Sarah E.
Rothenberg
*a,
Susan A.
Korrick
bc,
Donald
Harrington
d,
Sally W.
Thurston
de,
Sarah E.
Janssen
f,
Michael T.
Tate
f,
YanFen
Nong
g,
Hua
Nong
g,
Jihong
Liu
h,
Chuan
Hong
i and
Fengxiu
Ouyang
j
aCollege of Health, Oregon State University, 103 Milam Hall, Corvallis, OR 97331, USA. E-mail: sarah.rothenberg@oregonstate.edu; Tel: +1 541-737-3732
bChanning Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA 02115, USA
cDepartment of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA
dDepartment of Biostatistics and Computational Biology, University of Rochester Medical Center, Rochester, NY 14642, USA
eDepartment of Environmental Medicine, University of Rochester Medical Center, Rochester, NY 14642, USA
fU.S. Geological Survey Upper Midwest Water Science Center, Madison, WI 53726, USA
gMaternal and Child Health Hospital, Daxin County, China
hDepartment of Epidemiology & Biostatistics, Arnold School of Public Health, University of South Carolina, Columbia, SC 29208, USA
iDepartment of Environmental Health Sciences, Arnold School of Public Health, University of South Carolina, Columbia, SC 29208, USA
jMinistry of Education and Shanghai Key Laboratory of Children's Environmental Health, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
First published on 28th August 2024
Background. Fish and rice are the main dietary sources of methylmercury (MeHg); however, rice does not contain the same beneficial nutrients as fish, and these differences can impact the observed health effects of MeHg. Hence, it is important to validate a biomarker, which can distinguish among dietary MeHg sources. Methods. Mercury (Hg) stable isotopes were analyzed in hair samples from peripartum mothers in China (n = 265). Associations between mass dependent fractionation (MDF) (δ202Hg) and mass independent fractionation (MIF) (Δ199Hg) (dependent variables) and dietary MeHg intake (independent variable) were investigated using multivariable regression models. Results. In adjusted models, hair Δ199Hg was positively correlated with serum omega-3 fatty acids (a biomarker for fish consumption) and negatively correlated with maternal rice MeHg intake, indicating MIF recorded in hair can be used to distinguish MeHg intake predominantly from fish versus rice. Conversely, in adjusted models, hair δ202Hg was not correlated with measures of dietary measures of MeHg intake. Instead, hair δ202Hg was strongly, negatively correlated with hair Hg, which explained 27–29% of the variability in hair δ202Hg. Conclusions. Our results indicated that hair Δ199Hg can be used to distinguish MeHg intake from fish versus rice. Results also suggested that lighter isotopes were preferentially accumulated in hair, potentially reflecting Hg binding to thiols (i.e., cysteine); however, more research is needed to elucidate this hypothesis. Broader impacts include 1) validation of a non-invasive biomarker to distinguish MeHg intake from rice versus fish, and 2) the potential to use Hg isotopes to investigate Hg binding in tissues.
Environmental significanceThere is evidence that hair mercury (Hg) isotopes, specifically mass independent fractionation (MIF), can be used as a conservative tracer for dietary methylmercury (MeHg) intake; however, prior studies had small sample sizes (n = 5–45) and most studies focused on fish consumers. To the best of our knowledge, this is the first large-scale study to investigate the utility of hair Hg isotopes as non-invasive biomarkers of dietary MeHg intake (n = 265 peripartum mothers). Most mothers ingested rice daily while a majority of mothers also ingested some fish. Using a robust modeling approach (multivariable regression), our results support the notion that MIF can distinguish MeHg intake primarily from rice versus fish, providing a potentially valuable non-invasive MeHg exposure biomarker for health effects studies. |
Mercury (Hg) has seven stable isotopes, which can be used to trace sources and geochemical pathways of Hg in the environment.4 All Hg isotopes undergo mass-dependent fractionation (MDF, represented by δ202Hg), which is produced during biological, chemical, and physical transformations of Hg.4 In addition, Hg is one of the few elements that undergoes mass-independent fractionation (MIF), which is a deviation from the predicted MDF for an isotope.4 Hg-MIF of odd isotopes (represented by Δ199Hg and Δ201Hg) mainly occurs during photochemical processes,5,6 while Hg-MIF of even isotopes (represented by Δ200Hg and Δ204Hg) has been reported for gaseous Hg(0) processes in the atmosphere.7,8
To date, just 10 studies have reported Hg isotopes in human hair, and all were limited to small sample sizes (n = 5–45) (Table S1†).9–18 Two main findings have been reported. First, unlike MDF, the MIF induced by photochemical reactions was conserved during trophic transfer and was used to track dietary MeHg intake from seafood and rice.10–16 However, some discordance (i.e., lack of association) was noted between Δ199Hg and estimates of fish/rice consumption in our prior study among 21 mothers,15 suggesting other factors contributed to these associations (discussed below). Second, higher hair δ202Hg values relative to the dietary source (∼2‰) were hypothesized to reflect MeHg metabolism (i.e., demethylation) by gut microbiota.13,19 Given that MeHg's isotopic composition (specifically MDF) is altered during metabolism, other factors that influence MeHg metabolism may also impact the MDF and MIF isotopic composition of hair; however, these associations have not been investigated in prior studies. For example, fruit and fiber intake were both associated with lower uptake of MeHg in hair and tissues,20,21 which potentially reflected more efficient metabolism and excretion of MeHg. Factors associated with altered structure or function of the gut microbiome, including body mass index (BMI)22 and dietary omega-3 fatty acid intake,23 may also impact MeHg metabolism. Higher BMI, itself, was also associated with lower blood Hg in adults suggesting potential BMI-related differences in metabolism.24 Thus, dietary fiber/fruit intake and BMI may be important covariates to consider when assessing the relationship of dietary MeHg intake with MDF, while seafood intake (a rich source of omega-3 fatty acids)25 may influence both MDF and MIF.
We leveraged our birth cohort study in rural China, including 265 peripartum mothers,26–28 to determine the utility of hair Hg isotopes as non-invasive biomarkers for dietary MeHg intake and metabolism, the first large-scale study to date.
Hg isotopes were analyzed at the U.S. Geological Survey Mercury Research Laboratory (Madison, WI), as previously described.31,32 Isotope ratios were measured using a Neptune Plus multicollector-inductively coupled plasma-mass spectrometer (MC-ICP-MS, Thermo Finnigan Neptune) (ESI,† detailed methods).31,32 MDF (represented by δ202Hg) was calculated using eqn (1). MIF was expressed as ΔxxxHg (represented by Δ199Hg and Δ201Hg), which was defined as the difference between the measured δxxxHg value and the value predicted based on the kinetic MDF law derived from transition state theory (i.e., δ202Hg value multiplied by a constant) (eqn (2) and (3)).4,33
δ202Hg (‰) = [(δ202Hg/δ198Hg)sample/(δ202Hg/δ198Hg)NIST 3133 − 1] × 1000 | (1) |
Δ199Hg = δ199Hg − (δ202Hg × 0.2520) | (2) |
Δ201Hg = δ201Hg − (δ202Hg × 0.7520) | (3) |
A secondary standard (UM-Almadén) was run every five samples to ensure accuracy and precision of measurements. UM-Almadén and IAEA-086 isotope values agreed with established values (Table S2; ESI, and Dataset S1†).
Multivariable regression models were used to investigate associations between hair MDF (δ202Hg) and hair MIF (Δ199Hg) (dependent variables) and dietary MeHg intake (independent variable). Four potential estimates of dietary MeHg intake were identified to evaluate whether the MeHg source (rice or fish) impacted the hair Hg isotopic composition:
Model A: weekly fish ingestion (3 categories) and daily rice ingestion (yes/no); Model B: %MeHg intake from rice; Model C: log10 rice MeHg intake (μg per day) and log10 fish MeHg intake (μg per day); and Model D: log10 serum omega-3 fatty acids (mg mL−1) and log10 serum n-6 fatty acids (mg mL−1).
Each isotopic endpoint (δ202Hg or Δ199Hg) was examined in relation to these four sets of variables (ESI,† detailed methods). Prior studies have reported a positive association between Δ199Hg and weekly fish ingestion (Model A), and an inverse association between Δ199Hg and %MeHg intake from rice (Model B).10,15 Percent MeHg intake from rice was derived as the ratio of rice MeHg intake to the sum of rice and fish MeHg intake, which was used in Model C. Fish tissue is a rich source of omega-3 fatty acids, which compete with omega-6 fatty acids for the same enzymes in biosynthetic pathways;25 thus both were included in Model D.
We fit both unadjusted and covariate-adjusted models. The latter were adjusted for the same set of covariates, which were selected based on inclusion in previous studies investigating dietary MeHg exposure and metabolism.19,20,23,27,28 Additional covariates included maternal age at study enrollment (years), maternal pre-pregnancy BMI (3 categories: underweight, normal weight, and overweight/obese), log10 maternal energy intake (kcal), maternal proportion of calories (%) from protein, log10 maternal hair THg, and whether one or both parents was a farmer (yes/no). Energy intake (kcal) served as a proxy for fruit intake (servings per day) and fiber intake (g per day) (Spearman's rho: 0.52 and 0.85, respectively). A log10-transformation was applied to continuous variables that were skewed right to improve normality of the residuals. Missing data on covariates were imputed as the mean from multiple imputation based on the multivariate normal distribution,37 conditional on parental and child characteristics, and maternal biomarker concentrations, as previously described.27,28 Cook's distance was used to assess influential observations, and assumptions for model residuals were checked (no evidence of non-linearity, constant variance, normal distribution).
As sensitivity analyses, we investigated regression models: (1) excluding mothers with imputed data; (2) including hair samples corresponding to exposures during the second trimester only; (3) excluding 21 hair samples previously measured;15 and (4) excluding households where one or both parents were farmers because parental occupation (farmer/non-farmer) correlated with rice/fish ingestion (Table S3†). Lastly, as a sensitivity analysis, associations between hair Hg isotopes and hair MeHg concentrations were investigated in a subset of 15 mothers for which hair MeHg concentrations were previously analyzed.26
An alpha-level of 0.05 was selected as a guide for significance. Analyses were performed using Stata (Version 17.0, College Station, TX, USA), and the R-platform (Version 4.3.0, 21 April 2023).
Fig. 1 Comparison of hair mercury (Hg) isotopes, including (a) Δ199Hg versus δ202Hg (all studies), (b) Δ199Hg versus Δ201Hg (all studies), (c) Δ199Hg versus δ202Hg (this study only), and (d) Δ199Hg versus Δ201Hg (this study only). Representative values for 2 standard deviations of analytical uncertainty measured for this study (c and d) were 0.08 (δ202Hg), 0.05 (Δ199Hg) and 0.05 (Δ201Hg). For a and b, studies were ranked from highest to lowest median hair Δ199Hg values (Table S1†). |
Compared to nearly all other study populations (11 of 14 populations) in which hair Hg isotopes were reported (Fig. 1 and Table S1†),9–14,16–18 the median values for δ202Hg, Δ199Hg, and Δ201Hg from the present study were lower. There were just three exceptions; lower median values for one or more hair Hg isotopes were reported in Guizhou province, China10,14 and among Ghanaian artisanal gold miners using Hg amalgamation,17 compared to the present study.
Hair Δ199Hg (‰) | Hair δ202Hg (‰) | |||||
---|---|---|---|---|---|---|
Beta | (95% confidence interval) | p-Value | Beta | (95% confidence interval) | p-Value | |
a *p < 0.05 **p ≤ 0.01 ***p ≤ 0.001 p-value is for the Beta coefficient. BMI (body mass index), Hg (mercury), THg (total mercury). | ||||||
Unadjusted | ||||||
Fish weekly servings | ||||||
Never or rarely | (Referent) | (Referent) | ||||
0 < servings per weekly ≤ 2 | 0.018 | (-0.017, 0.054) | 0.31 | −0.042 | (-0.19, 0.11) | 0.58 |
≥ 2 servings per weekly | 0.082 | (0.024, 0.14) | 0.005** | −0.0015 | (-0.24, 0.24) | 0.99 |
Daily rice ingestion (yes) | 0.00052 | (-0.048, 0.049) | 0.98 | −0.034 | (-0.24, 0.17) | 0.74 |
Adjusted | ||||||
Fish weekly servings | ||||||
Never or rarely | (Referent) | (Referent) | ||||
0 < servings per weekly ≤ 2 | 0.0052 | (-0.031, 0.042) | 0.78 | −0.031 | (-0.16, 0.10) | 0.65 |
≥ 2 servings per weekly | 0.050 | (-0.016, 0.12) | 0.14 | 0.079 | (-0.16, 0.32) | 0.52 |
Daily rice ingestion (yes) | −0.0015 | (-0.062, 0.059) | 0.96 | −0.0013 | (-0.22, 0.22) | 0.99 |
log 10 hair THg (μg g −1 ) | −0.11 | (−0.19, −0.032) | 0.006** | −1.7 | (−1.8, −1.1) | <0.001*** |
Maternal age (years) | 0.00034 | (−0.0026, 0.0033) | 0.82 | 0.016 | (0.0047, 0.026) | 0.005** |
Pre-pregnancy BMI (kg m− 2 ) | ||||||
Underweight | (Referent) | (Referent) | ||||
Normal weight | 0.0068 | (-0.033, 0.047) | 0.74 | 0.052 | (-0.094, 0.20) | 0.49 |
Overweight or obese | 0.0042 | (-0.048, 0.056) | 0.87 | 0.12 | (-0.073, 0.31) | 0.23 |
Mother or father is a farmer (yes) | −0.054 | (−0.095, −0.013) | 0.01** | −0.097 | (-0.25, 0.053) | 0.21 |
log 10 maternal energy intake (kcal) | 0.0039 | (-0.15, 0.14) | 0.96 | −0.042 | (-0.58, 0.49) | 0.88 |
Maternal %calories from protein | 0.0034 | (-0.0045, 0.011) | 0.40 | −0.014 | (-0.043, 0.014) | 0.33 |
Hair Δ199Hg (‰) | Hair δ202Hg (‰) | |||||
---|---|---|---|---|---|---|
Beta | (95% confidence interval) | p-Value | Beta | (95% confidence interval) | p-Value | |
a *p < 0.05 **p ≤ 0.01 ***p < 0.001 p-value is for the Beta coefficient. BMI (body mass index), Hg (mercury), THg (total mercury). b Two mothers did not eat rice or fish, reducing the sample size from 265 to 263. | ||||||
Unadjusted | ||||||
% MeHg intake from rice | −0.00060 | (−0.0011, −0.00011) | 0.02* | 0.00035 | (−0.0017, 0.0025) | 0.74 |
Adjusted | ||||||
% MeHg intake from rice | −0.00031 | (−0.00086, 0.00023) | 0.26 | −0.00019 | (−0.0022, 0.0018) | 0.86 |
log 10 hair THg (μg g −1 ) | −0.12 | (−0.19, −0.039) | 0.003** | −1.4 | (−1.7, −1.2) | <0.001*** |
Maternal age (years) | 0.00033 | (-0.0025, 0.0032) | 0.82 | 0.016 | (0.0050, 0.026) | 0.004** |
Pre-pregnancy BMI (kg m− 2 ) | ||||||
Underweight | (Referent) | |||||
Normal weight | 0.0013 | (−0.037, 0.040) | 0.95 | 0.046 | (−0.099, 0.19) | 0.54 |
Overweight or obese | 0.0038 | (−0.047, 0.055) | 0.88 | 0.12 | (−0.073, 0.31) | 0.23 |
Mother or father is a farmer (yes) | −0.058 | (−0.098, −0.018) | 0.004** | −0.092 | (−0.24, 0.056) | 0.22 |
log 10 maternal energy intake (kcal) | 0.023 | (−0.091, 0.14) | 0.69 | −0.091 | (−0.51, 0.33) | 0.67 |
Maternal %calories from protein | 0.0031 | (−0.0045, 0.011) | 0.42 | −0.0081 | (−0.036, 0.020) | 0.57 |
Hair Δ199Hg (‰) | Hair δ202Hg (‰) | |||||
---|---|---|---|---|---|---|
Beta | (95% confidence interval) | p-Value | Beta | (95% confidence interval) | p-Value | |
a *p < 0.05 **p < 0.01 ***p < 0.001 p-value is for the Beta coefficient. BMI (body mass index), Hg (mercury), THg (total mercury). | ||||||
Unadjusted | ||||||
log 10 rice MeHg intake (μg per day) | −0.031 | (-0.063, 0.0024) | 0.052 | −0.15 | (−0.28, −0.019) | 0.03* |
log 10 fish MeHg intake (μg per day) | 0.025 | (0.0068, 0.043) | 0.007** | −0.035 | (−0.11, 0.040) | 0.36 |
Adjusted | ||||||
log 10 rice MeHg intake (μg per day) | −0.038 | (−0.075, −0.00098) | 0.04* | −0.096 | (−0.23, 0.040) | 0.17 |
log 10 fish MeHg intake (μg per day) | 0.014 | (−0.0066, 0.034) | 0.19 | −0.014 | (−0.088, 0.061) | 0.72 |
log 10 hair THg (μg g −1 ) | −0.10 | (−0.18, −0.020) | 0.01* | −1.4 | (−1.7, −1.1) | <0.001*** |
Maternal age (years) | 0.00021 | (−0.0027, 0.0031) | 0.89 | 0.015 | (0.0046, 0.026) | 0.005** |
Pre-pregnancy BMI (kg m− 2 ) | ||||||
Underweight | (Referent) | |||||
Normal weight | 0.0063 | (−0.033, 0.046) | 0.75 | 0.048 | (−0.096, 0.19) | 0.51 |
Overweight or obese | 0.0072 | (−0.045, 0.059) | 0.78 | 0.12 | (−0.067, 0.31) | 0.20 |
Mother or father is a farmer (yes) | −0.057 | (−0.098, −0.017) | 0.006** | −0.11 | (−0.26, 0.038) | 0.15 |
log 10 maternal energy intake (kcal) | 0.061 | (−0.074, 0.20) | 0.38 | 0.13 | (−0.36, 0.63) | 0.60 |
Maternal %calories from protein | 0.0017 | (−0.0060, 0.0094) | 0.67 | −0.015 | (−0.043, 0.014) | 0.31 |
Hair Δ199Hg (‰) | Hair δ202Hg (‰) | |||||
---|---|---|---|---|---|---|
Beta | (95% confidence interval) | p-Value | Beta | (95% confidence interval) | p-Value | |
a *p < 0.05 **p < 0.01 ***p ≤ 0.001 p-value is for the Beta coefficient. BMI (body mass index), Hg (mercury), N-3 fatty acids (docosahexaenoic acid, eicosapentaenoic acid, and alpha-linolenic acid), N-6 fatty acids (linoleic acid and arachidonic acid), total mercury (THg). | ||||||
Unadjusted | ||||||
log 10 serum N-3 fatty acids (mg mL −1 ) | 0.25 | (0.13, 0.38) | <0.001*** | −0.11 | (−0.66, 0.44) | 0.70 |
log 10 serum N-6 fatty acids (mg mL −1 ) | 0.11 | (-0.19, 0.41) | 0.49 | 0.12 | (−1.2, 1.4) | 0.86 |
Adjusted | ||||||
log 10 serum N-3 fatty acids (mg mL −1 ) | 0.27 | (0.14, 0.39) | <0.001*** | 0.17 | (−0.30, 0.64) | 0.47 |
log 10 serum N-6 fatty acids (mg mL −1 ) | −0.0059 | (-0.29, 0.30) | 0.97 | −0.49 | (−1.6, 0.64) | 0.40 |
log 10 hair THg (μg g −1 ) | −0.13 | (−0.20, −0.052) | 0.001*** | −1.5 | (−1.7, −1.2) | <0.001*** |
Maternal age (years) | 0.00040 | (−0.0024, 0.0032) | 0.78 | 0.016 | (0.0056, 0.027) | 0.003** |
Pre-pregnancy BMI (kg m− 2 ) | ||||||
Underweight | (Referent) | |||||
Normal weight | −0.0034 | (-0.042, 0.035) | 0.86 | 0.043 | (−0.10, 0.19) | 0.56 |
Overweight or obese | 0.0025 | (-0.053, 0.048) | 0.92 | 0.11 | (−0.078, 0.30) | 0.25 |
Mother or father is a farmer (yes) | −0.046 | (−0.086, −0.0068) | 0.02* | −0.10 | (−0.25, 0.048) | 0.18 |
log 10 maternal energy intake (kcal) | −0.0020 | (-0.11, 0.11) | 0.97 | −0.077 | (−0.49, 0.34) | 0.72 |
Maternal %calories from protein | 0.0049 | (−0.0019, 0.012) | 0.16 | −0.0094 | (−0.035, 0.016) | 0.47 |
In Models C and D, in the adjusted models, hair Δ199Hg was negatively associated with rice MeHg intake (beta: −0.038‰, 95% CI: −0.075‰, −0.00098‰) (Table 3), and positively associated with serum omega-3 fatty acids (beta: 0.27‰, 95% CI: 0.14‰, 0.39‰) (Table 4), respectively. Moreover, in all four models, hair Δ199Hg was negatively associated with hair THg (Fig. 2b), and whether one or both parents were farmers (Tables 1–4). The latter association likely reflected lower fish consumption among farmers, compared to non-farmers (Table S3†).
Fig. 2 Partial regression plots from the fully adjusted regression models relating hair mercury (Hg) isotopes versus log10 hair total mercury (THg) for a) hair δ202Hg, and (b) hair Δ199Hg. These results are for Model D (Table 4); similar results were observed for Models A–C (Tables 1–3) (n = 265 mothers). For Model D, models were adjusted for log10 serum n-3 fatty acids (mg mL−1), log10 serum n-6 fatty acids (mg mL−1), maternal age (years), pre-pregnancy body mass index (kg m−2) (3 categories), whether at least one parent was a farmer (yes/no), log10 maternal energy intake (kcal), and %calories from protein. |
In the unadjusted model (Model C), hair δ202Hg was negatively associated with rice MeHg intake (beta: −0.15‰, 95% CI: −0.28‰, −0.019‰) (Table 3). However, in the adjusted models (Models A–D), hair δ202Hg was not associated with estimates for rice or fish ingestion (Tables 1–4). In all four models, hair δ202Hg was strongly, negatively associated with hair THg (Fig. 2a), and positively associated with maternal age (Tables 1–4).
Regression models were also re-evaluated excluding those participants with imputed data (Tables S10–S13†), including those participants with hair Hg isotopes corresponding to exposures only during the second trimester (Tables S14–S17†), and excluding 21 maternal hair samples previously analyzed (Tables S18–S21†).15 Similar results were observed as in the main analyses, including associations between hair Δ199Hg and serum omega-3 fatty acids (Tables S13, S17, and S21†), and between hair δ202Hg and hair THg and maternal age (Tables S10–21†). Additionally, the same direction of association was observed between hair Δ199Hg and hair THg and whether one or both parents were farmers (Tables S10–S21†). Lastly, the same direction of association was observed between hair Δ199Hg and rice MeHg intake (Model C) (Tables S12, S16, and S20†), as in the main analyses.
For the subset of 15 mothers with paired hair measurements reflecting second and third trimester exposures, positive correlations were observed for hair δ202Hg and hair Δ199Hg (Spearman's rho: 0.84 and 0.56, respectively) (Fig. S1†). Hair δ202Hg values corresponding to the third trimester were lower (median hair δ202Hg: 0.18‰), compared to hair δ202Hg corresponding to second trimester exposures (median hair δ202Hg: 0.43‰) (paired t-test, p = 0.0003) (Fig. 3 and S1†). Conversely, no differences were observed for Δ199Hg values between the second and third trimester exposure periods (paired t-test, p = 0.11) (Fig. 3 and S1†).
In contrast to previously reported results,10,15 in the adjusted models, hair Δ199Hg was not associated with the number of fish meals ingested weekly, nor with %MeHg intake from rice (Models A and B, respectively). The three estimates used in Models A–C were derived from the mother's self-reported FFQ, and (for two of the estimates) from analyses of MeHg (in rice) and THg (in fish). Several factors potentially contributed to more uncertainty in the FFQ estimate for fish ingestion (which was used in Models A and B), compared to rice ingestion. First, recall bias is often higher for foods ingested seasonally or infrequently.39 In this population, 82% of the mothers ingested rice daily and 42% of the mothers rarely or never ingested fish; therefore, recall bias for fish consumption was likely higher, compared to rice ingestion. Second, MeHg concentrations were measured in rice samples brought from each participant's home, whereas THg concentrations in ocean fish and shellfish were estimated from a literature review (ESI,† detailed methods).26 Lastly, each mother estimated her rice ingestion rate (g per serving) by selecting bowls with known quantities of rice, whereas fish serving sizes were assumed to be the same for all mothers (ESI,† detailed methods). Although all three measures were derived in part from the FFQ, the estimate for MeHg intake from rice potentially had less measurement error compared to the estimate for MeHg intake from fish.
In the adjusted models, hair Δ199Hg was strongly, positively associated with serum omega-3 fatty acids (Model D), explaining 7.4% of the variability in hair Δ199Hg (compared to 1.6% of the variability, explained by daily MeHg intake from rice). Serum omega-3 fatty acids are a biomarker for fish consumption.25 In this cohort, serum omega-3 fatty acids were weakly, positively correlated with fish MeHg intake (Spearman's rho: 0.15), and the number of fish meals ingested weekly (Kruskal–Wallis, p = 0.03) (Tables S22–S23†). As noted above, biomarkers have less measurement error compared to estimates derived from self-report FFQs.39 Thus, it was not surprising that serum omega-3 fatty acids were more predictive of hair Δ199Hg, compared to other measures of dietary MeHg intake for fish and rice, which were derived from the self-report FFQ.
Unlike hair Δ199Hg, hair δ202Hg was not associated with rice or fish ingestion in the adjusted models (Models A–D), which was consistent with previous studies.10–13,15,16 Instead, hair δ202Hg was strongly, negatively associated with hair THg, explaining 27–29% of the variability in hair δ202Hg values (Fig. 2a). The associations between hair δ202Hg and hair THg were not likely attributed to the dietary source of MeHg because all models were adjusted for rice/fish ingestion. Moreover, in a subset of 15 hair samples corresponding to the third trimester, the same negative trends were observed between hair δ202Hg and MeHg (Fig. S2a†).
In prior studies, hair δ202Hg fractionation was thought to mainly reflect MeHg metabolism, in particular, demethylation by gut microbiota.13,19 However, fractionation of stable isotopes also results from metal binding.40 Bonds involving high oxidation states [e.g., iron (Fe3+) and copper (Cu2+)] prefer heavy isotopes, whereas bonds with sulfur (S), including cysteine and methionine, favor light isotopes.41 Note that the terms “light” and “heavy” are analogous to “depleted” and “enriched,” respectively. For example, in mice, zinc (Zn) bonding with phosphates in bones resulted in more enriched δ66Zn, while δ66Zn in the liver was isotopically lighter, reflecting fractionation with cysteine-rich bonds.42 In human subjects, blood δ65Cu and δ34S were isotopically lighter among hepatocellular carcinoma patients, compared to healthy controls, potentially reflecting reallocation of Cu from the liver and binding to cysteine-rich proteins.40
Like other metals, Hg has a high affinity for thiols.43 Lower δ202Hg has been observed in the presence of thiols,44,45 which was consistent with the notion that cysteine (with its thiol side-chain) favors lighter isotopes.41 Using thiol resins, stable Hg isotopes were analyzed in Hg-thiol complexes and in dissolved Hg species, in nitrate and chloride solutions.45 In both solutions, thiol-bound Hg had lower δ202Hg, compared to dissolved Hg species.45 Similarly, in soil the cysteine-bound fraction of Hg had lower δ202Hg, compared to the bulk soil.44 Moreover, cysteine in soil was also associated with higher uptake and translocation of MeHg in rice roots and stems, compared to other thiol-containing compounds, including glutathione and penicillamine.46
Due to the high affinity of MeHg for thiols, transport processes in biological systems typically act on MeHg conjugates with either cysteine or glutathione.1 For example, in the human body, MeHg enters the cells as a complex with the amino acid L-cysteine and exits the cell as a complex with reduced glutathione.1 Similarly, in human hair, MeHg enters the keratinocytes as a MeHg-cysteine complex.47,48 In the present study, because hair δ202Hg decreased as hair THg increased, our results suggested that hair THg (most of which was MeHg)26 was likely bound to cysteine, which favors light isotopes,41,42 and that light isotopes were preferentially transported into the hair follicle. However, further research could help test this hypothesis.
Hair Δ199Hg was also negatively associated with hair THg, explaining 1.8–3.6% of the variability in the adjusted models (Tables 1–4 and Fig. 2b). As noted above, a previous study reported lower δ202Hg for Hg-thiol complexes compared to dissolved Hg species; however, the same was not observed for MIF (Δ199Hg and Δ201Hg).45 In fact, slightly higher MIF was observed in the presence of thiols, compared to dissolved Hg species.45 More research is needed to understand the potential mechanisms contributing to the observed negative association between hair Δ199Hg and hair THg.
For the subset of 15 mothers with paired hair measurements reflecting both second and third trimester exposures, hair δ202Hg was significantly lower during the third trimester compared to the second trimester, while hair Δ199Hg did not differ between trimester exposure periods (Fig. 3 and S1†). Our results suggested no changes in the dietary sources of MeHg (rice versus fish) between the second and third trimesters. However, as noted above, changes in isotopic abundances of hair δ202Hg may reflect metabolism, potentially via the gut microbiome (i.e., demethylation).13,19 This explanation was consistent with prior studies reporting overall changes in the maternal gut microbiome between early/last gestation.49 For example, among 24 pregnant mothers in South Carolina, changes in gut microbiota diversity and associations with MeHg concentrations in biomarkers (after controlling for hemodilution) were observed between early/late gestation.50 Findings from the present study were limited to just 15 mothers, and longitudinal studies within a larger cohort could improve our understanding of associations between MeHg concentrations in biomarkers and gestation.
Although our study has several strengths, there are some limitations to note. Among non-farmers (n = 62), who ingested more fish compared to farmers (Table S3†), we did not observe the same associations between hair Δ199Hg and rice MeHg intake that were observed among the entire cohort (Tables 3 and S8†), potentially because of lower rice consumption or because this sensitivity analysis was underpowered. However, we observed a strong, positive association between hair Δ199Hg and omega-3 fatty acids among the entire cohort (Tables 4 and S9†), suggesting this association was valid regardless of the primary dietary MeHg source (fish or rice). Additionally, we did not analyze whether hair Hg was bound to cysteine, although the MeHg-cysteine complex is the dominant form of MeHg in human hair.47 It is important to measure MeHg complexes in hair to confirm our findings regarding uptake of lighter Hg isotopes in human hair. Our study focused on peripartum mothers; however, more studies are needed in the general population to assess these findings. There were likely other factors, including other components of the diet, that contributed to MeHg exposure and metabolism, which were not measured in this study.
Footnote |
† Electronic supplementary information (ESI) available: Detailed methods, Fig. S1 (second/third trimester comparison), Fig. S2 (Hg isotopes, THg, and MeHg in the third trimester), Table S1 (all studies reporting hair Hg isotopes), Table S2 (quality assurance/quality control), Table S3 (farmer/non-farmer comparison of biomarkers), Table S4 (with/without Hg isotopes comparison of characteristics), Table S5 (with/without Hg isotopes comparison of biomarkers), Tables S6–S21 (four sets of sensitivity analyses), Table S22 (Spearman's correlation for continuous maternal characteristics), Table S23 (Associations for categorical maternal characteristics), legend for Dataset S1 and references. All isotope data and hair Hg data used in this study are available in the Electronic Supporting Information associated with this work (Dataset S1), and other data are available upon reasonable request. See DOI: https://doi.org/10.1039/d4em00231h |
This journal is © The Royal Society of Chemistry 2024 |