Arsenic as an Endocrine Disruptor: Effects of Arsenic on Estrogen Receptor–Mediated Gene Expression In Vivo and in Cell Culture

Abstract

Arsenic (As) contamination of drinking water is considered a serious worldwide environmental health threat that is associated with increased disease risks including skin, lung, bladder, and other cancers; type 2 diabetes; vascular and cardiovascular diseases; reproductive and developmental effects; and neurological and cognitive effects. Increased health risks may occur at as low as 10–50 ppb, while biological effects have been observed in experimental animal and cell culture systems at much lower levels. We previously reported that As is a potent endocrine disruptor, altering gene regulation by the closely related glucocorticoid, mineralocorticoid, progesterone, and androgen steroid receptors (SRs) at concentrations as low as 0.01μM (∼ 0.7 ppb). Very low doses enhanced hormone-mediated gene transcription, whereas slightly higher but still noncytotoxic doses were suppressive. We report here that As also disrupts the more distally related estrogen receptor (ER) both in vivo and in cell culture. At noncytotoxic doses (1–50 μmol/kg arsenite) As strongly suppressed ER-dependent gene transcription of the 17β-estradiol (E2)–inducible vitellogenin II gene in chick embryo liver in vivo. In cell culture, noncytotoxic levels (0.25–3μM, ∼ 20–225 ppb) of As significantly inhibited E2-mediated gene activation of an ER-regulated reporter gene and the native ER-regulated GREB1 gene in human breast cancer MCF-7 cells. While the effects of As on ER-dependent gene regulation were generally similar to As effects on the other SRs, there were specific differences, particularly the lack of significant enhancement at the lowest doses, that may provide insights into possible mechanisms.

Arsenic is ubiquitous in the environment, and contamination of groundwater by inorganic As is widespread and naturally occurring as a result of geological formations and geochemistry conditions that are common in many areas. Arsenic in drinking water is considered one of the top environmental health threats both in the United States and worldwide, based on the extent of the population's potential exposure and the numerous diseases with which it has been associated (Abernathy et al., 2003; NRC, 1999, 2001; Smith et al., 2002; Tapio and Grosche, 2006; Watanabe et al., 2003). In places like Bangladesh and Taiwan, where this public health problem first surfaced, millions of people are threatened by As poisoning. High levels and extensive population exposures are found in many other countries around the world. In the United States there are also regions of extensive groundwater contamination including New Hampshire, Maine, and other Northeastern states; Michigan and other northern plain states; vast areas of the Southwest; areas of the Rocky Mountains; and regions of California. Generally the levels in the United States are lower than those in other areas of the world such as Bangladesh, with contaminated wells typically containing 10–100 ppb, although some wells test as high as 800–1000 ppb (Peters et al., 1999). Moreover, the problem is quite extensive, particularly in rural areas. For example, as many as one in five private wells in New Hampshire and Maine are contaminated with As (Peters et al., 1999), and perhaps 25 million people in the United States are drinking excess As. Chronic intake of inorganic As is strongly associated with an increased risk of skin, lung, bladder, and other cancers; type 2 diabetes; vascular and cardiovascular diseases; reproductive and developmental problems; and most recently with neurological and cognitive deficits (Abernathy et al., 2003; NRC, 1999, 2001; Smith et al., 2002; Tapio and Grosche, 2006; Wasserman et al., 2004; Watanabe et al., 2003). Recent studies suggest that there may be increased health risks at levels beginning as low as 10–50 ppb (Karagas et al., 2001, 2002; Wasserman et al., 2004). Arsenic is not a direct-acting genotoxin or mutagen but may increase DNA damage and mutations indirectly, such as by altering DNA repair (Andrew et al., 2003, 2006), and can also act as a cocarcinogen and/or tumor promoter and progressor (Rossman, 2003). A number of other mechanisms for the ability of As to increase disease risk have been postulated and have supporting data including increased reactive oxygen and oxidative signaling, altered cell signaling through other mechanisms, altered cell cycling and apoptotic and differentiation responses, and other mechanisms (Aposhian and Aposhian, 2006; Bode and Dong, 2002; Kitchin, 2001; Rossman, 2003).

We first reported that As can act as a potent endocrine disruptor, altering gene regulation by the closely related steroid hormone receptors for glucocorticoids (GRs), mineralocorticoids (MRs), progesterone (PR), and androgen (AR) in a similar manner (Bodwell et al., 2004, 2006; Kaltreider et al., 2001). These receptors all showed a similar and a strikingly complex dose-dependent response to As, with enhancement of hormone-dependent gene transcription at very low doses (0.1–1μM As) and suppression at higher but still noncytotoxic doses (1–5μM As). Using GR as a representative model for all four of these steroid receptors (SRs), we examined the mechanistic basis for these effects in detail. Arsenic altered DNA-dependent but not DNA-independent GR regulation of transcription, suggesting that the target for these effects was the transcriptional machinery required for SR-mediated transcription (Bodwell et al., 2004). Detailed mutational analysis of GR indicated that the receptor itself is not the causal target for As effects. For example, the entire N-terminal and C-terminal domains can be removed from GR without altering the As effect, indicating that the central DNA-binding domain (DBD) is the primary mediator of the response (Bodwell et al., 2004). However, mutation of virtually all the predicted As-binding sites within this DBD that did not abolish function also did not ablate the As effect (Bodwell et al., 2004). Further, GR DBD mutants that are constitutively in a DNA-bound conformation also displayed the suppressive effects of the higher concentrations of As but lacked the enhancement seen at lower As doses (Bodwell et al., 2006). This not only indicated that these two effects can be separated but also suggested that the enhancement by As may involve activation steps prior to DNA binding, whereas the suppressive effects may involve steps downstream of activation and DNA binding. Moreover, although the four SRs share overall structural and functional aspects within the DBD, they appear to lack complete conservation of any specific residue or domain structure that might confer similar As effects to all members by a direct interaction of As within the DBD.

These observations, despite the striking similarity in response among these four receptors, suggested that the actual target for these As effects is a part of the shared machinery they use to regulate gene expression rather than the receptors per se. In order to test this hypothesis further, we examined effects of As on estrogen receptor (ER)–mediated gene expression. ER, although sharing some common structural features, is the most divergent of the five SRs in primary and secondary structure, with only 15%, 52%, and 30% identities in the regulatory domains, DBDs, and hormone ligand–binding domains (LBD), respectively, compared to GR. In contrast, GR, MR, PR, and AR share 90–94% identities within the DBD and 55–57% identities within the LBD. Likewise, although ER is understood to share much of the coregulatory and transcriptional machinery used by the other four receptors, it also has been reported to have divergent coregulatory interactions and may also undergo unique steps in its activation, SRE binding and gene activation (Dennis and O'Malley, 2005; Stenoien et al., 2000). Other previous studies have suggested that As can interfere with or effect the actions of ER, although the mechanism is unclear, and the results suggested that this may be an indirect effect (Chen et al., 2002; Chow et al., 2004a,b; Stoica et al., 2000). Thus, the goals of the current study were to determine whether ER-mediated transcription is directly affected by As, whether such effects are similar to those of the other SRs, and whether any similarities or differences might provide further insights into the mechanisms by which As acts as an endocrine disruptor.

MATERIALS AND METHODS

Reagents.

All reagents and instruments were from Sigma (St Louis, MO) or Fisher Scientific (Pittsburgh, PA) unless otherwise noted. The 17β-estradiol (E2) was dissolved in ethanol (for cell experiments) or in acetone (for chick embryo experiments), the concentration was verified by spectrophotometer (1950 per M at 280 nm), and the stock was stored at − 20°C. Arsenic (As) was prepared as 1mM stocks of sodium arsenite (inorganic As³⁺) dissolved in sterile water, aliquoted, and kept frozen until day of use. Charcoal-stripped serum was used to remove endogenous steroids and was prepared as previously described (Hu et al., 1997).

Chick embryo treatments.

Studies were performed in compliance with Institutional Animal Care and Use guidelines approved by Dartmouth Medical School. Fertile White Leghorn chicken eggs were obtained (Truslow Farms, Inc., Chestertown, MD) and incubated as previously described (Hamilton et al., 1988) in an automated incubator at 37.5°C, 85% humidity, and rotated every 2 h. At 14 days, viable embryos were treated with indicated concentrations of As administered in 100 μl water and/or E2 administered in 10 μl acetone. Compounds were pipetted onto the inner shell membrane as previously described (Hamilton et al., 1983). Treatment with As was always 1 h before E2 treatment, and chick embryo livers were collected 3 h after E2 treatment. RNA was immediately collected from liver tissue using the following procedure: disruption of tissue with a dounce homogenizer, further homogenization with the Qiashredder kit (Qiagen, Valencia, CA), and RNA extraction with the Qiagen RNeasy kit per kit instructions. RNA was stored at − 80°C, and vitellogenin expression was measured via semiquantitative real-time PCR. Each treatment group was repeated in at least four experiments with 21–40 total embryos per treatment group. Statistical significance was determined with Prism (Graphpad, San Diego, CA) as determined by standard t-tests with indicated p-values.

MCF-7 cell culturing, transfections, and treatments.

Human breast cancer MCF-7 cells obtained from American Type Culture Collection (Manassas, VA) were cultured at 37°C and 5% CO₂ for a maximum of 30 passages in complete media: minimum essential media α (α-MEM, Invitrogen, Carlsbad, CA) containing 10% Premium Select Fetal Bovine Serum (FBS; Atlanta Biologicals, Norcross, GA), and 100 IU penicillin plus 100 μg/ml streptomycin (Mediatech, Herndon, VA). Confluent cells were split at a ratio of 1:4 twice weekly, using 0.25% trypsin/EDTA (Atlanta Biologicals). For transient transfections of the 1ERE/TK/luciferase construct (ERE-luc), the cells were split into 6-well plates at 3 × 10⁵ cells per well and grown overnight. The construct used for transfections in MCF-7 cells was generously donated by James DiRenzo (Dartmouth Medical School). Briefly, it contains one classic vitellogenin estrogen response element (ERE) sequence upstream of a thymidine kinase promoter and the firefly luciferase–coding region with a pBR322 backbone (DiRenzo et al., 2000). The entire construct was sequenced for verification. Cells were transfected with 1 μg of ERE-luc, using Fugene transfection reagent (Roche Diagnostics, Basel, Switzerland) per manufacturer's recommended protocol in phenol red–free complete α-MEM with 10% charcoal-stripped serum. Treatment with E2 and As was performed 24 h after transfection with the doses and for the durations described in the “Results” section. Each treatment and reverse transcription–real time polymerase chain reaction (real time RT-PCR) experiment (see below) with ERE-luc was repeated three to four times with six replicates per treatment group per experiment. For real time RT-PCR experiments of GREB1 and ERα in MCF-7 cells, the cells were split into 6-well plates at 3 × 10⁵ cells per well and grown for 3 days in phenol red–free complete α-MEM with 10% charcoal-stripped serum. Transfection efficiency was consistent, with little variability among the biological repeats in each treatment group and low SDs between replicate experiments. Treatment with E2 and As was performed with the doses and for the durations described in the “Results” section. Unless indicated, E2 and As were added simultaneously. Each treatment in the GREB1 or ERα real time RT-PCR experiments (see below) was repeated three to four times with six replicates per treatment group per experiment. Statistical significance for GREB1 and ERα experiments was determined with Prism's standard t-test with Welch's correction.

ICP–mass spectrometry analysis of total arsenic in cell culture media.

Total arsenic levels in cell culture media were measured by the Dartmouth Superfund Trace Element Analysis (TEA) Core Facility using collision cell ICP–mass spectrometry (MS) (Agilent octopole/reaction cell 7500c ICP-MS, Palo Alto, CA), employing He as the collision gas. The media were diluted and analyzed by the method of standard additions. Method detection limits were 0.37 ppb, and the overall method uncertainty was 8%.

Clonogenic cell survival cytotoxicity assays.

MCF-7 cells were grown to confluency in complete α-MEM media (see above) until clonogenic assays were started. All steps of clonogenic assays were performed using phenol red–free complete α-MEM with 10% charcoal-stripped serum. Approximately 50 cells per well were plated into 6-well plates and allowed to grow overnight. They were then exposed to 0.01–500μM As with and without added 50pM E2 for 24 h. At the end of the 24 h, the media were changed, but samples with added E2 continued to get 50pM E2 throughout. Cell growth was checked daily. When colonies of 25 cells or more were formed in control wells, the media were removed, the cells were fixed with 100% methanol for 5 min, methanol was removed, and cells were stained with 0.3% Giemsa stain (in 70% methanol) for 30 min. The stain was rinsed away with Hank's buffer, and colonies were counted. Results are given as a percentage of the no arsenic control, and each condition was repeated in triplicate. Three independent experiments of each type were performed. The sigmoidal dose-response curve and 24-h LD₅₀ were produced with Prism's dose-response curve analysis using average value for each dose.

Luciferase and protein assays.

Luciferase assays were performed on cell lysate supernatants with the Promega Luciferase Assay System kit (Madison, WI) following the manufacturer's recommended protocol. Briefly, after transfection with ERE-luc, followed by E2 and/or As treatment, MCF-7 cells were rinsed twice with cold phosphate-buffered saline, covered with 150 μl per well (of 6-well plates) of Promega lysis buffer, and scraped from the plate. Cell lysates were frozen at − 80°C for at least 30 min, thawed, vortexed for 30 s, spun at 12,000 × g for 5 min, and finally the supernatants were transferred to a new tube and stored at − 80°C. A Dynatech Laboratories (Chantilly, VA) luminometer was used to determine the light units per sample. Protein assays were also performed on lysate supernatants to normalize the total level of protein in samples. Pierce BCA Protein Assay Reagent Kit (Rockford, IL) was utilized for these assays per manufacturer's recommended protocol. Protein assay results were determined using a Thermomax microplate reader (Molecular Devices, Sunnyvale, CA). Each treatment and luciferase experiment with ERE-luc was repeated three times with six replicates per treatment group per experiment. The sigmoidal dose-response curves for effects of E2 or arsenic on ERα were produced with Prism's dose-response curve analysis using average value for each dose.

Isolation of RNA and relative quantitation real-time RT-PCR.

Chick embryo livers were treated and obtained as described above, and total RNA was extracted and homogenized using the Qiagen RNeasy and Qiashredder kits. RNA from treated MCF-7 cells was extracted using Trizol reagent (Invitrogen), following the manufacturer's recommendations. RNA was treated with DNase (DNA-free kit, Ambion, Austin, TX) for 20 min at 37°C, according to the manufacturer's instructions, to remove contaminating DNA. Purified total RNA was quantitated by spectrophotometric absorbance at 260 nm (NanoDrop, NanoDrop Technologies, Rockland, DE). RNA quality was assessed by 18S/28S analysis via electrophoresis on an agarose gel or by analysis using the RNA 6000 LabChip kit on an Agilent 2100 Bioanalyzer (Agilent Technologies, Wilmington, DE).

RNA samples extracted from cells contain significant levels of transiently transfected ERE-Luc DNA. Thus it was necessary to remove the contaminating construct DNA from the RNA sample. Therefore, we digested one half of the total RNA sample with 10 U of the frequent-cutting restriction enzyme, AluI (Invitrogen), per sample for 1 h at 37°C to linearize the plasmid DNA. AluI cuts within the targeted amplicon. This was followed with a DNAse digestion for a full 30 min at 37°C following the manufacturer's procedure (DNA free, Ambion). DNA-free 10× buffer was used for both enzyme reactions.

Two-step real-time RT-PCR was performed. Total RNA (2 μg) was reverse transcribed according to the instructions provided with the Qiagen Omniscript Reverse Transcriptase kit. For the RT step, random hexamers (Applied Biosystems, Foster City, CA) were used for chick vitellogenin II and for eukaryotic 18S-rRNA controls (see below). Gene-specific RT primers were used for GREB1 (5′-CGGAGTTTGACAAGATACCT-3′), ERα (5′-CAGGCTGTTCTTCTTAGAG-3′), and ERE-luc (5′-GGACTCTGGTACAAAATCGT-3′). Samples were reverse transcribed for 60 min at 37°C, and the reaction was terminated by heating to 93°C for 5 min. Controls with no reverse transcriptase were created during the RT process for each sample: no reverse transcriptase was included in these cDNA controls, so any amplicon formed in the PCR experiment with them would verify the level of contaminating DNA (found to be zero in all data shown).

Expression of chick vitellogenin II (GenBank ID M18060), GREB1 (GenBank gene ID NM_014668), ERα (GenBank gene ID NM_000125), and the ERE-luc construct were assessed by real-time PCR using TaqMan gene-specific primers/probes and reagents (Applied Biosystems). Primers and probes for real-time RT-PCR were designed using Applied Biosystems Primer Express software. Each probe target was set to a predicted exon-exon splice junction when possible. Primers and VIC-labeled probe (proprietary sequences) for eukaryotic 18S-rRNA were purchased directly from Applied Biosystems. All other probes were 6-FAM dye labeled. Probes for GREB1, ERα, and ERE-luc contained a minor groove-binding modification, to provide a nonfluorescent quencher on the 3′ end, whereas probes for vitellogenin II and 18S contained the TAMRA fluorescent quencher on the 3′ end. Sequences of the primers and probes used (5′ to 3′) are as follows—chick vitellogenin II sense: CAAAGGGACCACCGCTTTC; chick vitellogenin II antisense: CCAGTGGGCTCATTGAATGG; chick vitellogenin II TaqMan-labeled probe sequence: 6-FAM-ATACACCTGCCGCGACGAGACATCTG-TAMRA; human GREB1 sense: AATGGGTCCGGCTGTTTTC; human GREB1 antisense: CCAGTTGTTGGCACTTCGG; human GREB1 TaqMan-labeled probe sequence: 6-FAM-ACGGCAAAGATTC; human ERα sense: GGGAATGATGAAAGGTGGGATAC; human ERα antisense: TCATCTCTCTGGCGCTTGTGT; human ERα TaqMan-labeled probe sequence: 6-FAM-TTCAACATTCTCCCTCCTCTTCGGTCTTTT; ERE-luc sense: AATTGCTTTTACAGATGCACATATCG; ERE-luc antisense: TGCCAACCGAACGGACAT; and ERE-luc TaqMan-labeled probe sequence: 6-FAM-TGAACATCACGTACGCGGAATACTTCGA.

The TaqMan primers and probes were combined with TaqMan Universal PCR Master Mix (Applied Biosystems) and with cDNA (diluted in RNase-free water) according to manufacturer's recommendations in a 20-μl total reaction. Real-time RT-PCR was also performed with the eukaryotic 18S-rRNA primer/probe kit from Applied Biosystems for each sample to normalize the target gene PCR results. Reactions were placed in 96-well Optical MicroAmp plates (Applied Biosystems). The Applied Biosystems PRISM 7700 sequence detection system and software was used. Duplicates of each sample were incubated at 95°C for 10 min, followed by 40 cycles of 15 s at 95°C and 1 min at 60°C.

With each real-time RT-PCR experiment, relative quantitation was performed using the appropriate standard curve of the eukaryotic 18S-rRNA control (Applied Biosystems) and of each target gene reactant, consisting of serial dilutions of pooled sample cDNA from the same source as the test RNA within each plate (known to contain the appropriate transcript). Relative expression levels of each gene were normalized to the eukaryotic 18S-rRNA control to allow the determination of relative, normalized, semiquantitative results.

RESULTS

Effects of As on ER-Dependent Gene Expression in Chick Embryo In Vivo

We had previously shown that As can alter GR-mediated gene expression in the chick embryo in vivo at low, nonovertly toxic doses. Here we investigated whether similar As exposures would alter ER-dependent gene expression in a similar manner in this in vivo model. Chick embryo development and physiology has been well characterized, and the chick embryo has been widely used in physiology and toxicology investigations. The ER is expressed in liver as early as day 7 of chick embryo development and can mediate E2-induced gene expression as early as day 10 (Elbrecht et al., 1981, 1984). Interestingly, there is no difference between male and female embryos in this regard until after hatching, thus male and female embryos can be used equivalently for such experimental studies. Liver vitellogenin expression can be dramatically induced by exogenous E2 treatment in the 14-day chick embryo, increasing mRNA levels by up to 600-fold within hours. We used this robust E2 response to test the effects of As on ER-mediated gene activation. The embryos were exposed to As and/or E2 for 4 and 3 h prior to sacrifice, respectively, and RNA was immediately collected, and expression levels of vitellogenin were assessed with real-time PCR.

An E2 dose-response curve was determined for the induction of vitellogenin transcripts in the 14-day chick embryo liver (see Supplemental Fig. A). An EC₅₀ of 3.5 μmol/kg was calculated from this curve. We used an E2 dose of 10 μmol/kg (representing submaximal stimulation of approximately 60–80% depending on the experiment) in further studies which was sufficient for a robust vitellogenin induction yet allowing us to potentially detect both increases and decreases in vitellogenin mRNA. The chick embryos were then exposed to E2 plus various doses of As. A previous study had indicated that 100 μmol/kg dose of As was the lowest dose that caused observable toxicity in 14-day embryos (Hamilton and Kaltreider, unpublished results). As shown in Figure 1, As alone had no significant effect on basal expression. However, we observed a profound and dose-dependent suppression of E2-induced vitellogenin expression by As in chick embryo liver with an approximately 80% percent suppression in induction at 50 μmol/kg. Given that induction of this gene has been shown to be an E2-dependent, ER-mediated activation, this indicates that As is an endocrine disruptor of ER-mediated gene regulation in vivo.

Effects of As on ER-Dependent Gene Expression in MCF-7 Cells

We next examined the effects of As on ER-mediated gene expression in a cell culture system in order to be able to manipulate the conditions more extensively and to apply molecular biology approaches to examining possible mechanisms. Human breast cancer–derived, ER-positive MCF-7 cells were chosen for these experiments since they have been widely used to test the effects of E2-inducible gene expression mediated by EREs and ERα. We initially used an ERE-luc genetic construct (described in the “Materials and Methods” section) which was transiently transfected into cells followed by treatment with E2 and/or As.

All MCF-7 experiments were done in phenol red–free media using charcoal-stripped serum in order to minimize potential background stimulation of ERα. Initial dose-response experiments were performed to determine the optimal dose of E2 needed for robust E2-dependent, ER-mediated expression of the transfected ERE-luc construct, and the EC₅₀ was determined to be 10pM (see Supplemental Fig. B). Doses of either 50 or 100pM E2 were used in subsequent experiments, equivalent to the approximately EC₈₀ and EC₉₀ concentrations, respectively, for ERE-luc induction. As a negative control, a construct lacking the ERE was transfected; it was not significantly affected by E2 or As treatments (data not shown).

The cytotoxicity of As was determined in the MCF-7 cells in the presence or absence of E2 using a clonogenic assay, since it is important to determine the dose-response for each cell line due to widely varying susceptiblity to As toxicity of different cell lines. We typically use doses with equivalent toxicities when comparing responses of different cell lines rather than absolute As concentrations. As shown in Figure 2, MCF-7 cells are remarkably resistant to As cytotoxicity, with little or no change in colony-forming ability at or below 5μM As (in the presence of E2), a concentration which produces substantial toxicity in many other cell lines, and with 5–10% colony-forming ability remaining at 50μM As, a dose which is normally completely lethal in other cell lines. Also interestingly, there was a shift in the dose-response with the addition of E2, which increased survival and proliferation of these ER-positive MCF-7 cells in all conditions including the As-free control group. The LC₅₀ for As alone was approximately 15μM, whereas it was increased to approximately 25μM in the presence of E2, indicating a protective effect of E2 in these cells. Given that As is ubiquitous and we were testing extremely low As doses, we also examined the level of total As in our media, which was determined by ICP-MS to be less than 0.5 ppb (0.007μM) in complete media and in phenol red–free complete α-MEM with 10% charcoal-stripped serum containing 10% stripped FBS (Dartmouth College TEA Core Facility).

We then examined the dose-dependent effects of As on E2-induced expression of the ERE-luc construct using an E2 dose of 50pM which is equivalent to approximately the EC₈₀ concentration. Cells were transfected as above and treated with E2 in the presence or absence of various concentrations of As for 24 h prior to measurement of luciferase expression as shown in Figure 3. Concentrations of As as low as 0.25μM suppressed E2-stimulated ERE-luc expression, and the EC₅₀ was approximately 2.5μM. Similar suppression was seen using an E2 concentration of 100pM (data not shown). We also measured expression of this construct at the mRNA level using RT-PCR rather than luciferase expression (Fig. 4), with similar results, indicating that this suppression was a result of decreased gene expression. These results indicate that As strongly suppresses E2-dependent, ER-mediated gene expression in MCF-7 cells, similar to what we observed for the vitellogenin gene in chick embryo liver in vivo.

The time course for the suppressive effects of As on ER-mediated transcription was determined. As shown in Figure 5, As treatment suppressed E2-stimulated ERE-luc expression progressively over a 24-h period. Three hours after E2 alone there was a twofold increase in ERE-luc expression and a modest, nonsignificant effect of As (8% decrease). At 6 h, E2 induced a threefold increase in expression and As treatment significantly suppressed this induction by 19%. At 15 and 24 h there was approximately 6- and 12-fold induction by E2, respectively, and a progressively greater suppression by As of approximately 40% and 50%, respectively. We next examined the timing of As addition relative to a 24-h E2 stimulation. As shown in Figure 6, As suppressed ER-mediated gene expression whether added up to 4 h before E2 addition or up to 4 h after E2 addition, indicating that there neither is a prerequisite step nor is E2-stimulated expression protected from subsequent As effects even though a 4-h E2-alone treatment prior to As addition would result in a three- to fourfold increase in ERE-luc expression prior to As exposure (see Fig. 5). These results are comparable to those of a previous experiment with a similar design examining effects of As on GR (Kaltreider et al., 2001) and suggest that As affects ongoing transcription rather than ER activation per se.

Transcript levels of the endogenously expressed ER-regulated genes, GREB1 and ERα were also assessed with RT-PCR. GREB1 is an E2-responsive gene which is induced through an ERE in its promoter (Rae et al., 2005). GREB1 was recently identified (and named) as a gene that is upregulated in breast cancer (Ghosh et al., 2000) and appears to play a central role in ER-dependent cell proliferation (Rae et al., 2005). GREB1 has been reported to be expressed in all ER-dependent cell lines tested to date and has been found in other hormone-responsive cell lines (Rae et al., 2005). It has also been reported to play a role in AR-dependent proliferation in prostate cancer (Rae et al., 2006) and likely plays a role in other cancers as well. The GREB1 promoter contains regulatory elements for all five SRs, and the ERE and Androgen response element ARE elements have been shown to be functional by chromatin immunoprecipitation analysis (Rae et al., 2005, 2006). GREB1 mRNA exhibited clear induction with a 100pM E2 treatment following either a 6- (data not shown) or 24-h exposure (Fig. 7B). A 24-h treatment of MCF-7 cells with As alone caused a significant suppression in basal GREB1 mRNA expression as measured by real-time PCR (Fig. 7A) with an approximately 50% suppression at 5μM As. The effects of As on E2 induction of GREB1 mRNA are shown in Figure 7B. E2 alone increased GREB1 expression by approximately fivefold. As suppressed E2 induction of GREB1 in a dose-dependent manner to approximately 50% that of the E2 control at 5μM As.

ERα transcript levels were also measured by real-time PCR in the same experiment. ERα has been reported by others to be downregulated at both the transcript and protein levels by E2 alone (Stoica et al., 2000). However, we did not observe a significant decrease in ERα mRNA expression in our 24-h E2-alone experiments, probably due to differences in culture conditions (Read et al., 1989; Stoica et al., 1998). Similar to what was observed for GREB1, As alone repressed ERα mRNA expression, and effects were observed at concentrations as low as 0.01μM with maximal suppression of approximately 50% at 5μM As (Fig. 8). However, E2 added simultaneously with As at 100pM had no effect and even at concentrations as high as 1000pM (data not shown) had little additional effect over As alone on ERα expression.

DISCUSSION

In the current studies we report that As at low, environmentally relevant concentrations has profound effects on ER-mediated gene regulation both in vivo and in a model cell culture system. These effects are similar in many respects to previously reported effects of As on gene regulation by other SRs, particularly the strong suppression in hormone-stimulated, ER-dependent gene expression at doses as low as 1 μmol/kg in vivo and as low as 0.25μM (approximately 20 ppb) in cell culture. However, there were also some interesting differences in response between ER and the other SRs, particularly the lack of significant enhancement in hormone-stimulated gene expression by ER at very low As doses, as compared to the other SRs which typically show a two- to threefold enhancement in hormone-activated gene transcription in the range of 0.05–1μM As (Bodwell et al., 2004, 2006). There were also quantitative and qualitative differences in response between the transiently transfected construct and the native genes in MCF-7 cells and differences among the native genes in their response to As alone or in the presence of hormone.

Previous reports examining effects of As on ER are conflicting. All the previous studies examined effects of As were in cell culture, but these studies used different cell lines; different transiently transfected constructs and native genes; and different experimental conditions including different doses, time points, and endpoints. Thus, it is difficult to directly compare these results. Importantly, we observed profound effects of As on ER-dependent gene regulation in an in vivo system, the chick embryo, and also examined the effects of As on a native gene using physiologically relevant concentrations of As and hormone. We then replicated this observation in a cell culture model that is well characterized and that has been widely used to examine ER and its regulation of gene expression, the ER-positive human breast cancer MCF-7 cell line. In addition, the ERE sequence used in our genetic construct in the MCF-7 studies is identical to the native ERE sequence found in the chick vitellogenin promoter (Burch, 1984) and is a canonical ERE, which strongly binds ER and robustly transactivates gene transcription in the presence of E2. Others using similar experimental systems have seen similar results (Chen et al., 2002; Chow et al., 2004b), but some other investigators have reported no effect of As using different experimental systems (Stoica et al., 2000). It is not clear why some of these systems are negative, but this is likely a result of one or more differences in experimental variables with the system we used. One study that reported negative findings used a system that was very different than ours, i.e., COS-1 cells, an ERE-CAT construct (as compared to our ERE-luc construct), transfection of an ER expression vector (as compared to the native ER of MCF-7 cells because COS-1 cells are essentially ER-negative), and different doses and time points for As (Stoica et al., 2000). This group also used 100nM E2 which is 1000 times the level we used in MCF-7 cells to obtain optimal ER activation. It is not clear which, if any, of these factors may be the most important for understanding the lack of an As effect, but many of these conditions are far removed from the physiological condition, and it has previously been shown that culture conditions can profoundly affect ER and ER-mediated responses (Read et al., 1989).

We specifically chose the MCF-7 cell line because it is ER positive and well characterized with respect to ER regulation of gene transcription. We used the lowest doses of E2 necessary for robust ER-mediated gene activation, in a range that is physiological and also produced 70–90% of maximal stimulation. We also observed effects of As on two native genes, GREB1 and ERα, that have been reported to be regulated by E2. GREB1 is known to be regulated via ER and an ERE in its promoter (Rae et al., 2005), whereas ERα expression is more indirectly regulated by E2, most likely at a posttranscriptional step involving E2-mediated degradation (Dennis and O'Malley, 2005; Fan et al., 2004; Reid et al., 2002). The effect of As on E2 regulation of GREB1 was similar to what we observed for other SR-dependent genes in other cell lines (Bodwell et al., 2004, 2006; Kaltreider et al., 2001), suggesting a common mechanism of action. We believe that it is critically important to choose cell lines that are most relevant to the specific receptor pathway under study, even though this often means examining different SRs in different cell lines. But it is also important to confirm any observed effects in vivo, preferably with native genes and under physiological conditions, in order to demonstrate the relevance of the cell culture observations.

We also believe that it is critically important to use low, nonovertly cytotoxic doses of As for such studies if such results are to be considered relevant to chronic human exposures. The doses of As used here in which we observed significant effects on ER-mediated gene expression are well below the cytotoxic range for this cell line and in the range of environmental exposures, i.e., 10–200 ppb As (0.13–2.67μM) which is the range typically found in contaminated drinking water supplies in the United States and elsewhere in the world and of concern for human health effects. In this regard it is important to note that different cell lines demonstrate quite different ranges of tolerance. The MCF-7 cells used in these studies are quite resistant to As toxicity, with little or no effect on colony-forming ability below 5μM and an LC₅₀ of 15–25μM. In contrast, other cell lines such as the GH3 and NT2 cell lines we have used to examine effects on thyroid hormone and retinoic acid receptors, respectively, are highly sensitive to As with LC₅₀ values of approximately 3μM in both lines (Davey and Hamilton, manuscript in preparation), and H4IIE rat hepatoma cells used to investigate GR are in-between with an As LC₅₀ of 5–10μM (Bodwell et al., 2004, 2006; Kaltreider et al., 2001). Thus, we believe that it is important to determine and report the cytotoxic dose response for each cell line in relation to the doses of As used in the study and also to use toxic-equivalent doses of As when comparing results among different cell lines. It is also interesting and potentially important to note that all the receptor-positive cell lines we have tested are more sensitive to As cytotoxicity when cultured in stripped serum without added hormone than in the presence of complete serum or stripped serum plus hormone.

The precise mechanism by which As alters hormone-stimulated gene regulation by ER and the other SRs remains to be determined. As is clearly not only an endocrine-disrupting chemical (EDC) but also acts by a unique mechanism quite distinct from those of previously characterized organic EDCs, most of which act as hormone mimetics and as either agonists or competitive antagonists. Our previous studies with As and SRs, in combination with the current results, indicate that neither is As an agonist of SR activation nor does it act as a competitive or noncompetitive antagonist (Bodwell et al., 2004, 2006). However, As clearly alters the ability of hormone-activated, DNA-bound SRs to regulate gene transcription. These effects of As on SR activity appear to be restricted to DNA-dependent gene regulation (Bodwell et al., 2004). However, while effects of As on ER and all four of the other SRs—i.e., GR, AR, PR, and MR—are remarkably similar, they share little absolute sequence or structural identity within the DBD that could be attributed to a common As target (Bodwell et al., 2006). These results, taken together with other recent studies in our laboratory showing similar effects of As on thyroid hormone receptor and retinoic acid receptor, suggest to us that the receptors themselves are not the actual target for As effects, but rather that other proteins or regulatory pathways that are common to these diverse nuclear receptors represent the true causal target.

ER is the most ancestral SR (Thornton, 2001) and the most distally related to the other four SRs which are believed to have more recently diverged and share much closer homologies with each other. For example, GR, MR, AR, and PR share 90–94% identity within the DBD and almost 60% identity within the LBD. In contrast, ER is only about 50% identical to GR within the DBD, and less than 30% identical within the LBD. Aspects of ER biochemistry and cell biology are also divergent from the other forms. For example, ER is predominantly nuclear in localization regardless of ligand status due to a constitutively active nuclear localization signal, whereas GR is principally cytosolic in the resting state and translocates to the nucleus following ligand binding and activation (Dennis and O'Malley, 2005; King and Greene, 1984; Stenoien et al., 2000; Thornton, 2001; Welshons et al., 1984). We observed that ERα transcript levels were repressed by As alone or As plus E2 over a 24-h period. Similar repression of both ER transcript and ER protein levels was reported in three other studies (Chen et al., 2002; Chow et al., 2004b; Stoica et al., 2000). However, the decrease in ER expression was seen after long treatment times (24–72 h), yet we observed effects of As on ER-mediated transcription within 3–4 h of treatment. Moreover, we observed no effect of As on GR levels despite profound effects on GR-mediated gene expression (Bodwell et al., 2006). Thus, we believe that this suppressive effect is unlikely to explain the shorter term effects of As on ER-regulated processes we observed in the current studies. However, long-term suppression of ER levels might result in long-term effects of As on ER-regulated processes.

Given the striking similarity of effects we observed among all five SRs plus RAR and TR, which share only distal homology with the SRs and which employ different activation steps, it is likely that As targets one or more steps that are common and shared among these nuclear receptors, such as common coregulators, common signaling pathways that regulate them, or the shared transcription machinery itself. We suggest that the target for the suppressive effects of As is one or more common coregulators that mediate transcription of hormone-activated nuclear receptors. SRs and nuclear receptors share many corepressors and coactivators, several of which require specific activation or inactivation steps such as phosphorylation, acetylation, and methylation. One or more of these is likely to be the actual target that leads to the downstream effects on SR function that we have observed. In previous experiments, we observed that very low levels of As (0.05–1μM in rat hepatoma cells) actually enhanced hormone-mediated gene regulation by other SRs, by as much as two to threefold at maximum (Bodwell et al., 2004, 2006). In these experiments, slightly higher As concentrations suppressed hormone-mediated transcription. In our experiments with ER over a wide range of conditions, we have not observed the striking lower dose enhancement seen with GR and the other SRs, although the suppression seen at higher doses is very similar. This suggests that there may be a fundamental difference between ER and the other SRs that may be indicative of the mechanism for those low-dose effects. However, our previous results also suggest that the mechanism for the low-dose enhancement is distinct from that causing the higher dose suppression, since in certain mutants of GR that are in a constitutively DNA-bound conformation, we can abolish the lower dose effect while they retain the higher dose suppression (Bodwell et al., 2006). This further suggests that the suppression by As involves the DNA-bound form and its ability to mediate transcription, whereas the lower dose enhancement may be more related to activation of the other SRs to the DNA-bound, transcriptionally active form. Current studies are underway to differentiate these mechanisms. It is also interesting to note that As could alter ER-regulated gene expression even when added up to 4 h after ER gene activation by E2, suggesting that As is able to affect ongoing processes.

It has previously been proposed that As can act as a ligand for ER (Stoica et al., 2000), activating it in the absence of hormone. However, Chow et al. (2004b) saw no effect of As as a ligand mimetic for ER in their studies. We also observed no induction effect of noncytotoxic concentrations of As alone on ER-mediated gene activation in vivo or in cell culture and have also seen little or no effect of As alone on activation of other SRs. Previous studies have also reported that As could activate GR in a hormone-independent manner, but all these studies employed concentrations of As that were completely lethal to cells and which induced a heat-shock–mimetic response (Shen et al., 1993). Likewise, previous studies had also shown that high doses of As could block hormone binding to GR, but only at concentrations that are lethal to cells (Lopez et al., 1990; Simons et al., 1990; Stancato et al., 1993). These effects were not observed at lower, survivable concentrations that are relevant to typical human exposures of concern. In fact, in a previous study with GR we demonstrated that a constitutively active mutant of GR in which the entire ligand-binding domain had been deleted demonstrated similar effects of As as did the wild-type GR (Bodwell et al., 2004), suggesting that this domain is not involved in mediating the As effect. We also observed no effect of As alone on activation or nuclear localization of GR using a fluorescently labeled GR and confocal microscopy under conditions where As suppressed GR-mediated gene expression (Kaltreider et al., 2001).

In addition to mechanistic information, there are important human health implications for effects of As on ER function. ER has been shown to play an important role in the etiology and therapeutic responsiveness of breast, ovarian, and uterine cancers as well as many other cancers. ERα has been shown to directly mediate the growth and aggressiveness of breast cancer, and there are numerous examples of both ER agonists and antagonists demonstrating a role in either promoting or inhibiting breast cancer and other ER-dependent cancers. Thus, any chemical that can disrupt ER activities could contribute to the etiology, progression, or regression of the disease. To our knowledge, there are currently no reports of increased incidence of breast cancer or other known ER-dependent cancers in populations exposed to As in drinking water. However, this issue has not been thoroughly investigated, and so this possible link remains an open question. ER has also been shown to play a role in the normal biology and the pathobiology of other tissues in both males and females, and thus As disruption of ER signaling may contribute to other disease risks including other cancers and certain noncancer diseases such as heart disease. For example, a recent study by Waalkes and coworkers demonstrated that inorganic As exposure in utero in mice increased the incidence of carcinogenesis in the adult offspring and that there was hypomethylation of the ERα receptor promoter in these in utero–exposed mice, leading to overexpression of ERα (Chen et al., 2004; Liu et al., 2006a,b; Shen et al., 2006). ER is known to play an important role in normal liver development and function and has also been previously shown to play a role in liver carcinogenesis in experimental systems. Since As exposure in drinking water has been associated with increased risk of liver cancer, this raises the possibility that As effects on ER could play a role in this disease incidence. Moreover, the implication in the Waalkes study that fetal As exposure might lead to ER-related imprinting that subsequently plays a role in adult disease risk further supports the need for understanding As effects on ER function.

In summary, these results indicate that As is a potent endocrine disruptor of ER-mediated gene regulation, strongly suppressing ER-mediated, E2-stimulated gene expression in chick embryo liver in vivo and showing similar effects in a cell culture model system, the ER-positive human breast cancer MCF-7 cell line. The suppression of ER-mediated gene transcriptional activation is similar to what was observed previously for the other four SRs for GRs, MRs, AR, and PR. Given the critical role of these receptors in normal biology as well as many pathophysiological processes, this may explain, at least in part, how chronic exposure to As in drinking water has been associated with a large and continually growing list of serious and fatal illnesses including many different cancers, vascular and cardiovascular diseases, diabetes, reproductive and developmental problems, and neurological and cognitive problems. An important role of one or more hormone receptors has been shown or implicated for each of these diseases. But it will be important to determine, for each specific disease process, the precise causal role that As disruption of these receptors and the pathways they regulate may play in their overall etiology.

This work was supported by National Institutes of Health–National Institute of Environmental Health Sciences grants R01 ES011819 (J.E.B.) and P42 ES07373 (J.W.H., Superfund Basic Research Program [SBRP] Project, Project 2). J.A.G. was supported by a fellowship from P42 ES07373 (SBRP, Training Core). We gratefully acknowledge the assistance of Dr Brian Jackson and Ms Angela LaCroix-Fralish of the Dartmouth TEA Core Facility for their assistance in the trace analysis of our samples. The TEA Core is partially supported by P42 ES07373 (SBRP, Core B) and by an instrument grant from National Science Foundation (MRI-0215913).

References