.07, 10.14	0.04		4.34	1.35, 13.94	0.01

^*CLR unadjusted Model 3: logit P(ASD=1) = β1(BMC quartile 2) + β2(BMC quartile 3) + β3(BMC quartile 4) + β4(GSTT1 DD) + β5(GSTP1 Val/*) + β6(BMC quartile 2*GSTT1 DD) + β7(BMC quartile 3 *GSTT1 DD) + β8 (BMC quartile 4*GSTT1 DD) + β9 (GSTT1 DD * GSTP1 Val/*). (under a dominant model for GSTP1)

^aGSTT1 missing for 3 ASD cases and 4 TD controls, and GSTP1 genotype was missing for 3 ASD cases.

^bIn addition to the variables in the unadjusted model we adjusted for socioeconomic status.

^cP-Values are based on the Wald’s test in conditional logistic regression models.

Results

We compared the distributions of sociodemographic and socioeconomic characteristics of the children and their parents, as well as GST genotypes of the children between ASD cases and TD controls (Table 1). The mean age of ASD cases and TD controls at the time of enrollment was 63.5 and 64.0 months, respectively. As expected, about 81% of the ASD cases and TD controls were male. Nearly all of the ASD cases (95.5%) and TD controls (97%) were Afro-Caribbean. Similarly, 96.8% of mothers and 96.4% of fathers were Afro-Caribbean. Compared to TD controls, a significantly higher proportion of both the mothers (18.9% vs. 12.3%) and fathers (46% vs. 31.0%) of ASD cases were age 35 years or greater at the time of the child’s birth. Similarly, a higher proportion of the mothers (48.5%) and fathers (41.4%) of ASD cases had more than a high school education compared to the mothers (37.6%) and fathers (22.3%) of TD controls. A higher proportion of ASD cases were from families with higher SES compared to TD controls, with 56.4% of case families owning a car versus 39.9% car ownership by control families. The genotype frequencies for GSTP1, GSTT1, and GSTM1 were not significantly different between ASD cases and TD controls (all three P > 0.11). Furthermore, the GSTP1 allele frequency in the TD controls was in agreement with Hardy-Weinberg equilibrium expectations (P = 0.28).

Table 1.

Characteristics of children and their parents by ASD case status (266 matched pairs)

Variables	Categories	ASD Case (n=266) N (%)	TD Control (n=266) N (%)	P-Value*
Child’s sex	Male	217 (81.6)	217 (81.6)	1.00
Child’s age (months)	Age < 72	185 (69.5)	190 (71.4)	0.18
	Age ≥ 72	81 (30.5)	76 (28.6)
Child’s race	Afro-Caribbean	254 (95.5)	258(97.0)	0.37
Maternal age a (at child’s birth in years)	Age < 35	215 (81.1)	229 (87.7)	0.03
	Age ≥ 35	50 (18.9)	32 (12.3)
Paternal age b (at child’s birth in years)	Age < 35	141 (54.0)	176 (69.0)	<0.01
	Age ≥ 35	120 (46.0)	79 (31.0)
Maternal race	Afro-Caribbean	256 (96.2)	259 (97.4)	0.47
Paternal race c	Afro-Caribbean	254 (95.9)	255 (97)	0.49
Maternal education d (at child’s birth)	Up to high school †	137 (51.5)	164 (62.4)	0.02
	Beyond high school††	129 (48.5)	99 (37.6)
Paternal education e (at child’s birth)	Up to high school †	150 (58.6)	195 (77.7)	<0.01
	Beyond high school††	106 (41.4)	56 (22.3)
Parental educationf (at child’s birth)	Both up to high school †	98 (37.7)	132 (52.0)	<0.01
	At least one beyond high school††	162 (62.3)	122 (48.0)
Socioeconomic status (SES)	Car ownership	150 (56.4)	106 (39.9)	<0.01
GSTP1g	Ile/Ile	68 (25.9)	65 (24.4)	0.58
	Ile/Val	144 (54.8)	139 (52.3)
	Val/Val	51 (19.4)	62 (23.3)
GSTM1h	DD i	78 (29.7)	62 (23.6)	0.11
	I/I or I/Dj	185 (70.3)	201 (76.4)
GSTT1 k	DD i	70 (26.6)	65 (24.8)	0.70
	I/I or I/Dj	193 (73.4)	197 (75.2)

^*P-Values are based on Wald’s test in conditional logistic regression models.

^†Up to high school education means attended Primary/Jr. Secondary, and Secondary/High/Technical schools.

^††Beyond high school education means attended a Vocational, Tertiary College, or University.

^aMaternal age was missing for 1 ASD case and 5 TD controls.

^bPaternal age was missing for 5 ASD cases and 11 TD controls.

^cPaternal race was missing for 1 ASD case and 3 TD controls.

^dMaternal education was missing for 3 TD controls.

^ePaternal education was missing for 10 ASD cases and 15 TD controls.

^fParental education was missing for 6 ASD cases and 12 TD controls.

^gGSTP1 genotype was missing for 3 ASD cases.

^hGSTM1 genotype was missing for 3 ASD cases and 3 TD controls.

ⁱDD indicates the null alleles for GSTT1 and GSTM1.

^jI/I or I/D indicate the homozygote (I/I) or a heterozygote (I/D) for GSTT1 and GSTM1

^kGSTT1 was missing for 3 ASD cases and 4 TD controls.

In unadjusted Model 1 that we used to investigate a possible interaction between the categorized BMC (in 4 quartiles) and GSTT1 in relation to ASD, we found a significant interaction between GSTT1 genotypes and BMC in relation to ASD (overall interaction P = 0.02) suggesting potential effect modification by GSTT1 genotype when assessing the association between BMC and ASD status (Table 2). Specifically, using BMC ≤ 25^th percentile as the referent category, we found that among children with the DD genotype, the odds of having BMC > 75^th percentile (> 12 μg/L) in ASD cases was 3.32 times that of the TD control group, [MOR (95% CI) = 3.32 (1.08, 10.17), P = 0.04], while among children with the I/I or I/D genotypes, although not statistically significant, the odds of having BMC > 75^th percentile in ASD cases was lower than in the TD control group [MOR (95% CI) = 0.91 (0.47, 1.75), P = 0.77].

In other CLR models we also accounted for the interaction between GSTT1 and GSTP1 while assessing the possible modifying effect of GSTT1 genotype on the BMC-ASD association. The findings from these models also indicated a significant interaction between GSTT1 genotypes and BMC in relation to ASD when GSTP1 genotype was considered, either using a co-dominant (unadjusted Model 2, Table 3A) or dominant (unadjusted Model 3, Table 3B) genetic model, (overall interaction P = 0.03 and P = 0.02 for the co-dominant and dominant models, respectively). Specifically, considering BMC ≤ 25^th percentile as the referent category, we found that among children with the DD genotype, the odds of having BMC > 75^th percentile (> 12 μg/L) in ASD cases was 3.27 and 3.30 times that of the TD control group, when we accounted for the interaction of GSTT1 genotype with GSTP1 using co-dominant and dominant genetic models, respectively [MOR (95% CI) = 3.27 (1.06, 10.14), P = 0.04; and MOR (95% CI) = 3.30 (1.07, 10.14), P = 0.04], while this association was not statistically significant among children with the I/I or I/D genotypes in the same interactive models [MOR (95% CI) = 0.90 (0.47, 1.74), P = 0.76; and MOR (95% CI) = 0.91 (0.47, 1.75), P = 0.78].

In separate multivariable CLR interactive models we assessed a potential modifying effect of parental education, as well as SES measured by ownership of a car by the family in the association between BMC (in 4 quartiles) and ASD status. We did not find any significant interactions between parental education and BMC (overall interaction P = 0.83), or between family SES and BMC (overall interaction P = 0.18) in relation to ASD. However, when we included SES as a covariate while assessing the possible modifying effect of GSTT1 genotype on the association of BMC and ASD, SES remained significant in the model (Wald test P <0.01). Therefore, in all interactive models we also adjusted for SES as a covariate (adjusted Model 1, 2, and 3 in Tables 2, 3A, and 3B).

The significant interaction between GSTT1 genotypes and BMC in relation to ASD remained consistently significant after accounting for SES (overall interaction P = 0.01, adjusted Model 1, Table 2), as well as for SES and the gene-gene interaction between GSTT1 and GSTP1 using either co-dominant (adjusted Model 2, Table 3A) or dominant (adjusted Model 3, Table 3B) genetic models for GSTP1 (overall interaction P = 0.02 and P = 0.01 for the co-dominant and dominant models, respectively). Specifically, considering BMC ≤ 25^th percentile as the referent category, we found that among children with the DD genotype, the odds of having BMC > 75^th percentile (> 12 μg/L) in ASD cases was 4.33 and 4.34 times that of the TD control group, when we accounted for the family SES and the interaction of GSTT1 genotype with GSTP1 using co-dominant and dominant genetic models, respectively [adjusted MOR (95% CI) = 4.33 (1.34, 14.04), P = 0.02; and adjusted MOR (95% CI) = 4.34 (1.35, 13.94), P = 0.01], while this association was not statistically significant among children with the I/I or I/D genotypes in the same interactive models [adjusted MOR (95% CI) = 0.95 (0.48, 1.87), P = 0.88; and adjusted MOR (95% CI) = 0.96 (0.48, 1.88), P = 0.90].

Discussion

In this study, we investigated the possible interactive effects of GSTT1 genotypes and BMC divided into quartiles in relation to ASD in Jamaican children, and report that GSTT1 could be an effect modifier for the association between BMC and ASD status (overall interaction P = 0.02). Specifically, considering BMC ≤ 25^th percentile (≤ 8.3 μg/L) as the referent category, we found that among children with the DD genotype, the odds of having BMC > 75^th percentile (> 12 μg/L) in ASD cases was 3.32 times that of the TD control children, while among children with the I/I or I/D genotypes, there was no significant association between BMC and ASD. This finding remained consistent when we further accounted for the gene-gene interaction between GSTT1 and GSTP1 in relation to ASD that we reported previously (Rahbar et al. 2015b) and SES of the family. Specifically, considering BMC ≤ 25^th percentile (≤ 8.3 μg/L) as the referent category, we found that among children with the DD genotype, the odds of having BMC > 75^th percentile (> 12μg/L) in ASD cases was 4.33 and 4.34 times that of the TD control children, when we accounted for SES of the family and the interaction of GSTT1 genotype with GSTP1 using co-dominant and dominant genetic models, respectively.

In our previous studies, we reported a significant interaction between binary BMC (BMC ≥ 75^th percentile (12 μg/L) vs. <75^th percentile) and GSTP1 indicating that among children who had the Ile/Ile genotype for GSTP1, the odds of having BMC ≥ 12 μg/L vs. BMC< 12 μg/L in ASD cases was about 4 to 6 times that of the TD control children using either co-dominant or dominant genetic models (Rahbar et al. 2015a). In a more recent paper, we reported that the interaction between Mn and GSTP1 in relation to ASD remained significant with similar magnitude of associations after adjusting for the mixture of four other metals (Pb, Hg, As, and Cd) based on estimated mixture score using WQS (Rahbar et al. 2018). However, our findings of a significant interaction between GSTT1 and BMC divided into quartiles in relation to ASD reported here are new. To our knowledge, we are the first to report on interaction between GSTT1 and BMC in quartiles in relation to ASD. As mentioned earlier, the mechanisms for Mn-induced neurotoxicity could be due to enhanced production of reactive oxygen species in the affected areas of the brain that could lead to oxidative stress (Gavin et al. 1999; Lucchini et al. 2017). On the other hand, chronic Mn accumulation in the mitochondria has been shown to affect MnSOD activity and could result in an abnormal protective response against oxidative stress in the mitochondria that may lead to endothelial dysfunction (Koh et al. 2014). Therefore, susceptibility to Mn-induced neurotoxicity may vary by mitochondrial anti-oxidant function as the primary route of action against oxidative stress (Gavin et al. 1990). Absence of properly functioning GST genes in individuals with the GSTT1 null genotype (DD) may decrease the cell capacity to protect against oxidative stress, and subsequently cause mitochondrial damage, decrease effectiveness of detoxification, and increase susceptibility to neurological diseases (Pagano et al. 2014). In addition, there is evidence suggesting that mitochondrial dysfunction is more likely among children with ASD than in TD children (Bjørklund et al. 2020; Giulivi et al. 2010). All these findings from previous studies could provide explanations for our results suggesting that the GSTT1 gene may play a significant role in the pathway to oxidative stress and mitochondrial dysfunction in relation to ASD, hence serving as an effect modifier in the association between BMC and ASD. However, further investigation in other populations is warranted to support this possible role of GSTT1.

GST genes are a family of genes that encode phase II enzymes with similar functions in catalyzing detoxification of various endogenous metabolites of oxidative stress or exogenous xenobiotics. The GST gene polymorphisms may affect the detoxification capacity and the ability of cells to protect against oxidative stress, and therefore may contribute to the development of a wide range of diseases (Nebert and Vasiliou 2004). However, genetic variants may contribute to different levels of enzymatic activity. For example, while the GSTT1 null genotype (DD) results in complete enzyme inactivity (Hayes and Strange 2000), for the GSTP1 Ile105Val polymorphism, the impact of the single amino acid change at codon 105 on enzymatic function is not well-established and findings of previous studies are inconsistent. For example, a study by Williams et al. (2007) indicated that children of mothers with the GSTP1 Ile/Val genotype are more likely to have ASD (Williams et al. 2007). Another study by Aynacioglu et al. (2004) reported that the Val/Val homozygous genotype may be protective against developing asthma (Aynacioglu et al. 2004). Therefore, genetic polymorphisms in almost all classes of GST superfamily genes, may potentially interact with each other and contribute to individual susceptibility to oxidative stress (Frustaci et al. 2012), and with environmental factors associated with disorders such as ASD (Matelski and Van de Water 2016).

We have previously reported that for children with the Ile/Val GSTP1 genotype, those with the GSTT1 null genotype (DD) have almost three times higher odds of ASD when compared to children with either the GSTT1 I/I or I/D genotype (Rahbar et al. 2015b). Findings of a recent age- and sex-matched case-control study in Serbia suggested a similar gene-gene interaction, indicating an 83% lower odds of ASD in the presence of the combination of GSTT1 active (I/I or I/D) with the GSTP1 Ile/Ile genotype [OR (95% CI) = 0.17 (0.03, 0.99), P = 0.049] (Mandic-Maravic et al. 2019). They further explored the possibility of a gene-environment interaction between different GST gene variants and maternal tobacco smoking during pregnancy in relation to ASD, while accounting for the individual genotypes of GST genes and all of their possible two-way interactions. However, they did not find any significant association between maternal tobacco smoking during pregnancy as an environmental exposure and ASD, neither individually nor interactively with GST genes (Mandic-Maravic et al. 2019). Similar gene-gene interactions (GSTT1*GSTP1) have been reported in a recent case-control study that investigated other oxidative stress-related health outcomes such as lung cancer (Ritambhara et al. 2019). Their findings suggest increased odds of lung cancer among those who have the GSTT1 null (DD) genotype in combination with the GSTP1 Ile/Ile genotype [OR (95% CI) = 2.03 (1.20–3.45)]. In addition, they have reported a significant gene-environment interaction between GSTT1 null genotype and smoking status in relation to lung cancer (Ritambhara et al. 2019). However, they have not accounted for the gene-gene (GSTT1*GSTP1) interaction while assessing the gene-environment (GSTT1*smoking) interaction in relation to lung cancer. To the best of our knowledge, this is the first study that accounted for the gene-gene interaction between GSTT1 and GSTP1, while assessing the gene-environment interaction between the GSTT1 gene and BMC in relation to ASD. In the current study, the interactive effects of BMC and GSTT1 genotype remained significant with a very similar MOR magnitude (i.e., among children with the GSTT1 DD genotype, MOR for having BMC > 12 vs. ≤ 8.3 μg/L was 3.27 and 3.30 in ASD cases compared to TD controls, using a co-dominant or dominant genetic model for GSTP1, respectively) after controlling for the GSTT1-GSTP1 gene-gene interaction. These findings remained consistently significant when we adjusted for SES as an important covariate that may impact the associations between Mn exposure and neurodevelopment based on previous literature (Chiu et al. 2017; de Water et al. 2019; Mora et al. 2015). In addition, our findings did not indicate that SES was an effect modifier for the association between BMC and ASD status (overall interaction P = 0.18). Replication of our findings in other populations and in larger studies that are initially designed with sufficient power to explore these gene-gene and gene-environmental interactive effects is warranted.

The frequency of the GSTT1 null genotype is about 35% in the Jamaican population (Taioli et al. 2011), which appears to be higher than in our study (about 25%). Our findings suggest that for this subgroup of children, the odds of having higher levels of BMC [BMC in quartiles > 75^th percentile (> 12 μg/L) vs. ≤ 25^th percentile (≤ 8.3 μg/L)] among ASD cases is more than 3 times that in TD controls. The primary source of exposure to Mn is through diet (e.g., consumption of root vegetables, salt water fish and drinking water) (ATSDR 2012a, 2012b). In addition, the levels of Mn found in Jamaican soil are approximately twice the average for soils in other countries as reported by the International Atomic Energy Agency (Lalor 1996; Howe et al. 2005). Although genetic risk factors are not modifiable, implementation of interventions focused on dietary and environmental factors that could help to moderate exposure to Mn in Jamaican children could be considered, especially in those with genetic variants that make them more susceptible to adverse outcomes related to Mn exposures.

Limitations

We acknowledge several limitations in this study. First, we acknowledge that BMC, which we used as the biomarker of Mn exposure, may not be the best indicator of long term exposure as compared to tissue markers of exposure such as nervous tissues. Therefore, we were not able to account for timing of Mn exposure. However, since Jamaican soils contain a unique distribution of a variety of trace elements, including Mn, that can accumulate in crops and seafood (Howe et al. 2005; Lalor 1996) and because the Jamaican population consumes large amounts of locally grown fruits and vegetables along with large amounts of seafood (Howe et al. 2005), the residents may have continuous exposure to Mn, resulting in bioaccumulation (St-Pierre et al. 2001). Hence, our Mn levels could represent a long term exposure. Nevertheless, we acknowledge that BMC measurements for children represent “current” exposure, implying that the Mn exposure for ASD cases may only reflect BMC after the child is already diagnosed with ASD. In addition, our BMC data were not differentiated by type of Mn exposures (i.e., organic and inorganic), so we could not provide any discussion regarding this distinction in our data. We also lack information on maternal diet or Mn levels during pregnancy. Additionally, although we have conducted multiple testing for two different genes and a number of different genetic models, the reported p-values were not adjusted for multiple comparisons. Therefore, our findings should be replicated in other populations.

Conclusions

In this article, we have shown that among Jamaican children with the GSTT1 DD genotype, the odds of having higher BMC [i.e., BMC in quartiles > 75^th percentile (> 12 μg/L) vs. ≤ 25^th percentile (≤ 8.3 μg/L)] in ASD cases was about 4.3 times that of TD controls after accounting for SES of the family and the interaction between GSTT1 and GSTP1 under both a co-dominant and dominant genetic model for GSTP1. These findings suggest a possible synergistic effect of BMC categorized as quartiles and GSTT1 in relation to ASD. Although replication of our findings in other populations is warranted, these results support the development of targeted interventions focused on dietary and environmental factors that could help to moderate exposure to Mn in Jamaican children. These interventions could be effective especially among those with genetic variants that make them more susceptible to adverse outcomes related to Mn exposures. However, since appropriate quantities of Mn are essential for healthy development, we emphasize that regulating dietary intake of Mn should be supervised and discussed by qualified health professionals.