nature biotechnology • VOLUME 20 • NOVEMBER 2002 • www.nature.com/naturebiotechnology

RESEARCH ARTICLE

1140

Engineering tolerance and hyperaccumulation ofarsenic in plants by combining arsenate reductase

and γ-glutamylcysteine synthetase expressionOm Parkash Dhankher1, Yujing Li1, Barry P. Rosen2, Jin Shi2, David Salt3, Julie F. Senecoff1, Nupur A. Sashti1,

and Richard B. Meagher1*

Published online 7 October 2002; doi:10.1038/nbt747

We have developed a genetics-based phytoremediation strategy for arsenic in which the oxyanion arsenate istransported aboveground, reduced to arsenite, and sequestered in thiol–peptide complexes.The Escherichia coliarsC gene encodes arsenate reductase (ArsC), which catalyzes the glutathione (GSH)-coupled electrochemicalreduction of arsenate to the more toxic arsenite. Arabidopsis thaliana plants transformed with the arsC geneexpressed from a light-induced soybean rubisco promoter (SRS1p) strongly express ArsC protein in leaves, butnot roots, and were consequently hypersensitive to arsenate. Arabidopsis plants expressing the E. coli geneencoding γ-glutamylcysteine synthetase (γ-ECS) from a strong constitutive actin promoter (ACT2p) were moder-ately tolerant to arsenic compared with wild type. However, plants expressing SRS1p/ArsC and ACT2p/γ-ECStogether showed substantially greater arsenic tolerance than γ-ECS or wild-type plants. When grown on arsenic,these plants accumulated 4- to 17-fold greater fresh shoot weight and accumulated 2- to 3-fold more arsenic pergram of tissue than wild type or plants expressing γ-ECS or ArsC alone.This arsenic remediation strategy shouldbe applicable to a wide variety of plant species.

Arsenic is an extremely toxic metalloid pollutant that adversely affectsthe health of millions of people worldwide1. Inorganic arsenic species,which are considered very hazardous for human health, are classified asgroup A human carcinogens and cause skin lesions, lung, kidney, andliver cancers, and also damage to the nervous system2,3 (US Envir-onmental Protection Agency (EPA), 1996: www.epa.gov/ogwdw/ars/arsenic.html). Hundreds of Superfund sites in the United States are list-ed on the National Priority List (NPL) (www.epa.gov/superfund/sites/nl/info.html) as having unacceptably high levels of arsenic and recom-mended for cleanup. In the majority of cases, arsenic-contaminatedsites are not cleaned up because the cost in both dollars and environ-mental damage is too high. Physical remediation methods involvingsoil removal and burial are expensive, impractical on the scale that isneeded, and environmentally destructive.

Higher plants can extract pollutants from the soil or water throughtheir normal root uptake of nutrients4. They can store and concentratepollutant in their cells and/or convert toxic pollutants to less toxicforms5. Plant-based phytoremediation strategies for heavy metals relyon plant roots to extract, plant vascular systems to transport, and leavesas a sink to concentrate arsenic aboveground for harvest and process-ing. A native fern indigenous to the southern United States has recentlybeen characterized that hyperaccumulates arsenic to very high levels6.However, the genetic basis for its activity is unknown, and hence theenzymes responsible for hyperaccumulation are not yet available formanipulation into other species with wider geographic and ecologicaldistribution and greater biomass.

Most arsenic in surface soil and water exists primarily in its oxidizedform, the oxyanion arsenate (AsO4

3–), which is an analog of phosphate.Arsenate can potentially be taken up from soil and translocated up theplant vascular system along with phosphate7,8. In contrast, the reducedform of arsenic, arsenite (AsO3

3–), has a strong affinity toward thiolgroups and, once formed aboveground in leaf and stem tissues, shouldbe trapped as peptide–thiol complexes such as those formed by γ-glutamylcysteine (γ-EC; refs 9,10) as shown in Figure 1. Our workinghypothesis was, therefore, that controlling the electrochemical state ofarsenic in aboveground tissues and increasing thiol sinks throughoutthe plant would result in both resistance and hyperaccumulation ofarsenic. As a first step toward testing this hypothesis, we examined theeffects of coexpressing two bacterial genes, arsenate reductase (arsC)and γ-glutamylcysteine synthetase (γ-ECS), in Arabidopsis plants.

ResultsTransgenic plants expressing ArsC aboveground. Bacterial resistanceto arsenic is acquired by first reducing arsenate to arsenite using thearsenate reductase (ArsC) enzyme in a glutathione-dependent reduc-tion11, as shown in Figure 1. Because the prokaryotic arsC gene hasbeen previously shown to confer resistance in the eukaryoteSaccharomyces cerevisiae12, it was reasonable to consider that it wouldfunctionally express in plants. To test the first part of our workinghypothesis that deals with controlling the electrochemical state ofarsenic aboveground, the gene encoding this same bacterial ArsCenzyme was expressed under control of the regulatory sequences from

1Department of Genetics, University of Georgia, Athens, GA 30602. 2Department of Biochemistry and Molecular Biology, Wayne State University, Detroit, MI 48201.3Center for Plant Environmental Stress Physiology, 1165 Horticulture Building, Purdue University, West Lafayette, IN 47907.

*Corresponding author (meagher@arches.uga.edu).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

www.nature.com/naturebiotechnology • NOVEMBER 2002 • VOLUME 20 • nature biotechnology 1141

The strong aboveground specificity of ArsC expression and the lackof belowground expression were important in the proposed arsenicremediation strategy to allow arsenate to be transported aboveground.A transcriptional fusion between SRS1p and the β-glucuronidase(GUS) reporter gene also supported the aboveground leaf-specificexpression of the SRS1p (Fig. 3A). Very strong SRS1p/GUS expressionwas seen in all leaf and most stem tissues, but no expression was detect-ed in any hypocotyl or root tissues even after very close inspection ofseveral independent transgenic lines stained for GUS activity for pro-longed periods.

ArsC plants show enhanced arsenate sensitivity. Arabidopsis wild-type and three independent SRS1p/ArsC transgenic lines with low(SRS1p/ArsC2), medium (SRS1p/ArsC7), and high (SRS1p/ArsC9)levels of ArsC protein expression were tested for arsenic resistance orsensitivity by germination on medium containing 0, 75, and 150 µMarsenate. All three ArsC transgenic lines were hypersensitive toarsenic, whereas wild-type plants remained relatively healthy at theseconcentrations, as shown for SRS1p/ArsC9 in Figure 3B and C.Leaves of transgenic plants grown on 75 µM arsenate developedmore slowly than wild-type controls and turned yellow. When chal-lenged with 150 µM arsenate, the leaves expressing the SRS1p/ArsCtransgene were extremely stunted, whereas root growth was not sig-nificantly inhibited, as shown in Figure 3C (right panel). All of theArsC plants were extremely chlorotic after three weeks on 150 µMarsenate. Although wild-type plants grew at reduced rates, they allsurvived at these two concentrations of arsenate and had two- tothreefold higher fresh weights than the transgenic ArsC plants afterthree weeks of growth (Fig. 4A). There was no significant differencein fresh weight between transgenic lines and wild type when grownon medium not supplemented with arsenate. The sensitivity of theSRS1p/ArsC transgenic leaves to arsenate suggests that the bacterialarsenate reductase arsC gene is functional and, considering its activ-ity in bacteria, it makes leaves more sensitive to arsenate (AsO4

3–) byelectrochemically reducing it to the more toxic thiol-reactive formarsenite (AsO3

3–)14. The control 35Sp/ArsC lines expressing ArsCconstitutively were also more sensitive to arsenate than wild type, butgenerally less sensitive than the SRS1p/ArsC lines (not shown).

Plants coexpressing ArsC and γ-ECS. Another system used by bacte-ria and other organisms to resist thiol-reactive ions like arsenite

the well-characterized soybean ribulose bisphosphate carboxylasesmall-subunit (rubisco) SRS1 gene13, which shows strong light-inducedexpression in leaves and stems. The arsC gene was cloned as a transla-tional fusion to the ATG initiation codon of SRS1p and nos terminatorto make SRS1p/ArsC. For comparison, arsC was also cloned into a well-characterized constitutive plant viral cauliflower mosaic virus (CaMV)35S promoter and nos terminator (35Sp/ArsC). Arabidopsis thalianawas transformed with both constructs using vacuum infiltration, andthe T1 generation seeds were screened for a linked kanamycin resistancemarker. Five kanamycin-resistant Arabidopsis plants containing eacharsC construct were randomly selected after transformation. The linescontaining SRS1p-driven arsC were designated SRS1p/ArsC2,SRS1p/ArsC7, SRS1p/ArsC8, SRS1p/ArsC9, and SRS1p/ArsC10, andthe lines containing 35Sp-driven arsC were designated 35Sp/ArsC3,35Sp/ArsC4, 35Sp/ArsC7, 35Sp/ArsC8, and 35Sp/ArsC13. These trans-genic plants did not show any phenotypic differences from the untrans-formed Arabidopsis plants when grown without arsenic.

The organ-specific expression of ArsC protein was examined onwestern blots for both SRS1p/ArsC and 35Sp/ArsC constructs. Three ofthe five lines for both SRS1p and 35Sp promoter constructs are shownin Figure 2A and B, respectively. ArsC protein from the SRS1p/ArsCtransgene was expressed only in leaf tissues and not in roots (Fig. 2A),whereas protein from the 35Sp/ArsC construct was expressed equiva-lently in leaves and roots (Fig. 2B). Prolonged exposure of these andadditional blots analyzing SRS1p/ArsC expression still did not detectArsC in the roots.

Figure 1. ArsC- and γ-ECS-catalyzed reactions.The bacterial arsenatereductase (ArsC) catalyzes the electrochemical reduction of arsenate toarsenite.The bacterial γ-glutamylcysteine synthetase (γ-ECS) catalyzes theformation of γ-glutamylcysteine (γ-EC) from the amino acids glutamate andcysteine and is the committed step in the synthesis of glutathione (GSH)and phytochelatins (PCs; indicated by three arrows). Reduced arsenite canbind organic thiols (RS) such as those in γ-EC, GSH, and PCs.

Figure 2. Immunodetection of ArsC and γ-ECS proteins in transgenicplants. (A) Strong ArsC (16 kDa) protein expression was observed onwestern blots from leaves, but not roots of three independent linestransformed with the SRS1p/ArsC construct. Protein extracts from E. coliexpressing ArsC from pNA1 plasmid and wild-type (WT) plant extractsserve as positive and negative controls, respectively. (B) Strongexpression of ArsC protein was observed in both leaves and roots of threeindependent transgenic lines transformed with the 35Sp/ArsC construct.(C) Both γ-ECS (top panel) and ArsC (bottom panel) are assayed onwestern blots of protein extracts from transgenic ArsC9 parental plant lineexpressing ArsC alone, a ACT2p/ECS1 line expressing γ-ECS (57 kDa)alone, and ten lines generated by transforming the ArsC9 parental linewith ACT2p/γ-ECS (lines ArsC9 + ECS1–10). The western membranewas cut into two strips and reacted separately with ArsC- and γ-ECS-specific antisera. For (A–C), equal amounts (10 µg) of total protein wereresolved on a 12% (wt/vol) polyacrylamide gel by SDS–PAGE and blottedto membrane. Western blots of plant extracts were developed asdescribed in Bizily et al.34 after reacting with polyclonal antisera to ArsC35

and monoclonal antibodies to γ-ECS31. Equal loading of samples wasconfirmed by Coomassie staining of parallel samples on a separate gel.

A

B

C

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

nature biotechnology • VOLUME 20 • NOVEMBER 2002 • www.nature.com/naturebiotechnology1142

involves increasing the thiol-rich peptide content of cells and tissues. Inparticular, the enzymes in the pathway leading to the biosynthesis ofglutathione (GSH), such as γ-glutamylcysteine synthetase (γ-ECS),shown in Figure 1, play a role in metal ions resistance. For example,arsenite-resistant Leishmania often show amplification of the γ-ECSlocus along with other arsenic resistance genes15. Related work hasshown that transgenic plants expressing bacterial γ-ECS containedhigher levels of both peptides γ-EC (γ-glutamylcysteine) and GSH (glu-tathione) and accumulate more of the thiol-reactive metal ion Cd(II)than wild-type plants16,17. Building on this work for cadmium, we setout to test the second part of our working hypothesis, enhancing thearsenite-binding thiol–peptide sink in plants by expressing γ-ECS.

It was recently shown that plants expressing the bacterial γ-ECS geneunder the control of a strong constitutive actin (ACT2) gene promoter(ACT2p/γ-ECS) are moderately resistant to mercury and arsenic (Y. Li,O.P. Dhankher, and R.B. Meagher, manuscript in preparation). Wehave repeated these results for arsenate, selecting for the kanamycinresistance marker on the ACT2p/γ-ECS construct. One of the strongarsenate-resistant lines, ACT2p/ECS1, was selected as a control for fur-ther experiments. When the ACT2p/ECS1 line was grown on medium

containing 200 µM arsenate, these plants grew substantially better thanwild type, as shown in Figure 5B (left lower quadrant on each plate).

The expression of arsC and γ-ECS was combined in a number ofplant lines by retransforming the highly arsenate-sensitive arsC trans-genic plant line showing the highest levels of ArsC protein,SRS1p/ArsC 9, with ACT2p/γ-ECS construct. T3 generation homozy-gous SRS1p/ArsC9 plants were vacuum-infiltrated withAgrobacterium containing the ACT2p/γ-ECS construct. The seedsfrom transformed individual lines were germinated on MS mediumsupplemented with 250 µM arsenate for four weeks, and the trans-genic plants with both transgenes were selected for arsenate resistance.Transgenic plants (ACT2p/γ-ECS) expressing the bacterial γ-ECSenzyme survived and showed resistance to arsenate, whereas plantsexpressing only ArsC were sensitive to arsenate and died. As shown inFigure 5A, a few seedlings from each transformation showed signifi-cant resistance to arsenate. No arsenic-resistant seedlings wereobtained on 250 µM arsenate from control transformations with otherunrelated DNA constructs. Ten arsenate-resistant plants (ArsC9 + γ-ECS-1, ArsC9 + γ-ECS-2, etc.) were recovered from the arsenateselection, and these T1 plants were grown in soil as a source of T2 gen-eration seeds. Western analysis of these ten independent T1 plantsshowed strong expression of γ-ECS protein from the newly acquired

Figure 3. Arsenic-sensitive phenotype of ArsC-expressing Arabidopsis.(A) Leaf-specific expression of SRS1p/GUS reporter fusion (A3 and A4) iscompared to GUS-stained wild type (WT: A1 and A2), confirming thestrong aboveground light-induced expression from the rubisco SRS1promoter in two-week-old plants. Stained enlarged roots (inset box) fromWT and SRS1p/GUS plants are shown in panels A2 and A4, respectively.(B, C) ArsC-expressing Arabidopsis are hypersensitive to arsenate incomparison with wild type. (B) Arsenate sensitivity of the transgenic lineSRS1p/ArsC9 expressing ArsC from the SRS1p promoter compared withwild type (WT) on increasing concentrations of sodium arsenate (0, 75,and 150 µM) in 0.5× MS medium. (C) Enlargement of wild-type andtransgenic line SRS1p/ArsC9 grown vertically on 150 µM arsenate. Plantsin (B) and (C) were grown for three weeks on half-strength MS mediumwith the arsenic concentrations indicated.

A

B

C

Figure 4. Relative growth inhibition and arsenic speciation of ArsC-overexpressing plants. (A) Comparative growth inhibition of threeSRS1p/ArsC lines (SRS1p/ArsC2, SRS1p/ArsC7, and SRS1p/ArsC9)and the wild-type plants grown on the indicated concentrations (theaverage and s.e. values of three replicates of 30 seedlings for each line).(B) The percentage relative concentration of free AsO4

3–, AsO33–, and

As(III) tris-glutathione in leaf samples as determined by XANES at theStanford Synchrotron Radiation Laboratory (SSRL) following themethods described by Pickering et al9. Leaf tissues of wild-type plantsand the four SRS1p/ArsC Arabidopsis lines 2, 7, 8, and 9 grown for threeweeks on 75 µM arsenate were examined.

A

B

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

www.nature.com/naturebiotechnology • NOVEMBER 2002 • VOLUME 20 • nature biotechnology 1143

ACT2p/γ-ECS transgene (Fig. 2C). As expected, all these lines showedsimilar levels of strong ArsC expression to that in the commonparental SRS1p/ArsC9 line. In the absence of arsenate, these plantsappeared similar to wild-type plants, suggesting that neither insertionof these two foreign genes nor their expression was deleterious.

Increased arsenate resistance of ArsC/γ-ECS plants. Several doublytransgenic lines (ArsC9 + γ-ECS1, -2, -10) showed substantially moreresistance than wild-type or the γ-ECS-expressing plants. On 200 µMarsenate, they appeared dark green and vigorous with healthy shootsand roots and grew approximately at the same rate as unchallengedcontrols. Plants expressing γ-ECS alone showed much less resistance,whereas plants expressing ArsC alone were hypersensitive to arsenate,appearing extremely chlorotic after three weeks. Wild-type plants werestrongly inhibited, chlorotic, and died a week or two later than thoseexpressing ArsC alone when plated on arsenate (Fig. 5B and C). Afterthree weeks of growth on 200 µM arsenate, the hybrid plants (ArsC9 +γ-ECS) attained 3-fold more biomass than plants with γ-ECS alone, 6-fold more than wild type, and 10-fold more than ArsC alone (Fig. 6A).Additionally, when the plants were allowed to grow for up to four weeks

(Fig. 5C), the hybrid plants accumulated 4-fold more biomass thanγ-ECS alone and ∼ 17-fold more than ArsC alone.

Arsenic hyperaccumulation. The arsenic concentrations in shoots ofhybrid plants were determined from plants grown at 125 µM arsenate.This mid-range concentration of arsenate and high plant density wasused to avoid the severe inhibition of growth of wild-type control andSRS1p/ArsC9 plants. Two hybrid lines (ArsC9 + γ-ECS1 and ArsC9 +γ-ECS10), along with SRS1p/ArsC9, ACT2p/ECS1, and wild-type con-trols, were grown on 125 µM arsenate for three weeks. Both ArsC9 +γ-ECS hybrid lines showed substantially higher concentrations ofarsenic in their shoots than wild-type, ACT2p/ECS1, and SRS1p/ArsC9plants alone. The shoot arsenic concentrations in ArsC9 + ECS1 andArsC9 + ECS10 plants were threefold higher than wild-type controlsand twofold higher than plants expressing γ-ECS or ArsC alone, asshown in Figure 6B.

DiscussionEngineering effective hyperaccumulation of arsenic depends uponmaking plants that are highly tolerant to arsenic so as to accumulatesignificant biomass and plants that have the capacity to store morearsenic aboveground. We coexpressed two bacterial genes, arsC andγ-ECS, under control of a light-regulated, leaf-specific rubisco small-subunit promoter SRS1p, and a constitutively expressed actin pro-moter/terminator cassette, ACT2pt, respectively. The bacterial arsCgene product directs leaf- and stem-specific reduction of arsenate toarsenite. This confers on plants the potential to trap more arsenic inarsenite–thiol complexes aboveground (Fig. 1). The γ-ECS gene prod-uct directs the synthesis of the dipeptide γ-EC, the first and proposedlimiting enzyme in the phytochelatin (PC) synthetic pathway18. Alldownstream thiol–peptide compounds in the phytochelatins path-way (e.g., γ-EC, GSH, PCs) can bind arsenite and potentially con-tribute to arsenic tolerance and accumulation. ACT2pt-driven consti-tutive expression of γ-ECS enzyme should have resulted in increasedthiol-sink peptides in all major vegetative organs (e.g., roots, stem,leaves, petals, sepals)19 and resulted in plants that can grow onarsenic-contaminated media. The coexpressed SRS1p/ArsC andACT2p/ECS transgenes complemented each other’s activities result-ing in increased arsenic resistance.

Early in this research, we made an attempt to quantify the activity ofArsC in transgenic plants by examining arsenic speciation using X-rayabsorption near-edge spectroscopy (XANES). Leaf tissues of wild-typeplants and the SRS1p/ArsC Arabidopsis lines grown for three weeks on75 µM arsenate were compared. The K-edge XANES spectra were ana-lyzed to determine the relative concentration of AsO4

3–, AsO33–, and

As(III) tris-glutathione in leaf samples, as shown in Figure 4B. In bothwild-type and transgenic plants exposed to arsenate, 96–100% of the

Figure 5. Arsenic resistance of plants expressing ArsC9 and γ-ECS.(A) Selection of SRS1p/ArsC9 (ArsC9) transgenic plants retransformedwith ACT2p/γ-ECS (ECS1) after growth for four weeks on half-strengthMS medium supplemented with 250 µM arsenate. Only the doubletransformants are sufficiently arsenic resistant to grow at this arsenateconcentration. (B) Arabidopsis lines overexpressing both ArsC and γ-ECS(ArsC9 + ECS1 and ArsC9 + ECS10) show increased resistance toarsenate compared with the transgenic SRS1p/ArsC9 (ArsC9) parentalline expressing ArsC alone, an ACT2p/ECS1 line (ECS1) expressing γ-ECS alone, and wild-type seedlings grown for three weeks on half-strength MS medium without (left) and with (right) 200 µM sodiumarsenate. The two sets of plates show examples of independentexperiments examining two independent doubly transformed linesexpressing both enzymes. (C) The doubly transformed lines expressingboth ArsC and γ-ECS (ArsC9 + ECS1 and ArsC9 + ECS10), and singlytransformed SRS1p/ArsC9 parental line (ArsC9), and an ACT2p/ECS1line (ECS1) expressing γ-ECS alone grown on 200 µM sodium arsenatefor four weeks. In all the above experiments ArsC is expressed from therubisco SRS1p small-subunit promoter and γ-ECS from the actin ACT2ptpromoter and terminator (see Experimental Protocol).

A

B

C

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

nature biotechnology • VOLUME 20 • NOVEMBER 2002 • www.nature.com/naturebiotechnology1144

arsenic in leaves was reduced to arsenite, similar to observations madepreviously in Indian mustard9. With these high endogenous levels ofarsenate reduction to arsenite, even in control plants, we concludedthat it would be very difficult to quantify the added effect of ArsCexpression on arsenic speciation. Therefore, no further attempt wasmade to monitor arsenic speciation.

The hypersensitivity to arsenic of SRS1p/ArsC plants expressingArsC in leaves, when only the roots were contacting the arsenate-containing medium, is in itself quite remarkable and suggests at leasttwo possible models for sensitivity. Both models require that residualarsenate in roots is transported to leaves as shown previously forIndian mustard9. In the first model, sensitivity comes about as a directconsequence of ArsC activity on arsenate. A fraction of the residualarsenate in leaves and stems that was not electrochemically reducedby endogenous plant reductase(s) was reduced by ArsC. Smallamounts of free arsenite generated by ArsC must be toxic to leaves.The root growth from SRS1p/ArsC plants was less inhibited than thatof the 35Sp/ArsC plants (not shown). This is presumably because therubisco promoter did not express ArsC in roots, and hence there wasless reactive arsenite in roots. In the second model, arsenate sensitivi-ty comes about indirectly as a result of ArsC activity. ArsC catalyzesthe oxidation of GSH (Fig. 1) coupled to the reduction of arsenate. Inthe presence of arsenate, ArsC activity lowers the pool of GSH avail-able to endogenous plant protection mechanisms. GSH is both a pre-

cursor in the synthesis of PCs and a required biochemical cofactorrecognized by glutathione S-conjugate pumps (GCP) that dispose oftoxic xenobiotics such as arsenite20–22. Damaging either PC synthesisor GCP activity could bring about arsenic sensitivity. These two mod-els for sensitivity are not mutually exclusive.

Plants overexpressing ArsC aboveground from the SRS1p/ArsCconstruct were hypersensitive to arsenic, whereas plants overexpress-ing γ-ECS not only were resistant to arsenic but accumulated slightlymore arsenic than wild type. γ-ECS overexpression complementedthe hypersensitivity of SRS1p/ArsC plants. In addition, the overex-pression of both ArsC and γ-ECS proteins together in hybrid plantsfurther enhanced arsenic tolerance and hyperaccumulation of arsenicin the aboveground parts, far beyond what γ-ECS expression couldachieve alone. We suggest this is due to more electrochemical reduc-tion of arsenate to arsenite by the activity of ArsC, as proposed by thefirst model, and binding of arsenite to the immediate or downstreamproducts of γ-ECS (e.g., γ-EC, GSH, PCs). Plants overexpressing γ-ECS alone were much less resistant to arsenate than hybrid plantsexpressing both ArsC and γ-ECS. This is presumably due to the toxic-ity of residual arsenate as a phosphate analog. As suggested by the sec-ond model, γ-EC synthesis compensates by diverting more glutamineand cysteine down the pathway to GSH synthesis, complementing theloss of GSH due to ArsC activity.

We can speculate with regard to the fate of the arsenic–thiol com-plexes and their effect on resistance. The excess of arsenic–thiol–peptide complexes are expected to concentrate in vacuoles, and thevacuole transport is a likely target for further improvement ofarsenic remediation. In yeast, As(III)–GS–metal complexes are trans-ported into vacuoles by the action of a GCP, encoded by the YCF1gene22. Arabidopsis has an exceptionally large family of potentialsequence homologs23. One Arabidopsis sequence homolog of theYCF1 transporter, AtMRP3, has been shown to complement the lossof Cd(II) tolerance in yeast ycf1 mutants24. We assume that thearsenic–thiol–peptide complexes in the Arabidopsis plants overex-pressing ArsC and γ-ECS are translocated into vacuoles by the activ-ity of endogenous GCP, such as AtMRP3, and cause the hyperaccu-mulation of arsenic in leaves. We anticipate that the overexpressionof a GCP under a light-regulated promoter along with ArsC and γ-ECS will further enhance hyperaccumulation of arsenic in leaves.These results suggest that a multigene strategy holds great potentialfor remediation of arsenic-contaminated soil and water by trans-genic plants.

The doubly transformed plants expressing ArsC and γ-ECS exam-ined in this study grew as well as unchallenged controls on 200 µMarsenate and were slightly retarded in their growth by 300 µM arsenate(not shown). Because arsenate is extremely soluble in water and the lev-els of arsenate chelators are insignificant in MS medium, most of the200 µM concentration (15,000 p.p.b. (parts per 109) or 15 mg/L)should have been available for uptake by plant tissues. Most of thearsenic-contaminated sites listed on the US EPA’s website(www.epa.gov/superfund/sites/nl/info.html) contain levels on theorder of 100,000–500,000 p.p.b. Most of this arsenic is bound to soiland sediment, and these sites show pore water concentrations of only15,000–50,000 p.p.b. The EPA has suggested that the current limit fordrinking water of 50 p.p.b. (0.05 mg/L) be lowered to between 3 p.p.b.(0.003 mg/L) and 20 p.p.b. (0.02 mg/L)25. This would bring the UnitedStates into the same range as that recommended by the World HealthOrganization (WHO; 10 p.p.b., 0.01 mg/L)26. In some areas of WestBengal (India) and Bangladesh, the contaminated-water arsenic con-centrations reach 2,000 p.p.b. (2 mg/L)27. The resistance of the plantsengineered herein to 15,000 p.p.b. of soluble arsenate suggests that sim-ilarly engineered conservation species would grow on soil and watercontaminated with these environmentally relevant concentrations ofarsenic and could be used to concentrate arsenic aboveground.

Figure 6. Growth inhibition and arsenic accumulation of wild-type andtransgenic Arabidopsis. (A) Comparative growth of two hybrid lines(ArsC9 + γ-ECS1 and ArsC9 + γ-ECS10), and a transgenic SRS1p/ArsC9parental line expressing ArsC alone, and a ACT2p/ECS1 expressing γ-ECS alone, and wild-type (WT) seedlings grown for three weeks onmedium containing the indicated concentrations of sodium arsenate (theaverage and s.e. values are presented for three replicates of 30 seedlingsfor each line). (B) Total arsenic accumulation in shoots of the wild-type(WT), ACT2p/ECS1, SRS1p/ArsC9, and two hybrid lines (ArsC9+ECS1and ArsC9+ECS10) Arabidopsis seedlings grown on 125 µM sodiumarsenate. Values shown are the average of three replicates of 50 plantsfor each line.

A

B

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

www.nature.com/naturebiotechnology • NOVEMBER 2002 • VOLUME 20 • nature biotechnology 1145

To our knowledge, such a substantial increase in arsenic toleranceand hyperaccumulation by transgenic plants has not been demon-strated previously. While overexpression of γ-ECS resulted in mod-erate resistance, the highest levels of resistance and accumulationrequired the additional expression of ArsC aboveground. We areexamining other genes that might complement arsC and γ-ECSfunction to further advance arsenic tolerance, uptake, and hyperac-cumulation. These transgenic technologies and the understanding ofthe physiology of arsenic behavior in plants should lead to the devel-opment of high-biomass, fast-growing arsenic hyperaccumulatorplants for field use.

Experimental protocolPlant expression of ArsC and γ-ECS genes. A 141-codon arsC gene (GenBankaccession no. J02591), found on the arsenicals resistance plasmid R773 (ref. 28),was PCR-amplified from pAlterC (ref. 29) to introduce a BamHI site at the 5′ endand a HindIII site at the 3′ end. The sense primer consisted of the 56 nt sequence5′-TACGTCGGATCCGAATTCGTCGACTAAGGAGGAGCCACAATGAGCAACATCACTAT -3′; an antisense primer had the 44 nt sequence 5′-TAGGTCGGATCCGAATTCAAGCTTATTATTTCAGCCGTTT-3′. A 45-cycle PCR (94°C for 1 min, 48°C for 1 min, and 72°C for 1 min) was run with pAlterC as template.The amplified ArsC fragment was cleaved by BamHI and HindIII and ligatedinto BamHI-HindIII sites of pBluescript-SKII to make plasmid pNA1.

For plant expression, the arsC gene was subcloned under control of two dif-ferent promoters, SRS1p and CaMV’s 35Sp. Both constructs used the nopalinesynthase (nos) 3′ terminator. A 1.5 kb fragment of SRS1 promoter and a 300 bpfragment of nos terminator were cloned into pBluescript-SKII by using XbaI-BamHI and HindIII-XhoI sites, respectively, to create pSRS1p/nos. The arsCgene was placed under the SRS1p using BamHI and HindIII sites, and theresulting construct was designated as pSRS1p/ArsC/nos. The entire cassettecontaining SRS1 promoter, arsC coding sequence, and nos 3′ terminator was

1. Nriagu, E. (ed.) Arsenic in the Environment, Part I: Cycling and Characterization(Wiley, New York, 1994).

2. Kaiser, J. Toxicologists shed new light on old poisons. Science 279, 1850–1851(1998).

3. Chen, C.J., Chen, C.W., Wu, M.M. & Kuo, T.L. Cancer potential in liver, lung, blad-der and kidney due to ingested inorganic arsenic in drinking water. Br. J. Cancer66, 888–892 (1992).

4. Meagher, R.B. & Rugh, C.L. Phytoremediation of heavy metal pollution: ionic andmethyl mercury. Biotechnology for Water Use and Conservation: The Mexico `97Workshop 306–321 (OECD Publications, Paris, 1997).

5. Meagher, R.B. Phytoremediation of toxic elemental and organic pollutants. Curr.Opin. Plant Biol. 3, 153–162 (2000).

6. Ma, L.Q. et al. A fern that hyperaccumulates arsenic. Nature 409, 579 (2001).7. Meharg, A.A. & Macnair, M.R. Genetic correlation between arsenate tolerance

and the rate of influx of arsenate and phosphate in Holcus lanatus. Heredity 69,336–341 (1992).

8. Meharg, A.A. & Macnair, M.R. An altered phosphate uptake system in arsenate-tolerant Holcus lanatus L. New Phytol. 116, 29–35 (1990).

9. Pickering, I. J. et al. Reduction and coordination of arsenic in Indian mustard. PlantPhysiol. 122, 1171–1177 (2000).

10. Cobbett, C.S. Phytochelatins and their roles in heavy metal detoxification. PlantPhysiol. 123, 825–832 (2000).

11. Gladysheva, T., Oden, K.L. & Rosen, B.P. Properties of the arsenate reductase ofplasmid R773. Biochemistry 33, 7288–7293 (1994).

12. Mukhopadhyay, R., Shi, J. & Rosen, B.P. Purification and characterization ofACR2p, the Saccharomyces cerevisiae arsenate reductase. J. Biol. Chem. 275,21149–21157 (2000).

13. Shirley, B.W. & Meagher, R.B. A potential role for RNA turnover in the light regula-tion of plant gene expression: ribulose-1,5-bisphosphate carboxylase small sub-unit in soybean. Nucleic Acids Res. 18, 3377–3385 (1990).

14. Tsutsumi, M. Comparative toxicity of arsenite and arsenate to rice seedlingsunder various levels of phosphate supply. Soil Sci. Plant Nutr. 29, 63–69 (1983).

15. Grondin, K., Haimeur, A., Mukhopadhyay, R., Rosen, B.P. & Ouellette, M. Co-amplification of the γ-glutamylcysteine synthetase gene gsh1 and of the ABCtransporter gene pgpA in arsenite-resistant Leishmania tarentolae. EMBO J. 16,3057–3065 (1997).

16. Arisi, A.C., Noctor, G., Foyer, C.H. & Jouanin, L. Modification of thiol contents inpoplars (Populus tremula × P. alba) overexpressing enzymes involved in glu-tathione synthesis. Planta 203, 362–372 (1997).

17. Zhu, Y. L. et al. Cadmium tolerance and accumulation in Indian mustard isenhanced by overexpressing γ-glutamylcysteine synthetase. Plant Physiol. 121,1169–1177 (1999).

18. Hell, R. & Bergmann, L. γ-Glutamylcysteine synthetase in higher plants: catalyticproperties and localization. Planta 180, 603–612 (1990).

19. An, Y.-Q. et al. Strong, constitutive expression of the Arabidopsis ACT2/ACT8actin subclass in vegetative tissues. Plant J. 10, 107–121 (1996).

20. Szczypka, M.S., Wemmie, J.A., Moye-Rowley, W.S. & Thiele, D.J. A yeast metalresistance protein similar to human cystic fibrosis transmembrane conductanceregulator (CFTR) and multidrug resistance-associated protein. J. Biol. Chem. 269,22853–22857 (1994).

21. Li, Z.S., Szczypka, M., Lu, Y.P., Thiele, D.J. & Rea, P.A. The yeast cadmium factorprotein (YCF1) is a vacuolar glutathione S-conjugate pump. J. Biol. Chem. 271,6509–6517 (1996).

22. Ghosh, M., Shen, J. & Rosen, B.P. Pathways of As(III) detoxification inSaccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 96, 5001–5006 (1999).

23. Sanchez-Fernandez, R., Davies, T.G., Coleman, J.O. & Rea, P.A.The Arabidopsisthaliana ABC protein superfamily, a complete inventory. J. Biol. Chem. 276,30231–30244 (2001).

24. Tommasini, R. et al. An ABC-transporter of Arabidopsis thaliana has both glu-tathione-conjugate and chlorophyll catabolite transport activity. Plant J. 13,773–780 (1998).

25. Kaiser, J. Second look at arsenic finds higher risk. Science 293, 2189 (2001).26. WHO. Guideline for Drinking Water Quality, Recommendation 2nd edn. Vol. 1, 41

(World Health Organization, Geneva, 1992).27. Tondel, M. et al. The relationship of arsenic levels in drinking water and the preva-

lence rate of skin lesions in Bangladesh. Environ. Health Perspect. 107, 727–729(1999).

28. Chen, C.-M., Misra, T.K., Silver, S. & Rosen, B.P. Nucleotide sequence of thestructural genes for an anion pump. J. Biol. Chem. 261, 15030–15038 (1986).

29. Liu, J., Gladysheva, T.B., Lee, G.L. & Rosen, B.P. Identification of an essential cys-teinyl residue in the ArsC arsenate reductase of plasmid R773. Biochemistry 34,13472–13476 (1995).

30. Bevan, M.W. Binary Agrobacterium vectors for plant transformation. Nucleic AcidsRes. 12, 8711–8721 (1984).

31. Li, Y., Kandasamy, M.K. & Meagher, R.B. Rapid isolation of monoclonal antibod-ies: monitoring enzymes in the phytochelatin synthesis pathway. Plant Physiol.127, 711–719 (2001).

32. Bent, A. et al. RPS2 of Arabidopsis thaliana: a leucine-rich repeat class of diseaseresistance genes. Science 265, 1856–1860 (1994).

33. Jones, J.J. & Case, V.V. Sampling, handling, and analyzing plant tissue samples.in Soil Testing and Plant Analysis 2nd edn. SSSA book series, No. 3 (ed.Westerman, R.) 389–427 (Soil Science Society of America, Madison, WI, 1990).

34. Bizily, S., Rugh, C.L. & Meagher, R.B. Phytodetoxification of hazardousorganomercurials by genetically engineered plants. Nat. Biotechnol. 18, 213–217(2000).

35. Rosen, B.P., Weigel, U., Monticello, R.A. & Edwards, B.P.F. Molecular characteri-zation of an anion pump: purification of the ArsC protein. Arch. Biochem. Biophys.284, 381–385 (1991).

subcloned into pBIN19 (ref. 30) by using 5′ SacI and 3′ XhoI-SmaI blunt sites tocreate pSRS1p/ArsC. For cloning arsC under constitutive 35S promoter,BamHI-XhoI (blunt) ArsC/nos fragment from pSRS1p/ArsC/nos construct wascloned into pBI121 at BamHI-EcoRI (blunt) site, and the resulting constructwas designated as p35Sp/arsC. The cloning of γ-ECS gene to make constructACT2p/γ-ECS has been described in Li et al.31. The above constructs were intro-duced into A. thaliana (ecotype Columbia) by Agrobacterium-mediated trans-formation using the vacuum infiltration procedure32.

Arsenic extraction and quantification. Plants were grown in 0.5× MS mediumat 22°C with a regime of 16 h light/8 h darkness. The shoots from three-week-old seedlings were harvested, washed three to four times with deionized water toremove any shoot surface contamination, and extracted for determination oftotal arsenic. The plant samples were dried at 70°C for 48 h, digested in a mixtureof nitric and perchloric acids (7:1 vol/vol) following standard methods33, andwere analyzed for arsenic contents by inductively coupled plasma spectrometry(ICP-MS). Certified National Institute of Standard and Technology plant stan-dards (peach leaves) were digested and analyzed as well. In addition, reagentblanks and the internal standards were used, where appropriate, to insure accu-racy and precision in the analysis.

AcknowledgmentsWe thank Ingrid J. Pickering and Roger C. Prince for their help in collecting andanalyzing the XAS data at SSRL and Gay Gragson, Rebecca S. Balish, and M.K.Kandasamy for editorial comments. This research was supported by USDepartment of Energy (DOE) grant DE-FC09-93R18262, DOE EnvironmentalScience Management Program grant DE-FG07-96ER20257 and NIH grant RO1GM52216.

Competing interests statementThe authors declare that they have no competing financial interests.

Received 17 April 2002; accepted 23 August 2002

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy