Metal-Mixtures in Toenails of Children Living Near an Active Industrial Facility in Los Angeles County, California

Abstract

Background

Children residing in communities near metalworking industries are vulnerable to multiple toxic metal exposures. Understanding biomarkers of exposure to multiple toxic metals is important to characterize cumulative burden and to distinguish potential exposure sources in such environmental justice neighborhoods impacted by industrial operations. Exposure to metal mixtures has not been well-characterized among children residing in the United States, and is understudied in communities of color.

Methods

In this study we used toenail clippings, a non-invasive biomarker, to assess exposure to arsenic (As), cadmium (Cd), mercury (Hg), manganese (Mn), lead (Pb), antimony (Sb), selenium (Se), and vanadium (V). We used non-negative matrix factorization (NMF) to identify “source” signatures and patterns of exposure among predominantly working class Latinx children residing near an industrial corridor in Southeast Los Angeles County. Additionally, we investigated the association between participant demographic, spatial and dietary characteristics with identified metal signatures.

Results

Through NMF, we identified three groupings (source factors) for the metal concentrations in children’s toenails. A grouping composed of Sb, Pb, As, and Cd, was identified as a potential industrial source factor, reflective of known airborne elemental emissions in the industrial corridor. We further identified a manganese source factor primarily composed of Mn, and a potential dietary source factor driven by Se and Hg. We observed differences in the industrial source factor by age of participants, while the dietary source factor varied by neighborhood.

Conclusion

Utilizing an unsupervised dimension reduction technique (NMF), we identified a “source signature” of contamination in toenail samples from children living near metalworking industry. Investigating patterns and sources of exposures in cumulatively burdened communities is necessary to identify appropriate public health interventions.

Introduction

Industrial corridors emitting multiple toxic contaminants persist throughout the United States (US). Communities situated along industrial metalworking corridors, home to metal plating facilities, metal-finishing, and metal recyclers, may be disproportionately burdened by the pollution associated with these industrial activities [1]. Individuals are seldom exposed to one environmental contaminant at a time, particularly when considering industrial sources of exposure. Children are especially vulnerable to the effects of these contaminants, as their bodies are still developing [2,3].

Metalworking facilities release a multitude of contaminants into the environment, including lead (Pb), arsenic (As), cadmium (Cd), manganese (Mn), antimony (Sb), and hexavalent chromium (Cr6) [4–6]. Living in close proximity to industrial facilities, such as secondary lead smelters, incinerators, and other industrial plants, has been associated with higher levels of toxic metals, such as Mn, Pb, and Vanadium (V), in the blood, toenails and hair of children [7–14]. Soil has been identified as a primary exposure pathway for Pb and Mn, as emissions from the facilities settle into the soil and dust of the surrounding area [7,15]. Re-suspension of contaminated soil and hand-to mouth activities result in children being more highly exposed to these pollutants [16].

Exposure to metals and metallic elements, including Pb, As, Cd, and Mn, has been linked to a number of adverse health effects [17]. In children, early life exposures to individual metals, such as As, Pb and Mn, have been associated with neurocognitive deficits, immune dysfunction, respiratory, and cardiovascular effects [18–25]. However, fenceline communities are seldom exposed to one environmental contaminant at a time. Increasingly, evidence shows that multipollutant exposures may reflect synergistic or antagonistic adverse effects or in combination may exert effects at levels that individually do not produce observable impacts [26–29]. In recent years, several studies have reported that early life exposures to metal mixtures may lead to more profound neurodevelopmental and intellectual deficits, and reduced head circumference in newborns [30–35]. While there are many studies of children’s exposure to multiple metals, the health risks due to exposures to metal mixtures have not been well characterized, particularly in highly exposed U.S. urban communities of color, vulnerable populations that may be more susceptible to the adverse impacts of toxic metals. With the continued reliance on industrial facilities across the U.S. it is important to characterize exposures and assess sources in cumulatively burdened communities near industrial corridors.

While nationally toxic emissions have been declining, these reductions have been less evident in low-income and communities of color, who are more likely to be exposed to multiple sources of pollution and social stressors [36,37]. Facing environmental hazards, community organizations and the environmental justice movement have turned to community-driven biomonitoring research to understand environmental hazards [38]. As contamination has become ubiquitous and disproportionally impacts working class communities of color, it can be difficult to disentangle the role of exposure to local sources of contamination and health effects. The lack of data on toxic emissions, unequal enforcement of environmental regulations and impact to communities has prompted organizations to request tools for research and biomonitoring of exposures that can be leveraged as a tool for community empowerment [39].

The predominantly low-income Latinx community surrounding the City of Industry located in the San Gabriel Valley of Los Angeles County CA ranks among the top communities most burdened by toxic pollution, according to CalEnviroScreen 3.0 [40]. From 2017–2018 there were 4,825,578 lbs. of chemicals released from 21 industrial facilities dedicated to chemical storage, fabrication of metal goods, and recycling of metals situated in this residential neighborhood [41].

Throughout the County of Los Angeles, little to no buffer exists between homes, schools and recreational areas and industrial sites. Metal operations have existed in this community for nearly 100 years, with residential homes situated among metal plating facilities, metal-finishing, metal recyclers, metal manufacturing, and the associated diesel truck traffic [42] (Figure S1). For decades community-based environmental justice organizations in Los Angeles County have raised concerns over the health impacts of this highly concentrated and underregulated industry [43,44]. Utilizing a community-academic collaboration, this project aimed to investigate children’s exposure to metal mixtures possibly associated with metal processing facilities located in an industrialized corridor of the San Gabriel Valley, including the largest (and only) lead-acid battery recycling facility in the western U.S. The recycling facility has received multiple violations for failure to minimize dust emissions and has emitted Pb and As exceeding air quality rules set by the South Coast Air Quality Management District [45]. In this study, we analyzed children’s toenails, a non-invasive biomarker of multiple metals exposures, for a panel of elements and innovatively used non-negative matrix factorization (NMF) to understand sources of metal exposures as well as demographic and spatial predictors of metal mixtures in children living near industrial sources.

Methods

Study Area and Recruitment

Children residing in three neighborhoods (Avocado Heights, Hacienda Heights, or La Puente, in Los Angeles County CA) adjacent to the City of Industry in the southeastern part of Los Angeles County CA were recruited for participation in the study (Figure 1). These locations were identified for biomonitoring by our community partners. Children were eligible for the study if they had lived in one of the three neighborhoods, for at least the past 2 years, were between the ages of 4 to 17 years of age and spoke either English or Spanish. Recruitment was conducted by speaking at neighborhood meetings, outreach events, church gatherings and events hosted by our community partners. This pilot study was designed to develop and test the feasibility of measuring metals exposure biomarkers from community participants. In total a convenience sample of 95 participants were enrolled in the study. Study staff obtained written informed consent from parents and verbal assent from all children over the age of 7. All protocols and study materials were approved by the University of Southern California’s Institutional Review Board.

Figure 1. Map of study area.

Industrial Source Factor loadings by study participants.

Questionnaire and household visit

Upon enrollment, participants received a packet containing 1) a child toenail collection kit, 2) a health and lifestyle questionnaire, and 3) participated in height and weight measurements. All participants recruited between July 2017 and July 2018 who provided toenail specimens were included in the current analysis (N=95). A bilingual field coordinator conducted visits at individual households or community centers. At each visit parents of participants were asked to fill out a questionnaire, including sociodemographic information, residential history, child health history, and exposure and lifestyle variables (e.g. smoking habits, tap water intake, fish and rice intake). Standing height was measured three times using a Seca portable stadiometer (Perspectives Enterprises Model PE-AIM-101). Weight was measured three times to the nearest 0.1 kg using a Tanita instrument (HD-662). Height and weight measures were used to calculate body mass index (BMI; kg/m2) according to the Centers for Disease Control growth charts for children guidelines [46].

Toenail metal biomarker assessment

Toenail clippings are a useful biomarker of exposure to multiple metals, as they can be collected non-invasively and their growth rate can reflect exposures that occurred 6 to 12 months prior to collection [47–49]. Toenail collection kits instructed the participant (or participant’s parent in case of younger children) to cut and collect nails from all 10 toenails after bathing or showering and after removal of any nail polish. Nails were collected in small envelopes labeled with date/time of collection and a unique study ID. Participants were given pre-paid mailing materials to return the samples to the study office or returned to the study coordinator at one of the community collection events. Toenail samples were analyzed by the Dartmouth Trace Element Core Facility for a panel of metals, including As, Cd, mercury (Hg), Mn, Pb, Sb, selenium (Se), and V. Nails were cleaned and washed 5x in an ultrasonic bath using Triton X-100 and acetone. The samples were sonicated with deionized water, freeze-dried, weighed, and then digested in Optima HNO₃ by low-pressure microwave digestion [50]. Toenail samples were analyzed for each element μg total per gram (μg/g) by inductively coupled plasma-mass spectrometry (ICP-MS) on an Agilent 7700X (Agilent Technologies Headquarters, Santa Clara, California) [50]. In each analysis, quality control measures, included a duplicate analysis of digested toenail samples, spikes of digested samples, along with blank and fortified blank digests. Recovery criteria for all analytes ranged from 80%−120% of the spike amount. While there is no available toenail Certified Reference Material, Dartmouth Trace Element Core Facility participates in a proficiency-testing program (QEMQAS, Center for Toxicology, Quebec, Canada) [35,51]. Additionally, the Dartmouth Trace Element Core Facility follows quality control procedures outlined in EPA SW 846 and EPA method 6020 (Test Methods for Evaluating Solid Waste, Physical/Chemical Methods, EPA publication SW‐846, Third Edition, Final Updates I (1993), II (1995), IIA (1994), IIB (1995), III (1997), IIIA (1999), IIIB (2005), IV (2008), and V (2015)). The limits of detection were 0.01, 0.01, 0.01, 0.05, 0.01, 0.02, 0.02, 0.01 μg/g, for As, Cd, Hg, Mn, Pb, Sb, Se, and V, respectively. Toenail levels of As, Mn, Pb, and Se were detected in every participant. Toenail levels of Cd, Hg, Sb, and V were detected in 98.9%, 86.2%, 97.7%, and 81.6% of participants, respectively. Samples with concentration levels reported as non-detectable were imputed using the limit of detection divided by the square root of two [52].

Spatial measurements

Participants’ residences were geocoded using the R package ggmap. Distance to the battery recycling facility and highways (defined by categories A1–A3 of the Census Feature Class Codes (also called FCC)), were calculated for each home (ArcGIS 10.7, Environmental Systems Research Institute Inc., Redlands, CA). The primary wind direction in relation to the recycling facility was calculated to identify the upwind and downwind direction using climate data from the National Oceanic and Atmospheric Administration (NOAA) from 2012 to 2018 (Table S1, Figure S2) [53].

Statistical methods

We used non-negative matrix factorization (NMF), an unsupervised dimension reduction technique to capture the variability of metal concentrations in the toenail samples. NMF has been previously used for contaminant mixtures to identify patterns in the mixture of pesticides originating from the diet [54,55]. It has also been utilized to estimate lead exposure estimates from multiple biomarker matrices (i.e., blood, urine, nails) to better represent the overall body burden [56].

In this study the relationship between the individual observations of each metal concentration in each toenail B from all subjects M can be represented by

$$\text{X}\in {ℝ}^{M\times B.}$$ (1)

NMF will approximate X, such that

$$\text{X}\approx \text{WH}$$ (2)

where W and H are of lower matrix rank than X. These factor matrices are estimated to minimize ||X − WH||_F, where ‖·‖_F is the Frobenius norm. The dimensions of the new matrices after factorization (i.e. number of underlying factors), are set by the analyst. We used R Software package NMF for analysis of the toenail data (R Studio Version 1.2.5019, R Version 3.6.1) [57].

A commonly employed method to reduce dimensionality in datasets is Principal Component Analysis (PCA) [58,59]. Similar to NMF, PCA can be used to identify patterns and groupings of contaminants in biomarkers and has been extensively applied in environmental health studies [60–62]. While the intent of this paper is not to compare algorithm performance, we do present PCA results to demonstrate the reliability of the NMF approach and the interpretability of the “source factors” identified. In this study we conducted PCA using the R software function prcomp (R Studio Version 1.2.5019, R Version 3.6.1). Consistent with previous studies we will retain all principal components with eigenvalues ≥1.0 and will consider factor loadings >0.4 to suggest high loadings [60,61].

We summarized descriptive statistics for participant characteristics (i.e., age, ethnicity, mother’s education, father’s occupation, smoking at home, BMI and sex), spatial characteristics (i.e., neighborhood, distance, wind direction, and proximity to highway), and toenail biomarkers (i.e., source factors from NMF analysis). We investigated potential differences in source factors using a Mann-Whitney-Wilcoxon Test for child’s sex (female vs male), child’s ethnicity (Hispanic vs Non-Hispanic), mother’s education (high school vs some college+), father’s occupation (industrial vs other), presence of indoor environmental tobacco smoke (yes vs no), consumption of tap water, bottled water, fish, rice or rice products (non/low consumption vs med/high consumption), as well as wind direction (upwind vs downwind from the battery recycling facility), and proximity to highway (under 500 m vs greater than 500 m). For father’s occupation, industrial was defined as an occupation where metal exposure could occur and be introduced to the home via a take-home pathway. No mothers reported being employed in an occupation with potential metals exposure. These included employment at an industrial facility, construction work, autobody/automobile repair industry or pest control. For consumption categories, none/low consumption was defined as consumption of the item once weekly or less; med/high consumption was defined as consumption of the item more than once a week. Differences in source factors for age categories consistent with the U.S. EPA’s Child-Specific Exposure Factors Handbook were investigated using a Kruskal Wallis test (4 to <6, 6 to <11, and 11+) [63]. Differences in sources for BMI categories (normal, overweight, and obese), neighborhood surrounding the industrial city (Avocado Heights, Hacienda Heights, or La Puente, in Los Angeles County CA), and proximity to the industrial corridor (less than 1700 m, between 1700 and 2750 m, and more than 2750 m) were also investigated using a Kruskal Wallis test. Spearman correlations were assessed between source factors and individual toenail concentrations. In a sensitivity analysis, we excluded participants with a presence of indoor environmental tobacco smoke (n=11), and source factors were compared after including versus excluding these participants.

Results

In this study we investigated metal exposures in toenail samples collected from 95 children living near an industrial metal corridor in an urban environment (Figure 1). The average child’s age in this study was 9.8 years (range: 4 to 17). There were 52 females and 43 males. The majority of children identified as Latinx (N= 84, 88%). A majority of participants (N= 51, 53.7%) resided in the Avocado Heights neighborhood, an unincorporated community located to the northwest of the industrial area. The median residential distance from the facility was 2433 meters (range: 374 to 4827). Additional participant demographic and neighborhood characteristics are presented in Table 1. Descriptive statistics of metal concentrations in toenails are presented in Table 2. There were strong positive correlations between toenail As concentrations and Cd, Mn, Pb, Sb, and V (Table 3). We also found a positive correlation between Hg and Se (σ = 0.35, p-value <0.01).

Table 1.

Participant and spatial characteristics (N=95)

Variable	Mean ± SD or N (%)
Age, years	9.8 ± 3.5
Median (range)	10 (4–17)
BMI, kg/m2	19.9 ± 4.4
Median (range)	18.6 (13.6–32.9)
BMI category
Normal (5th to 85th percentile)	58 (61.1)
Overweight (85th to < 95th percentile)	16 (16.8)
Obese (≥95th percentile)	21 (22.1)
Ethnicity
Non-Hispanic	11 (11.6)
Hispanic	84 (88.4)
Mother’s Education
High School	26 (27.4)
Some College +	69 (72.6)
Sex
Male	43 (45.3)
Female	52 (54.7)
Father’s Occupation
Industry	19 (20)
Other	76 (80)
Smoking home
No	85 (89.5)
Yes	10 (10.5)
Tap water
None/Low consumption	78 (82.1)
Med/High consumption	15 (15.8)
Bottled water
None/Low consumption	20 (21.1)
Med/High consumption	73 (76.8)
Fish
None/Low consumption	38 (40.0)
Med/High consumption	55 (57.9)
Rice
None/Low consumption	48 (50.5)
Med/High consumption	45 (47.4)
Foods with rice
None/Low consumption	40 (42.1)
Med/High consumption	51 (53.7)
Neighborhood
Avocado Heights	51 (53.7)
Hacienda Heights	29 (30.5)
La Puente	15 (15.8)
Wind Direction
Upwind	29 (30.5)
Downwind	66 (69.5)
Distance Tertial
1 (<1700 m)	32 (33.7)
2 (1700 < 2750 m)	32 (33.7)
3 (2750 <4827 m)	31 (32.6)
Highway <500 m
Yes	39 (41.1)
No	56 (58.9)

Consumption of tap water, bottled water, fish, rice and rice products does not add up to 100% as there are missing responses by participants

None/Low consumption = consumed once weekly or less

Med/High consumption = consumed more than once weekly

Table 2.

Metal concentrations in toenails (μg/g)

			Percentile
	%detect	Mean ± SD	Min	5th	25th	50th	75th	95th	Max
As	100	0.23 ± 0.20	0.01	0.04	0.10	0.16	0.29	0.60	1.43
Cd	98.9	0.05 ± 0.08	ND	0.01	0.01	0.03	0.06	0.11	0.63
Hg	86.2	0.11 ± 0.15	ND	ND	0.03	0.05	0.11	0.37	0.86
Mn	100	1.72 ± 1.65	0.16	0.28	0.70	1.20	2.08	5.70	8.77
Pb	100	0.84 ± 0.89	0.07	0.11	0.26	0.54	1.10	2.38	4.86
Sb	97.7	0.18 ± 0.18	ND	0.03	0.08	0.13	0.23	0.43	1.14
Se	100	0.94 ± 0.52	0.45	0.58	0.76	0.88	1.00	1.18	5.49
V	81.6	0.15 ± 0.13	ND	ND	0.06	0.12	0.19	0.38	0.66

ND = not detected

Table 3.

Spearman correlations between source factors and individual metals in toenails

	Mn Source Factor	Dietary Source Factor	Industry Source Factor	As	Cd	Hg	Mn	Pb	Sb	Se
Dietary Source Factor	-0.40 **
Industry Source Factor	0.66 **	-0.29 *
As	0.70 ***	-0.21 *	0.68 ***
Cd	0.76	−0.19	0.55 ***	0.57 ***
Hg	0.08	0.32 **	0.05	0.01	0.14
Mn	1.00 ***	-0.35 **	0.55 **	0.70 ***	0.76 ***	0.08
Pb	0.82 ***	-0.25 *	0.73 ***	0.62 ***	0.77 ***	0.14	0.82 ***
Sb	0.64 ***	−0.11	0.83 ***	0.60 ***	0.60 ***	0.10	0.64 ***	0.71 ***
Se	0.18	0.75 ***	0.15	0.20	0.24	0.35 **	0.18	0.19	0.23
V	0.83 ***	-0.26 *	0.44 **	0.49 ***	0.71 ***	0.21	0.83 ***	0.77 ***	0.51 ***	0.19

^*<0.05,

^**<0.01,

^***<0.001

Three groupings or “source factors” were selected to represent the pattern of metals observed in the children’s toenail samples. This selection was made after comparing two to six “source factor” results using the R NMF Package factor sparseness and residual sum of squares values (Figures S3:S4) [57]. This comparison suggested that two factors did not improve the data set variance, and more than three factors resulted in splitting of existing factors into new factors. Based on prior studies on local industrial emissions, we identified a source factor comprised of As, Sb, Cd and Pb, which may be indicative of an “industrial source factor” [4,64]. The two additional factors were identified as a “manganese source factor”, which was primarily composed of Mn, and a “dietary source factor”, primarily composed of Se and Hg. Figure 2 includes our interpretation of the source factor origins based on comparisons to identified sources from previously published studies as well as emissions information obtained from the surrounding facilities. Additionally, there was a strong correlation (σ = 0.66, p-value <0.001) between the manganese source factor and the industrial source factor. We observed a moderate negative correlation (σ = −0.44, p-value <0.01) between the dietary source factor and the manganese source factor and weaker correlation (σ = −0.29, p-value <0.05) between the dietary source factor and the industrial source factor (Table 3).

Figure 2. NMF analysis results.

Interpretation of Source Factors identified utilizing NMF.

We identified a statistically significant difference in the industrial source factor by child age, such that younger children had higher factor loading than older children (Table 4). While not statistically significant, younger children were observed to have higher manganese factor loading scores than older children. We also observed a statistically significant difference for the dietary source factor by neighborhood, with the northeast community having the lowest dietary source factor loadings. The industrial and manganese source factor loadings were higher for participants residing downwind of the battery recycling facility, but not significantly different from participants living upwind (Table 4). There were no significant differences in any of the source factors by ethnicity, BMI, mothers’ education, gender, father’s occupation, presence of indoor environmental tobacco smoke, consumption of tap water, bottled water, fish, rice or rice products, proximity to industrial sites, or proximity to highway (Table 4).

Table 4.

Differences in participant and spatial characteristics by factors (N=95)

	Manganese Source Factor	p-value	Dietary Source Factor	p-value	Industrial Source Factor	p-value
Variable	Mean ± SD		Mean ± SD		Mean ± SD
Age‡		0.10		0.98		0.0001
4 to <6	0.40 ± 0.37		0.25 ± 0.05		0.18 ± 0.19
6 to <11	0.38 ± 0.22		0.26 ± 0.07		0.14 ± 0.09
11+	0.32 ± 0.36		0.29 ± 0.28		0.07 ± 0.11
BMI		0.79		0.72		0.54
Normal	0.36 ± 0.34		0.29 ± 0.23		0.12 ±0.14
Overweight	0.33 ± 0.18		0.26 ± 0.05		0.12 ± 0.14
Obese	0.42 ± 0.48		0.24 ± 0.08		0.18 ± 0.22
Ethnicity		0.44		0.99		0.35
Hispanic	0.37 ± 0.36		0.27 ± 0.19		0.14 ± 0.17
Non-Hispanic	0.31 ± 0.32		0.26 ± 0.05		0.12 ± 0.08
Mother’s Education		0.20		0.84		0.50
High School	0.29 ± 0.24		0.26 ± 0.07		0.13 ± 0.18
Some College +	0.40 ± 0.39		0.27 ± 0.21		0.14 ± 0.16
Sex		0.22		0.66		0.17
Male	0.40 ± 0.37		0.29 ± 0.26		0.16 ± 0.19
Female	0.34 ± 0.34		0.25 ± 0.07		0.11 ± 0.14
Father’s occupation		0.18		0.97		0.67
Other	0.34 ± 0.20		0.27 ± 0.20		0.13 ± 0.16
Industry	0.47 ± 0.43		0.25 ± 0.09		0.14 ± 0.17
Smoking home		0.18		0.94		0.13
No	0.38 ± 0.37		0.27 ± 0.19		0.14 ± 0.17
Yes	0.22 ± 0.12		0.25 ± 0.05		0.06 ± 0.05
Tap water		0.24		0.54		0.69
None/Low consumption	0.38 ± 0.37		0.27 ± 0.20		0.14 ± 0.18
Med/High consumption	0.30 ± 0.32		0.29 ± 0.06		0.12 ± 0.09
Bottled water		0.37		0.28		0.72
None/Low consumption	0.26 ± 0.15		0.28 ± 0.05		0.15 ± 0.16
Med/High consumption	0.39 ± 0.39		0.27 ± 0.21		0.13 ± 0.17
Fish		0.85		0.17		0.10
None/Low consumption	0.35 ± 0.34		0.30 ± 0.27		0.18 ± 0.22
Med/High consumption	0.38 ± 0.37		0.25 ± 0.08		0.10 ± 0.11
Rice		0.89		0.74		0.53
None/Low consumption	0.39 ± 0.42		0.28 ± 0.24		0.12 ± 0.13
Med/High consumption	0.34 ± 0.28		0.26 ± 0.09		0.15 ± 0.19
Foods with rice		0.70		0.42		0.42
None/Low consumption	0.41 ± 0.44		0.28 ± 0.26		0.10 ± 0.11
Med/High consumption	0.34 ± 0.27		0.26 ± 0.09		0.16 ± 0.20
Neighborhood*		0.85		0.0002		0.53
Avocado Heights	0.35 ± 0.34		0.28 ± 0.08		0.15 ± 0.18
Hacienda Heights	0.36 ± 0.33		0.29 ± 0.31		0.12 ± 0.13
La Puente	0.44 ± 0.42		0.20 ± 0.08		0.11 ± 0.18
Wind Direction*		0.87		0.04		0.91
downwind	0.37 ± 0.36		0.27 ± 0.09		0.14 ± 0.18
upwind	0.35 ± 0.33		0.29 ± 0.31		0.12 ± 0.13
Distance Tertial		0.61		0.36		0.83
1 (<1700 m)	0.38 ± 0.36		0.29 ± 0.29		0.12 ± 0.15
2 (1700 < 2750 m)	0.33 ± 0.35		0.26 ± 0.07		0.15 ± 0.19
3 (2750 <4827 m)	0.38 ± 0.35		0.26 ± 0.10		0.14 ± 0.16
Highway <500 m		0.85		0.71		0.81
No	0.38 ± 0.37		0.26 ± 0.08		0.14 ± 0.17
Yes	0.35 ± 0.33		0.29 ± 0.27		0.12 ± 0.15

Kruskal-Wallis for age, neighborhood, and proximity to industrial facility as a tertial

Mann-Whitney for ethnicity, mother’s education, sex, father’s occupation, smoking, wind direction, and proximity to highway

^‡Manganese and Industrial source factors significantly different for age categories

^*Dietary source factor significantly different by neighborhood and wind direction

Given that exposure to tobacco smoke may be related to metals exposure [65], we performed a sensitivity analysis to exclude participants with a presence of indoor environmental tobacco smoke at home (n=11). The exclusion of participants from the NMF analysis did not impact our source factor findings or observed differences by demographic, lifestyle or spatial characteristics.

As a comparison, we also performed a principal components analysis, which identified similar source groupings to those determined by NMF. Based on the eigenvalue ≥1.0 only two principal components (PCs) would be retained, which together explained 60% of the total variance (Table S2). However, four PCs would be retained in order to explain 81% of the variance in the data (Table S2) [58]. The four PCs (%variance explained) were V-Mn-As-Pb (44%), Se-Hg (16%), Cd-Sb (11%), and Sb (10%). The Se-Hg component closely matches the composition of the dietary source factor identified by NMF. The V-Mn-As-Pb component is a mixture of both the manganese and industrial source factor identified by NMF. PC 3 contained Cd-Sb, while PC 4 contained only Sb, potentially indicating different potential source signatures. While not identical these groupings are similar to those identified by NMF.

Discussion

In this study of children living near industrial facilities in urban Los Angeles County, we measured exposures to metal mixtures in toenail samples and identified three potential source signatures of exposure. Using NMF to group these metals, three source factor groupings representing the mixture of metals in the toenails of children were identified. We identified an industrial source factor, composed primarily of As, Sb, Cd, and Pb in the children in this study. A second source factor, composed primarily of Mn, was highly correlated with the industrial factor and third factor, driven by Se and Hg, was identified as a potential dietary source factor. Child’s age was observed to be an important determinant of exposure, as age was related to differences in the industrial source factor, with younger children having higher source factor loadings. The strong positive correlations between the industrial source factor and individual toenail metal components (As, Sb, Cd, and Pb) may indicate that children exposed to high levels of one metal may be more likely to be exposed to multiple metals. This is also supported by the strong positive correlation between the industrial source factor and toenail manganese concentrations (σ = 0.55, p-value <0.01).

In this study we identified a source factor composed of Sb, Pb, As, and Cd. This factor likely represents an industrial source, given that these toxic metals have been associated with airborne emissions, which may be associated with the metal smelter facility in the industrial corridor [4,64]. For the past three years, total emissions from the 21 facilities in and around the City of Industry mandated to report to the TRI program have exceeded 2 million pounds per year [41]. The battery recycling facility in this industrial corridor is responsible for 96% of the TRI releases, with average annual on-site airborne emissions of As, Cd, Pb, Mn, Hg, and Se reported at 3.9, 1.7, 10.1, 3.4, 8.6, and 6.8 lbs./year by the South Coast Air Quality Management District (Table S3) [41,66]. A study of surface soil around and on the facility found that while As, Pb, and Cd soil concentrations were higher inside the facility, there is evidence for increasing Pb contamination with decreasing distance to the facility [67]. In this study, we found a difference in the industrial source factor by age, with younger children having higher industrial source factor values (Table 4). This finding is consistent with prior studies that reported higher Cd and Pb exposures in younger children due to ingestion of contaminated dust and soil [68,69]. A previous study of Pb levels from the surrounding community reported slightly higher blood Pb levels in the Hacienda Heights neighborhood (South of the battery recycling facility) compared to West Covina (North East from the battery recycling facility) [64]. A recent survey of the Hacienda Heights neighborhood by the Los Angeles County Department of Public Health revealed that residents continue to be concerned about their health due to their proximity to the battery recycling facility [70]. Lead has been the primary contaminant of focus in this community, but as Cd, Sb, and As are also associated with this battery recycling facility [41,66], efforts should be made to investigate the full extent of exposure in this community.

The manganese source factor from the NMF analysis is dominated by Mn but includes other metals, such as V, As, and Se, in relatively low loadings. The battery recycling facility in the area emits approximately 8.6 lbs. of Mn a year [66]. Living near industrial facilities has been associated with higher levels of Mn in nearby populations [6,10,71,72]. It is possible that Mn exposure in our study population is also related to industrial sources in this area, which is supported by the high correlation between the Mn and industrial source factors. Additionally, the toenails of children in this study from parents employed in occupations that could potentially lead to take-home exposures of toxic metals (e.g., construction) were higher for the Mn source factor but not significantly different from other participants (Table 4). Studies have reported an association between Mn in drinking water and Mn in toenails [73–78], and with occupational respirable airborne Mn exposures [79]. The Mn levels in the drinking water of these Los Angeles neighborhoods have not exceeded secondary drinking water standards [80], however the battery recycling facility was issued violations for inadequate groundwater monitoring that could potentially release Mn into the groundwater [81]. Direct ingestion of the groundwater is unlikely to occur as residents in this area do not rely on private wells.

The dietary source factor is composed of Se and Hg, which tend to be influenced by dietary sources [82]. Toenail Se levels have been associated with Se in drinking water [73–78], but have also been associated with intake of dietary supplements [83–89]. Mercury in toenails is primarily indicative of methyl-Hg exposure, which is largely attributable to dietary sources [90–92]. Selenium and methyl-Hg both have been reported to be associated with fish intake [93,94]. While we did not observe differences in the dietary source factor by consumption of fish (Table 4), we cannot rule out exposure through supplements as we did not include a question regarding supplement intake in the questionnaire. Further, diet was only minimally assessed and may not fully capture specific fish and seafood consumption habits or other foods that may contribute to metals exposures. Additionally, we cannot rule out the possibility that the toenail Hg and Se levels observed in this study may be from exposure to emissions from the nearby battery recycling facility which emits approximately 4.6 lbs. of Se a year, and 8.6 lbs./year of Hg [66].

In a novel application of the NMF method, we explored potential source signatures of exposure using toenail clippings, as biomarker of multiple metals exposure. NMF has successfully been used for pollutant source apportionment of non-methane hydrocarbons to unconventional oil and gas exploration [95]. NMF has also been utilized to identify pesticide and chemical mixtures associated with dietary patterns [54,55]. While there is a lack of environmental health studies utilizing NMF, it has been identified as an unsupervised dimension reduction technique that can be useful in exposome research [96]. Recently NMF was used to identify factors from Pb in blood, urine, hair and nails to represent an integrated exposure estimate, these factors were then used to find an association with children intelligence quotient (IQ) scores [56]. While NMF has not been applied in the context of source apportionment of contaminants in toenails, it is a statistical tool that can help us understand the complexity of mixtures and multiple exposure sources particularly amongst environmental justice communities.

PCA and NMF are both unsupervised reduction methods [57,58]. While PCA has been extensively used in biomonitoring studies, the NMF approach has only recently been identified as a promising technique in mixture pattern analysis [96]. One of the main differences between the two approaches is that NMF requires the inputs to be positive. Alternatively, PCA will group both positive and negative components together, based on the strength of their correlation. NMF may provide more interpretable results for exposure assessment, as concentrations of contaminants are not naturally negative in real world environments. Comparison of these approaches and their applications to different datasets are beyond the scope of this study. However, previous studies in other disciplines have compared these methods and concluded that both offer comparable results [56,97,98], and the method employed should be determined by the research question [99]. Future studies to compare and identify best performing algorithms under various assumptions and scenarios applicable to environmental health data should be conducted.

A limited number of studies have utilized toenails as a biomarker of metals exposure in children in the United States, despite their utility as a non-invasive biomarker of past metals exposure [47,49]. We found that mean toenail levels of As, Mn, Pb, Se and V among children in our study are within the same order of magnitude as those reported in other studies conducted in U.S. populations of children living in urban areas or near legacy toxic sites [100–103] (Table S4). We only identified one other peer-reviewed U.S. study that measured toenail Cd in children living in one of the most polluted urban areas in Michigan and Cd concentrations in our study population were observed to be lower than those children [100]. To our knowledge, Sb toenail levels in children have not been reported by previous peer reviewed U.S. studies. There has been one previous peer reviewed study on toenail metal levels in children living near small scale Pb-battery recyclers in Vietnam [10]. While our As toenail levels are comparable, Cd, Hg, Mn, and Pb concentrations are exposures in our study are of similar magnitude to those reporting a decrease in intelligence among children participants [103,104]. However, the majority of these prior studies focus on single pollutants, which do not fully capture the reality of the cumulative burdened experiences by the community.

Low-income and communities of color across the United States are overburdened by exposure to toxic contaminants [105]. These communities have limited influence on land-use decisions and policies that determine location of industrial polluters within residential neighborhoods which contribute to environmental injustice and health disparities [106–109]. While biomonitoring began as tool of industry and government to test occupational workers for evidence of chemical poisoning, it is now utilized by social justice organizations to document the burden of chemical exposures in communities [39]. Previous community-drive research has identified an association between Mn levels and proximity to a ferromanganese refinery [11], elevated lead levels and proximity to a legacy smelting facility [110], and early life exposure to lead by utilizing teeth biomarkers [111]. In this research, we demonstrate an example of extending biomonitoring approaches at the community scale to identify patterns and potential sources of exposures.

The participants in this study live in communities that rank among the top 10% of the most pollution burdened neighborhoods in California [40]. The industrial facilities in the area contribute to the multipollutant exposures and cumulative impacts of environmental and social factors affecting this vulnerable community. Engaging and collaborating with these communities who typically have less access to resources for conducting research into the relationships between industry, environmental quality, and health conditions, can help increase environmental health literacy, raise awareness, inform policymakers and contribute to decisions that improve public health [112]. Regulating and policy making organizations continue the practice of ‘a one-chemical at a time / one facility at a time approach’ [113]. The most polluting industrial facility in the City of Industry has operated since 1959, producing ~600 tons of recycled Pb from Pb-acid batteries daily. The operation releases other toxic metals, such as Sb, As, and Cd, which are components of the source signature in our results [4]. From 2009 to 2018 the facility received 24 violations as a result of Pb, As, and 1,3-butadiene ambient air emissions exceeding air quality rule limits set by the South Coast Air Quality Management District [45]. These fenceline industries are un- or under-regulated, with the surrounding communities lacking access to resources for conducting research into the association between industry, environmental pollution, and health effects [114]. Public health campaigns in the area continue to primarily focus on lead contamination, ignoring the cumulative impacts of the other contaminants released by industrial polluters [70].

Until 2015, another secondary Pb smelter in Los Angeles, CA processed ~11 million batteries per year in a densely populated neighborhood of predominantly working-class Latinx residents [111,115]. The termination in production in this facility brings up questions about transferring the burden of contamination by continuing production elsewhere. While it is estimated that 8% of all toxic emissions in the United States can be attributed to the metals sector [41,116], this is likely to vary geographically as in the City of Industry the primary metals sector contributes to 90% of the toxic emissions in the area. Decades of contamination and under-regulation of these industrial polluters have contributed to the complex ‘riskscape’ of exposures in fenceline communities [114,117,118]. The lack of data linking individual exposures to health effects for these overburdened communities living near industrial corridors continues to be a top concern for community members. Identifying contamination priorities using community-based participatory research approaches are needed to ensure health equity amongst these communities is achieved [38,44]. For community organizations, particularly fenceline communities overburdened by industrial pollution this ensures their participation strengthens the ‘relevance, rigor, and reach’ of science [112].

One of the strengths of this study is the usage of toenails as a biomarker, which are widely accepted indicator of multiple metals exposure [82,119]. Exposures though air, water, dust and diet can contribute to metals in toenails, which reflect these exposures in the past 6–12 months. However, we only collected toenails once from children, and we did not collect detailed information of diet or time activity patterns. One of the limitations of utilizing techniques such as NMF and PCA are that while the identification of source factors is a robust process the interpretation of the results is dependent on expert knowledge [55,60,98,99]. Identified patterns and groupings require careful interpretation based on likely sources and routes of exposure, anthropogenic sources of contamination, and community specific behaviors. Following this approach, our interpretations of the possible sources for the factors are based on knowledge of contamination in environmental media in the surrounding communities [66,67,70], previous studies on metal sources [10,82,102,103], and differences identified by demographic and spatial characteristics in our participants (Table 4). However, it is possible that unidentified sources of metals in the community may also be related to the source groupings we observed. Another limitation of this study is its small sample size, and thus potentially limiting our power to observe significant differences within the community. Another strength of this study is that we collected in-depth information on many covariates including spatial and lifestyle information through questionnaires, allowing us to account for various risk factors of exposure. However, we did not measure any contaminants in environmental media. Future studies should investigate sources of exposure by implementing a multi-pathway multi-media research design. Utilizing a community-driven research approach and a new approach to assess mixtures and sources of exposure, this study identified patterns of toxic metal exposures in children. Investigating how these sources contribute to exposure and their influence on metal mixtures is important for assessing health outcomes.

In conclusion, our study is the first to identify “source” signatures and patterns of exposure in toenails of children residing in a community at the fenceline of industry. Although toenails have been widely used among U.S. adults to study metal exposures, there is a lack of information on this biomatrix in children. This study adds to the limited literature on children’s toenail metals levels in the United States and provides information on utilizing an unsupervised dimension reduction technique (NMF) to identify a source factor “signature” of contamination. As different communities experience various exposure mixtures it is unlikely that the sources factors identified by methods such as NMF or PCA will be identical across populations. Utilizing and adapting these methods is particularly important to characterize exposures in communities of color across the United States, who continue to be disproportionately impacted by exposure to toxic chemicals [105,120]. Identifying these exposure patterns would allow for targeted interventions and regulations aimed at reducing anthropogenic emissions. Upholding agreements, stricter enforcement of emissions, and clean sustainable technologies are key to ensuring the improvement of health equity in one community does not come at the cost of another.