RESEARCH ARTICLE

Inorganic arsenic causes fatty liver and interacts with ethanol tocause alcoholic liver disease in zebrafishKathryn Bambino1, Chi Zhang2, Christine Austin1, Chitra Amarasiriwardena1, Manish Arora1, Jaime Chu3 andKirsten C. Sadler2,*

ABSTRACTThe rapid increase in fatty liver disease (FLD) incidence isattributed largely to genetic and lifestyle factors; however,environmental toxicants are a frequently overlooked factor thatcan modify the effects of more common causes of FLD. Chronicexposure to inorganic arsenic (iAs) is associated with liver diseasein humans and animal models, but neither the mechanism of actionnor the combinatorial interaction with other disease-causing factorshas been fully investigated. Here, we examined the contribution ofiAs to FLD using zebrafish and tested the interaction with ethanol tocause alcoholic liver disease (ALD). We report that zebrafishexposed to iAs throughout development developed specificphenotypes beginning at 4 days post-fertilization (dpf ), includingthe development of FLD in over 50% of larvae by 5 dpf.Comparative transcriptomic analysis of livers from larvae exposedto either iAs or ethanol revealed the oxidative stress response andthe unfolded protein response (UPR) caused by endoplasmicreticulum (ER) stress as common pathways in both these models ofFLD, suggesting that they target similar cellular processes. This wasconfirmed by our finding that arsenic is synthetically lethal with bothethanol and a well-characterized ER-stress-inducing agent(tunicamycin), suggesting that these exposures work togetherthrough UPR activation to cause iAs toxicity. Most significantly,combined exposure to sub-toxic concentrations of iAs and ethanolpotentiated the expression of UPR-associated genes, cooperated toinduce FLD, reduced the expression of as3mt, which encodes anarsenic-metabolizing enzyme, and significantly increased theconcentration of iAs in the liver. This demonstrates that iAsexposure is sufficient to cause FLD and that low doses of iAs canpotentiate the effects of ethanol to cause liver disease.

This article has an associated First Person interview with the firstauthor of the paper.

KEY WORDS: Arsenic, Ethanol, Fatty liver disease, Environmentalexposure

INTRODUCTIONFatty liver disease (FLD) is the most common liver pathology in theworld (Levene and Goldin, 2012). The dramatic rise in incidence inthe past several decades has prompted intense investigation into thebiological basis for this observation. A high fat, high sugar diet(Adams and Lindor, 2007), genetic predisposition (Anstee and Day,2015; Liu et al., 2013) and alcohol abuse (Magdaleno et al., 2017) areclear risk factors for FLD, but these risks alone do not account for thesteep rise in FLD incidence, nor do they provide an explanation for allFLD cases. Epidemiological studies have shown that multipleenvironmental and anthropogenic toxicants cause liver disease inhumans (Das et al., 2010; Islam et al., 2011;Mazumder, 2005; Santraet al., 1999), and work in rodents (Ditzel et al., 2016) and zebrafish(Cheng et al., 2016) have demonstrated a direct, causative relationshipbetween some environmental toxicants and FLD (Al-Eryani et al.,2015;Wahlang et al., 2013). The combination of epidemiological andbasic research on environmental toxicants and metabolic disease israpidly advancing, yet the scope of the problem and the mechanismsof toxicity are not yet clear.

Chronic exposure to inorganic arsenic (iAs) is aworldwide publichealth concern as it is associated with a broad range of healthproblems (Centeno et al., 2002; Naujokas et al., 2013). iAs is anaturally occurring element in the earth’s crust and both humans andwildlife are exposed to iAs through food and water. Estimates ofover 100-million people worldwide are exposed to levels exceedingWorld Health Organization (WHO)-established limits (Adamset al., 2016; Farzan et al., 2016; Yang et al., 2009). The firstprospective cohort study of people chronically exposed to arsenichas revealed an increase in all-cause and chronic disease mortalities(Ahsan et al., 2006; Argos et al., 2010). Notably, frequent co-morbidities of FLD, including diabetes (Kuo et al., 2015; Wanget al., 2014b), cardiovascular disease (Moon et al., 2013) and livercancer (Wang et al., 2014a), are significantly associated with iAsexposure. A study in the arsenic-endemic regions of Bangladeshand West Bengal, India, where obesity and alcohol abuse are low,found that chronic exposure to iAs via drinking water is associatedwith liver damage and fibrosis (Das et al., 2012; Islam et al., 2011).This was confirmed by the finding of a high prevalence of FLD andother liver diseases in this same region (Das et al., 2010). Together,the epidemiological data suggests that arsenic is a liver toxicant inhumans. Whether it can potentiate the effects of other causes of liverdisease, such as alcohol abuse, remains to be investigated.

Work using animal models has demonstrated that iAs can causeFLD. In some mouse studies, chronic iAs exposure induceslipogenic gene expression in the liver (Adebayo et al., 2015) andFLD (Santra et al., 2000b). Strikingly, exposure to iAs in utero andpost-weaning leads to adults that have a higher rate of FLD when feda high-fat diet (Ditzel et al., 2016). In zebrafish, arsenic exposure inembryos causes widespread developmental defects (Adeyemi et al.,2015; Bambino and Chu, 2017; Li et al., 2009, 2012;Ma et al., 2015;Received 27 July 2017; Accepted 7 December 2017

1Department of Environmental Medicine and Public Health, Icahn School ofMedicine at Mount Sinai, New York, New York 10029, USA. 2Program in Biology,New York University Abu Dhabi, Saadiyat Island Campus, PO Box 129188 AbuDhabi, United Arab Emirates. 3Department of Pediatrics, Division of PediatricHepatology, Icahn School of Medicine at Mount Sinai, New York, New York 10029,USA .

*Author for correspondence (Kirsten.Edepli@nyu.edu)

K.B., 0000-0001-8541-973X; J.C., 0000-0002-9291-8630; K.C.S., 0000-0002-1100-4125

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2018. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

McCollum et al., 2014; Wang et al., 2006), and exposing adultzebrafish to arsenic causes a range of gene and protein expressionchanges in the liver related to lipid metabolism (Carlson and VanBeneden, 2014; Hallauer et al., 2016; Li et al., 2016; Xu et al., 2013);one study reported FLD in adult zebrafish acutely exposed to iAs (Liet al., 2016). Thus, across species, iAs causes liver damage. Thesedata indicate that iAs alone causes FLD and can also predispose toFLD susceptibility, and we propose that lower doses of iAs interactwith more common risk factors to promote FLD.The importance of iAs as a toxicant has generated significant

interest in deciphering how iAs exposure causes disease. iAsmetabolism via arsenic 3 methyltransferase (AS3MT) utilizes thesame methyl donor that is used for DNA methylation (Hamdi et al.,2012; Thomas et al., 2004) and the iAs methylation reactionproduces reactive oxygen species (ROS) (Jomova et al., 2011; Santraet al., 2000a; Shi et al., 2004; Xu et al., 2017). Thus, reduction inDNA methylation and increased oxidative stress are two leadingtheories for the mechanism of iAs toxicity. A third possibility isbased on the finding that iAs impairs protein folding: iAs binds thiolgroups, and it is well established that the basis for acute arsenicpoisoning is iAs acting as a reducing agent for sulfhydryl groups inkey metabolic enzymes (Hughes, 2002; Sattar et al., 2016).Similarly, at lower doses, iAs acts to reduce sulfhydryl groups oncysteine residues in nascent peptides, which prevents disulfide bondformation (Jacobson et al., 2012; Ramadan et al., 2009) and preventsaccurate protein folding. In addition, ROS generated via arsenicmetabolism can disrupt the redox balance required for disulfide bondformation and protein folding in the endoplasmic reticulum (ER).The finding that the unfolded protein response (UPR), the pathwayinduced by ER stress, is activated with iAs treatment of some celltypes (Doudican et al., 2012; Weng et al., 2014) supports thehypothesis of ER stress as a mechanism for iAs-induced toxicity.The UPR is a central pathway in FLD pathophysiology across

species (Goessling and Sadler, 2015; Wang and Kaufman, 2014).Robust ER stress is sufficient to cause FLD in mice and zebrafish(Cinaroglu et al., 2011; Ji et al., 2011; Ozcan et al., 2004; Thakuret al., 2011; Vacaru et al., 2014; Yamamoto et al., 2010), andwe haveshown that activation of Atf6, a main upstream player in the UPR, isnecessary and sufficient to cause FLD (Cinaroglu et al., 2011;Howarth et al., 2014). This provides a direct and mechanistic linkbetweenUPR activation and fatty liver. ROS generation from ethanolmetabolism is a central mechanism of alcoholic liver disease (ALD)(Louvet and Mathurin, 2015). ROS alone can induce the UPR, andwe (Tsedensodnom et al., 2013) and others (Lu and Cederbaum,2008) have shown that the UPR activation and FLD caused byalcohol is mediated by ROS. Given that UPR activation is a centralmechanism of ALD (Cinaroglu et al., 2011; Howarth et al., 2012,2014), we hypothesize that other toxicants that disrupt ER function,such as iAs, could collaborate with ethanol to cause liver disease.In this study, we use zebrafish to identify the mechanism by

which iAs causes liver disease and to investigate whether iAsexposure interacts with ethanol to induce ALD. We found that iAsalone can cause FLD, and that sub-toxic concentrations of iAs andethanol interact to induce the UPR and cause FLD. Transcriptomicanalysis revealed that the ER- and oxidative-stress responses areactivated in the liver of larvae treated with either iAs or ethanol, but,strikingly, although a few UPR target genes were common to both,these two toxicants induced unique UPR signatures. We found thatlarvae exposed to iAs are sensitized to ethanol, showing an increasein larval mortality and increased iAs concentration in the liver.Importantly, we show that iAs and ethanol interact to cause FLD.This provides evidence that environmental toxicants can modify the

effect of more common risk factors for liver disease and thus mayact as coordinating events in the multistep process of progressiveliver disease (Browning and Horton, 2004; Day and James, 1998;Dowman et al., 2010).

RESULTSZebrafish are susceptible to arsenic toxicityMultiple theories have been proposed to explain how chronic iAscauses disease; however, none have yet been conclusively identified.Studies using zebrafish to investigate the developmental toxicity ofiAs (Li et al., 2009; Ma et al., 2015; Seok et al., 2007) andconsequences of iAs exposure in adults (Carlson and Van Beneden,2014; Hallauer et al., 2016; Lam et al., 2006; Xu et al., 2013) haveestablished this as an excellent model for investigating themechanism of iAs toxicity. We extended this work by establishinga protocol to expose zebrafish embryos to iAs throughoutdevelopment at a dose that would not impair liver development, toallow analysis of liver disease in older larvae. Zebrafish embryoswereexposed to a range of sodium (meta)arsenite (iAs) concentrationsbeginning at 4 hpf and monitored daily for mortality. Treatment with2.0 mM (260 ppm) iAs resulted in death of all exposed fish by 6 dpf(n=2 clutches, 80 embryos per condition), whereas concentrations<1.0 mM were well tolerated during this time (Fig. 1A, Table S1).Using this data, we calculated the concentration at which 50% ofexposed fish died by 6 dpf (LC50) to be 1.25 mM (Fig. S1A). Short-term exposure to 2 mM iAs prior to 48 hpf resulted in fewerdevelopmental defects than longer exposures (Fig. S1B), suggestingthat the effects of iAs exposure during development are cumulative.Exposure to the highest dose of iAs (2 mM) beginning at 72 or 96 hpfuntil 120 hpf also resulted in significant abnormalities and larvaldeath (data not shown), raising the possibility that laterdevelopmental stages may also be susceptible; however, thisremains an area for further exploration.

Embryos exposed to concentrations of iAs that caused minimallethality by 5 dpf (i.e. <1.4 mM) developed a range of specificphenotypes by 5 dpf, including a shortened body length, edema,clustering of pigment cells, under-consumption of yolk and a smallhead (arrows, Fig. 1B). Of the larvae that survived exposure to 0.6,1.0, 1.4 and 2.0 mM iAs until 5 dpf, 19.6, 37.1, 64.1 and 100%,respectively, displayed at least one of these phenotypes at 5 dpf(Fig. 1C and Fig. S1C). These data are consistent with other studiesthat show that iAs at concentrations of 1.0 mM and above causeddevelopmental abnormalities, whereas 0.5 mM was well toleratedby developing zebrafish embryos (Li et al., 2009).

Zebrafish metabolize iAs in the liverPrevious studies have demonstrated that zebrafish expressaquaglyceroporins and the trivalent arsenic-specificmethyltransferase (As3mt) (Hamdi et al., 2009, 2012), which arerequired for the uptake and metabolism of iAs, respectively. Weasked whether zebrafish embryos expressed as3mt throughoutdevelopment and determined whether, like in mammals (Thomaset al., 2004; Tseng, 2007), as3mt was enriched in the liver. Throughboth quantitative real-time PCR (qRT-PCR) analysis and miningtranscriptomic data from Array Express (www.ebi.ac.uk/arrayexpress), it is clear that as3mt mRNA is maternallycontributed and zygotically expressed (Fig. 2A, Fig. S2), and, at5 dpf, expression was enriched by over 8-fold in the liver comparedto the liver-less carcass (Fig. 2A).

We next asked whether iAs accumulated in zebrafish using laser-ablation–inductively-coupled-plasma–mass spectroscopy (LA-ICP-MS) (Hare et al., 2012) and inductively-coupled-plasma–mass

2

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

spectroscopy (ICP-MS), highly sensitive microanalytical techniquesused to detect elements in biological or geological samples. ICP-MSis best suited to provide a highly sensitive quantitative measurementof iAs levels and LA-ICP-MS is used for mapping the tissues whereiAs accumulates (Sussulini et al., 2017). We found highconcentrations of iAs in the skin, eye, gut and, most notably, liverin larvae exposed to 1 mM iAs from 4 to 120 hpf (Fig. 2B). ICP-MSperformed on livers dissected from iAs-exposed and control larvaeshowed a highly consistent and dose-dependent increase in theamount of iAs with increasing exposure concentrations (Fig. 2C).These data show that zebrafish larvae express the enzyme required tometabolize iAs in the liver and that this is a site of iAs accumulation,suggesting this as a target tissue for iAs toxicity.

iAs causes fatty liverChronic exposure to iAs is associated with altered liver function andliver disease (Das et al., 2012; Islam et al., 2011; Mazumder, 2005;Santra et al., 1999; Wang et al., 2014a,b). Previous studies in bothmice and adult zebrafish showed that arsenic exposure can causesteatosis (Ditzel et al., 2016; Li et al., 2016; Tan et al., 2011) and canalso sensitize mice to other factors that cause FLD, such as a high-fatdiet (Tan et al., 2011). We tested the efficacy of iAs to directly causeFLD in zebrafish larvae treated with 1.0 mM iAs from 4 to 120 hpfusing Oil Red O (ORO), which we have previously demonstrated isa straightforward means of assessing steatosis incidence acrossmultiple clutches of larvae (Passeri et al., 2009; Vacaru et al., 2014)(Fig. 3A). The percent of control larvae with steatosis was anaverage of 15.2±3.0% (ranging from 0 to 42% across 15 clutches;Fig. 3B, Table S1). The incidence was significantly higher in theiAs-exposed larvae, with a mean of 46.9±5.2% (ranging from 16 to85.7%, n=15 clutches, P<0.001) (Fig. 3B, Table S1). These datashow that iAs is sufficient to cause FLD in this model, facilitatinginvestigation of the cellular mechanisms of this phenotype.

ER- and oxidative-stress responses are common pathwaysactivated in response to either iAs or ethanolTo identify cellular pathways activated in response to iAs and to gaininsight into those processes that contribute to iAs-induced steatosis,we performed RNAseq analysis on livers dissected from 3 clutchesof 5 dpf Tg(fabp10a:nls-mCherrymss4Tg) larvae exposed to 1 mMiAs from 4 to 120 hpf and from unexposed controls. In total, 5118genes were differentially expressed between iAs-exposed andcontrol siblings [adjusted P-value (padj)<0.05]; 2629 genes wereupregulated (537 log2 fold change>2) and 2489 genes weredownregulated (412 log2 fold change<−2) (Fig. 4A, Table 1,Table S3). In addition, some genes with read counts greater than orequal to 20 in one condition and less than or equal to 5 in the otherwere identified by our analysis. We designated these as ‘on-off’genes, and found that 186 genes turned on and 214 genes turned offin the liver in response to iAs (Table 1). We identified pathwaysinduced in the liver by iAs exposure using Ingenuity PathwayAnalysis (IPA) of the upregulated genes in iAs samples.Metabolism(mitochondrial dysfunction and oxidative phosphorylation) andNrf2-mediated oxidative stress pathways were highly enriched in thelivers of iAs-exposed larvae (Fig. 4B), consistent with findings fromother systems (Li et al., 2015; Sinha et al., 2013).

Reanalysis of an RNAseq dataset we previously generated fromlivers dissected from Tg(fabp10a:nls-mCherrymss4Tg) zebrafishexposed to 2% ethanol from 96 to 132 hpf (Howarth et al., 2014)was carried out to identify genes and pathways common to both iAsand ethanol treatment. We found significantly fewer differentiallyexpressed genes (1094; padj<0.05), with 619 genes upregulated(302 log2 fold change>2) and 475 genes downregulated (178 log2fold change<−2), with 153 genes on and 65 genes off in onlyethanol-treated samples (Fig. 4C, Table 1, Table S3). IPA of allupregulated genes in the ethanol-treated group revealed thatcholesterol biosynthesis (Superpathway of CholesterolBiosynthesis, Cholesterol Biosynthesis I, Mevalonate Pathway),oxidative stress and UPR/ER-stress responses were among the mostsignificantly enriched processes (Fig. 4D). This is consistent withour previous finding that UPR activation and FLD in response toalcohol was mediated by oxidative stress (Howarth et al., 2014;Tsedensodnom et al., 2013).

We found a high correlation between control samples in the twodatasets, with correlation coefficients ranging from 0.87 to 1.0(Fig. S3A). This allowed us to compare these two datasets to identify

Fig. 1. Exposure to iAs is toxic to zebrafish embryos. (A) Survival analysisof zebrafish treated with iAs from 4 hpf through 6 dpf. Fish were scored daily formortality (n≥2 clutches, >80 embryos per condition, Table S1). Red arrowindicates addition of iAs. Concentration of iAs used throughout the manuscriptis shown in red and indicated by an asterisk. (B) Bright-field images ofrepresentative wild-type control (top) and arsenic-exposed (bottom) 5 dpfzebrafish larvae. Arrows indicate arsenic-induced phenotypes, includingshortened body length, edema, clustering of pigment cells, under-consumption of yolk, and a small head. (C) Exposure to increasing doses of iAsfrom 4 hpf to 5 dpf led to an accumulation of phenotypes. The proportion ofsurviving embryos at 5 dpf that were affected increased with increasingconcentrations of iAs (n=2 clutches exposed to 0.1 mM or 0.3 mM, n=3clutches exposed to 0.6 mM, 1.0 mM or 1.4 mM, >30 fish exposed pertreatment condition, Table S1). hpf, hours post-fertilization; dpf, days post-fertilization; iAs, inorganic arsenic.

3

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

pathways that were commonly regulated. We identified 555 genes thatwere significantly differentially expressed in both the iAs and ethanoldatasets (Fig. 4E, Table 1), of which 479 changed their expression inthe same direction, with 215 genes upregulated and 264 genesdownregulated in both datasets (Table 1, Table S3). IPA of thecommon upregulated genes revealed that oxidative stress, ER stressand UPR pathways were the most highly represented (Fig. 4F).Oxidative stress generated during iAs metabolism is a proposedmechanism of arsenic toxicity (Huang et al., 2004; Santra et al., 2000a,2007). Decades of research in mammalian systems (Lu andCederbaum, 2008; Magdaleno et al., 2017) and our work inzebrafish (Tsedensodnom et al., 2013) have shown that oxidativestress is also a primary cellular mechanism of ALD. Interestingly, 3 ofthe commonly regulated pathways (Pentose Phosphate Pathway,Glycolysis, and Gluconeogenesis) were previously found to beinduced in transcriptome ormetabolomics analysis of livers from adultzebrafish acutely exposed to iAs (Li et al., 2016; Xu et al., 2013).

Different stressors induce distinct UPRs (Vacaru et al., 2014) andeach branch of the UPR modulates protein folding in a unique way.We analyzed the iAs and ethanol RNAseq datasets to determinewhether these toxicants also activated distinct UPRs. We generated alist of 254 genes that are targets of the UPR or function in the UPRbased on annotations in AmiGo and NCBI. Of these UPR-associatedgenes, the majority (227 genes) were expressed in both the iAs andethanol RNAseq datasets and nearly half of these (103 genes) weresignificantly differentially expressed in one or both datasets (Fig. 5A,Table S4), with 89 differentially expressed in the iAs dataset and 37differentially expressed in the ethanol dataset. Interestingly, thepattern of UPR gene expression was largely distinct in these twomodels, as only 10% of all UPR-associated genes and 20% of genesdifferentially expressed in one of the datasets showed a sharedexpression pattern in both datasets (Fig. 5A,B). The UPR-associatedgenes affected by both stressors include the bona fideUPRgenes atf3,hspa5, hyou1 and dnajc3, as well as heat-shock-related genes

Fig. 2. Zebrafish are capable of arsenic metabolism and accumulate iAs in their tissues. (A) Expression of the zebrafish as3mt transcript is dynamic duringzebrafish development, as determined by qRT-PCR. as3mt is maternally provided. Expression is enriched in the liver at 120 hpf. Error bars correspond tomean±s.d. L, liver; C, carcass. (B) Representative images of LA-ICP-MS analysis of 10-μm sections of control and iAs-exposed larvae. Following exposure from 4to 120 hpf, iAs accumulated in the eye (white arrows, white circle in enlarged image), liver (yellow arrows, yellow circle in enlarged image) and in the gut (greenarrows, green circle in the enlarged image). Refer to Table S2 for operating parameters. (C) Quantification of total arsenic content in the livers of 5-dpf larvaeby ICP-MS. Livers dissected from larvae exposed to 0, 0.1, 0.5 and 1.0 mM iAs from 4 to 120 hpf showed a dose-dependent increase in the total arsenic contentper liver (n=3 clutches). Error bars correspond to mean±s.d. hpf, hours post-fertilization; LA-ICP-MS, laser-ablation–inductively-coupled-plasma–mass-spectroscopy; iAs, inorganic arsenic; ICP-MS, inductively-coupled-plasma–mass-spectroscopy.

Fig. 3. Exposure to iAs causes steatosis in zebrafish larvae.(A) Representative bright-field images of 5-dpf Oil Red O (ORO)-stained control and iAs-exposed (1.0 mM from 4 to120 hpf) larvae. Thearea around the liver (outlined in black) is enlarged. (B) The percent oflarvae with steatosis analyzed by ORO staining of 15 clutches, with anaverage of 20 larvae per treatment per clutch. The total number of larvaeanalyzed in each clutch is listed in Table S1. Statistical significancewasdetermined by unpaired, 2-tailed Student’s t-test (n=15 clutches,P<0.001, Table S1). Error bars correspond to mean±s.d.

4

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

(Fig. 5B, Table S4). Additionally, 67 of the UPR-associated geneswere only changed in the iAs dataset and 15 genes were only changedin the ethanol dataset (Fig. 5A, Table S4). Because most of the UPR-

associated transcripts were detected in both datasets at similar levelsbased on FPKM (fragments per kilobase of transcript per millionmapped reads) values (Fig. S3B), we concluded that the unique

Fig. 4. Exposure to iAs induces expression of genes involved in metabolic processes and the UPR.MA plot of normalized gene expression in livers fromlarvae exposed to 1.0 mM iAs from 4 to 120 hpf compared to unexposed siblings (A) and in livers from larvae exposed to 2% ethanol from 96 to 132 hpf comparedto unexposed siblings (B). P-values and fold-changes for all significantly differentially expressed genes are included in Table S1. (B) IPA analysis of biologicalprocesses based on upregulated genes identified through RNAseq analysis. (C) MA plot of normalized gene expression in larvae exposed to 2% ethanol from 96-132 hpf compared to unexposed larvae. (D) IPA of biological processes based on upregulated genes identified through RNAseq analysis. (E) Plot of genessignificantly upregulated in both the iAs and ethanol datasets (upper right quadrant), upregulated in the iAs dataset and downregulated in the ethanol dataset(lower right quadrant), downregulated in both the iAs and ethanol datasets (lower left quadrant), and upregulated in the ethanol dataset and downregulated iniAs dataset (upper left quadrant). (F) IPA of biological processes based on upregulated genes in both iAs and ethanol RNAseq datasets. iAs, inorganicarsenic; IPA, Ingenuity Pathway Analysis.

5

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

patterns of UPR gene expression in these two datasets reflect distincttypes of UPRs [i.e. UPR subclasses (Vacaru et al., 2014)]. We usedqRT-PCR to confirm that the UPR genes we previously showed to be

associatedwith FLD [xbp1, xbp1s, bip, chop, dnajc3, edem1, atf4 andatf6 (Vacaru et al., 2014)] were upregulated in the liver of iAs-exposed larvae, and found that all except xbp1 and xbp1s weresignificantly induced (n=9 clutches, P<0.01) (Fig. S4A). Wefunctionally assessed the relevance of ER stress to iAs-inducedtoxicity by co-exposing larvae to iAs and tunicamycin, a well-characterized and widely used ER stressor that we have shown causesFLD in zebrafish (Cinaroglu et al., 2011; Howarth et al., 2013; Vacaruet al., 2014), and found them to be synergistically lethal (Fig. S4B),suggesting that they act through sharedmechanisms to cause lethality.

Together, these data indicate that (1) similar pathways arederegulated by iAs and ethanol, or ethanol alone, includingoxidative stress and the UPR, (2) both iAs and ethanol induce theUPR, that (3) a subset of UPR genes are targeted by both iAs andethanol, and that (4) each of these toxicants induces a unique UPRsignature, extending our previous findings of distinct UPRsubclasses in response to other ER stressors (Vacaru et al., 2014).We speculate that the subset of UPR genes deregulated in response

Table 1. Summary of gene expression changes in RNAseq datasets

Category of expressionUpregulatedgenes

Downregulatedgenes Total

1.0 mM iAs vs controlpadj<0.05 and LFC2>0 2629 2489 5118padj<0.05 and LFC2>2 537 412 949On-off 186 on in iAs 214 on in siblings2% ethanol vs controlpadj<0.05 and LFC2>0 619 475 1094padj<0.05 and LFC2>2 302 178 480On-off 153 on in 2% ethanol 65 on in 0% ethanolOverlappadj<0.05 and LFC2>0 215 264 479

LFC2, log2 fold change; padj, adjusted P-value.

Fig. 5. Arsenic and ethanol regulate common cellular pathways. (A) Plot of UPR-associated gene expressions in iAs and ethanol RNAseq datasets. (B) Heatmap of the expression of 103 significantly differentially regulated UPR genes. Refer to Table S4 for P-values and fold-changes. iAs, inorganic arsenic.

6

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

to both stressors reflects some commonality in their mechanism ofUPR induction.

Ethanol potentiates iAs toxicity, ER stress and fatty liverOur finding that there are shared pathways targeted by both ethanoland iAs suggests that combined exposure may interact to cause liverdisease. To test this, embryos were treated with a range ofconcentrations of iAs (0.1 mM, 0.5 mM, 1.0 mM) starting at 4 hpfand subsequently exposed at 96 hpf to a range of ethanolconcentrations (0.5%, 1.0%, 1.5%) which we have previouslyshown are below and above the threshold for causing lethality andFLD (Tsedensodnom et al., 2013). We found that iAs significantlyreduced survival of larvae exposed to ethanol concentrations that wereotherwise well tolerated on their own (n=3 clutches, P<0.001)(Fig. 6A), showing that they synergize to cause lethality andsuggesting that this results from their convergence on the samepathways.To investigate the nature of this interaction, we reduced exposure

concentrations to ensure survival of co-exposed larvae to allowfurther analysis. qPCR analysis of samples from larvae exposed to0.5% ethanol or 0.5 mM iAs alone or in combination revealed thatco-exposure enhanced expression of bip, chop and atf6, comparedto that in samples treated with either toxicant alone (Fig. 6B). Thisdemonstrates that exposure to a dose of iAs that does not cause anymorphological or gene expression changes can predispose zebrafishlarvae to a dramatic response to a low dose of ethanol. This isimportant as it suggests that co-exposure to sub-toxic concentrationsof these two toxicants can cause disease.We next tested our hypothesis that a potential mechanism by

which iAs and ethanol could interact would be via alteredexpression of the enzymes responsible for iAs metabolism andclearance. We found that as3mt expression was significantlyreduced in fish treated with iAs and ethanol compared to thoseexposed to either toxicant alone (n=5 clutches, 0.5% ethanol vs0.5 mM iAs+0.5% ethanol P=0.0013, 0.5 mM iAs vs 0.5 mM iAs+0.5% ethanol P=0.0158) (Fig. 6C). Because reduced expression ofas3mt can impair clearance of iAs from the liver, we hypothesizedthat ethanol exposure would increase the iAs levels in the liver. ICP-MS analysis showed that iAs levels were significantly higher in thelivers of larvae exposed to 0.5 mM iAs plus 0.5% ethanol than inlarvae exposed to 0.5 mM iAs alone (n=6 clutches, P=0.0003)(Fig. 6D). This demonstrates that the clearance of arsenic from theliver is hindered upon co-exposure to these two toxicants.We next hypothesized that accumulation of iAs can increase the

toxic sequelae of iAs exposure, including exacerbating ER stressand potentially increasing FLD incidence. To test this, we exposedzebrafish to each toxicant alone or in combination at doses which,when given independently, do not cause steatosis. Zebrafishembryos were treated with 0.5 mM iAs at 4 hpf, followed byaddition of either 0 or 0.5% ethanol at 96 hpf and collected at120 hpf for ORO staining. We found that the incidence of steatosiswas significantly higher in larvae exposed to both toxicantscompared to those exposed to either agent alone (n=7 clutches,0.5% ethanol vs 0.5 mM iAs+0.5% ethanol P=0.0025; 0.5 mM iAsvs 0.5 mM iAs+0.5% ethanol P=0.0172) (Fig. 6E, Table S5). Todetermine whether we could detect an interaction between iAs andethanol in the development of steatosis, we modeled the riskdifference between the exposure conditions. The combinedincidence of steatosis across all 7 clutches was determined bydividing the number of larvae with steatosis in each group by thetotal population of that group. This yielded a background incidenceof steatosis of 20.45% (27/132) in control larvae, 19.85% (27/136)

in larvae exposed to 0.5% ethanol alone, 25.19% (33/131) in larvaeexposed to 0.5 mM iAs alone, and 48.97% (71/145) in larvae co-exposed to 0.5 mM iAs+0.5% ethanol (Fig. 6E, Table S5).Controlling for the fact that some of the larvae were from thesame clutch, neither exposure to 0.5% ethanol nor 0.5 mM iAsalone significantly increased the risk of developing steatosisrelative to control (0.5% ethanol P=0.98, 0.5 mM iAs P=0.36).However, co-exposure to 0.5 mM iAs+0.5% ethanol resulted in anincreased risk of developing steatosis at an alpha of 0.1, commonlyused to assess the risk of interaction in epidemiological studies(P=0.08). These data support the conclusion that iAs and ethanolinteract to increase the concentration of iAs in the liver, aggravateER stress and enhance FLD (Fig. 7).

DISCUSSIONGenetic and lifestyle factors are assumed to be primary causes ofFLD; however, environmental toxicants can serve importantdisease-modifying roles by targeting cellular pathways orprocesses that render them more susceptible to other stimuli.Here, we show that high doses of iAs causes FLD and can modifythe risk of ALD. Importantly, the interaction between iAs andethanol is observed at doses of these toxicants which alone do notcause any detectable phenotypic or cellular changes. This suggeststhat exposure to a sub-toxic level of iAs could serve as a risk factorfor more common causes of liver disease, such as ALD.Additionally, our unbiased approach to identify pathways thatserve as a nexus for these toxicants revealed oxidative and ERstress as points of convergence. We hypothesize that these cellularstress responses are the mechanisms for the increased toxicity weobserve in the presence of iAs and ethanol. Finally, we presentmechanistic data demonstrating that ethanol acts to decrease as3mtexpression, thereby reducing iAs metabolism, leading to sustained,high concentrations of iAs in the liver and potentiating its toxicimpact.

Oxidative stress is a major cause of hepatic injury causedby alcohol because ROS are generated from alcohol metabolismby the cytochrome P450 system (CYP2E1) (Cederbaum, 2010;Lu and Cederbaum, 2008). Previous work from our lab(Tsedensodnom et al., 2013) and others (Ashraf and Sheikh,2015; Cederbaum, 2010) has shown that ROS generated duringethanol metabolism leads to ER stress. Protein folding in the ER isdependent on a delicate redox balance in order to form thedisulfide bonds that are required for proper protein structure. Byaltering the redox balance in the ER, oxidative protein folding isimpaired (Hudson et al., 2015), leading to UPR induction tomitigate the accumulation of unfolded proteins (Malhotra andKaufman, 2007). An important insight into arsenic-mediatedpathology is the finding that iAs and its metabolic derivatives actas strong inhibitors of oxidative protein folding (Jacobson et al.,2012; O’Connell et al., 2014; Sapra et al., 2015) and bind tounfolded proteins (Ramadan et al., 2009), and that trivalent iAsspecies directly prevent the proper folding of nascent peptides dueto strong affinity for thiol groups on cysteine residues (Ramadanet al., 2009; Watanabe and Hirano, 2013). Our working model(Fig. 7) proposes that iAs exposure increases the unfolded proteinload in the ER by directly acting as a reducing agent, and indirectlyby disrupting the redox balance in the ER through ROS generatedduring iAs metabolism by As3mt. Given that our RNAseq analysisfound that a CYP450 enzyme that metabolizes ethanol in zebrafish(cyp2y3; Tsedensodnom et al., 2013) is downregulated in responseto iAs, it is unlikely that iAs treatment potentiates ethanolmetabolism; if anything, it may reduce it.

7

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

Our model suggests that iAs induces FLD by activating apathological or ‘stressed’ UPR subclass (Howarth et al., 2014;Tsedensodnom et al., 2013; Vacaru et al., 2014; Wang andKaufman, 2016). We speculate that some aspects of this stressedUPR are shared with the response to ethanol. We demonstratedthat Atf6 is necessary and sufficient for fatty liver in zebrafish(Cinaroglu et al., 2011; Howarth et al., 2014) and, because atf6

is one of the few genes that is targeted by both iAs and ethanol,we speculate that FLD caused by iAs is mediated by Atf6activation. However, it is also possible that the unique aspects ofthe UPR which are induced by iAs and by ethanol could createunique proteostatic environments that are only minimallyoverlapping, as found in mammalian cells in culture(Shoulders et al., 2013).

Fig. 6. Arsenic potentiates alcohol-induced liver disease in zebrafish larvae. (A) Survival of zebrafish larvae at 120 hpf. Zebrafish larvae were exposed to arange of concentrations of iAs and/or ethanol. Survival was assessed at 120 hpf. Statistical significance was determined by unpaired, 2-tailed Student’s t-test,correcting for multiple comparisons (n=3 clutches, P<0.001). Data are presented as mean±s.d. (B) qRT-PCR data from liver cDNA. Data are presented as fold-change vs control. Statistical significance was determined by one-way ANOVA (n=5 clutches, bip: 0.5% ethanol vs 0.5 mM iAs+0.5% ethanol *P=0.0072, 0.5 mMiAs vs 0.5 mM iAs+0.5% ethanol P=0.078; chop: 0.5% ethanol vs 0.5 mM iAs+0.5% ethanol *P=0.0021, 0.5 mM iAs vs 0.5 mM iAs+0.5% ethanol *P=0.0008;atf6: 0.5% ethanol vs 0.5 mM iAs+0.5% ethanol *P=0.0205, 0.5 mM iAs vs 0.5 mM iAs+0.5% ethanol *P=0.0241). (C) qRT-PCR data from liver cDNA. Data arepresented as fold-change vs control. Statistical significance was determined by one-way ANOVA (n=5 clutches, 0.5% ethanol vs 0.5 mM iAs+0.5% ethanol*P=0.0013, 0.5 mM iAs vs 0.5 mM iAs+0.5% ethanol *P=0.0158). (D) Quantification of total arsenic content in the livers of 5-dpf larvae by liquid ICP-MS.Statistical significance was determined by one-way ANOVA (n=6 clutches, 0.5 mM iAs vs 0.5 mM iAs+0.5% ethanol *P=0.0003). (E) Steatosis incidence in120-hpf larvae exposed to 0.5 mM iAs, 0.5% ethanol and 0.5 mM iAs+0.5% ethanol. Statistical significance was determined by one-way ANOVA (n=7 clutches,0.5% ethanol vs 0.5 mM iAs+0.5% ethanol *P=0.0025, 0.5 mM iAs vs 0.5 mM iAs+0.5% ethanol *P=0.0172). Error bars in all panels correspond to mean±s.d.

8

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

Other work has suggested that iAs impacts mitochondrialfunction (Luz et al., 2016; Santra et al., 2007) and our RNAseqanalysis also highlighted mitochondrial dysfunction as a response toiAs exposure. It is possible that mitochondrial damage in hepatocytescould reduce lipid oxidation and cause lipid accumulation. Anotherinteresting possibility is that communication between themitochondria and ER could impair the function of both of theseorganelles if one is damaged. We found that ero1a, a component ofmitochondrial-associated ER membranes (MAMs) and a regulator ofoxidative protein folding and ER redox homeostasis (Anelli et al.,2011;Benhamet al., 2013; Seervi et al., 2013), is among the genes thatare significantly upregulated in both iAs and ethanol datasets. MAMsare key signaling centers that act at the interface betweenmitochondriaand the ER, and are linked by their key functions in redox homeostasisand calcium storage (Carreras-Sureda et al., 2017). Thus, it is possiblethat defects in the function of multiple organelles contribute both tothe ER stress and FLD phenotypes caused by iAs exposure.Other mechanisms proposed to mediate iAs toxicity – namely

changes in DNA methylation (Huang et al., 2004; Hughes, 2002;Kile et al., 2012; Mauro et al., 2016) – were not identified by ourstudies in zebrafish (not shown). In fact, there is nearly no overlap inthe gene expression changes induced by iAs and bymutants that havea defect in DNA methylation (Chernyavskaya et al., 2017; Jacobet al., 2015), and our preliminary studies found neither an impact onglobal DNA methylation nor synergy with mutants that have DNAmethylation loss (not shown). Thus, we conclude that oxidative andER stress are the major mechanisms of iAs hepatotoxicity.By comparing the iAs and ethanol RNAseq datasets, we found

several additional enriched pathways that may contribute to thedevelopment of FLD. Both glycolysis and gluconeogenesis pathwayswere enriched in both datasets, and these pathways have previouslybeen demonstrated to be induced by iAs in the liver of guinea pigs andadult zebrafish (Li et al., 2016; Reichl et al., 1988). We speculate thatthese pathways are important for maintaining the energetic balancerequired for efficient protein folding and hepatic metabolism, and thatthe accumulation of lipids in hepatocytes of toxin-stressed cells mayserve as an adaptive function to store energy.

Although our RNAseq analysis and other data provide evidencethat iAs and ethanol are converging on the same cellular pathways, itis important to note that there are differences in the gene expressionprofiles in response to the two toxicants. It is therefore possible thatthe cumulative effects of independent cellular responses areconverging to yield the phenotypes analyzed in this study (i.e.FLD and death). The same zebrafish line, Tg(fabp10a:nls-mCherrymss4Tg), was used to generate both RNAseq datasets, andconsistency across sample maintenance (i.e. light:dark cycle,culture conditions and time of collection) were maintained toprovide a basis for comparison. Moreover, qPCR analysis ofadditional clutches that were treated in parallel confirmed that thegene expression changes detected by RNAseq were reproducible.

An important finding from our work is that co-exposure to iAsand ethanol reduces the expression of as3mt, which catalyzes themethylation of iAs and results in an increase in the totalaccumulation of arsenic in the liver. Polymorphisms in the humanAS3MT gene are predictive of arsenic metabolism across multiplestudy populations (Engstrom et al., 2011, 2013). Indeed, reducedcapacity for arsenic methylation has previously been shown tocorrelate with adverse outcomes following arsenic exposure (Chenet al., 2013; Das et al., 2016). Human genome-wide associationstudies for mutations promoting arsenic susceptibility identified anAS3MT haplotype that leads to reduced expression of AS3MT andincreased toxicity (Das et al., 2016). Recently it was shown thatC57BL/6J mice are more susceptible to arsenic-induced oxidativeliver injury than 129X1/SvJ mice, and that this difference is relatedto their reduced capacity for arsenic methylation (Wu et al., 2017).In addition, one study found that alcohol use may decreasemethylation of iAs (Hopenhayn-Rich et al., 1996; Tseng, 2009).Future research will seek to determine the mechanisms by whichethanol reduces expression of as3mt.

Although zebrafish are at the forefront of toxicology research(Bambino and Chu, 2017; Garcia et al., 2016), there are somelimitations in the extension of this study to understanding the effectsof arsenic and alcohol on liver disease in humans. As with otherexperimental models, translation of the results to the effects ofhuman exposures are complicated by differences in arsenicmetabolism (Vahter, 1999) and the doses used under laboratoryconditions (States et al., 2011). The concentration of iAs used heresurpasses the levels that are associated with disease in humans,suggesting that, in this system where animals are immersed in iAs,low levels are not acutely toxic, but instead may have an impact overlonger-term exposures. We have used exposures in the low parts permillion (ppm) range, whereas humans are typically exposed toarsenic concentrations in the parts per billion (ppb) range, with aWHO standard of 10 ppb. However, samples of drinking water fromArgentina, Bangladesh, China, Taiwan and the United States haveall been found to contain iAs at concentrations greater than 2 mg l−1

(2 ppm) (Naujokas et al., 2013), indicating the possibility ofexposure at levels higher than previously presumed. We examinedthe effect of arsenic exposure beginning just prior to gastrulation tomimic the effects of early-life exposure; however, this does notcompletely recapitulate the predominant route of human exposure,as the zebrafish do not begin to swallow water until 3 dpf.

Collectively, these data demonstrate that combined exposure toiAs and ethanol led to a reduced survival of zebrafish larvae,enhanced induction of the UPR and an increase in the risk of fattyliver disease. Given the high conservation of common disease-related genes and pathways between zebrafish and humans(Goessling and Sadler, 2015), we hypothesize that the mechanismof iAs-induced FLD and the cellular pathways targeted by both

Fig. 7. Model of the mechanism by which iAs and ethanol interact tocause ER stress. iAs and ethanol interact to increase the concentration ofiAs in the liver, aggravate ER stress and enhance FLD. This figure wasdrawn by J. Gregory (Mount Sinai Health System).

9

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

toxicants will also be conserved. With the increasing globalincidence of liver disease, the risk of liver damage caused by iAsexposure may be more significant than previously presumed.

MATERIALS AND METHODSZebrafish maintenance and treatment of embryosProcedures were performed in accordance with the Icahn School ofMedicine at Mount Sinai Institutional Animal Care and Use Committee(IACUC). Adult wild-type (WT; AB, Tab14 and TAB5) and Tg(fabp10a:nls-mCherrymss4Tg) (Mudbhary et al., 2014) fish were maintained on a 14:10light:dark cycle at 28°C. Fertilized embryos from natural spawning of groupmatings were collected and cultured in embryo water (0.6 g/l Crystal SeaMarinemix; Marine Enterprises International, Baltimore, MD) containingMethylene Blue at 28°C. Embryos were treated with sodium (meta)arsenite(Sigma, S7400) beginning at 4 h post fertilization (hpf ). Sodium(meta)arsenite and/or ethanol were diluted from stock solutions in 10 mlof embryo water. After addition of exposure medium, 35-mm dishes weresealed with Parafilm and returned to the incubator. Mediumwas not replacedduring the exposure period unless otherwise noted. For co-exposures,sodium (meta)arsenite was removed and replaced with embryo watercontaining sodium (meta)arsenite and tunicamycin at 72 hpf or sodium(meta)arsenite and ethanol at 96 hpf at the indicated concentrations. Imagesof anesthetized larvae were taken by mounting in 3% methyl cellulose usinga Nikon SMZ1500 stereomicroscope.

Gene expression analysis by qRT-PCRLivers were microdissected in from anesthetized 5 dpf zebrafish larvae thatwere immobilized in 3% methyl cellulose using 30 gauge needles andcollected in 20 µl RLT Buffer (Qiagen), and RNAwas isolated by extractionin TRIzol Reagent (Life Technologies) as described (Passeri et al., 2009).RNA (250 ng) was reverse transcribed using the SuperScript cDNAsynthesis kit (Quanta) as per the manufacturer’s instructions. qRT-PCR wasperformed using PerfeCTa SYBR Green Fast Mix (Quanta). Samples wererun in triplicate on the Roche LightCycler 480 (Roche), with at least threeindependent samples analyzed for each experiment. Target gene expressionwas normalized to ribosomal protein large P0 (rplp0) using the comparativethreshold cycle (ΔΔCt) method. Expression in treated animals wasnormalized to untreated controls from the same clutch. Primer sequencesare listed in Table S6.

mRNA sequencing (RNAseq)Total RNAwas extracted from livers dissected from 5 dpf Tg(fabp10a:nls-mCherrymss4Tg) zebrafish larvae using the Zymo Quick-RNA Micro Kit(R1050, Zymo Research) as per the manufacturer’s instructions. RNAquality was determined by Agilent 2100 BioAnalyzer. RNAseq librarieswere prepared according to Illumina TruSeq RNA sample preparationversion 2 protocol, and library quality was analyzed on the Agilent 2100BioAnalyzer. cDNA libraries were sequenced on the Illumina Hiseq 2500platform to obtain 100-base-pair paired-end reads. Sequencing quality wasassessed by using FASTQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and reads were quality trimmed using Trimmomatic(Bolger et al., 2014) to remove low Q scores, adapter contamination andsystematic sequencing errors. Reads were aligned to the Danio rerioGRCz10 reference genome assembly with Tophat 2.0.9 (Trapnell et al.,2009). To estimate gene expression, read counts were calculated by HTSeqwith ensemble annotation (Aken et al., 2016; Anders et al., 2015; Trapnellet al., 2009). Normalization and test of differential expression using ageneralized linear model were implemented in DESeq2 (Gentleman et al.,2004; Love et al., 2014). Adjusted P-value (FDR) <0.05 was treated assignificantly different expression. We established a method to identify ‘on-off’ genes as those that were expressed in controls but not in theexperimental condition, or vice versa. These were designated as geneswith read counts greater than or equal to 20 in one condition and less than orequal to 5 in the other. RNAseq datasets were submitted to Gene ExpressionOmnibus (GEO) with the access number GSE104953. RNAseq from liversof Tg(fabp10a:nls-mCherrymss4Tg) larvae exposed to 2% ethanol from 96 to132 hpf was previously described (Howarth et al., 2014; GSE56498), as

were the unexposed 5 dpf controls (GSE52605; Mudbhary et al., 2014). Thedata was realigned and reanalyzed for this study.

Qualitative and quantitative assessment of iAs content inzebrafish larvaeTotal iAs analysis was carried out by ICP-MS on pools of 20 dissectedlivers from 5-dpf larvae in 20 μl deionized water followed by 0.5 ml ofconcentrated double-distilled nitric acid. After an overnight digestion atroom temperature, samples were diluted to 5 ml and analyzed using ICP-MS (Agilent 8800-QQQ, Wilmington, DE). iAs concentration wasmeasured in MS/MS mode using oxygen as the cell gas and tellurium asthe internal standard. Quality control (QC) and quality assuranceprocedures included analyses of procedural blanks and QC standards.Lab recovery rates for QC standards with this method were 90-110% and<5% precision (given as %RSD). The limit of detection for arsenic was0.02 ng ml−1.

For tissue mapping by LA-ICP-MS, larvae were washed with embryowater and fixed in 4% paraformaldehyde (PFA) in phosphate buffered saline(PBS) overnight at 4°C. They were transferred to 30% sucrose in PBS andembedded in a 2:1 mixture OCT (Tissue Tek):30% sucrose. Cryosections of10 μmwere cut on a Leica CM3050S cryostat, and were thawed and warmedto room temperature prior to analysis.

A New Wave Research NWR-193 (ESI, OR, USA) laser ablation unitequipped with a 193 nm ArF excimer laser was connected to an AgilentTechnologies 8800 triple-quad ICP-MS (Agilent Technologies, CA, USA).Helium was used as the carrier gas from the laser ablation cell and mixedwith argon via Y-piece before introduction to the ICP-MS. The system wastuned daily using NIST SRM 612 (trace elements in glass) to monitorsensitivity (maximum analyte ion counts), oxide formation (232Th16O+/232Th+, <0.3%) and fractionation (232Th+/238U+, 100±5%). The laserwas scanned across the tissue sections at 20 μm spot size, 40 μm/s scanspeed, 40 Hz repetition rate and approximately 0.2 J cm2 power. ICP-MSintegration times for analytes were adjusted so that total scan time was 0.5 s,maintaining the dimensions of the tissue in data analysis. LA-ICP-MSoperating parameters are given in Table S2.

Lipid stainingLarvaewere fixed in 4% PFA overnight at 4°C and stained with Oil Red O aspreviously described (Passeri et al., 2009; Vacaru et al., 2014). Sampleswere blinded before scoring as positive (presence of 3 or more lipid dropletsper liver) or negative for steatosis. The average steatosis incidence wascalculated for at least 3 clutches per condition and, on average, 15-20 larvaeper condition per experiment were scored (Tables S3 and S6).

Statistical analysisData were analyzed using GraphPad Prism 7 software, and interactionbetween iAs and ethanol in steatosis was modeled in SAS version 9.4 (SASInstitute Inc., Cary, NC). Data are expressed as means±s.d. Differencesbetween experimental groups were analyzed by unpaired Student’s t-test(with correction for multiple comparisons where applicable), or byone-way analysis of variance (ANOVA) followed by Dunnet’s or Tukeypost hoc correction when more than two groups were compared. P-values<0.05 were considered statistically significant unless otherwise noted.Differentially expressed genes from RNAseq data were analyzed usinga package DESeq2 in Bioconductor (Gentleman et al., 2004; Love,2014), which embedded a generalized linear model for test of differentialexpression. Adjusted P-value (FDR) <0.05 was treated as significantlydifferential expression.

AcknowledgementsThe authors thank members of the Sadler and Chu laboratories for their closereading of this manuscript; Dr Jeanette Stingone for statistical expertise; ShikhaNayar, Liam Gordon, Leonore Wunsche and Priyanka Lakhiani for technicalassistance; Jill Gregory for graphic design of our working model; Dr ChristophBuettner, Dr Scott L. Friedman and Dr Robert O. Wright for scientific input; and theCCMS staff at Icahn School of Medicine at Mount Sinai for zebrafish facilitymaintenance. We thank the NYUAD High Performance Computing, SequencingCenter and Core Bioinformatics team at theGenomics and Systems Biology Institutefor assistance with sequencing of the arsenic samples.

10

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: K.B., J.C.; Methodology: K.B., C.Z., C. Austin,C. Amarasiriwardena, M.A., K.C.S.; Formal analysis: K.B., C.Z., C. Austin,C. Amarasiriwardena; Investigation: K.B., C. Austin, C. Amarasiriwardena, K.C.S.;Resources: K.C.S.; Writing - original draft: K.B., K.C.S.; Writing - review & editing:K.B., C.Z., C. Austin, C. Amarasiriwardena, M.A., J.C., K.C.S.; Visualization: K.B.,C. Austin; Supervision: M.A., J.C., K.C.S.; Project administration: K.C.S.; Fundingacquisition: K.B., K.C.S.

FundingThis work was funded by National Institute on Alcohol Abuse and Alcoholism(5RO1AA018886 to K.C.S.), Eunice Kennedy Shriver National Institute of ChildHealth and Human Development (T32 HD049311-09) and National Institute ofEnvironmental Health Sciences (P30ES023515 to K.B.).

Data availabilityRNAseq datasets were submitted to Gene Expression Omnibus (GEO) with theaccess number GSE104953.

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.031575.supplemental

ReferencesAdams, L. A. and Lindor, K. D. (2007). Nonalcoholic fatty liver disease. Ann.Epidemiol. 17, 863-869.

Adams, S. V., Barrick, B., Christopher, E. P., Shafer, M. M., Song, X., Vilchis, H.,Newcomb, P. A. and Ulery, A. (2016). Urinary heavy metals in hispanics 40–85years old in don a ana county, new mexico. Arch. Environ. Occup. Health 71,338-346.

Adebayo, A. O., Zandbergen, F., Kozul-Horvath, C. D., Gruppuso, P. A. andHamilton, J. W. (2015). Chronic exposure to low-dose arsenic modulateslipogenic gene expression in mice. J. Biochem. Mol. Toxicol. 29, 1-9.

Adeyemi, J. A., da Cunha Martins-Junior, A. and Barbosa, F.Jr. (2015).Teratogenicity, genotoxicity and oxidative stress in zebrafish embryos (danio rerio)co-exposed to arsenic and atrazine. Comp. Biochem. Physiol. C Toxicol.Pharmacol. 172-173, 7-12.

Ahsan, H., Chen, Y., Parvez, F., Argos, M., Hussain, A. I., Momotaj, H., Levy, D.,Van Geen, A., Howe, G. and Graziano, J. (2006). Health effects of arseniclongitudinal study (heals): Description of a multidisciplinary epidemiologicinvestigation. J. Expo. Sci. Environ. Epidemiol. 16, 191-205.

Aken, B. L., Ayling, S., Barrell, D., Clarke, L., Curwen, V., Fairley, S., FernandezBanet, J., Billis, K., Garcıa Giron, C., Hourlier, T. et al. (2016). The ensemblgene annotation system. Database (Oxford) 2016, baw093.

Al-Eryani, L., Wahlang, B., Falkner, K. C., Guardiola, J. J., Clair, H. B., Prough,R. A. and Cave, M. (2015). Identification of environmental chemicals associatedwith the development of toxicant-associated fatty liver disease in rodents. Toxicol.Pathol. 43, 482-497.

Anders, S., Pyl, P. T. and Huber, W. (2015). Htseq–a python framework to workwith high-throughput sequencing data. Bioinformatics 31, 166-169.

Anelli, T., Bergamelli, L., Margittai, E., Rimessi, A., Fagioli, C., Malgaroli, A.,Pinton, P., Ripamonti, M., Rizzuto, R. and Sitia, R. (2011). Ero1α regulatesca2+ fluxes at the endoplasmic reticulum–mitochondria interface (mam). AntioxidRedox Signal. 16, 1077-1087.

Anstee, Q. M. and Day, C. P. (2015). The genetics of nonalcoholic fatty liverdisease: Spotlight on pnpla3 and tm6sf2. Semin. Liver Dis. 35, 270-290.

Argos, M., Kalra, T., Rathouz, P. J., Chen, Y., Pierce, B., Parvez, F., Islam, T.,Ahmed, A., Rakibuz-Zaman, M., Hasan, R. et al. (2010). Arsenic exposure fromdrinking water, and all-cause and chronic-disease mortalities in bangladesh(heals): a prospective cohort study. Lancet 376, 252-258.

Ashraf, N. U. and Sheikh, T. A. (2015). Endoplasmic reticulum stress and oxidativestress in the pathogenesis of non-alcoholic fatty liver disease. Free Radic. Res.49, 1405-1418.

Bambino, K. and Chu, J. (2017). Chapter nine-zebrafish in toxicology andenvironmental health. In. Current Topics in Developmental Biology, Vol. 124 (ed.C. S. Kirsten), pp. 331-367. Oxford: Academic Press.

Benham, A. M., Van Lith, M., Sitia, R. and Braakman, I. (2013). Ero1–pdiinteractions, the response to redox flux and the implications for disulfide bondformation in the mammalian endoplasmic reticulum. Philos. Trans. R. Soc. B Biol.Sci. 368, 20110403.

Bolger, A. M., Lohse, M. and Usadel, B. (2014). Trimmomatic: a flexible trimmer forillumina sequence data. Bioinformatics 30, 2114-2120.

Browning, J. D. and Horton, J. D. (2004). Molecular mediators of hepatic steatosisand liver injury. J. Clin. Investig. 114, 147-152.

Carlson, P. and Van Beneden, R. J. (2014). Arsenic exposure alters expression ofcell cycle and lipid metabolism genes in the liver of adult zebrafish (danio rerio).Aquatic Toxicol. 153, 66-72.

Carreras-Sureda, A., Pihan, P. and Hetz, C. (2017). The unfolded proteinresponse: At the intersection between endoplasmic reticulum function andmitochondrial bioenergetics. Front. Oncol. 7, 55.

Cederbaum, A. I. (2010). Role of cyp2e1 in ethanol-induced oxidant stress, fattyliver and hepatotoxicity. Dig. Dis. 28, 802-811.

Centeno, J. A., Mullick, F. G., Martinez, L., Page, N. P., Gibb, H., Longfellow, D.,Thompson, C. and Ladich, E. R. (2002). Pathology related to chronic arsenicexposure. Environ. Health Perspect. 110 Suppl. 5, 883-886.

Chen, Y., Wu, F., Liu, M., Parvez, F., Slavkovich, V., Eunus, M., Ahmed, A.,Argos, M., Islam, T., Rakibuz-Zaman, M. et al. (2013). A prospective study ofarsenic exposure, arsenic methylation capacity, and risk of cardiovasculardisease in bangladesh. Environ. Health Perspect. 121, 832-838.

Cheng, J., Lv, S., Nie, S., Liu, J., Tong, S., Kang, N., Xiao, Y., Dong, Q., Huang, C.and Yang, D. (2016). Chronic perfluorooctane sulfonate (pfos) exposure induceshepatic steatosis in zebrafish. Aquatic Toxicol. 176, 45-52.

Chernyavskaya, Y., Mudbhary, R., Zhang, C., Tokarz, D., Jacob, V., Gopinath,S., Sun, X., Wang, S., Magnani, E., Madakashira, B. P. et al. (2017). Loss ofDNA methylation in zebrafish embryos activates retrotransposons to triggerantiviral signaling. Development 144, 2925-2939.

Cinaroglu, A., Gao, C., Imrie, D. and Sadler, K. C. (2011). Activating transcriptionfactor 6 plays protective and pathological roles in steatosis due to endoplasmicreticulum stress in zebrafish. Hepatology 54, 495-508.

Das, K., Das, K., Mukherjee, P. S., Ghosh, A., Ghosh, S., Mridha, A. R., Dhibar,T., Bhattacharya, B., Bhattacharya, D., Manna, B. et al. (2010). Nonobesepopulation in a developing country has a high prevalence of nonalcoholic fatty liverand significant liver disease. Hepatology 51, 1593-1602.

Das, N., Paul, S., Chatterjee, D., Banerjee, N., Majumder, N. S., Sarma, N., Sau,T.J., Basu, S., Banerjee, S., Majumder, P. et al. (2012). Arsenic exposurethrough drinking water increases the risk of liver and cardiovascular diseases inthe population of west bengal, india. BMC Public Health 12, 639.

Das, N., Giri, A., Chakraborty, S. and Bhattacharjee, P. (2016). Association ofsingle nucleotide polymorphism with arsenic-induced skin lesions and geneticdamage in exposed population of west bengal, india. Mutat. Res. 809, 50-56.

Day, C. P. and James, O. F. W. (1998). Steatohepatitis: a tale of two “hits”?Gastroenterology 114, 842-845.

Ditzel, E. J., Nguyen, T., Parker, P. and Camenisch, T. D. (2016). Effects ofarsenite exposure during fetal development on energy metabolism andsusceptibility to diet-induced fatty liver disease in male mice. Environ. HealthPerspect. 124, 201-209.

Doudican, N. A., Wen, S. Y., Mazumder, A. and Orlow, S. J. (2012). Sulforaphanesynergistically enhances the cytotoxicity of arsenic trioxide in multiple myelomacells via stress-mediated pathways. Oncol. Rep. 28, 1851-1858.

Dowman, J. K., Tomlinson, J. W. and Newsome, P. N. (2010). Pathogenesis ofnon-alcoholic fatty liver disease. QJM 103, 71-83.

Engstrom, K., Vahter, M., Mlakar, S. J., Concha, G., Nermell, B., Raqib, R.,Cardozo, A. and Broberg, K. (2011). Polymorphisms in arsenic(+iii oxidationstate) methyltransferase (as3mt) predict gene expression of as3mt as well asarsenic metabolism. Environ. Health Perspect. 119, 182-188.

Engstrom, K. S., Hossain, M. B., Lauss,M., Ahmed, S., Raqib, R., Vahter, M. andBroberg, K. (2013). Efficient arsenic metabolism–the as3mt haplotype isassociated with DNA methylation and expression of multiple genes aroundas3mt. PLoS ONE 8, e53732.

Farzan, S. F., Gossai, A., Chen, Y., Chasan-Taber, L., Baker, E. and Karagas, M.(2016). Maternal arsenic exposure and gestational diabetes and glucoseintolerance in the New Hampshire birth cohort study. Envir. Health 15, 106.

Garcia, G. R., Noyes, P. D. and Tanguay, R. L. (2016). Advancements in zebrafishapplications for 21st century toxicology. Pharmacol. Ther. 161, 11-21.

Gentleman, R. C., Carey, V. J., Bates, D. M., Bolstad, B., Dettling, M., Dudoit, S.,Ellis, B., Gautier, L., Ge, Y., Gentry, J. et al. (2004). Bioconductor: open softwaredevelopment for computational biology and bioinformatics. Genome Biol. 5,R80-R80.

Goessling, W. and Sadler, K. C. (2015). Zebrafish: an important tool for liverdisease research. Gastroenterology 149, 1361-1377.

Hallauer, J., Geng, X., Yang, H. C., Shen, J., Tsai, K. J. and Liu, Z. (2016). Theeffect of chronic arsenic exposure in zebrafish. Zebrafish 13, 405-412.

Hamdi, M., Sanchez, M. A., Beene, L. C., Liu, Q., Landfear, S. M., Rosen, B. P.and Liu, Z. (2009). Arsenic transport by zebrafish aquaglyceroporins. BMC Mol.Biol. 10, 104.

Hamdi, M., Yoshinaga, M., Packianathan, C., Qin, J., Hallauer, J., McDermott,J. R., Yang, H.-C., Tsai, K.-J. and Liu, Z. (2012). Identification of an s-adenosylmethionine (sam) dependent arsenic methyltransferase in danio rerio.Toxicol. Appl. Pharmacol. 262, 185-193.

Hare, D., Austin, C. and Doble, P. (2012). Quantification strategies for elementalimaging of biological samples using laser ablation-inductively coupled plasma-mass spectrometry. Analyst 137, 1527-1537.

11

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

Hopenhayn-Rich, C., Biggs, M. L., Smith, A. H., Kalman, D. A. and Moore, L. E.(1996). Methylation study of a population environmentally exposed to arsenic indrinking water. Environ. Health Perspect. 104, 620-628.

Howarth, D. L., Vacaru, A. M., Tsedensodnom, O., Mormone, E., Nieto, N.,Costantini, L. M., Snapp, E. L. and Sadler, K. C. (2012). Alcohol disruptsendoplasmic reticulum function and protein secretion in hepatocytes. Alcohol.Clin. Exp. Res. 36, 14-23.

Howarth, D. L., Yin, C., Yeh, K. and Sadler, K. C. (2013). Defining hepaticdysfunction parameters in two models of fatty liver disease in zebrafish larvae.Zebrafish 10, 199-210.

Howarth, D. L., Lindtner, C., Vacaru, A. M., Sachidanandam, R.,Tsedensodnom, O., Vasilkova, T., Buettner, C. and Sadler, K. C. (2014).Activating transcription factor 6 is necessary and sufficient for alcoholic fatty liverdisease in zebrafish. PLoS Genet. 10, e1004335.

Huang, C., Ke, Q., Costa, M. and Shi, X. (2004). Molecular mechanisms of arseniccarcinogenesis. Mol. Cell. Biochem. 255, 57-66.

Hudson, D. A., Gannon, S. A. and Thorpe, C. (2015). Oxidative protein folding:From thiol-disulfide exchange reactions to the redox poise of the endoplasmicreticulum. Free Radic. Biol. Med. 80, 171-182.

Hughes, M. F. (2002). Arsenic toxicity and potential mechanisms of action. Toxicol.Lett. 133, 1-16.

Islam, K., Haque, A., Karim, R., Fajol, A., Hossain, E., Salam, K. A., Ali, N., Saud,Z. A., Rahman, M., Rahman, M. et al. (2011). Dose-response relationshipbetween arsenic exposure and the serum enzymes for liver function tests in theindividuals exposed to arsenic: A cross sectional study in Bangladesh. Environ.Health 10, 64.

Jacob, V., Chernyavskaya, Y., Chen, X., Tan, P. S., Kent, B., Hoshida, Y. andSadler, K. C. (2015). DNA hypomethylation induces aDNA replication-associatedcell cycle arrest to block hepatic outgrowth in uhrf1 mutant zebrafish embryos.Development 142, 510-521.

Jacobson, T., Navarrete, C., Sharma, S. K., Sideri, T. C., Ibstedt, S., Priya, S.,Grant, C. M., Christen, P., Goloubinoff, P. and Tamas, M. J. (2012). Arseniteinterferes with protein folding and triggers formation of protein aggregates in yeast.J. Cell Sci. 125, 5073-5083.

Ji, C., Kaplowitz, N., Lau, M. Y., Kao, E., Petrovic, L. M. and Lee, A. S. (2011).Liver-specific loss of glucose-regulated protein 78 perturbs the unfolded proteinresponse and exacerbates a spectrum of liver diseases in mice. Hepatology 54,229-239.

Jomova, K., Jenisova, Z., Feszterova, M., Baros, S., Liska, J., Hudecova, D.,Rhodes, C. J. and Valko, M. (2011). Arsenic: Toxicity, oxidative stress andhuman disease. J. Appl. Toxicol. 31, 95-107.

Kile, M. L., Baccarelli, A., Hoffman, E., Tarantini, L., Quamruzzaman, Q.,Rahman, M., Mahiuddin, G., Mostofa, G., Hsueh, Y.-M., Wright, R. O. et al.(2012). Prenatal arsenic exposure and DNAmethylation in maternal and umbilicalcord blood leukocytes. Environ. Health Perspect. 120, 1061-1066.

Kuo, C.-C., Howard, B. V., Umans, J. G., Gribble, M. O., Best, L. G., Francesconi,K. A., Goessler, W., Lee, E., Guallar, E. and Navas-Acien, A. (2015). Arsenicexposure, arsenic metabolism, and incident diabetes in the strong heart study.Diabetes Care 38, 620-627.

Lam, S. H., Winata, C. L., Tong, Y., Korzh, S., Lim, W. S., Korzh, V. Spitsbergen,J., Mathavan, S., Miller, L. D., Liu, E. T. et al. (2006). Transcriptome kinetics ofarsenic-induced adaptive response in zebrafish liver. Physiol. Genomics 27,351-361.

Levene, A. P. and Goldin, R. D. (2012). The epidemiology, pathogenesis andhistopathology of fatty liver disease. Histopathology 61, 141-152.

Li, D., Lu, C., Wang, J., Hu, W., Cao, Z., Sun, D., Xia, H. and Ma, X. (2009).Developmental mechanisms of arsenite toxicity in zebrafish (danio rerio)embryos. Aquatic Toxicol. 91, 229-237.

Li, X., Ma, Y., Li, D., Gao, X., Li, P., Bai, N., Luo, M., Tan, X., Lu, C. and Ma, X.(2012). Arsenic impairs embryo development via down-regulating dvr1 expressionin zebrafish. Toxicol. Lett. 212, 161-168.

Li, J., Duan, X., Dong, D., Zhang, Y., Li, W., Zhao, L., Nie, H., Sun, G. and Li, B.(2015). Hepatic and nephric nrf2 pathway up-regulation, an early antioxidantresponse, in acute arsenic-exposed mice. Int. J. Environ. Res. Public Health 12,12628-12642.

Li, C., Li, P., Tan, Y. M., Lam, S. H., Chan, E. C. and Gong, Z. (2016). Metabolomiccharacterizations of liver injury caused by acute arsenic toxicity in zebrafish. PLoSONE 11, e0151225.

Liu, L. Y., Fox, C. S., North, T. E. and Goessling, W. (2013). Functional validationof gwas gene candidates for abnormal liver function during zebrafish liverdevelopment. Dis. Model. Mech. 6, 1271-1278.

Louvet, A. and Mathurin, P. (2015). Alcoholic liver disease: mechanisms of injuryand targeted treatment. Nat. Rev. Gastroenterol. Hepatol. 12, 231-242.

Love, M. I., Huber, W. and Anders, S. (2014). Moderated estimation of fold changeand dispersion for rna-seq data with deseq2. Genome Biol. 15, 550.

Lu, Y. and Cederbaum, A. I. (2008). Cyp2e1 and oxidative liver injury by alcohol.Free Radic. Biol. Med. 44, 723-738.

Luz, A. T., Godebo, T. R., Bhatt, D. P., Ilkayeva, O. R., Maurer, L. L., Hirschey,M. D. and Meyer, J. N. (2016). Arsenite uncouples mitochondrial respiration and

induces a warburg-like effect in caenorhabditis elegans. Toxicol. Sci. 154,195-195.

Ma, Y., Zhang, C., Gao, X. B., Luo, H. Y., Chen, Y., Li, H. H., Ma, X. and Lu, C. L.(2015). Folic acid protects against arsenic-mediated embryo toxicity by up-regulating the expression of dvr1. Sci. Rep. 5, 16093.

Magdaleno, F., Blajszczak, C. C. and Nieto, N. (2017). Key events participating inthe pathogenesis of alcoholic liver disease. Biomolecules 7, 9.

Malhotra, J. D. and Kaufman, R. J. (2007). Endoplasmic reticulum stress andoxidative stress: a vicious cycle or a double-edged sword? Antioxid Redox Signal.9, 2277-2294.

Mauro, M., Caradonna, F. and Klein, C. B. (2016). Dysregulation of DNAmethylation induced by past arsenic treatment causes persistent genomicinstability in mammalian cells. Environ. Mol. Mutagen. 57, 137-150.

Mazumder, D. N. (2005). Effect of chronic intake of arsenic-contaminated water onliver. Toxicol. Appl. Pharmacol. 206, 169-175.

McCollum, C. W., Hans, C., Shah, S., Merchant, F. A., Gustafsson, J.-A. andBondesson, M. (2014). Embryonic exposure to sodium arsenite perturbsvascular development in zebrafish. Aquatic Toxicol. 152, 152-163.

Moon, K. A., Guallar, E., Umans, J. G., Devereux, R. B., Best, L. G., Francesconi,K. A., Goessler, W., Pollak, J., Silbergeld, E. K., Howard, B. V. et al. (2013).Association between exposure to low to moderate arsenic levels and incidentcardiovascular disease. A prospective cohort study. Ann. Intern. Med. 159,649-659.

Mudbhary, R., Hoshida, Y., Chernyavskaya, Y., Jacob, V., Villanueva, A., Fiel,M. I., Chen, X., Kojima, K., Thung, S., Bronson, R. T. et al. (2014). Uhrf1overexpression drives DNA hypomethylation and hepatocellular carcinoma.Cancer Cell 25, 196-209.

Naujokas, M. F., Anderson, B., Ahsan, H., Aposhian, H. V., Graziano, J. H.,Thompson, C. and Suk, W. A. (2013). The broad scope of health effects fromchronic arsenic exposure: Update on a worldwide public health problem. Environ.Health Perspect. 121, 295-302.

O’Connell, J. D., Tsechansky, M., Royall, A., Boutz, D. R., Ellington, A. D. andMarcotte, E. M. (2014). A proteomic survey of widespread protein aggregation inyeast. Mol. Biosyst. 10, 851-861.

Ozcan, U., Cao, Q., Yilmaz, E., Lee, A. H., Iwakoshi, N. N. andOzdelen, E. (2004).Endoplasmic reticulum stress links obesity, insulin action, and type 2 diabetes.Science 306, 457-461.

Passeri, M. J., Cinaroglu, A., Gao, C. and Sadler, K. C. (2009). Hepatic steatosisin response to acute alcohol exposure in zebrafish requires sterol regulatoryelement binding protein activation. Hepatology 49, 443-452.

Ramadan, D., Rancy, P. C., Nagarkar, R. P., Schneider, J. P. and Thorpe, C.(2009). Arsenic(iii) species inhibit oxidative protein folding in vitro. Biochemistry48, 424-432.

Reichl, F.-X., Szinicz, L., Kreppel, H. and Forth, W. (1988). Effect of arsenic oncarbohydrate metabolism after single or repeated injection in guinea pigs. Arch.Toxicol. 62, 473-475.

Santra, A., Das Gupta, J., De, B. K., Roy, B. and Guha Mazumder, D. N. (1999).Hepatic manifestations in chronic arsenic toxicity. Indian J. Gastroenterol. 18,152-155.

Santra, A., Maiti, A., Chowdhury, A. and Mazumder, D. N. (2000a). Oxidativestress in liver of mice exposed to arsenic-contaminated water. IndianJ. Gastroenterol. 19, 112-115.

Santra, A., Maiti, A., Das, S., Lahiri, S., Charkaborty, S. K. and Mazumder, D. N.and Guha Mazumder, D. (2000b). Hepatic damage caused by chronic arsenictoxicity in experimental animals. J. Toxicol. Clin. Toxicol. 38, 395-405.

Santra, A., Chowdhury, A., Ghatak, S., Biswas, A. and Dhali, G. K. (2007).Arsenic induces apoptosis in mouse liver is mitochondria dependent and isabrogated by n-acetylcysteine. Toxicol. Appl. Pharmacol. 220, 146-155.

Sapra, A., Ramadan, D. and Thorpe, C. (2015). Multivalency in the inhibition ofoxidative protein folding by arsenic(iii) species. Biochemistry 54, 612-621.

Sattar, A., Xie, S., Hafeez, M. A., Wang, X., Hussain, H. I., Iqbal, Z., Pan, Y., Iqbal,M., Shabbir, M. A. and Yuan, Z. (2016). Metabolism and toxicity of arsenicals inmammals. Environ. Toxicol. Pharmacol. 48, 214-224.

Seervi, M., Sobhan, P. K., Joseph, J., AnnMathew, K. and Santhoshkumar, T. R.(2013). Ero1α-dependent endoplasmic reticulum–mitochondrial calcium fluxcontributes to er stress and mitochondrial permeabilization by procaspase-activating compound-1 (pac-1). Cell Death Dis. 4, e968.

Seok, S.-H., Baek, M.-W., Lee, H.-Y., Kim, D.-J., Na, Y.-R., Noh, K.-J., Park, S.-H.,Lee, H.-K., Lee, B.-H. and Ryu, D.-Y. (2007). Quantitative gfp fluorescence as anindicator of arsenite developmental toxicity in mosaic heat shock protein 70transgenic zebrafish. Toxicol. Appl. Pharmacol. 225, 154-161.

Shi, H., Shi, X. and Liu, K. J. (2004). Oxidative mechanism of arsenic toxicity andcarcinogenesis. Mol. Cell. Biochem. 255, 67-78.

Shoulders, M. D., Ryno, L. M., Genereux, J. C., Moresco, J. J., Tu, P. G., Wu, C.,Yates, J. R., Su, A. I., Kelly, J. W. and Wiseman, R. L. (2013). Stress-independent activation of xbp1s and/or atf6 reveals three functionally diverse erproteostasis environments. Cell Rep. 3, 1279-1292.

Sinha, D., Biswas, J. and Bishayee, A. (2013). Nrf2-mediated redox signaling inarsenic carcinogenesis: a review. Arch. Toxicol. 87, 383-396.

12

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms

States, J. C., Barchowsky, A., Cartwright, I. L., Reichard, J. F., Futscher, B. W.and Lantz, R. C. (2011). Arsenic toxicology: translating between experimentalmodels and human pathology. Environ. Health Perspect. 119, 1356-1363.

Sussulini, A., Becker, J. S. and Becker, J. S. (2017). Laser ablation icp-ms:Application in biomedical research. Mass Spectrom. Rev. 36, 47-57.

Tan, M., Schmidt, R. H., Beier, J. I., Watson, W. H., Zhong, H., States, J. C. andArteel, G. E. (2011). Chronic subhepatotoxic exposure to arsenic enhanceshepatic injury caused by high fat diet in mice. Toxicol. Appl. Pharmacol. 257,356-364.

Thakur, P. C., Stuckenholz, C., Rivera, M. R., Davison, J. M., Yao, J. K.,Amsterdam, A. Sadler, K. C. and Bahary, N. (2011). Lack of de novophosphatidylinositol synthesis leads to endoplasmic reticulum stress andhepatic steatosis in cdipt-deficient zebrafish. Hepatology 54, 452-462.

Thomas, D. J., Waters, S. B. and Styblo, M. (2004). Elucidating the pathway forarsenic methylation. Toxicol. Appl. Pharmacol. 198, 319-326.

Trapnell, C., Pachter, L. and Salzberg, S. L. (2009). Tophat: discovering splicejunctions with rna-seq. Bioinformatics 25, 1105-1111.

Tsedensodnom, O., Vacaru, A. M., Howarth, D. L., Yin, C. and Sadler, K. C.(2013). Ethanol metabolism and oxidative stress are required for unfolded proteinresponse activation and steatosis in zebrafish with alcoholic liver disease. Dis.Model. Mech. 6, 1213-1226.

Tseng, C.-H. (2007). Arsenic methylation, urinary arsenic metabolites and humandiseases: current perspective. J. Environ. Sci. Health Part C 25, 1-22.

Tseng, C.-H. (2009). A review on environmental factors regulating arsenicmethylation in humans. Toxicol. Appl. Pharmacol. 235, 338-350.

Vacaru, A. M., Di Narzo, A. F., Howarth, D. L., Tsedensodnom, O., Imrie, D.,Cinaroglu, A., Amin, S., Hao, K. and Sadler, K. C. (2014). Molecularly definedunfolded protein response subclasses have distinct correlations with fatty liverdisease in zebrafish. Dis. Model. Mech. 7, 823-835.

Vahter, M. (1999). Methylation of inorganic arsenic in different mammalian speciesand population groups. Sci. Prog. 82, 69-88.

Wahlang, B., Beier, J. I., Clair, H. B., Bellis-Jones, H. J., Falkner, K. C., Mcclain,C. J. and Cave, M. C. (2013). Toxicant-associated steatohepatitis. Toxicol.Pathol. 41, 343-360.

Wang, S. and Kaufman, R. J. (2014). How does protein misfolding in theendoplasmic reticulum affect lipid metabolism in the liver? Curr. Opin. Lipidol. 25,125-132.

Wang, M. and Kaufman, R. J. (2016). Protein misfolding in the endoplasmicreticulum as a conduit to human disease. Nature 529, 326-335.

Wang, Y.-H., Chen, Y.-H., Wu, T.-N., Lin, Y.-J. and Tsai, H.-J. (2006). A keratin 18transgenic zebrafish tg(k18(2.9):Rfp) treated with inorganic arsenite revealsvisible overproliferation of epithelial cells. Toxicol. Lett. 163, 191-197.

Wang, W., Cheng, S. and Zhang, D. (2014a). Association of inorganic arsenicexposure with liver cancer mortality: a meta-analysis. Environ. Res. 135, 120-125.

Wang, W., Xie, Z., Lin, Y. and Zhang, D. (2014b). Association of inorganic arsenicexposure with type 2 diabetes mellitus: a meta-analysis. J. Epidemiol. CommunityHealth 68, 176-184.

Watanabe, T. and Hirano, S. (2013). Metabolism of arsenic and its toxicologicalrelevance. Arch. Toxicol. 87, 969-979.

Weng, C.-Y., Chiou, S.-Y., Wang, L., Kou, M.-C., Wang, Y.-J. and Wu, M.-J.(2014). Arsenic trioxide induces unfolded protein response in vascular endothelialcells. Arch. Toxicol. 88, 213-226.

Wu, R., Wu, X., Wang, H., Fang, X., Li, Y., Gao, L. Sun, G., Pi, J. and Xu, Y. (2017).Strain differences in arsenic-induced oxidative lesion via arsenic biomethylationbetween c57bl/6j and 129x1/svj mice. Sci. Rep. 7, 44424.

Xu, H., Lam, S. H., Shen, Y. and Gong, Z. (2013). Genome-wide identification ofmolecular pathways and biomarkers in response to arsenic exposure in zebrafishliver. PLoS ONE 8, e68737.

Xu, M., Rui, D., Yan, Y., Xu, S., Niu, Q., Feng, G., Wang, Y., Li, S. and Jing, M.(2017). Oxidative damage induced by arsenic in mice or rats: A systematic reviewand meta-analysis. Biol. Trace Elem. Res. 176, 154-175.

Yamamoto, K., Takahara, K., Oyadomari, S., Okada, T., Sato, T., Harada, and A.Mori, K. (2010). Induction of liver steatosis and lipid droplet formation in atf6alpha-knockout mice burdenedwith pharmacological endoplasmic reticulum stress.Mol.Biol. Cell 21, 2975-2986.

Yang, Q., Jung, H. B., Culbertson, C. W., Marvinney, R. G., Loiselle, M. C.,Locke, D. B., Cheek, H., Thibodeau, H. and Zheng, Y. (2009). Spatial pattern ofgroundwater arsenic occurrence and association with bedrock geology in greateraugusta, maine. Environ. Sci. Technol. 43, 2714-2719.

13

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm031575. doi:10.1242/dmm.031575

Disea

seModels&Mechan

isms