Environmental arsenic exposure and serum matrix metalloproteinase-9

Abstract

The objective of this study was to evaluate the relationship between environmental arsenic exposure and serum matrix metalloproteinase (MMP)-9, a biomarker associated with cardiovascular disease and cancer. In a cross-sectional study of residents of Arizona, USA (n=215) and Sonora, Mexico (n=163), drinking water was assayed for total arsenic, and daily drinking water arsenic intake estimated. Urine was speciated for arsenic and concentrations were adjusted for specific gravity. Serum was analyzed for MMP-9 using ELISA. Mixed model linear regression was used to assess the relation among drinking water arsenic concentration, drinking water arsenic intake, urinary arsenic sum of species (the sum of arsenite, arsenate, monomethylarsonic acid and dimethylarsinic acid), and MMP-9, controlling for autocorrelation within households. Drinking water arsenic concentration and intake were positively associated with MMP-9, both in crude analysis and after adjustment for gender, country/ethnicity, age, body mass index, current smoking and diabetes. Urinary arsenic sum of species was positively associated with MMP-9 in multivariable analysis only. Using Akaike’s Information Criterion, arsenic concentration in drinking water provided a better fitting model of MMP-9, than either urinary arsenic or drinking water arsenic intake. In conclusion, arsenic exposure was positively associated with MMP-9 using all three exposure metrics evaluated.

Introduction

Arsenic exposure is linked to a variety of illnesses including cardiovascular disease (1-5), cancer (1, 2, 5-7), respiratory disease (8-10), diabetes (11, 12) and other diseases (13, 14). However, potential toxic mechanisms by which arsenic exposure causes chronic disease are not well understood.

Proposed mechanisms of arsenic toxicity, with a particular emphasis on cancer, include genotoxicity, oxidative stress, altered gene expression, and changes in cell cycle control, differentiation and apoptosis (15). In cellular and non-human animal models, a partial list of biomarkers of arsenic effect have included down-regulation of antioxidant gene expression (16), altered microRNA expression (17), decreased protein methylation (18), enzyme deactivation (19), altered protein concentrations (20) and altered wound healing (21). Information on biomarkers of effect in humans is more limited. Studies of the effects of ingested arsenic on human peripheral blood lymphocytes have demonstrated a wide array of modifications in gene expression (22), cytogenetic changes (23, 24), and decreased DNA repair (25). Arsenic exposure in drinking water has also been associated with reduced concentration of soluble receptor for advanced glycation end products (sRAGE) in sputum (20).

Matrix metalloproteinase 9 (MMP-9), also known as Collagenase Type IV-B, Collagenase Type IV, 92-Kd, and Gelatinase B, is a secreted zinc metalloprotease which degrades collagen in the extracellular matrix. Many studies link MMP-9 with cardiovascular disease (26-30), cancer (31-34), chronic respiratory disease (35, 36) and diabetes (37, 38). In earlier work, we demonstrated arsenic-related alterations in MMP-9 in cultured lung cells (21) and changes in MMP-9/tissue inhibitor of metalloproteinase-1 concentrations in sputum of arsenic-exposed human populations in Arizona (39).

Studies in the southwestern United States (Arizona) and the contiguous state of Sonora in Mexico identified populations exposed to arsenic in drinking water at concentrations exceeding the current US Environmental Protection Agency (EPA) maximum contaminant level (MCL) of 10 μg/L (39, 40). We hypothesized that among residents in communities predominantly exposed to low to moderate drinking water arsenic concentrations (< 50 ppb), increased arsenic exposure would be accompanied by higher levels of serum MMP-9. A relationship between arsenic exposure and serum MMP-9 would provide a potential mechanism for arsenic-induced chronic disease, including but not limited to cardiovascular disease and cancer. We carried out the present study in Arizona and Sonora to test this hypothesis.

Methods

The study was approved by the University of Arizona Institutional Review Board and all subjects provided informed consent. Detailed methods are described elsewhere (41). In brief, communities with relatively high and low water arsenic levels in Arizona and Sonora were selected for a cross-sectional study and households within these communities randomly selected from within census tracts and neighborhoods, respectively. Subjects were recruited using random digit dial telephone calls in Arizona and door-to-door visits in Sonora. Eligible households had at least one individual 18 years or older who had resided in the household for at least the previous year. Participants completed questionnaires in English or Spanish addressing residential, occupational and health histories, water sources and water usage. Water samples from all reported drinking water sources (e.g., indoor and outdoor tap, well, filtered, bottled, etc.), first morning void urine samples and blood samples were collected, and height and weight were measured.

Water and urine samples were stored at −20°C and transported to the University of Arizona (Tucson, AZ) for analysis. Drinking water samples were subjected to microwave digestion. After digestion, samples were brought up to volume with Milli-Q water for inductively coupled plasma extraction and mass spectroscopy (ICP-MS) analysis for total arsenic, with a detection limit of 0.1 μg/L. Urine total and speciated arsenic analyses were performed based on established methods (42) changing from a gradient to isocratic elution and adjusting the methanol concentration. Urine total arsenic was measured using an Agilent 7500ce ICP-MS (Agilent Technologies, Inc., Santa Clara, CA) and arsenite (As^III), arsenate (As^V), monomethylarsonic acid ^V (MMA^V) and dimethylarsinic acid ^V (DMA^V) analyzed using an Agilent 1100 HPLC (Agilent Technologies, Inc., Santa Clara, CA) with an anion exchange column (PRP-X100, 10um, 250×4.1mm, Hamilton, Reno, NV). The detection limits were as follows: total arsenic 0.1 μg/L, As^III 0.12 μg/L, As^V 0.21 μg/L, MMA^V 0.12 μg/L, and DMA^V 0.12 μg/L.

A weighted average water arsenic concentration was determined from reported usage of all water sources consumed for drinking (e.g., bottled water, filtered water, unfiltered tap water), weighted by the frequency of use (rare, moderate or frequent) of each source (41). Daily arsenic exposure from drinking water (drinking water arsenic intake) was then calculated from the weighted average water arsenic concentration (μg/L) multiplied by the average volume of drinking water consumed per day (L/day), and reported as μg/day. Urinary arsenic sum of species was calculated as the sum of As^III, As^V, MMA^V and DMA^V and reported in μg/L. Percent MMA was calculated as MMA^V divided by urinary arsenic sum of species multiplied by 100.

Non-fasting blood samples were collected for MMP-9 analysis in 10 mL red-top (serum) tubes (Becton Dickinson, Franklin Lakes, NJ), allowed to clot at room temperature for 15 minutes, and stored at 0-8°C for a maximum of 4 hours before centrifuging for 15 minutes at 1000 × gravity. Serum was separated into 2.0 ml aliquots and stored at −80°C until assayed for MMP-9. For MMP-9 analysis by ELISA (R&D Systems, Minneapolis, MN), standards and controls were assayed in duplicate using an automated microplate reader, Model ELx808 (BioTek, Instruments, Inc., Winooski, VT). The assay limit of detection was 0.156 ng/ml. MMP-9 concentration in samples was determined from the standard curve using a four parameter algorithm for best fit (KC4, BioTEK, Winooski, VT). Testing of all samples was done at a central laboratory in Arizona.

Stata 11.2 (StataCorp, College Station, TX) was used for statistical analyses. The concentration of arsenic in drinking water, drinking water arsenic intake, urinary arsenic sum of species, and serum concentration of MMP-9 were log (10)-transformed for parametric analysis. Pearson chi-square statistic, analysis of variance (ANOVA) using the Bonferroni test for post hoc comparisons, and Kruskal-Wallis tests were used to evaluate differences in population characteristics and risk factors by self-described country/ethnic background. Univariate and multivariable linear mixed models were used to assess the marginal relations between serum MMP-9 and arsenic exposure (weighted average arsenic concentration in drinking water, arsenic intake from drinking water, and urinary arsenic sum of species, adjusted for specific gravity) and other potential risk factors (age, body mass index , gender, country/ethnic background, current smoking status, and diabetes). Likelihood ratio tests were used to compare nested models and Akaike’s information criterion (AIC) was used to assess goodness of fit in non-nested models (43). A P-value < 0.05 was considered statistically significant.

Results

The total study population was composed of 487 participants, including 225 residents of Arizona and 262 residents of Sonora. For the analyses presented here, only those 378 individuals for whom both urine and blood samples were available were included. Characteristics of the study population are shown in Table 1 by country of residence and ethnic background: Mexican Hispanic, US Hispanic, and US non-Hispanic. These categories defined all but five people in the total study population, and these five subjects were excluded from analyses. Participants ranged in age from 19-88 years old, and the mean age was significantly higher in US non-Hispanics than US Hispanics or Mexicans. The study population was 56% female in Arizona and 74% in Sonora. Frequent use of alcohol (> 4x/week) was reported in over 18% of non-Hispanics in the US, but never in Mexico. Reported physician-diagnosed diabetes and BMI were substantially higher in US Hispanics than in the other groups.

Table 1.

Characteristics of the study population by country and self-described ethnicity.

	Total Population	Arizona, USA Non-Hispanic	Arizona, USA Hispanic	Sonora, Mexico Hispanic	P-value†
Number of participants (n, %)	377	165 (43.8)	50 (13.3)	162 (43.0)
Gender (n, %)
Male	133 (35.2)	68 (41.2)	23 (46.0)	42 (25.8)
Female	245 (64.8)	97 (58.8)	27 (54.0)	121 (74.2)	0.007
Age in years (mean, SD)	50.0 ± 16.3	57.2 ± 14.7	48.9 ± 15.5	43.0 ± 14.9	<0.001
BMI (mean, SD)	28.8 ± 5.9	28.1 ± 5.9	31.6 ± 6.5	28.8 ± 5.5	0.001
Current smoker	64 (18.3)	27 (17.6)	3 (7.0)	34 (22.2)	0.120
Frequent alcohol use (> 4x/wk)	37 (7.6)	32 (18.5)	5 (9.6)	0 (0)	<0.001
Physician-diagnosed diabetes	40 (10.6)	20 (12.2)	11 (22.0)	9 (5.8)	0.017
Drinking water As, geometric mean (95% CI)
Weighted concentration# (μg/L)	7.65 (6.80-8.63)	8.66 (6.93-10.81)	4.56 (3.77-5.51)	7.91 (6.90-9.07)	<0.002
Total volume consumed (L/day)	0.63 (0.55-0.72)	1.37 (1.20-1.58)	1.03 (0.48-1.37)	0.24 (0.20-0.29)	<0.001
Drinking water intake (μg/day)	2.47 (1.99-3.07)	5.83 (4.23-8.02)	3.37 (2.37-4.80)	0.93 (0.68-1.28)	<0.001
Urinary arsenic, adjusted for specific gravity (μg/L), geometric mean (95% CI)
Total	41.19 (37.65-45.07)	33.57 (28.92-38.98)	24.86 (21.81-28.33)	59.17 (52.63-66.51)	<0.001
As+3	1.10 (0.98-1. 23)	0.72 (0.60-0.85)	0.54 (0.42-0.69)	2.10 (1.86-2.37)	<0.001
As+5	0.82 (0.71-0.93)	0.61 (0.49-0.76)	0.47 (0.34-0.65)	1.29 (1.08-1.55)	<0.001
MMA	2.04 (1.85-2.25)	1.66 (1.40-1.96)	1.06 (0.87-1.30)	3.06 (2.75-3.41)	<0.001
DMA	13.67 (12.50-14.95)	10.38 (8.93-12.06)	7.89 (6.67-9.35)	21.39 (19.44-23.52)	<0.001
Sum of species	18.44 (18.86-20.17)	13.99 (12.01-16.28)	10.60 (9.05-12.43)	28.91 (26.31-31.77)	<0.001
Urine specific gravity, geometric mean (95% CI)
	1.016 (1.015-1.017)	1.015 (1.015-1.017)	1.015 (1.014-1.017)	1.017 (1.016-1.018)	0.035
Serum MMP-9 (ng/mL), geometric mean (95% CI)
	295.60 (274.4-318.5)	347.4 (311.6-387.3)	250.0 (204.1-306.3)	264.2 (235.4-296.5)	<0.001

^†P-values based on chi-square test, analysis of variance, or Kruskal-Wallis test

^#Weighted concentration = $\frac{1}{{\mathit{frequency}}_1+{\mathit{frequency}}_2+\dots +{\mathit{frequency}}_n}{\sum }_{i=1}^n({\mathit{frequency}}_i\times {\mathrm{concentration}\mathrm{of}\mathrm{source}}_i)$ (where n = the number of drinking sources used by the individual, and frequency_i =the frequency of use of each source i).

Weighted average arsenic concentration in household drinking water ranged from 0.132 μg/L to 1,004 μg/L, with a geometric mean concentration of 7.65 μg/L. It was significantly lower in US Hispanic households than in either of the other groups. The Mexican subjects reported consuming substantially less drinking water, contributing to markedly lower estimates of drinking water arsenic intake. However, urinary total arsenic and speciated arsenic concentrations were significantly greater in Mexican than in either US Hispanic or US non-Hispanic subjects, as was urine specific gravity. Serum concentration of MMP-9 ranged from 15-1,503 ng/ml. The geometric mean MMP-9 concentration was significantly higher among US non-Hispanics than in either Mexican or US Hispanic subjects.

The univariate relation between MMP-9 and exposure to arsenic estimated from 1) weighted mean drinking water arsenic concentration (μg/L), 2) drinking water intake (μg/day), and 3) urinary arsenic sum of species are shown for the total population and stratified by country/ethnic group in Figures 1, 2 and 3, respectively. In the total population, drinking water concentration and intake, but not urinary arsenic, showed a statistically significant positive relationship with MMP-9. All three exposures were each significantly associated with MMP-9 in the US non-Hispanic population, but none showed a significant association in other ethnic subgroups.

Figure 1.

Serum MMP-9 in relation to drinking water arsenic concentration by population subgroup: A) Total population, B) Mexicans, C) U.S. Hispanics, and D) U.S. non-Hispanics. P-values show statistical significance of univariate relationships.

Figure 2.

Serum MMP-9 in relation to drinking water arsenic intake by population subgroup: A) Total population, B) Mexicans, C) U.S. Hispanics, and D) U.S. non-Hispanics. P-values show statistical significance of univariate relationships.

Figure 3.

Serum MMP-9 in relation to urine arsenic sum of species by population subgroup: A) Total population, B) Mexicans, C) U.S. Hispanics, and D) U.S. non-Hispanics. P-values show statistical significance of univariate relationships.

MMP-9 concentration was evaluated against drinking water arsenic concentration, drinking water arsenic intake, urinary arsenic sum of species, gender, country/ethnic group category, age, BMI, smoking status, diabetes and alcohol use through univariate regression analysis (Table 2, column A). Drinking water arsenic concentration and intake, but not urinary arsenic sum of species, were significantly associated with MMP-9. Compared with Mexicans, US non-Hispanics but not US Hispanics had higher MMP-9 concentrations. Gender, age, BMI, current smoking status, diabetes and alcohol use were not significantly associated with MMP-9.

Table 2.

Predictors of serum (log10) MMP-9: A) Univariate models. Multivariable regression models with: B) drinking water arsenic (As) weighted concentration, C) drinking water As intake, and D) urinary As sum of species, adjusted for specific gravity.

	A) Univariate Models	B) Drinking Water As Concentration Model	C) Drinking Water As Intake Model	D) Urinary As Sum of Species Model
Predictor variables	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)
Log (10) drinking water As  concentration (μg/L)	0.162 (0.034) ‡	0.135 (0.035) ‡
Log (10) drinking water As  intake (μg/day)	0.073 (0.020) ‡			0.072 (0.022) ‡
Log (10) urinary As sum of  species (μg/L), adjusted	0.047 (0.044)			0.121 (0.050) *
Country/Ethnicity
Mexican/Hispanic	(ref)	(ref)	(ref)	(ref)
US/Hispanic	−0.050 (0.054)	0.003 (0.056)	−0.065 (0.057)	0.013 (0.060)
US/Non-Hispanic	0.092 (0.037) *	0.108 (0.040) †	0.059 (0.044)	0.155 (0.044) ‡
Male	0.048 (0.029)	0.037 (0.031)	0.034 (0.032)	0.043 (0.032)
Age (years)	−0.001 (0.001)	−0.001 (0.001)	−0.001 (0.001)	−0.002 (0.001)
BMI (kg/m2)	−0.005 (0.003)	−0.004 (0.003)	−0.005 (0.003)	−0.004 (0.003)
Current smoker	0.060 (0.042)	0.047 (0.042)	0.057 (0.043)	0.033 (0.043)
Diabetes	−0.079 (0.050)	−0.051 (0.052)	−0.058 (0.053)	−0.039 (0.053)
Model fit, AIC (df)		122.5 (11)	127.7 (11)	132.5 (11)

Abbreviations: SE, standard error; AIC, Akaike’s Information Criterion; df, degrees of freedom

^*P<0.05,

^†P<0.01,

^‡P<0.001

In adjusted linear mixed models, drinking water arsenic concentration and US non-Hispanic ethnicity were statistically significant independent predictors of MMP-9 (Table 2, column B). A separate model using estimates of drinking water arsenic intake, instead of concentration, was also predictive of MMP-9 (Table 2, column C), but this model provided a less parsimonious fit based on the AIC. A fourth model using urinary arsenic sum of species adjusted for specific gravity as a predictor of MMP-9 found that urinary arsenic was also significantly associated with MMP-9, but provided a poorer fit to the data (Table 2, column D). In the urinary arsenic model, US non-Hispanic ethnicity was also significantly related to MMP-9 in the model.

To assess the impact of adjusting urinary arsenic for specific gravity, models of the relation between MMP-9 and unadjusted urinary arsenic sum of species were also analyzed. Urinary arsenic sum of species unadjusted for specific gravity was borderline significant in the univariate model (P=0.080) (not shown), and both adjusted and unadjusted urinary arsenic were statistically significant in the multivariable models. In addition, multivariable models were constructed that included percent urinary MMA as an additional independent variable to evaluate the effect of urinary arsenic methylation, specifically, the formation of MMA, on MMP-9. Percent MMA was not a significant confounder of the relation between drinking water arsenic concentration, water arsenic intake, or urinary arsenic sum of species and MMP-9 (data not shown).

Discussion

In the present study three measures of arsenic exposure, the weighted mean drinking water arsenic concentration, the estimated daily intake from drinking water and the measured urinary arsenic sum of species, were independently associated with serum MMP-9 concentration. These results are consistent with the previously reported alterations in sputum MMP-9/TIMP-1 ratio in an Arizona population exposed to arsenic up to 20 ppb in their tap water (39) and increases in MMP-9 mRNA expression from cultured lung cells exposed to arsenic for five days at concentrations of 60 ppb and above (21). Furthermore, arsenic exposure has been associated with changes in MMP-9 in prostate cells (44) and keratinocytes (45), as well as in the lungs of mice exposed to arsenite in drinking water (46).

The exact mechanism for the association between arsenic exposure and increased MMP-9 is unknown. Possibilities include, but are not limited to, activation of activating protein-1 (AP-1) receptor sites in the promoter region of MMP-9 (47) or altered gene methylation (48). Compared with the limited knowledge concerning the toxic mechanism of arsenic, there are multiple postulated mechanisms by which increased MMP-9 can lead to cardiovascular disease and cancer. MMP-9 activity is increased in atheromatous plaque and increased extracellular matrix metabolism, with its consequent vascular remodeling, may contribute to further plaque development (49). Serum MMP-9 levels are associated with increased arterial stiffness (50). Plasma MMP-9 levels in patients with premature stable cardiovascular disease are positively associated with low density lipoprotein-cholesterol (LDL-C) and negatively associated with high density lipoprotein-cholesterol (HDL-C) (51). MMP-9 can also activate inflammatory mediators, such as interleukin 1-beta (52), and MMP-9 levels have been associated with other inflammatory biomarkers including C-reactive protein (53). For cancer endpoints, MMP-9 facilitates cancer cell growth, migration, invasion, metastasis and angiogenesis (54-56). There are, of course, other mechanisms not directly involving MMP-9 by which arsenic may cause or exacerbate a variety of chronic diseases, including other inflammatory pathways, oxidative stress, altered DNA methylation patterns, inhibition of DNA repair and modulation of signal transduction pathways (57-59).

Although our hypothesis that increased arsenic exposure would be associated with higher MMP-9 concentration was supported by these study results, the original expectation was that measurement of urinary arsenic sum of species would be a better predictor of inorganic arsenic exposure than drinking water arsenic (concentration or intake). This expectation was based on the reasoning that drinking water arsenic concentration would not reflect all sources of arsenic exposure and that estimates of water ingestion are subject to recall bias, while first morning void urinary arsenic concentrations would reflect all sources of exposure and would have more limited variability assuming a steady-state exposure such as consistent daily water arsenic consumption Calderon et al., 1999). One potential explanation for the finding that drinking water arsenic concentration yielded the best fitting model of MMP-9 is that these study populations may not have been at steady-state exposure relative to arsenic. Given the rapid absorption and excretion of arsenic from the body (60), urinary arsenic sum of species measured on a single occasion will only reflect recent intake of inorganic arsenic from water and food. Drinking water arsenic concentration, assuming consistent exposure, might be a better predictor of intermediate biological effects such as increased serum MMP-9. Although in our study drinking water arsenic concentration was a better predictor of serum MMP-9 than estimated intake or urinary arsenic sum of species, additional research is needed to see if such a relationship persists with other biomarkers of toxic effect.

One finding that is difficult to interpret is the elevated mean urinary arsenic sum of species concentration in the Mexican subjects. Even after adjusting for urine specific gravity, and despite low water consumption and therefore low drinking water arsenic intake, urinary arsenic sum of species was greatest in Mexico. Potential explanations for this apparent contradiction include greater arsenic exposure from water used in the preparation of food and beverages, greater dietary arsenic intake, and/or possible underreporting of the quantity of drinking water ingested.

There were a number of limitations to our study. The recruitment methods used in Mexico and the US were different. The proportion of females was markedly higher in the Mexican population as compared with the US, and the mean age was lower. Also, while we adjusted our models for self-reported, physician-diagnosed diabetes, the prevalence of undiagnosed diabetes is a concern. We also sought to determine whether the relation between As exposure and MMP-9 in US Hispanics was more similar to that in Mexicans or in US non-Hispanics, but the number of US Hispanics (n=50) was relatively small, limiting our power to demonstrate a statistically significant association between arsenic exposure and serum MMP-9 in this subpopulation. It was impossible to determine whether our subjects were at steady state in regards to arsenic exposure; however the literature suggests that exposure is fairly stable and significantly correlated over time (61-64), and subjects were eligible only if they had been living in their current residence for a minimum of one year. Because the transport time for the blood samples was relatively long and the ELISA used for MMP-9 analysis measures both the precursor (pro-MMP-9) and active forms, we were unable to determine the concentration of the active form alone. In addition, other variables affecting MMP-9 concentrations were not assessed in this study; including hormone therapy (65, 66), blood lead (67) and weight loss (68). Although alcohol abuse has been linked to high levels of serum MMP-9 and cardiovascular disease (69), alcohol consumption was not associated with MMP-9 in our study. Finally, although the association of arsenic exposure with serum MMP-9 was consistent across populations in both Arizona and Sonora, this association may not be generalizable elsewhere and the study needs to be replicated in other populations.

In summary, our study found an association between increased environmental arsenic exposure and increased concentrations of serum MMP-9 across populations in both Arizona and Sonora. These findings provide support for the hypothesis that alterations in MMP-9 could serve as one mechanism explaining the epidemiologic associations between arsenic exposure and select chronic diseases. In addition, serum MMP-9 was significantly associated with all three estimates of arsenic exposure evaluated, including drinking water arsenic concentration, drinking water arsenic intake and urinary arsenic concentrations.