Arsenic Exposure Is Associated with Decreased DNA Repair in Vitro and in Individuals Exposed to Drinking Water Arsenic


The mechanism(s) by which arsenic exposure contributes to human cancerrisk is unknown; however, several indirect cocarcinogenesis mechanismshave been proposed. Many studies support the role of As in altering oneor more DNA repair processes. In the present study we used individual-levelexposure data and biologic samples to investigate the effectsof As exposure on nucleotide excision repair in two study populations, focusingon the excision repair cross-complement 1 (ERCC1) component. Wemeasured drinking water, urinary, or toenail As levels and obtainedcryopreserved lymphocytes of a subset of individuals enrolled in epidemiologicstudies in New Hampshire (USA) and Sonora (Mexico). Additionally, incorroborative laboratory studies, we examined the effects ofAs on DNA repair in a cultured human cell model. Arsenic exposure wasassociated with decreased expression of ERCC1 in isolated lymphocytesat the mRNA and protein levels. In addition, lymphocytes from As-exposedindividuals showed higher levels of DNA damage, as measured by acomet assay, both at baseline and after a 2-acetoxyacetylaminofluorene (2-AAAF) challenge. Insupport of the in vivo data, As exposure decreased ERCC1 mRNA expression and enhanced levelsof DNA damage after a 2-AAAF challenge in cell culture. These data providefurther evidence to support the ability of As to inhibit the DNArepair machinery, which is likely to enhance the genotoxicity and mutagenicityof other directly genotoxic compounds, as part of a cocarcinogenicmechanism of action.


Arsenic is an established lung, skin, and bladder carcinogen [International Agency for Research on Cancer (IARC) 2004]; however, the carcinogenic mechanisms are currently under investigation. Basedprimarily on studies of highly exposed populations inTaiwan and elsewhere, the U.S. Environmental Protection Agency (EPA) recentlyreduced the maximum contaminant level (MCL) standard for arsenicin drinking water from 50 μg/L to 10 μg/L (U.S. EPA 2006). Atthe lower end of the dose–response curve, the biologiceffects and magnitude of disease risk in the human population remainunknown (Abernathy et al. 1999). However, a growing number of laboratory studies, both in cell culturesand in experimental animals, have demonstrated biologic effects ofAs at very low levels equivalent to those below the new 10 μg/Lstandard. These effects include endocrine disruption, altered cell signaling, alteredcell cycle kinetics, alterations in proliferative response, andother effects that may be associated with carcinogenesis andother disease processes (reviewed by Rossman 2003). Thus, it is important to understand the potential adverse effects ofsuch exposure in the human population.

An estimated 2% of the drinking water serving U.S. households contains 2 μg/L As (National Research Council 2001). Approximately 40% of households in the state of New Hampshireare served by unregulated private wells, with homeowner-financed, optionalcontaminant testing. Moreover, until recent studies revealed theextent of geologic As contamination of drinking water in the state (Peters et al. 1999), As was not part of the standard laboratory water testing panel. Case–controlstudies of bladder and skin cancer conducted in the NewHampshire population have detected evidence of elevated cancer risks. Forbladder cancer, an excess risk was observed primarily among smokersexposed to As in the drinking water, supporting hypotheses that theselevels of As are cocarcinogenic (Karagas et al. 2001, 2004). Likewise, drinking water in the southwestern United States and northernMexico contains As at concentrations above the new MCL of 10 μg/L (Meza et al. 2004).

The precise mechanism of As cocarcinogenesis is unknown. It has been difficultto detect genotoxic effects of As per se at environmental levels [Agency for Toxic Substances and Disease Registry (ATSDR) 1999; IARC 2004; Rossman 2003]. However, many studies support the role of As in altering oneor more DNA repair processes [World Health Organization (WHO) 2001]. Arsenic has been shown to potentiate the genotoxicity of otherorganic mutagen-carcinogens, particularly poly-cyclic aromatic hydrocarbons (PAHs), includingbenzo[a]pyrene (BAP) and ultraviolet radiation (UVR) (ATSDR 1999; Rossman 2003). Rats treated with As and BAP sustained adduct burdens longer than didrats treated with BAP alone, suggesting impairment of DNA repair byAs as a possible mechanism (Tran et al. 2002). A study using human fibroblasts found that low (2.5 μM, ~ 180 μg/L) concentrations of arsenite reduced nucleotide excisionrepair efficiency, and incision frequency in particular, after UVR exposure (Hartwig et al. 1997). Results of another study in cultured human fibroblasts indicated thatAs exposure reduced DNA repair capacity as measured by the comet assay (Curnow et al. 2001). The effects of As are strongly dose, time, and species dependent (Barchowsky et al. 1999; WHO 2001). In particular, several As-induced effects exhibit a biphasic characteristic. Forexample, low (≤ 1–2 μM) doses ofAs in cell culture increase cell proliferation and enhance endocrine signaling, whereashigher but still noncytotoxic doses (2–5 μM) suppressthese same pathways (Barchowsky et al. 1999). Likewise, patterns of altered gene expression, as detected by DNA microarraystudies, demonstrated very different patterns at low versus higherdoses (Andrew et al. 2003b). Thus, although animal and cell culture studies provide controlled modelsystems for mechanistic studies of As, it is critical to understandwhich of these findings translate into cellular, molecular, and clinicaleffects in actual human exposure situations, and the role of thesein pathophysiologic processes such as carcinogenesis. In our preliminarystudy of human lymphocytes from individuals exposed to drinking waterAs, As exposure was correlated in a strongly dose-dependent mannerto decreased expression of three nucleotide excision repair genes: ERCC1, XPB, and XPF (Andrew et al. 2003a).

The objective of this investigation was to evaluate our preliminary observationof an association between As exposure, focusing on ERCC1 gene expression levels in a larger number of individuals with exposuredata and biologic samples. In addition to gene expression, we investigatedthe effects of As exposure at both the protein and DNA repair functionallevels. We then extended our investigation into another populationexposed to similar levels of As in Mexico and also performed in vitro As experiments to validate our results in a controlled system.

Materials and Methods

Human populations

Cell line

We used Jurkat lymphoblast cells as a controlled in vitro system to evaluate the effects of As on DNA damage and repair. Cells weregrown in suspension in RPMI medium containing l -glutamine with 10% fetal bovine serum (Atlanta Biologicals, Norcross, GA, USA) and 1% penicillin-streptomycin (Mediatech Inc., Herndon, VA, USA). Cells were exposed to 0.01–10 μMsodium arsenite (Sigma, St. Louis, MO, USA), which is equivalent to 0.75–750 μg/L, for a period of 24 hr before harvestingand RNA isolation for gene expression analysis. Cells were exposed inculture to or 1 μM (equivalent to 75 μg/L) As as sodiumarsenite for 24 hr before the comet assay was performed, as describedbelow.

Gene expression analysis

RNA was harvested using Trizol reagent (Gibco/BRL Life Technologies, Gaithersburg, MD, USA) followed by DNase digestion using DNAfree (AmbionInc., Austin, TX, USA) according to the manufacturer’s instructionsand quantitated by spectrophotometric absorbance at 260 nm. Weperformed real-time reverse-transcription polymerase chain reaction (RT-PCR) usinggene-specific primers and reagents (Applied Biosystems, FosterCity, CA, USA) and the ABI PRISM sequence detection system and software (AppliedBiosystems). Briefly, total RNA (0.5 μg) wasreverse transcribed using 100 U M-MLV (Maloney murine leukemia virus) reversetranscriptase in a mixture with oligo-dT and dNTPs (deoxyribonucleotidetriphosphates) according to the instructions provided with theQiagen Omniscript kit (Qiagen, Valencia, CA, USA). Samples were reversetranscribed in a PTC-100 thermocycler (MJ Research Inc., Watertown, MA, USA) for 60 minat 44°C, and the reaction was terminatedby heating to 95°C for 10 min. Expression of ERCC1 [excision repair cross-complementing rodent repair deficiency, complementationgroup 1; GenBank gene ID 2067 (GenBank 2006)] was assessed by real-time PCR using 10 ng total RNA, 400 nMprimers, 200 nM probe, and TaqMan Universal PCR Master Mix (Applied Biosystems). Thesequence for the ERCC1 primer probe set is as follows: forward, CAGGACTTCGTCTCCCGGT; probe, TCTGGAACAGCTCATCGCCGCA; reverse, GCATAAGGCCAGATCTTCT-CTTG. Relative quantitation was performed using astandard curve consisting of serial dilutions of pooled sample cDNA fromthe same source as the test RNA with each plate. Relative expressionlevels of each gene were normalized to 18s rRNA or GAPDH (Applied Biosystems).

Protein levels

We assessed the level of ERCC1 protein by immunoblotting using sodium dodecylsulfate–polyacrylamide gel electrophoresis (SDS-PAGE) toresolve proteins from whole-cell lysates. Lymphocytes were rinsed withice-cold stop buffer (10 mM Tris-HCl, pH 7.4, 10 mM EDTA, 5 mM EGTA, 100 mMNaF, 200 mM sucrose, 100 μM Na-orthovanadate, 5 mM pyrophosphate, 4 μg/mL leupeptin, 4 μg/mL soybean trypsininhibitor, 1 mM benzamidine, 20 μM calpain inhibitor 1, 100 mU/mLaprotinin, and 100 μM phenylmethyl-sulfonylfluoride). Thestop buffer was then replaced with a minimal volume of SDS-PAGEbuffer (62.5 mM Tris-HCl, pH 6.8, 10% glycerol, 2% SDS, 5% β-mercap-toethanol, 0.05% wt/vol bromophenolblue). The lysates were boiled for 5 min and clarified by centrifugationat 13,000 rpm for 10 min. Equal amounts of cell lysate wereresolved by electrophoresis on 8–12% SDS-polyacrylamidegels. Electrophoresis was performed at constant voltage (200 V), andthen the resolved proteins were transferred from the polyacrylamidegel to polyvinylidene difluoride membrane (PVDF; Immobilon-P; Millipore, Bedford, MA, USA) bysemidry transfer (Hoeffer Semiphor, San Francisco, CA, USA) for 1 hrat constant current (32 mA/minigel) using transferbuffer (25 mM Tris, 192 mM glycine, 20% vol/vol methanol, 0.01% SDS). Toeliminate non-specific interactions of antibodieswith the membrane, the PVDF membrane was blocked with TTBS (10 mMTris-HCl, pH 8.0, 150 mM NaCl, 0.05% Tween-20) containing 5% milk (wt/vol) for 1 hr at room temperature or overnight at 4°C. Themembrane was incubated with the primary ERCC1 antibody (Neomarkers; LabVision Co., Fremont, CA, USA) diluted 1:200 in TTBS overnightat 4°C. The membrane was washed three times with TTBSand incubated with horseradish peroxidase–linked sheep anti-mouseIgG (1:2,000 in TTBS) (Amersham Pharmacia Biotech, Piscataway, NJ, USA) for 0.5–1 hr at room temperature. After three washes withTTBS, protein bands were visualized by enhanced chemiluminescence usingthe Renaissance system (NEN Life Sciences, Boston, MA, USA) and film (Lumi-Film; RocheMolecular Biochemicals, Indianapolis, IN, USA).

DNA damage and repair assessment

The single-cell gel electrophoresis or comet assay is widely used to measureDNA damage and repair in primary human lymphocytes by measuringstrand breaks and apurinic sites (Baltaci et al. 2002; De Silva et al. 2000; Rajaee-Behbahani et al. 2001; Schmezer et al. 2001). We divided the lymphocyte sample from each individual into parts toassess damage at several time points. We assessed baseline DNA damagelevels as well as the capacity of the lymphocytes to repair damage inducedby an in vitro challenge with 2-acetoxyacetyl-aminofluorene (2-AAAF; Midwest ResearchInstitute, Kansas City, MO, USA), the reactive and genotoxic metaboliteof 2-acetyl-aminofluorene. Alkaline-labile 2-AAAF adducts are primarilyrepaired through the nucleotide excision repair pathway (van Steeg 2001). Aliquots of lymphocytes were challenged for 2 hr in vitro with 4 μM 2-AAAF. A subset of 2-AAAF–challenged lymphocyteswere incubated for an additional 4 hr to allow for DNA repair ofthe lesions. Comet analysis was performed using materials and protocolsfrom Trevigen Inc. (Gaithersburg, MD, USA). Briefly, cells were mixedwith agarose and spread over a warmed, precoated microscope slide. Theagarose was allowed to solidify for 30 min at 4°C, followedby immersion in prechilled lysis solution for 45 min or overnight. Next, theslides were placed in freshly prepared alkaline solution, pH > 13, for 30 minat room temperature. The slides were then washed twiceby immersion in Tris-Borate-EDTA buffer for 5 min. Electrophoresiswas carried out in alkaline buffer for 20 min at 1 V/cm (measuredelectrode to electrode) in the dark. Last, the slides were dippedinto 70% ethanol for 5 min, allowed to dry completely, andstained with SYBR green (Trevigen). Image analysis of each cell wasperformed to quantify the length of the comet and the intensity of staining. Allcells were analyzed using a fluorescence microscope coupledto the MD Biotech comet assay image analysis system (Morgantown, WV, USA). TheOlive tail moment is a unitless measure of DNA damage that wascalculated as described previously (Olive et al. 1990) using the quantity of migrated DNA multiplied by the distance betweenthe comet head and the center of gravity of the DNA in the comet tail. Thequantity of migrated DNA is the fraction of the DNA that has migratedfrom the head. The quantity of DNA is assessed as the DNA stainingintensity subtracted from the background intensity. We scored the tailmoment of all cells in a given well. Each point represents an averageof 50 lymphocytes per individual (for human studies) or culture (for in vitro studies) from at least three to nine individuals or six cultures.

Statistical analysis

We performed statistical analysis for gene expression and immunoblottingusing analysis of variance with Newman-Keuls posttest, unpaired t-test, or linear regression procedures in GraphPad PRISM software (GraphPadSoftware Inc., San Diego, CA, USA). We considered p-values < 0.05 to be statistically significant. Statistical computationsand graphics for comet analysis were performed using the S-Plus statisticalpackage (version 6.2; Insightful Corporation, Seattle, WA, USA). Weplotted the mean Olive tail moment with 95% confidenceinterval (CI) for each treatment group as a function of time. We performedan unpaired two-sided t-test to compare the low- and high-As groups at each time point. Linearregression was used to assess the slopes of the lines. Corresponding p-values are indicated.


Demographic characteristics of the study populations are shown in Table 1. A larger percentage of the subjects in Mexico were female. In addition, theMexican population tended to be younger, with a mean age of 37 years, comparedto a mean age of 64 years in New Hampshire subjects. Approximatelyone-third of each population consisted of smokers.

As shown in Figure 1, analysis of both As-exposed populations combined indicated that individualsexposed to drinking water As concentrations 6 μg/L (n = 11) had lower ERCC1 gene expression levels than those with < 6 μg/L As (n = 42; p < 0.05). The Mexican population alone had a lower ERCC1 level among individuals exposed to As 6 μg/L, althoughthis was not statistically significant. Likewise, a lower ERCC1 level was observed in the New Hampshire individuals exposed to 6 μg/LAs (statistically significant at p < 0.05).

We further assessed As exposure using available internal biomarkers ofAs level. However, different biomarkers were used in the two populations, whichprevented pooling. Measurements included toenail As for theNew Hampshire population and urinary As for the Mexican population. Thesemarkers correlate with drinking water As concentration (Karagas et al. 2000). Linear regression analysis of the New Hampshire population indicatedan inverse association between toenail As levels and ERCC1 expression (r2 = 0.4; p < 0.05). ERCC1 expression decreased with increasing inorganic urinary As level [As(III) + As(V)] but not total urinary As (which mayinclude organic As), although this was not statistically significant (r2 = 0.08; p = 0.3). In the New Hampshire population, we found no differencein ERCC1 expression level according to cancer status (p = 0.8). For both populations, ERCC1 expression did not significantly differ by smoking status, age, or sex (datanot shown).

We went on to investigate whether the decreased gene expression levelsthat we observed in lymphocytes of subjects exposed to As in New Hampshiretranslated to changes in protein levels. Immunoblots using an ERCC1 antibodyindicated lower levels of ERCC1 protein among individualsexposed to drinking water As levels > 5 μg/L (p < 0.05), although there was some interindividual variation in expression (representativeblot shown in Figure 2A, quantification shown in Figure 2B).

Additionally, we hypothesized that As exposure would be associated withcorrespondingly higher DNA damage levels and reduced DNA repair function. Weanalyzed human lymphocytes from a subset of New Hampshire residentsexposed to low (< 0.7 μg/L) or high (≥ 13–93 μg/L) levels of drinking water As using the comet assay (Figure 3). We detected higher levels of DNA damage, as indicated by larger Olivetail moments, for lymphocytes analyzed at baseline from individualsexposed to high levels of drinking water As in vivo compared with those from lower level exposures (time hr; p < 0.05). Analysis of these cells at 2 hr after 2-AAAF challenge demonstrateda dramatic increase in DNA damage but did not reveal any statisticallysignificant differences in the amount of damage at 2 hr byAs exposure status (p = 0.25). However, at the 6-hr time point, after the 4-hr repairperiod, significantly higher levels of DNA damage remained in lymphocytesfrom individuals exposed to high compared with low levels of As in vivo (p < 0.05) (Figure 4). Control lymphocytes that did not receive in vitro challenge showed similar levels of DNA damage at the 6-hr time point (Figure 4). The difference in slopes of the low- and high-As lines was not statisticallysignificant.

To further confirm the hypothesis that As exposure inhibits ERCC1 expression, we repeated these experimental treatments using a human lymphoblastcell line. As shown in Figure 5, As suppressed ERCC1 expression in the treated cells in a dose-responsive manner, beginningat the 0.1 μM (~ 7 μg/L) dose, with statistically significantdecreases at 1 μM (p < 0.05) compared with unexposed controls.

We further investigated the hypothesis that As exposure in vitro would decrease DNA repair function using the lymphoblast cell line. Arsenic-exposedand -unexposed cells had similar levels of baseline DNAdamage at time hr (Figure 6). Arsenic-exposed cells challenged with 2-AAAF for 2 hr showed significantlyhigher levels of DNA damage than did 2-AAAF–challengedcells that had not been exposed to As (time = 2 hr; p < 0.05; Figure 6). DNA damage for both As-exposed and -unexposed, 2-AAAF–challengedcells decreased during the 4-hr repair period (Figure 6). Nevertheless, DNA damage for As-exposed cells remained significantlyhigher than the nonexposed cells at the 6-hr time point (p < 0.05; Figure 6).


Elucidating the mechanism of As carcinogenicity has been challenging, inpart due to the dose, time, and species specificity of its biologiceffects (Barchowsky et al. 1999; WHO 2001). Our early study (Andrew et al. 2003a) supported previous in vitro work showing disruption of DNA repair gene expression by As. In the presentstudy, we extended our findings to two different human populationsat the gene expression, protein, and DNA repair functional levels. Thus, ourdata provide both human in vivo and in vitro support for the hypothesis that As inhibits DNA repair processes (ATSDR 1999; Hartwig 1998) and that this has the potential to affect subsequent exposure to othergenotoxic and mutagenic agents.

The effects of As on DNA damage and repair have been evaluated almost exclusivelyin experimental systems. Previous in vitro studies demonstrated that As specifically interferes with the repair ofDNA photolesions (Yang et al. 1992) and cross-linking agents (Yager and Wiencke 1993). In another study using human fibroblasts, Hartwig et al. (1997) found that low (2.5 μM) concentrations of arsenite reduced nucleotideexcision repair efficiency, and incision frequency in particular, afterUVR exposure. Results of additional studies in cultured humanfibroblasts indicated that As exposure reduced DNA repair capacity andspecifically inhibited the repair of UV-induced pyrimidine dimer-relatedDNA damage in lymphoblastoid cells as measured by the comet assay (Curnow et al. 2001; Danaee et al. 2004; Yager and Wiencke 1993).

Differences in gene expression results between these in vitro studies may be explained by differences in As dose because the effectsof As are highly dose dependent. In the present study, treatment of lymphocyteswith > 1 μM sodium arsenite in vitro decreased ERCC1 gene expression. The circulating levels of As achieved in mice after intraperitoneallyadministering 0.625 nM As/kg body weight are approximatelyequivalent to the 5 μM in vitro dose and do not cause overt signs of toxicity (Soucy et al. 2003). In contrast, acutely toxic doses of As induced stress-response-pathwaygenes as well as ERCC1 gene expression in the livers of mice injected with 100–300 μMAs/kg body weight (Liu et al. 2001). Low concentrations of As induce cell proliferation, angiogenesis, hormonesignaling, and nuclear factor κB–dependent transcriptionand do not appear to activate mitogen-activated protein kinase (MAPK) signalingor other stress response pathways. In contrast, highdoses of As induce apoptosis and activate MAPKs, extracellular signal-regulatedkinase (ERK), and p-38, as well as stress-mimetic and heat-shock–mimeticresponses, inhibition of proliferation, and apoptosis (Barchowsky et al. 1999). Although decreased expression of ERCC1 may be partly responsible forthe decreased DNA repair function associated with As exposure, we recognizethat other pathway members may be involved, and future investigationwill be needed to elucidate all factors involved. Because otherenvironmental and genetic factors can influence DNA repair, we would notexpect complete concordance between As exposure and expression or functionon an individual level. Nevertheless, our in vitro studies demonstrate the effects of As within a controlled experimentalsystem. Further work is needed to identify genotypes that modify theinfluence of As exposure on DNA repair.

In a human population, we previously found that drinking water As exposureat levels between 5 and 75 μg/L was associated with decreasedmRNA expression of nucleotide excision repair pathway genes in lymphocytesfrom exposed individuals (Andrew et al. 2003a). Based on those preliminary results, we then followed up with the presentstudy, which uses a larger human population in New Hampshire. Inaddition to enlarging the sample size, examination of a second populationin Mexico exposed to moderate levels of As, supported our New Hampshirepopulation results. Although we observed decreased ERCC1 expression in both populations, the difference was not statistically significantin the Mexican population. The New Hampshire study had moreextreme levels of exposure (Mexico, 5.5–43 μg/L; NewHampshire, 0.007–75 μg/L), but more likely the Mexicostudy had a smaller sample size and lacked statistical power. To our knowledge, noother studies have evaluated the association between As exposureand DNA repair gene expression or protein levels in human populations, particularlyat As levels that are in the range that is routinelyfound in the United States.

In addition, our comet analysis provides functional DNA repair data inan As-exposed human population. These data support previous observationsof decreased DNA repair capacity after As exposure in vitro (Hartwig 1998). Arsenite has been shown to inhibit DNA repair and act as a cogenotoxinfor the direct-acting alkylating agent methyl methane-sulfonate, theindirect-acting PAH BaP, and UV-induced pyrimidine dimers in white bloodcells and fibroblasts (Curnow et al. 2001; Danaee et al. 2004; Hartmann and Speit 1996; Okui and Fujiwara 1986). As observed by Danaee et al. (2004) in a previous study, our in vitro As exposure appeared to inhibit the fast component of DNA repair becausethe difference is observable after challenge (Figure 6, time = 2 hr). This difference in DNA migration remained significantlyhigher in the As-exposed group after the repair period (Figure 6, time = 6 hr). Thus, our study and others consistently reportthat As exacerbates DNA damage induced by other mutagens.

Whether inorganic As can directly induce DNA damage by itself is more controversial, andprevious studies of DNA damage and mutagenesis by physiologiclevels of inorganic As have been inconsistent (Mass et al. 2001; Schwerdtle et al. 2003; Sordo et al. 2001; Yih and Lee 2000). Our in vitro comet data did not show an increase in DNA migration after 24 hr treatmentwith 1 μM As alone (Figure 6, time = hr), but we did observe higher DNA damage levels atbaseline in cells harvested from individuals exposed to drinking waterAs at levels 13 μg/L compared to those with low levelsof As (< 1 μg/L). The basis for this increase remains tobe determined; however, higher levels of DNA damage in these lymphocytesafter 2-AAAF challenge, and the slower repair kinetics of this damage, suggestthat the higher baseline levels may be a result of As-inhibitedrepair and exposure to other DNA-damaging agents.

In summary, our in vitro studies of As exposure and our novel work using in vivo As exposure in two human populations support the hypothesis that As exposuredecreases DNA repair capacity. Further, our data demonstrate decreasedexpression of the nucleotide excision repair pathway member ERCC1 anddecreased repair after 2-AAAF challenge. These results supportthe theory that As can act through a cocarcinogenic mechanism of action, exacerbatingthe genotoxicity and mutagenicity of other compounds.

Figures and Tables