Toxicology and Applied Pharmacology 250 (2011) 117–129

Contents lists available at ScienceDirect

Toxicology and Applied Pharmacology

j ourna l homepage: www.e lsev ie r.com/ locate /ytaap

Arsenic- and cadmium-induced toxicogenomic response in mouse embryosundergoing neurulation

Joshua F. Robinson a,b, Xiaozhong Yu a,b, Estefania G. Moreira b,c, Sungwoo Hong a,b, Elaine M. Faustman a,b,d,e,f,⁎a Department of Environmental and Occupational Health Sciences, University of Washington, Seattle, WA 98195, USAb Institute for Risk Analysis and Risk Communication, Seattle, WA 98195, USAc Department of Physiological Sciences, State University of Londrina, Londrina, PR, Brazild Center for Ecogenetics and Environmental Health, Seattle, WA 98195, USAe Center on Human Development and Disability, Seattle, WA 98195, USAf Center for Child Environmental Health Risks Research, Seattle, WA 98195, USA

⁎ Corresponding author. Department of EnvironmeSciences, 4225 Roosevelt Way NE Suite 100, Universit98105, USA. Fax: +1 206 685 4696.

E-mail address: faustman@u.washington.edu (E.M. F

0041-008X/$ – see front matter © 2010 Elsevier Inc. Aldoi:10.1016/j.taap.2010.09.018

a b s t r a c t

a r t i c l e i n f o

Article history:Received 30 November 2009Revised 20 September 2010Accepted 22 September 2010Available online 29 September 2010

Keywords:ArsenicCadmiumMetabolismCell cycleToxicogenomicsNTD

Arsenic (As) and cadmium (Cd) are well-characterized teratogens in animal models inducing embryotoxicityand neural tube defects (NTDs) when exposed during neurulation. Toxicological research is needed to resolvethe specific biological processes and associatedmolecular pathways underlying metal-induced toxicity duringthis timeframe in gestational development. In this study, we investigated the dose-dependent effects of Asand Cd on gene expression in C57BL/6J mouse embryos exposed in utero during neurulation (GD8) to identifysignificantly altered genes and corresponding biological processes associated with embryotoxicity. Wequantitatively examined the toxicogenomic dose–response relationship at the gene level. Our results suggestthat As and Cd induce dose-dependent gene expression alterations representing shared (cell cycle, responseto UV, glutathione metabolism, RNA processing) and unique (alcohol/sugar metabolism) biological processes,which serve as robust indicators of metal-induced developmental toxicity and indicate underlyingembryotoxic effects. Our observations also correlate well with previously identified impacts of As and Cdon specific genes associated with metal-induced toxicity (Cdkn1a, Mt1). In summary, we have identified in aquantitative manner As and Cd induced dose-dependent effects on gene expression in mouse embryos duringa peak window of sensitivity to embryotoxicity and NTDs in the sensitive C57BL/6J strain.

ntal and Occupational Healthy of Washington, Seattle, WA

austman).

l rights reserved.

© 2010 Elsevier Inc. All rights reserved.

Introduction

Derived from natural and anthropogenic sources, widespreadenvironmental exposure to arsenic (As) and cadmium (Cd) remains ofpublic health concern due to their potential to cause adverse effects inthe human population. In animal models, As and Cd are well-characterized teratogens inducing embryotoxicity, including growtheffects, mortality and a range of congenital malformations (Ferm andCarpenter, 1967; Ferm, 1977; Salvatori et al., 2004). Exposures toeither As or Cd during neurulation result in neural tube defect (NTD)formation, dependent on several factors including, dose, timing ofexposure, strain, exposure route, ionic state (As), and nutritionalstatus (Nakashima et al., 1987; Hovland et al., 1999; Machado et al.,1999). In the C57BL/6J mouse strain and related substrains, a generaldose–response relationship has been established between As, Cd, andembryotoxicity using intraperitoneal (primarily) or subcutaneousdosing during this window in gestational development. Maternal

exposure to 5 mg/kg As (As3+, ip, GD8, C57BL/6J) induces modestincreases in malformations but not NTDs, while 10 mg/kg associateswith increased resorptions and NTD (exencephaly, 17%) formation(Machado et al., 1999). For Cd (C57BL/6J, GD8 or GD9, ip, acute) dosesranging from 0.5 to 1 mg/kg are relatively nontoxic while doses≥2 mg/kg induce developmental malformations (Hovland et al.,1999; Lutz and Beck, 2000; Lee et al., 2006; Robinson et al., 2009).For example, maternal exposure to 4 mg/kg Cd (ip, GD8.0) inducesexencephaly in 33% of C57Bl/6J fetuses (Hovland et al., 1999;Robinson et al., 2009). As and Cd pass across the placenta anddistribute to the embryo dependent on dose, time of observation, androute of administration (Dencker, 1975; Christley andWebster, 1983;Lindgren et al., 1984; Hanlon and Ferm, 1986; Hood et al., 1988), andthey have been identified in the neuroepithelium of the neural tube aswell as other tissues including the limb buds and gut (Christley andWebster, 1983; Lindgren et al., 1984).

Embryonic uptake of As and/or Cd results in changes at themorphological, cellular, and molecular levels in a time-dependentmanner. Cd studies (GD8.0, Cd 4 mg/kg) suggest that peak cellularmorphological changes (increased pyknotic nuclei) occur 10–12 h,correlating with subsequent alterations in neural tube development(Webster and Messerle, 1980). Under these conditions, As and Cd

118 J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

exposures can affect DNA damage, cell cycle arrest, oxidative stress,and cell death pathways (Pulido and Parrish, 2003; Beyersmann andHartwig, 2008), yet the means by which these metals interfere withsignaling pathways and subsequent morphogenesis remains unre-solved. Investigations suggest additional effects such as glucoseimpairment and disruption in ion (e.g., Zn) or folic acid regulationmay also be potential mechanisms of metal-induced embryotoxicity(Wlodarczyk et al., 2001; Fernandez et al., 2004, 2007; Hill et al., 2009;Robinson et al., 2009, 2010b).

Toxicogenomic approaches provide researchers with the ability tosimultaneously examine multiple biological processes or pathwaysacross dose and time, and thus, these approaches hold promise tomakecross comparisons between variables of interest (e.g., teratogens, time,dose, strain). Dose– and time–response assessments are needed todefine the mechanisms of action for both of these metals in order todevelop sensitive indicators of toxicity as well as characterize sharedand unique mechanisms of embryotoxicity. Emerging bioinformatictools, such as GO-Quant, allow assessment of toxicogenomic data in thecontext of risk assessment, enabling dose– and time–response calcula-tionswithin subsets of genes associatedwith specific functional families(i.e., Gene Ontologies) (Yu et al., 2006). While debate still remains overthe application of toxicogenomic data in the risk assessment arena(Freeman, 2004), examination of toxicogenomic data using tools, likeGO-Quant, can assist in determining robust and sensitive responses inrelation to environmental teratogens and developmental toxicity.

Building upon previous studies using a similar experimentaldesign in terms of exposure and gestational timing (acute exposureon GD8.0, ip, C57BL/6J strain), we examined the dose-dependenteffects of As and Cd on gene expression in association with increasedembryotoxicity in C57BL/6J mouse embryos undergoing neurulation.In these preliminary assessments, we identified overlapping andnonoverlapping metal-induced gene expression alterations, whichassociate with embryotoxicity and NTD development in vivo.

Methods

Animals and metal exposure. Following procedures detailed in ourprevious studies (Robinson et al., 2009), colonies of C57BL/6J (JacksonLabs, Bar Harbor, MA) were maintained at the University of Washington,Department of Environmental and Occupational Health Sciences, housedin filter-covered transparent plastic cages with water and food availablead libitum. Mice were bred in climate-controlled rooms under analternating 12-h light/dark cycle. Timed matings were generated byplacing single male mice into cages containing two females in the lateafternoon. C57BL/6J females used for mating were aged between 3 and5 months and weighed between 17 and 24 g. Copulatory plugs wereidentified in the early morning (8:00 am±1 h) the following day anddesignated as gestational day (GD) 0. Pregnant mice were administeredsingledoses via intraperitoneal injection (ip)onGD8.0 at 8:00am(±1 h)witheither sodiumarsenite (2.5, 5, or10 mg/kgBW;AlfaAesar,WardHill,MA) or cadmium chloride (0.5, 2, or 4 mg/kg/BW; Alfa Aesar, Ward Hill,MA) dissolved in deionized water or vehicle control (water, 10ul/g).Exposure groups were prearranged before mating. The administrationtime (GD8) and methodology (intraperitoneal injection) were selecteddue to previous dose–response studies assessing developmental toxicityin C57BL/6J mouse embryos with the primary end point of interest beingthe formation of NTDs (exencephaly) (Hovland et al., 1999; Machadoet al., 1999; Robinson et al., 2009). GD8 represents the optimal day ofexposure regarding the aims of this study due to factors including highNTD sensitivity, a relatively low resorption rate (less sensitive than earliertime points in neurulation (bGD8)) and the ability to acquire enoughembryonic material for gene expression analyses.

Pregnant C57BL/6J females were euthanized on 0, 8, and 12 h afterinjection (GD8.0). Contemporaneous controls were performed for thethree developmental times investigated (GD8+0, 8, and 12 h) for asample of the collection dates. However, it was not possible to have

contemporaneous controls for all combinations of developmental timeson all collection dates due to failed pregnancies and time constraintsregarding the extraction of the embryos. Whole embryos were isolatedfrom the uterus. All extraembryonicmembranes, including the yolk sac,were removed. Embryos were quickly examined for intended develop-mental stage of interest (GD8=0 h, Thelier stage (TS) ~12;GD8.3=8 h,TS ~12.5; andGD8.5=12 h, TS ~13). Litters observed to be the incorrectstage (N0.5 TS) due to suspected earlier mating were not used for thisstudy. Embryoswerewashed in coldCMF–PBS,placed in liquidnitrogen,and stored at −80 °C. Complete pooled litters were kept separate.

RNA isolation. Embryos were placed in 500 μl of RTL Cell LysisBuffer (Qiagen, Valencia, CA) and lysed with a 30-gauge needle tohomogenize the tissue. RNA was purified using the RNAeasy kit(Qiagen, Valencia, CA). Quality and quantity were assessed usingNanodrop ND-100 spectrophotometer (Nanodrop Technologies,Wilmington, DE) and the “6000” assay on the 2100 Bioanalyzer(Agilent Technologies, Palo Alto, CA). The quality of RNA samples wasassessed based on absorbance values ranging from 1.9 to 2.1 (A260/A240), with distinct peaks of 18S and 28S rRNA subunits and profilesshowing minimal degradation. As displayed in Supplemental Table 1,exposed and nonexposed litters were collected and used for geneexpression analysis at the following time points: 0 h (GD8.0) control(n=3); 8 h (GD8.0+8 h), control, 2.5, 5, or 10 mg/kg sodiumarsenite and 0.5, 2, or 4 mg/kg cadmium chloride (n=4, except0.5 mg/kg Cd, n=3); 12 h (GD8.0+12 h) control (n=5), 10 mg/kgsodium arsenite and 4 mg/kg cadmium chloride (n=4).

Oligonucleotide microarrays. Microarray hybridizations were com-pleted using the oligonucleotide-based Affymetrix platform (Mouse 4302.0), with 1 μg of total RNA from each litter. Prior labelingwas completedwith the GeneChip One-Cycle Target Labeling kit (Affymetrix, Inc., SantaClara, CA) following themanufacturer's protocol. The amount and qualityof the cRNA was assessed using a Nanodrop ND-100 spectrophotometerand an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA). cRNAwas fragmented and hybridized to the arrays (Affymetrix, Inc., SantaClara, CA) for 17 h prior to washing and scanning.

Data processing. Intensity values were extracted from scannedimages using GeneChip Operating Software (Affymetrix, Inc., SantaClara, CA) for all 45101 probes. Raw intensities were normalized usingGCRMA and transformed by log base 2 (BRBarraytools, NCI) (Simonet al., 2007).

Identification of significantly altered genes due to metal exposure.Fig. 1 represents the statistical approach used in this study to exploredose- and time-dependent effects of As and Cd exposure on geneexpression and corresponding GO biological processes. To identify Asand Cd dose-dependent alterations, we employed two discrete linearmodels to assess genes that were altered by dose (As or Cd) or changingacross time. Model 1 assessed the effect of As on gene expression(BDose_Arsenic) and the effect of time (BTime). Model 2 assessed the effectof Cd on gene expression (BDose_Cadmium) and the effect of time (BTime).Two separatemodels were created due to initial statistical assessments,suggesting differential distributions of significant effects between Asand Cd. Both dose and time effects were examined in our model due totheir relevance and significance.

Model 1: Log2[Expn]C57BL/6J=BDose_Arsenicx1+BTimex2Model 2: Log2[Expn]C57BL/6J=BDose_Cadmiumx1+BTimex2

[BDose_ArsenicX1] Coding: “0,” control; “1,” 2.5 mg/kg; “2,” 5 mg/kg;“3,” 10 mg/kg As

[BDose_CadmiumX1] Coding: “0,” control; “1,” 0.5 mg/kg; “2,” 2 mg/kg;“3,” 4 mg/kg Cd

[BTimex2] Coding: “0,” 0 h; “1,” 8 h; “2,” 12 h

Fig. 1. Analysis of genes altered by arsenic and cadmium. This flow diagram represents the statistical approach used to explore dose- and time-dependent effects of As and Cdexposure on gene expression and corresponding GO biological processes. Linear models assessing for dose (As or Cd) and time effects were used to identify genes that weresignificantly altered by metal exposure (ANOVA, BDose, pb0.0001). Enrichment analyses of GO biological processes were conducted for all genes significantly differentially expresseddue to As or Cd exposure to identify biological processes altered by metals. Quantitative analysis of enriched GO biological processes was employed using GO-Quant, enabling dose-and time-dependent calculations in genes within GO subsets. In addition, K-means cluster gene analysis was used to assess unique dose–response relationships due to As (or Cd)exposure and corresponding enrichment of GO biological processes. We assessed the 116 genes altered by both As and Cd for unique dose–response relationships using K-meanscluster gene analysis, and we determined the quantitative relationship within these clusters across dose and time (average FC). Enrichment of transcription factor binding sites andassociated GO biological processes were also evaluated.

119J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

We determined significant As or Cd effects using a cutoff value ofsignificance (pb0.0001) for the primary dose effect (BDose_Arsenic orBDose_Cadmium) within Models 1 and 2 to assess significant metaleffects across dose over time (Simon et al., 2007). Adjusted falsediscovery rates were below 1% for both gene subsets for As or Cd(Benjamini and Hochberg, 1995). We established our cutoff values toidentify roughly 2%–4% of the probes to be differentially expressedacross dose and time. Approximately 10%–20% of all significantlyaltered genes were represented by more than one probe. Due to therelatively small proportion of genes represented by more than oneprobe, we herein use the term “genes” instead of probes to enhanceease of readership. Redundant probes were assessed separately, andcorresponding fold change (FC) values were not averaged. Post-hocanalyses were conducted to examine the dose relationship betweeneach dose group and its comparative concurrent control using aStudent t-test (pb0.05) for all significantly altered genes (As or Cd,pb0.0001). To assess similarities in gene expression changes acrossdose and time for both metals, we conducted principal componentsanalysis (PCA) (Genemaths XT, Applied Maths, Sint-Martens-Latem,Belgium) and hierarchical clustering analysis (Eisen et al., 1998).PCA plots were constructed using Ward linkage and Euclideandistance to assess similarities between each exposure group(average). PCA was conducted in a supervised manner (all genesimpacted by As and/or Cd, pb0.0001) and across all 45,101 probes.Hierarchical clustering plots were performed by calculating ratios(log 2), between gene intensities on each single array versus theaverage intensity for each gene across all included exposure groups.Using average linkage and Euclidean dissimilarity methods, generelationships were clustered to show commonalities and differencesbetween single arrays.

Enrichment analysis of GO biological processes. We conducted GeneOntology (GO) analysis (Doniger et al., 2003) to identify enriched GOgene categories (classifiedbybiological process,molecular function, and

cellular component) within genes identified to be significantly alteredby As or Cd exposure (pb0.0001). Since our aims in this preliminarystudy were to explore overlapping and nonoverlapping impacts ofmetals on biological processes and relative molecular pathways, GOanalyseswithin this article focused primarily onGO categories classifiedby biological process which includes broad biological processes (e.g.,developmental process,metabolic process) andmoredefined categoriesrepresenting molecular pathways of interest (e.g., Wnt signalingpathway). Separate analyses were conducted for each metal. Signifi-cantly enriched GO categories were based on a set criteria ofpermutation value (pb0.05), Z-score (N2), and genes changed withineach specific GO gene category (N2). A ranking value (GO enrichmentscore) was calculated using −log(permutation p value)×Z score tocompare the enrichment of GO biological processes between As and Cd.

Quantitative analysis of enriched GO biological processes. To quan-titatively evaluate changes within gene expression-linked GO genecategories, we used GO-Quant (Yu et al., 2006) to calculate theabsolute average FC associated within each GO subset. Enriched GOgene categories were analyzed across dose and time.

Cluster gene analysis to examine dose–response relationship at the geneand GO levels. K-means cluster gene analysis was conducted toidentify similar expression patterns amongst significantly alteredgenes (TIGR MEV, average linkage, default parameters) (Saeed et al.,2003). Using genes identified to be significantly altered by As or Cd(pb0.0001), we began our analyses by clustering gene expressionpatterns across dose and time into 10 dose–response relationships tovisualize similar unique relationships. Clusters were reduced if wedetermined similar directionality in dose–response patterns betweenclusters. Enrichment analysis of GO biological processes wasconducted using both MappFinder (same criteria as above) andDAVID (≥3 genes, pb0.05) (Dennis et al., 2003). For analysescompleted using DAVID GO biological processes are separated by

120 J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

“levels” within the GO hierarchical system (1–5 levels, with 1representing most broad and 5 representing high specificity), withbiological level 4 representing categories with moderate-highspecificity within the GO hierarchy tree. GO analysis was conductedacross all levels; however, only GO categories identified withinDAVID's classification “biological level 4” were displayed to reduceredundancy of similar GO terms. We used the combination ofMappFinder and DAVID to improve GO assessments due to differencesin GO databases and the lack of linkage usingMappFinder in particularclusters. Using the GO-Quant methodology to quantitatively assessgene responses within each cluster, we calculated the average FCassociated within genes of each cluster across dose and time.

Enrichment analysis of transcription factor binding sites withinoverlapping metal-induced alterations. Similar to above, K-meanscluster gene analysis was conducted to identify similar expressionpatterns within genes significantly altered by both As and Cd(pb0.0001) across dose and time (TIGR MEV). Overrepresentedtranscription factor binding sites (TFBS) within genes identified to besignificantly altered by both As and Cd were identified usingoPPOSSUM (default criteria) (Ho Sui et al., 2005). Approximately70% of these genes were linkedwith the TFBS database. All TFBSwith aZN2.5 represented by more than 5 genes were selected as enriched. Inaddition, TFBS analysis was conducted within each cluster (I–III) toidentify if overrepresented TFBS were enriched within each cluster.Associations with selected GO biological processes were determinedusing MappFinder and DAVID.

Fig. 2. Dose-dependent metal-induced gene expression alterations. (A) The Venn diagramrespectively (F-test, pb0.0001), and overlapping genes intersecting between these twosimilarities in gene expression alterations between exposure groups across all genes altered(D) (F-test, pb0.0001), post-hoc analyses (t-test) were completed to identify the total numrespective concurrent control (8 h or 12 h). The percentages of genes differentially expressedbe significantly altered across all dose groups (black bar). In addition, the proportion of up

Results

As and Cd induce gene expression alterations in a dose-dependentmanner

As shown in Fig. 2A, we identified 1960 and 775 genes to besignificantly altered by As and Cd, respectively, across dose and time,with 116 of these genes identified to be significantly altered by bothAs and Cd (ANOVA, pb0.0001). Using PCA, we observed a dose-dependent relationship in the distance between exposure groups ofAs and Cd (8 h) in comparison with the 8 h control within genesaltered by As and/or Cd (Fig. 2B). Furthermore, we identified highdose groups of As (10 mg/kg) and Cd (4 mg/kg) to display similarextreme differences from control (8 or 12 h), irrespective of time.Within genes identified to be significantly altered by As or Cd, weidentified a monotonic dose-dependent relationship in the totalnumber of genes altered by metal exposure, specifically at 8 h (Figs.2C and D). With As (Fig. 2C), we identified a relatively small increase(6%) in the amount of genes altered at 8 h between 2.5 and 5 mg/kgcompared to the dramatic 84% increase in the amount of genesaltered at 8 h between 5 and 10 mg/kg. At 8 and 12 h, we observed asimilar response in the amount of genes altered and the percentageof up- and downregulated genes with 10 mg/kg As. With Cd at 8 h(Fig. 2D), we observed a 27% increase in the amount of genessignificantly altered between dose groups 0.5 and 2 mg/kg and 2 and4 mg/kg. With 4 mg/kg Cd, we observed a 12% increase in theamount of probes altered at 12 h compared to that at 8 h. At 8 h, with

illustrates 1960 and 775 genes identified to be significantly altered by As and Cd,populations (116 total). Using principal components analysis (PCA), we determinedby As and/or Cd (B). Within genes identified to be significantly altered by As (C) and Cdber of differentially expressed genes with each dose group in comparison with their(t-test, pb0.05) are displayed in comparison to the total number of genes identified to(red) and down (green)-regulated genes are indicated by color.

121J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

0.5 and 2 mg/kg Cd, a higher proportion of genes were identified tobe upregulated than downregulated. In contrast, with 4 mg/kg Cd,more genes were identified to be downregulated (8 h and 12 h).Comparing both concentrations, which were chosen since they havepreviously induced NTDs (high dose), we observed differentialproportions of up/down regulated genes, whereas with 10 mg/kgAs, we observed 54% and 55% of altered genes to be upregulated (8 hand 12 h) and with 4 mg/kg Cd, we observed only 14% and 18% ofaltered genes to be upregulated.

As and Cd impact gene expression-linked GO biological processes

Conducting enrichment analysis of GO biological processes withingenes found to be significantly altered by either As or Cd, we identified172 biological processes to be overrepresented, including 25biological processes identified to be enriched with both As and Cd(Fig. 3). Both As and Cd (quadrant I) significantly impacted genesenriched for GO biological processes related to general metabolism(e.g., cellular metabolic process), transcriptional and translationalactivities (e.g., RNA processing, translation), mitochondrial organiza-tion, and cell cycle and environmental stress response-related

Fig. 3. Enrichment of GO biological processes altered by As and Cd. Using a cross scatter-plot,by As and Cd. Genes significantly impacted by As and Cd (pb0.0001), were separately analybiological processes, we identified 159 (As) and 38 (Cd) GO biological processes meeting ourthe −log (p value) and Z score, we obtained a measure of ranking the significance of GO catminimum criteria to be identified as significant (~2.6). Symbols are differentiated by Genrichments scores N2.6 for both As and Cd (quadrant I). Enriched categories unique to As

categories (e.g., cell cycle regulation, glutathione metabolic process,response to UV) (GO enrichment score N2.6). In addition, specificbiological processes were uniquely enrichedwith As (quadrant II) andCd (quadrant IV), including alcohol, sugar, and sterol metabolism aswell as biopolymer modification, identified to be enriched by only Asand additional environmental stress-related biological processes (e.g.,apoptotic mitochondrial changes with As or response to DNA damagestimulus with Cd).

As and Cd impacts enriched GO biological processes in a dose- andtime-dependent manner

In Fig. 4, we quantitatively describe the effect of As (A) or Cd (B) onselected enriched GO biological processes across dose and time, usingthe absolute average FC (log 2 scale) of all significantly altered genes(pb0.0001) within each GO biological process. For both As and Cd, weidentified the GO biological process “response to UV” to display thegreatest FC response at 12 h (As, 1.3, Cd, 1.5, log 2 scale). At 8 h, thiscategory was also one of the most robust, in terms of the degree of FC,showing peak effects with 10 mg/kg As and 2 mg/kg Cd. Other relatedGO biological processes previously linked with environmental stress

we illustrate overlapping and non-overlapping enriched GO biological processes alteredzed for enrichment of GO biological processes (MAPPFinder). Assessing over 4500 GOcriteria of enrichment (pb0.02, Z score N2, and≥3 genes altered). Using the product ofegories to allow direct comparisons between As and Cd. The dashed line represents theO hierarchy. Overlapping altered GO biological processes are represented as havingor Cd are identified in quadrants II and IV, respectively.

Fig. 4. Quantitative analysis of dose- and time-dependent response in enriched GO biological processes impacted by As and/or Cd. Selected enriched GO-based biological processes/pathways were assessed across dose and time using GO-Quant. We quantitatively describe genes across dose and time within each GO category by showing the absolute average FCof impact for As (A) or Cd (B). The absolute average FC associated with all genes impacted by As or Cd (pb0.0001) is also included to make comparisons with all genes impacted.Overlapping enriched GO biological processes are labeled by the same symbol and color between panels A and B.

122 J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

(glutathione metabolic process, regulation of cell cycle) showed lessrobust, but similar dose–response relationships for both As and Cd.Enriched GO biological processes, such as translation or RNAprocessing, showed smaller changes in comparison with environ-mental stress-related biological processes as well as the average of allgene FC values with As or Cd (Figs. 4A and B). Specifically, with Asexposure (Fig. 4A), uniquely enriched processes related to sterol and

Fig. 5. Enriched GO biological processes altered by As dependent on the dose–response relatioalteredbyAs(pb0.0001) (TIGRMEV). Tworelationshipsweredetermined, clusters I and II. Sepaprocess level 4) determined using both MappFinder (M) and/or DAVID (D) are shown. Enricheminimumof 3 genes forMappFinder and pb0.05, N2 genes altered for DAVID. The total numberin parentheses. The absolute average ratio of gene expression response for genes within each ecolor using TIGRMEV (Fig. 4C). Color codingwas used to differentiate betweenGObiological prand this figure).

alcohol/glucose metabolism (i.e., glycogenmetabolic process, alcoholmetabolic process, glucose metabolic process, monosaccharide met-abolic process, carbohydrate metabolic process) displayed monotonicdose-dependent changes. For example, across doses, the absoluteaverage FC for the 22 significantly altered genes (pb0.0001) related toglucose metabolic process was 0.08, 0.12, 0.73 (8 h), and 0.90 (12 h)(log 2 scale) with doses 2.5, 5, 10 (8 h), and 10 mg/kg (12 h) As,

nship. K-means cluster analysis was conducted on all 1960 genes found to be significantlyrateGOanalyseswereconducted for eachcluster. SignificantGOgenecategories (biologicald categories were determined by the following criteria: p value (b0.02), Z score N2 and aof genes alteredwithin eachGOcategory as determinedbyMappFinder orDAVID is shownnriched GO category was calculated using GO-Quant across dose and time and depicted byocess families within the GOhierarchy and tomake comparisons between As and Cd (Fig. 4

123J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

124 J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

respectively. These GO biological processes were similarly affected or,in some cases, more affected as categories reflective of environmentalstress pathways (response to UV, apoptotic mitochondrial changes,regulation of cell cycle, glutathione metabolic process, cell cycle) forAs. Due to the number of enriched GO biological processes impactedby As and Cd, only selected GO categories are presented across doseand time.

Enriched GO biological processes altered by As or Cd dependent on thedose–response relationship

To further investigate the dose–response relationship betweenmetal exposure, gene expression alterations, and associated GOcategories, we conducted K-means clustering within genes signifi-cantly altered by As or Cd (pb0.0001) (Figs. 5 and 6). We determined2 common dose–response relationships with As exposure in C57BL/6Jembryos (clusters I and II; Fig. 5A). In general, cluster I represented1112 genes upregulated with 10 mg/kg at both 8 h and 12 h, whilecluster II represented 848 genes downregulated with 10 mg/kg at 8 hand 12 h (Fig. 5A). Only minimal changes were identified with 2.5 and5 mg/kg within these two clusters on average (Fig. 5C). Common GObiological processes were identified in genes within clusters I and II,including genes involved in protein metabolism/transport (red),cellular localization/organization (purple), DNA metabolism (grey),biopolymer modification (mocha), and cell cycle regulation (lightblue) (Fig. 5B). Unique categories in cluster I represented genesinvolved in alcohol, carbohydrate, and sugar metabolism; stressresponse (e.g., glutathione metabolic process, response to oxidativestress); programmed cell death; intracellular signaling, cellularhomeostasis; and heart development. Biological processes related toRNA metabolism, sterol metabolism, and DNA repair were uniquelyidentified in cluster II.

We identified 4 unique dose–response relationships withinsignificant gene expression alterations associated with Cd exposureacross dose and time (clusters I–IV; Fig. 6). Cluster I represented 211genes upregulated with 0.5 and 2 mg/kg Cd (8 h p.i.) and down-regulatedwith 4 mg/kg at 8 h and 12 h (Fig. 6A). The 147 geneswithincluster II showed minimal changes with 0.5 mg/kg Cd and down-regulation with 2 and 4 mg/kg. Greater changes were observed with2 mg/kg (−0.5 FC, log 2 scale) compared to 4 mg/kg (−0.2 FC, log 2scale) (Fig. 6C). In parallel, but directionally opposite to observationsin cluster II, genes within cluster IV (180 genes) showed upregulation,again with peak changes occurring in the 2 mg/kg group (Figs. 6A andC). Genes within cluster III showedminimal changes with lower doses(0.5, 2 mg/kg); however, with 4 mg/kg, this group of genes (137 total)showed, on average, reduced expression by ~50% at 8 h and 12 h.Strong similarities between 8 h and 12 h for 4 mg/kg Cd wereobserved in terms of magnitude and directionality for clusters I–IV(Fig. 6C). Common enriched GO biological processes were identifiedwithin more than one cluster, including RNA metabolism (I–III,green), transcription (I and II, green), and protein metabolism/translation (II and III, red) (Fig. 6B). Unique categories includedbiopolymer modification, negative regulation of programmed celldeath, cell development, and negative regulation of neurogenesis(cluster I); mitochondrial organization and regulation of cell cycle(Cluster II); chromosomal organization, DNAmetabolism, and mitosis(Cluster III); and cytoskeleton organization, intracellular signalingcascade, glutathione metabolism, negative regulation of progressionthrough the cell cycle, and response to stress (Cluster IV).

Fig. 6. Enriched GO biological processes altered by Cd dependent on the dose–response resignificantly altered with Cd. In general, four relationships were determined, clusters I–IV.(biological process level 4) determined using both MappFinder (M) and/or DAVID (D) are sZ score N2 and a minimum of 3 genes for MappFinder and pb0.05, N2 genes altered for DAVIDdetermined by MappFinder or DAVID is shown in parentheses. Color coding was used to difcomparisons between As and Cd (Figs. 4 and 5).

Overlapping gene expression responses of arsenic and cadmium

Using K-means clustering analysis, we identified 3 uniquerelationships across dose and time within the 116 genes identifiedto be significantly altered by both As and Cd (Fig. 7). Cluster Irepresented 49 genes that were upregulated with As (5 and 10 mg/kg) and Cd (2 and 4 mg/kg) (Figs. 7A–C). These genes were linkedwith GO biological processes such as: cell cycle process (Cdk1na,Mapre1, Re8l1, and Sash1), stress response (Ercc5, Cdk1na, Pmaip1,and Ubf3b), cellular homeostasis (Gsr, Mt1, and Txndd3), tubemorphogenesis (G6pdx and Fzd6) and transport (Slc19a2,26100029101Rik, Npc2, Stau2, Txndc13, and Fzd6). Genes withincluster III displayed a relationship opposite to cluster I; in general,genes were downregulated with As (5 and 10 mg/kg) and Cd (2 and4 mg/kg) (Figs. 7B and C) and included genes (55 total) linked withcell cycle process (Hus1 and Cdk9) and RNA processing(672048F09Rik, Gtpb3, Tyw3, Tsen2, and Dlmt1) (Fig. 7A). The 12genes represented in cluster II displayed similarities in response withcluster I, upregulated response with 2.5, 5, and 10 mg/kg As and 0.5and 2 mg/kg Cd; however, they showed striking dissimilarities with4 mg/kg Cd (downregulated versus upregulated) at both 8 h and 12 h.

Enrichment of transcription factor binding sites within As and Cdcommonly impacted genes

Within the 116 genes identified to be significantly altered by bothAs and Cd., we identified enrichment of 7 TFBS, including Foxa2(n=35 genes), Nfkb1 (n=13 genes), Nk2-5 (n=54 genes), Nr2f1(n=21 genes), Pax5 (n=10 genes), Sox9 (n=46 genes), and Stat1(n=11 genes) (Fig. 7). Upon further analysis, overrepresentation ofspecific TFBS was identified to be cluster-dependent. For example,Foxa2, Nk2-5, and Stat1 were found to be overrepresented in onlycluster I.

Discussion

In this study, using a toxicogenomic approach, we investigated thedose–response relationship associatedwith either As or Cd in exposedembryos during the neurulation period of development. We exam-ined overlapping and nonoverlapping GO biological processesimpacted by each metal across dose using a combination of data-driven methodologies, data mining, and bioinformatic resources. Wediscuss the effects of As and Cd on gene expression response acrossdose and time in context with the dose–response relationshipbetween metals and developmental toxicity and NTD formation assummarized in Supplemental Table 2.

Dose-dependent gene expression alterations in relation to thedose–response relationship of metals and developmental toxicity

A collection of developmental toxicity studies investigating theeffects of As and Cd on neurulation in mouse have established a dose–response relationship between As, Cd and embryotoxicity,with doses ofAs (~5 mg/kg) and Cd (~1 mg/kg) producing minimal embryotoxicityand doses of As (10 mg/kg) and Cd (4 mg/kg) inducing NTDs, otherstructural defects, growth alterations, and increasedmortality (Hovlandet al., 1999; Machado et al., 1999; Lutz and Beck, 2000; Lee et al., 2006;Robinson et al., 2009, 2010b). Using a similar experimental design interms of dose, administration of exposure, strain, and gestational timing

lationship. K-means cluster analysis was conducted on all 775 genes identified to beSeparate GO analyses were conducted for each cluster. Significant GO gene categorieshown. Enriched categories were determined by the following criteria: p value (b0.02),as indicated by letters M or D. The number of genes altered within each GO category as

ferentiate between GO biological process families within the GO hierarchy and to make

125J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

126 J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

as the aforementioned studies, in this study, we observed a monotonicdose–response relationship, with increased alterations in gene expres-sion (at 8 h) coinciding with increased dose (Fig. 2). Our study suggeststhat at this time point increased alterations in gene expression is ameasure of metal-induced response corresponding with increaseddevelopmental toxicity that precedesmorphological effects determinedlater in gestation.

Arsenic and cadmium induce environmental stress response-relatedimpacts in association with developmental toxicity

In general toxicological studies, four common predominant me-chanisms have been proposed regarding metal toxicity: 1) alterationsin cell proliferation pathways, 2) inhibition of major DNA repairsystems, 3) interference with cellular redox regulation/generation ofoxidative stress, and 4) induction of apoptosis or cell death (Pulido andParrish, 2003; Beyersmann and Hartwig, 2008). In this first attempttoward understanding the toxicogenomic dose–response relationshipbetween metals (As and Cd) and developmental toxicity during theneurulation window of sensitivity, we provide evidence that thesepathways are disrupted in the developing embryo during neural tubedevelopment. This disruption appears to be dose-dependent at thegene and GO biological process level in relation to developmentaltoxicity (Figs. 2–7). In addition, these results are supported by initialstudies assessing mechanisms of metal-induced developmentaltoxicity during neurulation as discussed below (Wlodarczyk et al.,1996; Wlodraczyk et al., 1996; Fernandez et al., 2003; Kultima et al.,2006; Paniagua-Castro et al., 2007; Robinson et al., 2010a,b,c).

In this study, enriched GO categories representative of biologicalprocesses controlling cell cycle regulation (i.e., cell cycle, cell cycleprocess, regulation of cell cycle) were observed with both As and Cd(Fig. 3). Metals impacted cell cycle-related genes in a dose-dependent manner (Fig. 4) and bidirectionally (Fig. 5, clusters Iand II, and Fig. 6, clusters I and IV). Specific cell cycle genes, such asCdkn1a(↑), Mapre(↑), Re9L1(↑), Sash1(↑), Cdk9(↓), and Hus1(↓),were identified to be altered by both As and Cd (Fig. 7). These genesreflect different mechanistic aspects of cell cycle control. Forexample, in response to stress, the p53 downstream mediatorCdkn1a inhibits cell cycle progression between G1–S and G2–Mcheckpoints by interacting with cdk/cyclin complexes (Kuribayashiand El-Deiry, 2008). Studies assessing metal exposure across adiverse array of in vivo- and in vitro-based models suggest As and Cdto induce RNA expression of Cdkn1a and p53 as well as other p53downstream mediators in association with increased cell cycleinhibition (Yu et al., 2008, 2010). Furthermore, previous RNA (RT–PCR and microarray) and protein expression studies have identifiedCdkn1a to be induced by Cd associated with increased developmen-tal toxicity and NTD formation (Wlodarczyk et al., 1996; Fernandezet al., 2003; Robinson et al., 2009, 2010b). Therefore, this studyprovides further evidence of metal-induced alterations in cell cyclecontrol as a mechanism of toxicity and suggests disruption of cellcycle regulation to be associated with increased developmentaltoxicity in embryos undergoing neurulation.

In addition, we observed overlapping effects between As and Cd ongenes associated with the GO biological process term Response to UV(Fig. 3), as well as enrichment of related GO categories DNA repair(Cd, Fig. 3; As, Fig. 5) and response to DNA Damage (Cd, Fig. 3). Genessignificantly altered with As or Cd (pb0.0001) involved in thesecategories showed robust changes in comparison with other enriched

Fig. 7. Dose-dependent gene expression alterations altered by both As and Cd. Genes idenanalysis across dose and time. Three dose–response relationships were identified (I–III) (A).score N2.5,≥10 genes represented) were selected as enriched (*). Approximately 70% of comconducted within each cluster (I–III) to identify if overrepresented TFBS were enriched wiusing DAVID. All enriched TFBS and GO associations are displayed for each gene directly to

GO biological processes for both metals (Fig. 5), suggesting that thesegenes are sensitive indicators of metal-induced response in associa-tion with developmental toxicity. Genes within the GO categorytesponse to UV, impacted by As and/or Cd, included genes involved inDNA repair (Ercc5, Xpa, and Rev1), cell cycle arrest (Cdkn1a andHus1), and apoptosis (Pmaip1). Shared dose-dependent increases ingenes such as Ercc5, previously characterized to be a primarycomponent of nucleotide excision repair due to DNA damage(Castorena-Torres et al., 2008), suggest that both of these metalsmay disrupt DNA processes in the developing embryo in associationwith increased toxicity. Furthermore, gene expression responsesrelative to DNA damage response in this study correlate and supportprevious classical assessments of DNA damage (e.g., Comet assay,TUNEL assay) caused by As and Cd during embryonic development(Fernandez et al., 2003; Unis et al., 2009).

Initial studies suggest that exposure to metals during embryonicdevelopment results in increased oxidative stress (Paniagua-Castroet al., 2007) and RNA changes in oxidative stress-related genes such asmetallothioneins and Hmox1 (Kultima et al., 2006; Robinson et al.,2010b).While GOanalyses performedof all genes specifically altered byAs or Cd (Fig. 3) did not reveal enrichment of GO biological processes,such as “response to stress,” enrichment analysis in clusters of genesspecifically displaying upregulated response showed overrepresenta-tion of stress response-related associations (cluster I, As and cluster IV,Cd). In addition, GO categories, such as glutathione metabolism wereidentified to be enriched within these clusters (Figs. 5 and 6). At thegene level, specific targets such as Mt1, responsible in maintainingcellular homeostasis with known roles in ROS detoxification, wereobserved to be upregulated by As and Cd in a monotonic dose-dependent manner at 8 h (Fig. 7). Epidemiological and animal-basedstudies suggest Mt1 to be a biomarker of human As and Cd exposure(Chang et al., 2006; Liu et al., 2007). Furthermore, Mt1 RNA expressioncorrelates with embryonic Cd uptake in exposedmouse embryos and isan indicator of embryonic sensitivity to Cd-induced developmentaltoxicity (Robinson et al., 2010b, c). In this study, we provide dose–response evidence regarding the upregulation of Mt1 as well as otherstress-related genes due to metal exposure in developing mouseembryos undergoing neurulation.

In addition, we observed that both As and Cd alter genes related toapoptosis (programmed cell death) in a dose-dependent manner(Figs. 5–7). Genes related to apoptosis (46 total) were shown to beupregulated with As (primarily with 10 mg/kg) (cluster I, Fig. 5B).WithCd, significant alterations in apoptosis genes (12 total), including 7genes associatedwith negative regulation (Son, Htt, Mapk8, Polb, Birc6,Mapk8ip1, and Rtn4) were evident in cluster I (Fig. 6B). Interestingly,these geneswere upregulatedwith 0.5 and 2 mg/kg and downregulatedwith 4 mg/kg Cd (Fig. 6C). The non-monotonic nature of the dose–response relationship within this subset of genes may implicatediffering mechanisms associated with exposure levels of Cd which door do not induce developmental toxicity (e.g., NTDs). Upregulation ofgenes within this subset suggests upregulation of processes that inhibitcell death at nontoxic doses. For example, Birc6, whichwas upregulatedwithin this subset,waspreviously identified to inhibit caspase activationand subsequent apoptotic signaling (Bartke et al., 2004) with doses of0.5 and 2 mg/kg which do not induce NTDs. In comparison, down-regulation of Birc6 with 4 mg/kg Cd suggests a shift in regulation ofapoptotic gene expression in promoting apoptosis in association withincreased NTD formation. These observations correlate with previousstudies assessing high dose effects of Cd (C57 and GD8, 4 mg/kg) using

tified to be altered by As and Cd (pb0.0001) were analyzed using K-means clusteringTranscription factor binding motifs (TFBS) were identified using oPPOSUM (B). TFBS (Zmonly altered genes were linked with the TFBS database. In addition, TFBS analysis wasthin each cluster. Associations with selected GO biological processes were determinedthe right of its name.

127J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

128 J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

similar experimental design indicating morphological and molecularapoptotic events (i.e., increased cleaved caspase3) to occur in theneuraltube of Cd-exposed embryos during the neurulation period (Websterand Messerle, 1980; Fernandez et al., 2003; Robinson et al., 2009).Furthermore, Robinson et al. (2010b) suggest that the expression ofcleaved caspase 3 at 12 h associateswith increasedNTD incidence in Cd-exposed sensitive (C57BL/6J) and resistant (SWV) embryos (Robinsonet al., 2010b). In this study, we provide preliminary context of dose–response-related mechanisms of metal-induced apoptosis in wholemouse embryos in support of previous investigations.

Glucose and alcohol metabolism pathways as a unique mechanism ofAs-induced developmental toxicity

Uniquely in this study, we observed genes involved in sugarmetabolism-related categories (glycolysis, and glucose, alcohol,pyruvate, and monosaccharide metabolism-related categories) to bedisrupted by As exposure in a dose-dependent manner (Figs. 3–5).These GO biological processes represented some of the most robustchanges, in terms of magnitude of impact (Fig. 4). For example, theabsolute average FC of 22 genes linked with the GO biological processglucose metabolic process was similar to categories previouslyidentified to be linked with environmental stress-related pathways(response to UV, apoptosis, regulation of cell cycle), indicating geneswithin this category to be prime candidates for markers of Asdevelopmental toxicity. Preliminary dose–response information fromthis study may inform future studies examining As-induced altera-tions in glucose metabolism pathways within the developing embryoand the relationship with developmental toxicity and NTDs.

Conclusion

These preliminary studies used As and Cd as model teratogens toexamine dose-dependent metal effects on gene expression in C57BL/6J mouse embryos during neurulation to describe alterations in genesand associated biological processes with developmental toxicity andNTD formation. We provide evidence of As and Cd to disrupt genesinvolved in environmental stress-related pathways in a dose-dependent manner, such as cell cycle arrest, DNA damage response,oxidative stress, and apoptosis-related biological processes. We havealso identified unique responses associated with As exposure in itsability to significantly alter multiple genes involved in glucosemetabolism. Within the aforementioned shared and unique biologicalprocesses and related pathways, we have identified gene expressionalterations commonly altered as well as gene expression alterationsspecific for As or Cd that may serve as biomarkers of metalteratogenicity. Global assessment of RNA changes induced by As andCd across multiple genes and pathways expand initial observations ofspecific targets observed in previous studies in relation withdevelopmental toxicity and propose future assessments of specificfunctional interactions.

Supplementarymaterials related to this article can be found onlineat doi:10.1016/j.taap.2010.09.018.

Acknowledgments

The authors wish to thank Dr Edward K. Lobenhofer and Cogenicsfor conducting microarray hybridization for all Affymetrix arrays usedin this study and Alison Laing, Alison Scherer, and Erica Finsness foreditorial support. This work was supported in part by the NationalInstitute of Environmental Health Sciences (NIEHS) Toxicogenomicsgrants U10 ES 11387 and R01-ES10613, the US EnvironmentalProtection Agency (EPA) and NIEHS University of Washington (UW)Center for Child Environmental Health Risks Research grants R826886and 1PO1ES09601, the NIEHS and National Science Foundation (NSF)Pacific Northwest Center for Human Health and Ocean Studies grants

P50 ES012762 and OCE-0434087, and the NIEHS UW Center forEcogenetics and Environmental Health grant 5 P30 ES07033.

References

Bartke, T., Pohl, C., Pyrowolakis, G., Jentsch, S., 2004. Dual role of BRUCE as anantiapoptotic IAP and a chimeric E2/E3 ubiquitin ligase. Mol. Cell 14, 801–811.

Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical andpowerful approach to multiple testing. J. R. Stat. Soc. B 57, 289–300.

Beyersmann, D., Hartwig, A., 2008. Carcinogenic metal compounds: recent insight intomolecular and cellular mechanisms. Arch. Toxicol. 82, 493–512.

Castorena-Torres, F., Bermudez de Leon, M., Cisneros, B., Zapata-Perez, O., Salinas, J.E.,Albores, A., 2008. Changes in gene expression induced by polycyclic aromatichydrocarbons in the human cell lines HepG2 and A549. Toxicol. In Vitro 22, 411–421.

Chang, X.L., Jin, T.Y., Zhou, Y.F., 2006. Metallothionein 1 isoform gene expressioninduced by cadmium in human peripheral blood lymphocytes. Biomed. Environ.Sci. 19, 104–109.

Christley, J., Webster, W.S., 1983. Cadmium uptake and distribution in mouse embryosfollowing maternal exposure during the organogenic period: a scintillation andautoradiographic study. Teratology 27, 305–312.

Dencker, L., 1975. Possible mechanisms of cadmium fetotoxicity in golden hamstersand mice: uptake by the embryo, placenta and ovary. J. Reprod. Fertil. 44, 461–471.

Dennis Jr., G., Sherman, B.T., Hosack, D.A., Yang, J., Gao, W., Lane, H.C., Lempicki, R.A.,2003. DAVID: Database for Annotation, Visualization, and Integrated Discovery.Genome Biol. 4, P3.

Doniger, S.W., Salomonis, N., Dahlquist, K.D., Vranizan, K., Lawlor, S.C., Conklin, B.R.,2003. MAPPFinder: using Gene Ontology and GenMAPP to create a global gene-expression profile from microarray data. Genome Biol. 4, R7.

Eisen, M.B., Spellman, P.T., Brown, P.O., Botstein, D., 1998. Cluster analysis and display ofgenome-wide expression patterns. Proc. Natl. Acad. Sci. U. S. A. 95, 14863–14868.

Ferm, V.H., 1977. Arsenic as a teratogenic agent. Environ. Health Perspect. 19, 215–217.Ferm, V.H., Carpenter, S.J., 1967. Teratogenic effect of cadmium and its inhibition by

zinc. Nature 216, 1123.Fernandez, E.L., Gustafson, A.L., Andersson, M., Hellman, B., Dencker, L., 2003.

Cadmium-induced changes in apoptotic gene expression levels and DNA damagein mouse embryos are blocked by zinc. Toxicol. Sci. 76, 162–170.

Fernandez, E.L., Svenson, C., Dencker, L., Gustafson, A.L., 2004. Disturbing endodermsignaling to anterior neural plate of vertebrates by the teratogen cadmium. Reprod.Toxicol. 18, 653–660.

Fernandez, E.L., Dencker, L., Tallkvist, J., 2007. Expression of ZnT-1 (Slc30a1) and MT-1(Mt1) in the conceptus of cadmium treated mice. Reprod. Toxicol. 24, 353–358.

Freeman, K., 2004. Toxicogenomics data: the road to acceptance. Environ. HealthPerspect. 112, A678–A685.

Hanlon, D.P., Ferm, V.H., 1986. Concentration and chemical status of arsenic in the bloodof pregnant hamsters during critical embryogenesis: 2. Acute exposure. Environ.Res. 40, 380–390.

Hill, D.S., Wlodarczyk, B.J., Mitchell, L.E., Finnell, R.H., 2009. Arsenate-induced maternalglucose intolerance and neural tube defects in a mouse model. Toxicol. Appl.Pharmacol. 239, 29–36.

Ho Sui, S.J., Mortimer, J.R., Arenillas, D.J., Brumm, J., Walsh, C.J., Kennedy, B.P.,Wasserman, W.W., 2005. oPOSSUM: identification of over-represented transcriptionfactor binding sites in co-expressed genes. Nucleic Acids Res. 33, 3154–3164.

Hood, R.D., Vedel, G.C., Zaworotko, M.J., Tatum, F.M., Meeks, R.G., 1988. Uptake,distribution, and metabolism of trivalent arsenic in the pregnant mouse. J. Toxicol.Environ. Health 25, 423–434.

Hovland Jr., D.N., Machado, A.F., Scott Jr., W.J., Collins, M.D., 1999. Differential sensitivityof the SWV and C57BL/6 mouse strains to the teratogenic action of singleadministrations of cadmium given throughout the period of anterior neuroporeclosure. Teratology 60, 13–21.

Kultima, K., Fernandez, E.L., Scholz, B., Gustafson, A.L., Dencker, L., Stigson, M., 2006.Cadmium-induced gene expression changes in the mouse embryo, and theinfluence of pretreatment with zinc. Reprod. Toxicol. 22, 636–646.

Kuribayashi, K., El-Deiry, W.S., 2008. Regulation of programmed cell death by the p53pathway. Adv. Exp. Med. Biol. 615, 201–221.

Lee, G.S., Liao, X., Cantor, R.M., Collins, M.D., 2006. Interactive effects of cadmium andall-trans-retinoic acid on the induction of forelimb ectrodactyly in C57BL/6 mice.Birth Defects Res. A Clin. Mol. Teratol. 76, 19–28.

Lindgren, A., Danielsson, B.R., Dencker, L., Vahter, M., 1984. Embryotoxicity of arseniteand arsenate: distribution in pregnant mice andmonkeys and effects on embryoniccells in vitro. Acta Pharmacol. Toxicol. (Copenh.) 54, 311–320.

Liu, J., Cheng, M.L., Yang, Q., Shan, K.R., Shen, J., Zhou, Y., Zhang, X., Dill, A.L., Waalkes, M.P.,2007. Bloodmetallothionein transcript as a biomarker formetal sensitivity: low bloodmetallothionein transcripts in arsenicosis patients from Guizhou, China. Environ.Health Perspect. 115, 1101–1106.

Lutz, J., Beck, S.L., 2000. Caffeine decreases the occurrence of cadmium-inducedforelimb ectrodactyly in C57BL/6J mice. Teratology 62, 325–331.

Machado, A.F., Hovland Jr., D.N., Pilafas, S., Collins, M.D., 1999. Teratogenic response toarsenite during neurulation: relative sensitivities of C57BL/6J and SWV/Fnn miceand impact of the splotch allele. Toxicol. Sci. 51, 98–107.

Nakashima, K., Wakisaka, T., Fujiki, Y., 1987. Dose–response relationship of cadmiumembryotoxicity in cultured mouse embryos. Reprod. Toxicol. 1, 293–298.

Paniagua-Castro, N., Escalona-Cardoso, G., Chamorro-Cevallos, G., 2007. Glycinereduces cadmium-induced teratogenic damage in mice. Reprod. Toxicol. 23, 92–97.

Pulido, M.D., Parrish, A.R., 2003. Metal-induced apoptosis: mechanisms. Mutat. Res.533, 227–241.

129J.F. Robinson et al. / Toxicology and Applied Pharmacology 250 (2011) 117–129

Robinson, J.F., Yu, X., Hong, S., Griffith, W.C., Beyer, R., Kim, E., Faustman, E.M., 2009.Cadmium-induced differential toxicogenomic response in resistant and sensitivemouse strains undergoing neurulation. Toxicol. Sci. 107, 206–219.

Robinson, J.F., Griffith, W.C., Yu, X., Hong, S., Kim, E., Faustman, E.M., 2010a.Methylmercury induced toxicogenomic response in C57 and SWV mouse embryosundergoing neural tube closure. Reprod. Toxicol. 30, 284–291.

Robinson, J.F., Yu, X., Hong, S., Zhou, C., Kim, N., DeMasi, D., Faustman, E.M., 2010b.Embryonic toxicokinetic and dynamic differences underlying strain sensitivity tocadmium during neurulation. Reprod. Toxicol. 29, 279–285.

Robinson, J.F., Guerrette, Z., Yu, X.Z., Hong, S., Faustman, E.M., 2010c. A systems-basedapproach to investigate dose- and time-dependent methylmercury-induced geneexpression response in C57BL/6 mouse embryos undergoing neurulation. BirthDefects Res. B Dev. Reprod. Toxicol. 89, 188–200.

Saeed, A.I., Sharov, V., White, J., Li, J., Liang, W., Bhagabati, N., Braisted, J., Klapa, M.,Currier, T., Thiagarajan,M., Sturn, A., Snuffin,M., Rezantsev, A., Popov, D., Ryltsov, A.,Kostukovich, E., Borisovsky, I., Liu, Z., Vinsavich, A., Trush, V., Quackenbush, J., 2003.TM4: a free, open-source system for microarray data management and analysis.Biotechniques 34, 374–378.

Salvatori, F., Talassi, C.B., Salzgeber, S.A., Spinosa, H.S., Bernardi, M.M., 2004.Embryotoxic and long-term effects of cadmium exposure during embryogenesisin rats. Neurotoxicol. Teratol. 26, 673–680.

Simon, R., Lam, A., Li, M.C., Ngan, M., Menenzes, S., Zhao, Y., 2007. Analysis of geneexpression data using BRB-array tools. Cancer Inform. 3, 11–17.

Unis, D., Osborne, C., Diawara, M.M., 2009. Arsenite exposure compromises earlyembryonic development in the golden hamster. Reprod. Toxicol. 28, 329–334.

Webster, W.S., Messerle, K., 1980. Changes in the mouse neuroepithelium associatedwith cadmium-induced neural tube defects. Teratology 21, 79–88.

Wlodarczyk, B.J., Bennett, G.D., Calvin, J.A., Finnell, R.H., 1996. Arsenic-induced neuraltube defects in mice: alterations in cell cycle gene expression. Reprod. Toxicol. 10,447–454.

Wlodarczyk, B., Spiegelstein, O., Gelineau-van Waes, J., Vorce, R.L., Lu, X., Le, C.X.,Finnell, R.H., 2001. Arsenic-induced congenital malformations in geneticallysusceptible folate binding protein-2 knockout mice. Toxicol. Appl. Pharmacol.177, 238–246.

Wlodraczyk, B., Bennett, G.D., Calvin, J.A., Craig, J.C., Finnell, R.H., 1996. Arsenic-inducedalterations in embryonic transcription factor gene expression: implications forabnormal neural development. Dev. Genet. 18, 306–315.

Yu, X., Griffith, W.C., Hanspers, K., Dillman III, J.F., Ong, H., Vredevoogd, M.A., Faustman,E.M., 2006. A system-based approach to interpret dose- and time-dependentmicroarray data: quantitative integration of gene ontology analysis for riskassessment. Toxicol. Sci. 92, 560–577.

Yu, X., Robinson, J.F., Gribble, E., Hong, S.W., Sidhu, J.S., Faustman, E.M., 2008. Geneexpression profiling analysis reveals arsenic-induced cell cycle arrest and apoptosisin p53-proficient and p53-deficient cells through differential gene pathways.Toxicol. Appl. Pharmacol. 233, 389–403.

Yu, X., Robinson, J.F., Sidhu, J.S., Hong, S., Faustman, E.M., 2010. A system-basedcomparison of gene expression reveals alterations in oxidative stress, disruption ofubiquitin–proteasome system and altered cell cycle regulation after exposure tocadmium and methylmercury in mouse embryonic fibroblast. Toxicol. Sci. 114,356–377.