The Methylation Capacity of Arsenic and Insulin Resistance are Associated with Psychological Characteristics in Children and Adolescents

The goal of the present study was to compare the influence of the methylation capacity of arsenic, as well as insulin resistance on psychological characteristics of school students from elementary and junior high school. 296 elementary and 318 junior high school students participated in health examinations, completed questionnaires and determined their concentrations of urinary arsenic species and psychological characteristics. Insulin resistance was determined by means of the Homeostatic Model Assessment of Insulin Resistance (HOMA-IR). We found that HOMA-IR values were significantly related to increased score of the depression and anger after adjusted for age, gender, schools, father’s educational levels, mother’s educational levels, BMI, body fat, and urinary creatinine in all students. Anxiety scores and depression scores of junior high school children were significantly higher compared to elementary school children, but lower in disruptive behavior scores. HOMA-IR levels were significantly inversely related to self-concept scores in junior high school students. A greater urinary inorganic arsenic percentage (iAs%) was marginally significantly related to a higher depression score in junior high school students. This is the first study to show a relationship between HOMA-IR levels or urinary arsenic profiles and psychological distress in school students from elementary and junior high school.

The variability of arsenic methylation in the body may differentiate between the susceptibility to arsenic toxicity. Absorbed arsenate is reduced to arsenite and undergoes methylation to form monomethylarsonic acid (MMA V ) and dimethylarsinic acid (DMA V ), which have low toxicity 8 and are excreted by the kidneys 9 . However, in vitro studies have suggested that intermediate metabolites of inorganic arsenic such as monomethylarsonous acid (MMA III ) and dimethylarsenious acid (DMA III ) are more toxic than inorganic arsenic 10,11 , although epidemiologic data are not available. Our previous prospective study found that, with age, the percentage of MMA V (MMA V %) increased and the percentage of DMA V (DMA V %) decreased significantly 12 , suggesting that a decrease in the methylation capacity of arsenic is associated with aging. Our recent study demonstrated that total urinary arsenic (sum of the inorganic and methylated arsenic species) is negatively associated with BMI in adolescents in Taiwan, and obese adolescents with high insulin levels had significantly higher MMA% and significantly lower DMA% compared to obese adolescents with low insulin 13 . This implies that obesity and high insulin levels were associated with a reduced methylation capacity of arsenic in adolescents. In addition, we also found the homeostasis model assessment of insulin resistance (HOMA-IR) value was significantly and positively related to total urinary arsenic concentrations and the body mass index (BMI) Z score. Higher BMI values and higher total urinary arsenic concentrations were associated with higher HOMA-IR values in children and adolescents in Taiwan 14 .
Overweight and obese children have been shown to have a high risk of developing physiological abnormalities 15,16 , as well as being prone to psychosocial distress including depression, anxiety and social withdrawal 17,18 , leading to a poor quality of life 19 , and behavioral problems 20,21 . A recent study reported that children who were overweight/obese had significantly lower self-concept and less disruptive behavior 22 . In addition, a recent study reported that a low mood/depression in healthy children was associated with high HOMA-IR levels 23 .
It remains to be determined whether urinary arsenic profiles and insulin resistance can influence psychosocial distress (depression, anger, self-concept, anxiety, and disruptive behavior) in adolescents, even with low arsenic exposure. Therefore, the goal of the present study was to explore the effect of arsenic exposure, and insulin resistance using the HOMA-IR index on psychosocial distress in elementary school and junior high school students in an area of Taiwan with low arsenic exposure. Table 1 shows the city areas and schools in which children were living and studying. Fathers and mothers of elementary school students had higher educational levels than those of junior high school students. Among elementary school students, body fat percent, cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), glutamate oxaloacetate transaminase (GOT), and urea levels were significantly higher compared to junior high school students, and blood sugar, serum insulin, HOMA-IR values, and urine creatinine were significantly lower ( Table 1). The total urinary arsenic concentrations (μg/g creatinine) and DMA% were significantly higher in elementary school students compared to junior high school students, and iAs% was significantly lower. It seems that the methylation capability of arsenic in elementary school students was more efficient than that of junior school students (Table 1). Table 2 shows that anxiety scores and depression scores of junior high school children were significantly higher than those of elementary school children. In contrast, disruptive behavior scores of junior high school children were significantly lower than those of elementary school children. The distribution of categorized strata of anxiety scores and depression scores in junior high school was significantly different from those in elementary school children (Table 2).

Variables
Elementary school students (N = 296) Junior high school students (N = 318) p value  The comparison of total urinary arsenic concentrations, arsenic methylation indices and HOMA-IR values between different psychological characteristics groups in elementary school students is presented in Table 3. Total urinary arsenic concentration in students with a mild depression score was significantly higher than students with a normal and moderate/severe depression score in elementary school students. In addition, Total urinary arsenic concentration in students with a mild depression score and moderate/severe depression score was significantly higher than students with a normal depression score. Regarding arsenic methylation indices, we found that iAs% was significantly lower, and DMA% was significantly higher in students with abnormal depression score and anxiety score than in students with a normal depression score and anxiety score.
The HOMA-IR value in students had a normal score or higher than normal score for self-concept were significantly lower than students with a lower than normal self-concept score among junior high school students. In addition, the HOMA-IR value in students with a mild depression score was significantly higher than students with a normal or moderate/severe depression score in junior high school students. On the other hand, iAs% in students with an abnormal depression score or abnormal disruptive behavior score was significantly higher than students with a normal depression score or normal disruptive behavior score in junior high school students (Table 4).  We explored the association between parents' smoking status and the urinary total arsenic and arsenic methylation capacity indices of children (Supplementary Table S1); we found that the fathers' smoking status only affected the MMA% of children in elementary school. However, MMA% was not related to psychological characteristics in elementary school students or junior high school students in this study; therefore, family smoking status would not act as a confounding factor for analyzing the relationship between urinary total arsenic, iAs%, DMA% and psychological characteristics in elementary school students or junior high school students. In addition, we compared the different status of body fat, vegetables and fruit intake, along with family history, on the HOMA-IR values in children (Supplementary Table S2). We found that body fat affected HOMA-IR values in both elementary and junior high students, and that the diabetes history of the mother only affected HOMA-IR values in junior high students.
A multiple linear regression analysis between urinary arsenic profile, HOMA-IR values and psychological characteristics is shown in Table 5. The HOMA-IR values were significantly positively related to the depression and anger score adjusted for age, gender, schools, father's educational levels, mother's educational levels, BMI, body fat, and urinary creatinine in all students. In junior high school students, the HOMA-IR values were significantly negatively related to the self-concept score, and positively correlated with the depression and anger score. In addition, urinary total arsenic concentration was significantly positively related to the self-concept score. Moreover, a greater iAs% was associated with a higher depression score, to a marginal degree. Conversely, iAs% values were significantly positively related to the self-concept score in elementary school students. In addition, DMA% values, and HOMA-IR values were positively related to the anxiety score in elementary school students.
HOMA-IR value and iAs% were correlated with psychological characteristics respectively, therefore we analyzed the combined effects of HOMA-IR value and iAs% on psychological characteristics. We did not find any relationship between the combined effects of HOMA-IR values and iAs% and psychological characteristics in all students. Therefore, we found a sequential decrease in the OR of anxiety, depression and anger among those with no risk factor, one risk factor, and two risk factors (iAs% and HOMA-IR value) in a dose-dependent manner in elementary school students; inversely a sequential increase in the OR of depression among those with no risk factor, one risk factor, and two risk factors (iAs% and HOMA-IR value) in a dose-dependent manner in junior high school students (Table 6).

Discussion
To the best of our knowledge, this is the first study to show an association between multiple psychological characteristics including self-concept, anxiety, depression, anger, and disruptive behavior and HOMA-IR values and urinary arsenic profiles in children and adolescents in Taiwan. In this study, we have shown that elementary school students had a more efficient arsenic methylation capacity (higher DMA% and lower iAs%) than junior high school students; these results were similar to our previous study 13,14 . One interesting finding was that high HOMA-IR values were significantly associated with an increased depression and anger score in junior high school students and all students. A high HOMA-IR value was significantly associated with a lower self-concept score in junior high school students; in contrast, total urinary arsenic concentration was significantly related to the self-concept score in junior high school students. In addition, a higher iAs% was associated with a higher self-concept score in elementary school students and with a higher depression score in junior high school students after covariate adjustment.
The educational system in Taiwan is quite different from that in Western society; ninth grade (age 14-15) students have to face academic tracking in early adolescence 24 . A study reported that students taking the senior high school joint entrance examination slept less hours at night and were less alert during the daytime compared to those who were not taking the examination 25 , suggesting that the academic pressures that adolescents faced could influence their behavior. Therefore, the anxiety and depression scores were significantly higher in senior high school students than in the elementary school students of this study. By contrast, disruptive behavior scores of senior high school students were lower than those of elementary school students. A further study should be performed to verify this inconsistency.
Our previous study found that HOMA-IR values were significantly increased in relation to the total urinary arsenic (μg/L) 14 . This finding indicates that arsenic exposure may be related to β-cell dysfunction, increasing the risk of diabetes in Korean adults 26 . A longitudinal study found that healthy children with low moods had higher HOMA-IR values than those without 23 . This implies that neuropsychiatric syndromes alter metabolic networks such as insulin-glucose homeostasis, immuno-inflammatory processes and adipokine synthesis, and secretion is a defining pathophysiological component 27 . Understanding how depressive symptoms are linked to metabolism during childhood and adolescence may be important for identifying risk factors for diabetes. However, in our study, children with high HOMA-IR values were associated with high anger scores and depression scores. Taken together, high total urinary arsenic may induce an increase in HOMA-IR values, and result in impaired glucose homeostasis, which could be related to depression and anger in children and adolescents in Taiwan; however, this hypothesis needs further studies to confirm the underlying mechanisms of the reported association.
Insulin resistance has been shown to be negatively related to cognitive performance in adult humans 28 , and insulin resistance at the level of the brain may be associated with the effects on glucose uptake, and could influence mediators of brain function in adults 29 . However, one study reported that HOMA-IR was significantly and negatively correlated with self-perceived scholastic competence in girls 30 . We also found that HOMA-IR was significantly negatively associated with the self-concept score in junior high school students; the underlying reason why this relationship was age specific was beyond the scope of this study, but could be the subject of further investigation. By contrast, total urinary arsenic was significantly related to the self-concept score, which is opposite to what we originally hypothesized. This cannot be explained by the present study and needs further investigation. On the other hand, urinary arsenic concentrations greater than 50 μg/L were associated with poor scores on tests measuring visual-spatial reasoning, language and vocabulary, memory, intelligence, and math skills 31 , as well as a modest association with hyperactive behavior 32 in 6-8 year-old children in Mexico. However, total urinary arsenic concentration was significantly related with the self-concept score in this study, which requires further investigation. However, an interesting result showed that iAs% was significantly related to increasing depression score in this study. This may suggest that arsenic depletes s-adenosylmethione (SAM)-methyl levels in the arsenic methylation pathway, leading to alterations in DNA methylation 33 , and this epigenetic modification to the DNA, which may result in aberrant gene expression, even in the brain 34 , increases the risk of impaired cognition and enhanced susceptibility for mood disorders. However, whether other arsenic methylation capacity indices can also influence psychological distress or the link between a particular gene of DNA methylation and cognition deficits has yet to be elucidated. Overall, arsenic methylation capacity leads to insulin resistance 35 , and may then disrupt the clearance of β-amyloid protein 36 ; it may also increase tau dephosphorylation and microtubule binding of tau by weakening the activity of phosphoinositide 3-kinase/protein kinase B and adenosine monophosphate, resulting in neuronal degeneration 37 related to cognitive dysfunction. Insulin resistance may be related to  Table 4. Distribution of urinary total arsenic and HOMA-IR values according to psychological characteristics in junior high school students. aSelf-concept higher than normal vs. Self-concept lower than normal, p < 0.05; b Normal self-concept vs. Self-concept lower than normal, p < 0.05; c Normal depression vs. Mild depression, p < 0.05; d Mild depression vs. Moderate or Severe depression, p < 0.05.
hypothalamic-pituitary adrenal axis dysfunction resulting in depressive symptoms 38 . Hippocampal neurogenesis reduction and increased advanced glycation end products were associated with depression 39 . This study had several limitations that need to be taken into consideration when interpreting these results. Firstly, a single-spot measurement of urinary arsenic species and blood biochemical indices may not provide enough information. In addition, the methylation of arsenic and HOMA-IR values may be influenced by nutrients, for which information was unavailable in this study. However, the values could be reliable if all participants had no change to their lifestyle and maintained their homeostatic metabolism. Secondly, as this study had a cross-sectional design, it could not determine the causality of the observed associations. We cannot exclude the possibility that the association between high total urinary arsenic concentrations or high HOMA-IR values and psychological indices could be the result rather than the cause of a change in psychological index values. Thirdly, the measurement of psychological characteristics was self-reported using a rating scale, and may be incomplete or subject to socially desirable effects, interpretation of results should be treated with caution. Fourthly, we did not collect any information of potential factors, such as emission of exhaust gases, waste water or home decoration, which would probably affect arsenic methylation capacity. In addition, data concerning exercise was unavailable, and eating habits and family history include missing data; these variables may be related to insulin resistance in this study. The parents' level of income, which may influence children's psychological characteristics, was also unavailable. The fact that they cannot be adjusted is another limitation of this study. In addition, all students were sampled in Taipei city and New Taipei City (urban areas) and, therefore, our sample is a limited representation of the entire population of Taiwan. In spite of these limitations, this study represents the first attempt to address the effect of change in urinary arsenic profiles or HOMA-IR values on psychological characteristics after adjustments for BMI, educational levels and anthropometric measurements.

Methods
SCiENTiFiC RePoRTS | 7: 3094 | DOI:10.1038/s41598-017-03084-2 took part in a second study, from October 2010 to November 2011, and recruited 318 students. The Research Ethics Committee of the Taipei Medical University, Taipei, Taiwan, approved the study, which was conducted in agreement with standards specified in the World Medical Association Declaration of Helsinki. All participants came from Taipei City or New Taipei City. All study participants provided either their parents' or their own written informed consent form before participating in questionnaire interviews, or providing biological specimens. Two research assistants who had received 6 hours of specialized training performed the anthropometric measurements of weight and height for all elementary school students and junior high school students according to standard guidelines. Participants removed their shoes, and wore light clothing for measurements of standing height and weight in a rigid vertical position, using a standard medical balance scale. Height was measured to the nearest 0.5 cm, and weight was measured to the nearest 100 g. Body mass index (BMI) was calculated as weight (kg)/height (m 2 ). Categories were divided into overweight, obese, and lower than normal weight, defined according to guidelines developed by the Ministry of Health and Welfare, Executive Yuan, Taiwan 40 based on WHO Child Growth Standards 41 , and a modified locally weighted method 42 designed for use with children and adolescents, based on BMI, age, and gender. Body fat as a percentage of weight was calculated, to obtain two measurements, by a commercially available bioelectrical impedance analyzer (Maltron BioScan 920 analyzer, Maltron International Ltd). All participants lived in Taipei City and New Taipei City, which are without obvious arsenic exposure. Participants drank tap water with arsenic levels less than the standard 10 μg/L provided by the Taipei Water Department of the Taipei City Government. The average arsenic concentration of tap water in Taipei City is 0.7 µg/L, and ranged from non-detectable to 4.0 µg/L.

Questionnaire interview.
Well-trained interviewers carried out face-to-face interviews to collect information using a structured questionnaire. The questionnaire included information on demographics and socioeconomic characteristics, lifestyle behavior of parents, such as cigarette smoking and alcohol consumption, and personal and family disease history.
Biological specimen collection. About 5 to 8 mL of peripheral blood was collected from participants using vacuumed syringes at the time of recruitment. Blood samples were then separated into red blood cells and serum, and frozen at −80 °C for subsequently measuring biochemical indices. Concurrently, spot urine samples of 20 ml were collected and immediately transferred to a −20 °C freezer until required for urinary arsenic species analyses.
Urinary arsenic species measurement. The analytical methods for determining urinary arsenic species have been described previously 43   and the detection limits for iAs III , DMA V , MMA V and iAs V were 0.02, 0.08, 0.05 and 0.07 μg/L respectively. The certified value of standard reference material (SRM 2670) was 480 ± 100 μg/L of inorganic arsenic obtained from the National Institute of Standards and Technology (NIST). SRM 2670 was used as a quality standard and analyzed along with urine samples. The mean value of SRM 2670 determined by our system was 507 ± 17 μg/L (n = 4). Both arsenobetain and arsenocholin are ingested from seafood and are excreted without metabolic transformation and, therefore, arsenobetain and arsenocholin were not measured by the HG-AAS method 44 .
Serum biochemical examination. Total cholesterol, triglyceride, high density lipoprotein (HDL) cholesterol, low density lipoprotein (LDL) cholesterol and insulin serum levels were determined by means of an autoanalyzer (Hitachi 737, USA) with reagents obtained from Boehringer Mannheim Diagnostics. An enzymatic assay for serum homocysteine was described by Chan et al. 45 . A close correlation (r > 0.9) was observed between the results from the enzymatic method and a high performance liquid chromatography procedure used as ref. 45 . HOMA-IR values were calculated using the formula: fasting insulin (μU/mL) × fasting glucose (mg/dL)/405 46 .
Measurement of psychological characteristics. We used the Beck Youth Inventories, second edition (BYI-II), to evaluate the children's reported thoughts, feelings, and behavior related to emotional and social dysfunction 47 . A child's experiences in the five psychological domains, namely self-concept (assessing cognitions of competence, potency, and positive self-worth), anxiety (assessing worries about school performance, the future, negative reactions of others, and fears), depression (assessing negative thoughts about the self, life, and the future, and feelings of sadness and guilt), anger (assessing thoughts of being treated unfairly by others, and feelings of anger and hatred), and disruptive behavior (assessing thoughts and behavior associated with conduct disorder and oppositional defiant behavior) were measured, and each inventory included 20 questions. Every question was scored using a 4-point Likert scale, with 0 indicating never and 3 indicating always. Total scores were summed for each inventory. Thus, higher scores indicated a stronger display of a particular psychological domain. To enable comparison of a child's score with those of other age-matched children, raw scores were converted (standardized) into T scores (with a mean of 50 and a standard deviation of 10). The internal consistency was indicated as Cronbach's α coefficients, and ranged from 0.86 to 0.96 for all age groups on all scales, good test-retest reliability was reported to range from 0.74 to 0.93 48 . Validity was supported by correlations with other instruments that assess similar characteristics. Both the reliability and validity were further established using the Chinese version of the BYI-II 43 . All Cronbach's α coefficients were clearly greater than 0.9 for each of the five inventories, and test-retest reliability ranged from 0.64 to 0.81. Criterion-related validity was also supported for each inventory. Although continuous T scores were generally preferable in statistical modeling, providing greater variability, we wanted to identify children with psychological difficulties above or below the "warning line", who deserved special attention. We used cutoff points suggested in the BYI-II Manual to categorize students into "lower than normal, normal" and "higher than normal" for each of the five psychological domains of depression, anxiety, anger, disruptive behavior, and self-concept 48,49 .
Statistical analysis. The sum of urinary inorganic arsenic (iAs III and iAs V ) and methylated arsenic (MMA V and DMA V ) concentrations (μg/L) was defined as the total urinary arsenic. The inorganic arsenic percentages (iAs%), MMA V %, and DMA V % were calculated by dividing the concentration of each species [(iAs III + iAs V ), MMA V and DMA V ] by the total urinary arsenic concentrations. The Student's t-test was used to compare differences in the variables of age, biochemical indices, psychological characteristics, and urinary arsenic profiles between elementary school and junior high school students. The χ 2 test was used to test for differences in categorical variables between elementary school and junior high school students. Analysis of variance (one-way ANOVA) and least significant difference test (post hoc test) with Bonferroni correction were used to compare the differences in variables of urinary arsenic profiles, and HOMA-IR levels among different psychological characteristics. Multiple linear regression models were used to estimate multivariate adjusted associations between the urinary arsenic profiles, HOMA-IR levels and psychological characteristics, and presented in regression coefficients (β) and standard error (SE). For dose-dependent relationships, trend analysis was performed by treating ordinal-score variables as continuous variables in the regression model. For the interaction analysis, the cutoff point for iAs% and HOMA-IR value was the median of the overall students, elementary school students and junior high school students, respectively.