The study site, the Navajo Nation, is located in the Four Corners area of the southwestern United States. In 2013 the NBCS began recruiting pregnant women between 14 and 45 years of age who had lived on the Navajo Nation for at least 5 years, were willing to deliver at a participating hospital, and have their child followed up for one year postnatally. At the time of enrollment, a blood and urine sample was collected from the participant. Socioeconomic, demographic, and lifestyle information was also collected in the home shortly after enrollment. A subset of the maternal enrollment urine samples was selected for oxidative stress analysis, which forms the basis of this investigation.

The NBCS was initiated to address Navajo community concerns about how chronic environmental exposure to uranium mine waste affects human health. The research team led by the University of New Mexico (UNM) Health Sciences Center Community Environmental Health Program included partnerships with Navajo Nation Department of Health, Navajo Area Indian Health Service, Southwest Research Information Center and the US Centers for Disease Control and Prevention Agency for Toxic Substances and Disease Registry (CDC/ATSDR) National Center for Environmental Health (NCEH). Trained Indian Health Service staff recruited pregnant women during prenatal visits at one of six participating Indian Health Service or Public Law 638 hospitals on the Navajo Nation. The enrollment protocol prioritized recruitment of women during their 1^st trimester but an open enrollment process was used, allowing women to enroll at any time during their pregnancy. For the present study, urine samples from NBCS participants were stratified by serum zinc concentration above and below the WHO level of sufficiency of 70 μg/dL. Additionally, we limited the sample population to individuals who also had their enrollment urine sample analyzed for uranium, total arsenic, arsenous (III) acid (arsenite, AsIII), and dimethylarsinic acid (DMA). Then we randomly selected 66 urine samples from each zinc group for inclusion in the present study out of 204 women enrolled in the NBCS at that time. Written informed consent was obtained from all study participants and the study protocol approved by the University of New Mexico Institutional Review Board (HRPO 11-310) and the Navajo Nation Human Research Review Board (NNR 11.323).

Trained hospital staff collected biospecimen samples using pre-screened metal-free collection cups, transfer pipettes, and Nalgene cryo-vials provided by CDC NCEH Division of Laboratory Sciences (DLS). During a prenatal hospital appointment trained hospital staff collected spot urine samples in sterile 50 mL urine collection cup. After collection, a laboratory staff member transferred a 1.8 mL aliquot of urine to separate 2.0 mL Nalgene cryo-vials for multi-element metals, total arsenic, speciated arsenic, and creatinine analysis. Hospital laboratory staff also collected peripheral blood via venipuncture and then allowed the blood to clot at room temperature for 30 to 40 minutes. Once clotted laboratory staff centrifuged the blood tube at 2,400 revolutions per minute for 15 minutes to separate the serum and then transferred 1.8 mL aliquot of serum to a 2.0 mL Nalgene cryo-vial. After processing urine and serum samples, hospital staff placed all cryo-vials in a −80°C freezer for storage and transferred on dry-ice to freezer storage facilities at UNM. Chain of Custody forms were completed, reviewed, and validated at each stage of collection, storage and analysis.

UNM staff shipped samples on dry ice to CDC DLS for analysis. NCEH laboratory staff prepared urine samples for uranium and other metals analyses using NCEH Method 3018.3; samples analyzed for total arsenic were prepared using Method 3018A.2 [44]; and samples analyzed for speciated arsenic were prepared using Method 3000.11 [45]. Chemical concentrations in urine and serum were measured using Inductively Coupled Plasma – Dynamic Reaction Cell – Mass Spectrometry (ICP-DRC-MS) [46]. Arsenic species concentrations were determined in separate aliquots using High Performance Liquid Chromatography (HPLC) and an anion exchange column to separate species prior to ICP-DRC-MS [47]. Urinary creatinine was determined using Roche/Hitachi Modular P Chemistry Analyzer [48].

Analysis of urinary F2-isoprostanes and prostaglandin F2α in urine was performed by the Linus Pauling Institute Oxidative and Nitrative Stress Core Laboratory by HPLC-tandem mass spectrometry as described previously [49]. Creatinine was quantified in each sample by the same laboratory. Briefly, urine was thawed at room temperature, mixed by inversion, and centrifuged (200x g, 5 min). Aliquots of the supernatant were mixed with methanol and internal standards followed by addition of 0.02 M bis-tris-HCl. Samples were pH adjusted to 6.0 −/+ 0.05. Strata X-AW cartridges (100 mg/3 mL, Phenomenex, Torrance, CA) were each pre-conditioned with methanol, then water. Diluted urine samples were loaded, the cartridges rinsed with sucsessively with methanol/water, acetonitrile and cartridges dried under a gentle vacuum (~5 mm Hg, 30 sec). Cartridges were then eluted 3-times with 1 mL methanol, and the eluants from each cartridge pooled and collected into glass tubes. Samples were dried under nitrogen gas, reconstituted in 200 μL methanol containing 0.1% formic acid (v:v), and injected onto the LC-MS-MS. Analytes were detected and quantified using SRM: F2-isoprostanes/prostaglandin F2α, m/z 353 to 193; 8-iso-PGF_2α-d₄ internal standard, m/z 357 to 197. Samples were analyzed by HPLC-tandem mass spectrometry against standard curves for authentic standards [49].

Summary statistics including median (interquartile range) for continuous variables and frequency (%) for categorical variables were used to describe the demographics, environmental characteristics, chemical exposures of urinary uranium, urinary total arsenic, DMA, AsIII, and serum zinc of the participants in the NBCS, overall and by zinc groups (> 70 vs < 70 μg/dL for high vs low categorical membership).

The urine chemical measurements were corrected for urine creatinine, and values below limit of detection (LOD) were replaced by the LOD value divided by 2. Wilcoxon rank-sum tests were used to compare continuous variables between zinc groups. Chi-squared tests and Fisher Exact tests for small expected numbers were performed to compare categorical variables between zinc groups. Pearson correlation coefficients along with the corresponding 95% confidence intervals (CIs) and scatterplots were used to summarize the correlation between log-transformed chemicals.

We compared the chemical exposures in the NBCS study to the national levels measured in women surveyed in the NHANES. The chemical exposure data were extracted from the NHANES year 2011-2012 database for women to represent the national population. Summary statistics of NHANES were calculated following the analysis guideline to account for the complex design features of NHANES including stratification, cluster sampling, and weighting [50].

Geometric mean along with 95% CIs were summarized for the chemical exposures in NBCS and the NHANES study. Statistical significance was determined using two-sample Welch's t-test comparing the mean (standard error) of the log transformed analytes, which took into account the NHANES design features.

The oxidative stress biomarkers as the primary outcomes include 8-iso-PGF_2α, PGF_2α, and the ratio of 8-iso-PGF_2α to PGF_2α to distinguish enzymatic versus chemical lipid peroxidation [41, 42]. Descriptive statistics were summarized for those variables by trimester at enrollment (1^st, 2^nd, 3^rd). The measurements were corrected for urine creatinine and variables with skewed distributions were log transformed.

Univariable linear regression analyses were used to examine the association between each demographic variable and chemical exposure and the oxidative stress outcome (prostaglandin ratio or 8-iso-PGF_2α). Multiple linear regression models for the oxidative stress biomarkers were used to evaluate the impact of chemical exposures while adjusting for other potential confounding covariates. Covariates evaluated in the analysis included: age at interview, pre-pregnancy body mass index (BMI), trimester at enrollment, educational attainment (above or below high school), household income (above or below $20,000), employment status (currently employed or not), alcohol in the past year (yes or no), tobacco smoking (never, current or former smoker), vitamin intake (yes or no), and use of wood or coal for home heating (yes or no). Linear models with a backwards variable selection method based on the Akaike information criterion (AIC) measure were used for assessing the effects of variables along with their interactions on the oxidative stress biomarker variables. We also performed multiple linear regression stratified by zinc groups to describe the different effects of urinary total arsenic on 8-iso-PGF_2α moderated by high vs low zinc.

To assess the potential for zinc to mediate the effect of each metal (uranium, total arsenic, AsIII and DMA) on oxidative stress biomarkers (prostaglandin ratio and 8-iso- PGF_2α), we used a quasi-Bayesian Monte Carlo causal mediation analysis [51, 52]. Covariates including trimester, education, household income, and tobacco were evaluated and adjusted for in the mediation analyses. We also then stratified this analysis by trimester to assess any confounding effects of changes in zinc across trimesters.

No adjustments for the multiple comparison tests were considered due to the discovery nature of this study. All analyses were conducted using the SAS 9.4 and R 3.4.1.

Selected demographic characteristics for the subjects in this study are shown in Table 1. Study participants were drawn from the NBCS and all 132 participants included in this study were pregnant women between the ages of 16 to 42 with a median age of 26.8 years old. Analysis of survey information indicated that 38.6% of women had education beyond high school, 57.6% were unemployed and 43.2% had an annual household income below $20,000 at the time of study enrollment. The majority of women (59.8%) were taking vitamin supplements and were overweight or obese based on pre-pregnancy BMI (52.2%). Current cigarette smoking is negligible among the NBCS pregnant women (<1%) which is consistent with low tobacco usage overall in the Navajo population as reported previously by Redwood et al. [53]. However, other exposures such as wood or coal heating or use of ceremonial tobacco were noted as potential contributors to oxidative stress. The survey questions used to generate Table 1 are provided in Supplemental Table 1.

Urinary total arsenic, AsIII, DMA, uranium and serum zinc, were measured by ICP-DRC-MS as described in Materials and Methods (Table 2). Values for the same metals in women surveyed for the NHANES [54] are shown in Table 2. Of the 132 individuals in the sample population, urinary metals were detected among 132 (100%) for urinary total arsenic, 105 (79.5%) for urinary DMA, 88 (66.7%) for urinary AsIII, and 128 for urinary uranium (97.0%) (Table 2). Serum zinc was detected in 132 (100%) individuals from the sample population (Table 2). Total arsenic concentrations had a weak to moderate correlation with AsIII levels within the population (Supplemental Figure 1A), however, as expected, total arsenic correlated strongly with the metabolite DMA (Supplemental Figure 1B).

The median concentration for serum zinc was 67 μg/dL and median concentrations for urinary total arsenic, DMA, AsIII, and uranium were 5.5 μg/g, 4.3 μg/g, 0.41 μg/g, and 0.016 μg/g creatinine respectively. To provide context for the biomonitoring values, information for women was extracted from the NHANES 2011-12. Compared to the median for women included in the NHANES 2011-12 survey, urinary uranium was 2.3 times greater for NBCS samples (Table 2) with 88.6% and 15.9% of the sample population exceeding the NHANES 50^th and 95^th percentiles, respectively (Supplemental Table 2). Urine uranium concentrations for NBCS participants were greater than observed values from the NHANES study (p-value <0.001).

Urinary concentrations of total arsenic were lower for NBCS participants compared to NHANES values (p-value <0.001) (supplemental Table 2). Only 40.2% and 0.8% of NBCS participants had total arsenic urine concentrations exceeding the NHANES 50^th and 95^th percentiles, respectively. Serum concentrations of zinc were also lower in NBCS participants when compared to NHANES (p-value<0.001) (supplemental Table 2). No statistically significant difference was observed for DMA. Concentrations of urinary AsIII and serum zinc decreased from the 1^st to 3^rd trimester while concentrations of urinary uranium, total arsenic, and DMA remained stable (Table 3).

We observed no significant difference in urinary uranium or arsenic concentrations between the low versus high zinc groups suggesting that zinc status did not influence urinary metal levels (Table 2). The median concentration for the high zinc group was 78 μg/dL, which is similar to the NHANES values of 80 μg/dL and above the WHO sufficiency standard of 70 μg/dL. The median for the low zinc group was 48 μg/dL and well below the NHANES values and zinc sufficiency standard. The high and low zinc groups did not differ in the demographic characteristics with the exception of smoker classification. There were more Never Smokers and fewer Former Smokers in the low zinc group. Additionally, the high zinc group included more women who enrolled in the NBCS during the 1^st trimester and low zinc group had more women enroll during the 3^rd trimester (Table 1).

Table 4 provides the summary statistics for oxidative stress biomarkers as measured by HPLC-Tandem Mass Spectrometry. The biomarker 8-iso-PGF_2α is a well-regarded marker for detection of chemical lipid peroxidation in human studies but does not take into account enzymatic contributions to total 8-iso-PGF_2α. The use of a prostaglandin ratio (8-iso-PGF_2α to PGF_2α) has been described to overcome this issue [41, 42]. Of the oxidative stress biomarkers measured in this study, PGF_2α and the prostaglandin ratio, but not 8-iso-PGF_2α, were significantly different based on pregnancy trimester (Table 4). When only zinc is considered, comparison between the low and high zinc status subsets revealed no difference in the oxidative stress biomarkers 8-iso-PGF_2α or the prostaglandin ratio (Figure 1). The urinary 8-iso-PGF_2α levels measured in this study are comparable to previous reports in pregnant populations [55, 56]. Based on findings that the isoprostane metabolites dinorF1 and F2 values were subject to variability between different batch analyses (data not shown), only 8-iso-PGF_2α, PGF_2α and the prostaglandin ratio were included for the regression models.

Ceremonial tobacco use was associated significantly with higher levels 8-iso-PGF_2α and prostaglandin ratio oxidative stress markers (Supplemental Tables 3 and 4). In addition, significantly lower 8-iso-PGF_2α levels were observed in populations with a household income above $20,000 and education above high school (Supplemental Table 3). The prostaglandin ratio was significantly affected by trimester stage (Supplemental Table 4).

In a univariable analysis of the linear regression model there were no significant associations with environmental metals and 8-iso-PGF_2α (Table 5). Of the potential confounders, low income and ceremonial tobacco use had a significant association with urinary 8-iso-PGF_2α in the univariable analysis (Table 5). A multivariable analysis revealed significant main effects and interaction effects between total arsenic and low zinc on the 8-iso-PGF_2α outcome (Table 5).

The findings differed when using the prostaglandin ratio as the oxidative stress biomarker. In the univariable analysis there was a positive association with the arsenic metabolite DMA, but not other metals (Table 6) and a marginal positive association between total arsenic and the prostaglandin ratio was observed in the multivariable analysis. Notably, different sets of confounding variables were identified for each of the oxidative stress biomarkers. For example, pregnancy trimester was identified as a confounding variable for the prostaglandin ratio but not for 8-iso-PGF_2α alone (Table 6). This can be accounted for by the lack of significant difference of 8-iso-PGF_2α by trimester whereas PGF_2α was significantly increased as pregnancy progressed from 1^st to 3^rd trimester (Table 4).

In order to better understand the arsenic and zinc interaction term from Table 5, we ran the stratified multivariable analysis by serum zinc group. We observed that the association between total arsenic and oxidative stress was modified by zinc group. Specifically, arsenic was positively associated with increased 8-iso-PGF_2α for the high serum zinc group, but not for the low serum zinc group (Table 7). In terms of the mediation effect of zinc on the association or metal/metalloids with oxidative stress, the causal mediation analysis (Materials and Methods) yielded no significant results of zinc mediation.