Organic acids and aluminum detoxification:

Considerable natural variation in Al tolerance in agriculturally important species, such as wheat, maize, and sorghum, has been observed. Recent studies investigating the mechanisms of Al tolerance have provided support for a mechanism in which organic acids, such as malate or citrate (Figure 1A), are extruded at the root apex and Al is consequently chelated in the rhizosphere. In support of this proposed mechanism are the following observations: the correlation between Al tolerance and Al induction of organic acid release in a number of species; the correlation between the degree of Al tolerance and the level of organic acid release in wheat; the genetic linkage of Al-induced release of malate and Al tolerance in wheat; and an association of decreased malate exudation with loss of Al-tolerance (Ma et al., 2001; Osawa and Matsumoto, 2001). Organic acids have also been proposed to be involved in the chelation and sequestration of a number of metal ions including Cd, Zn, and Ni. However, there is little evidence in support of these functions.

One approach to identifying components of the response to Al toxicity is to identify Al-inducible genes. A number of such genes have been identified in studies on a range of organisms (Snowden and Gardner, 1993; Richards et al., 1994; Ezaki et al., 1995; Snowden et al., 1995; Ezaki et al., 1996). These genes include phenylammonia lysase, Bowman-Birk protease inhibitors, pathogen-response (PR) proteins, a peroxidase, and a glutathione-S-tranferase (GST). A common feature of these genes is that they are all induced by other stresses, particularly oxidative stress. A study in Arabidopsis (Richards et al., 1998) also identified a peroxidase, a GST, a blue copper-binding protein, and a superoxide dismutase among others that were induced by Al. These studies indicate that oxidative stress is an important component of Al toxicity. In another study, a number of Al-inducible genes from Arabidopsis and other plants have been expressed in yeast, and some conferred an increase in Al-tolerance (Ezaki et al., 1999). Conversely, Al-induced genes from wheat, Arabidopsis, tobacco, and yeast have been over-expressed in transgenic Arabidopsis plants and several of these conferred a degree of Al-tolerance and reduced Al-accumulation (Ezaki et al., 2000). In addition, some of these transgenic lines had increased resistance to oxidative stress or other stresses, such as heavy metal exposure. It is likely then, that these Al-inducible genes are more general stress response genes that individually confer a level of tolerance to oxidative stresses, including exposure to Al. It is unlikely that they are involved in a specific Al-tolerance mechanism, such as the release of organic acids into the rhizosphere.

A number of mutants of Arabidopsis with increased or decreased tolerance to Al have been identified (Larsen et al., 1996; Larsen, 1997). Resistant mutants were identified on the basis of increased root growth in the presence of inhibitory Al levels. The mutant phenotypes were semi-dominant and mapped to two different chromosomal loci (Larsen et al., 1998). The locus on chromosome 1 may represent a cluster of genes or a single gene. Mutants at this locus accumulated lower levels of Al in root tips and released greater amounts of organic acids. The single mutant at the chromosome 4 locus did not exhibit increased organic acid secretion but, in contrast, resulted in an Al-induced increase in rhizosphere pH at the root tip (Degenhardt et al., 1998). Because Al-toxicity is a function of soil acidity, a resistance mechanism involving increased alkalinization of the rhizosphere has been proposed (Foy et al., 1965), although there have been no convincing demonstrations of this as a mechanism of resistance. The isolation and characterization of this Arabidopsis mutant indicates such a mechanism is possible, although in other species in the wild it does not appear to have undergone strong selection. As with other species, natural variation in Al tolerance can be detected among Arabidopsis ecotypes. The Col-0 and Ws ecotypes are more tolerant to Al than the Ler ecotype (Toda et al., 1999). The availability of recombinant inbred lines in Arabidopsis allows the genetic basis for this variation to be genetically mapped and the genes isolated.

Metallothionein genes and proteins:

Metallothioneins have the following general characteristics: 1) low molecular weight (less than 10 kD), 2) a large fraction of cysteine residues, 3) high metal content with coordination of metal ions in metal-thiolate clusters, as shown in Figure 1B. Class I MTs comprise those from mammals and other organisms that have a highly conserved arrangement of cysteine residues, while Class II comprises all other MT proteins, including all plant MTs described to date. Class III is a formal classification that includes phytochelatins and similar peptides that are not directly gene-encoded. These are considered separately below.

The Class II plant MTs have been sub-classified into types according to conserved features of amino acid sequences (Robinson et al., 1993). Type 1 plant MTs contain a total of 12 cysteine residues present in Cys-Xaa-Cys motifs in two clusters at the amino- and carboxy-termini of the protein. In the majority of Type 1 MTs, these two clusters are separated by a spacer of approximately 40 amino acids. However, Type 1 MTs in Arabidopsis and Brassica napus lack this large spacer and contain an additional cysteine residue. Type 2 MTs also contain amino- and carboxy-terminal clusters of cysteines but are distinguished from Type 1 MTs by the presence of cysteines in Cys-Cys and Cys-Xaa-Xaa-Cys motifs. The majority of plant MT genes described to date can be categorized as either Type 1 or Type 2. However, there appear to be at least two additional types of MTs present in many plant species. Type 3 MTs have only four cysteine residues in the amino-terminal domain, and at least six in the carboxy terminal domain. Examples of this Type 3 have been identified in Arabidopsis (Murphy et al., 1997) and other species (Ledger and Gardner, 1994; Clendennen and May, 1997). Type 4 MTs contain three distinct cysteine-rich domains and include the wheat Ec protein as well as MTs predicted from genes that have been identified in Arabidopsis and maize (White and Rivin, 1995). A comprehensive listing of MT genes identified in plants was given in Rauser (1999).

In the Arabidopsis genome, seven functional MT genes have been reported. However, recent exhaustive analysis of the genome sequence suggests the presence of at least four additional MT genes (Marie and Meagher, unpublished data). Of the seven reported genes, Types 1, 2, and 4 are represented by at least two expressed genes of each type, whereas there is only a single gene encoding a Type 3 MT (Zhou and Goldsbrough, 1995). There is at least one pseudogene related to the Type 1 MTs. The structure of Arabidopsis MT genes suggests that at least some share a common ancestral gene. Type 1 and 2 MT genes contain a single intron located just upstream of the last cysteine codon in the amino terminal domain. This intron position is also conserved in MT genes from pea, tomato, and cotton. While the intron positions in MT genes of other types are not conserved, there are still striking similarities in the cysteine-rich domains both within individual MTs and between different MTs, suggesting that this domain may have been duplicated initially within a single MT gene. Subsequent gene amplification has produced at least some members of the current MT gene family found in Arabidopsis and other plants. A number of MT genes have been cloned as tightly linked gene clusters; two Arabidopsis Type 1 MT genes are separated by less than 3 kb (Zhou and Goldsbrough, 1995). These observations indicate that recent gene duplication is an important process in the development of the MT gene family in Arabidopsis.

The wheat embryo Ec protein was the first MT to be purified from a plant tissue and was characterized as a Zn-binding protein (Tang et al., 1996). A number of plant MTs have been expressed in a variety of microbial hosts, both as native proteins or as translational fusions. However, few studies have involved the identification and characterization of MT proteins from vegetative tissues of plants. Murphy et al. (1997) purified a number of copper-binding proteins from Arabidopsis and showed that these included proteins with the amino acid sequences predicted for MT1a, MT2a, MT2b, and MT3. Antibodies raised against MT1a and MT2a proteins expressed in E. coli were used to monitor the purification of these proteins. In addition, the predominance of these proteins in leaf or root tissue, respectively, and their induction by Cu matched the levels observed for the corresponding mRNAs.

The ability of plant MTs to bind heavy metals has been demonstrated by expressing plant MT proteins in microbial hosts. For example, an MT-deficient mutant of S. cerevisiae cup1, has a copper- and cadmium-sensitive phenotype. Expression of either Arabidopsis MT1a or MT2a under a constitutive, high level promoter was able to restore copper and cadmium tolerance to this mutant (Zhou and Goldsbrough, 1994). Similarly, expression of Arabidopsis MT2a in a Zn-sensitive MT deletion mutant of Synechococcus partially restores tolerance to Zn(II) (Robinson et al., 1996).

Expression of MT genes in response to heavy metals:

Zhou and Goldsbrough (1994) demonstrated that MT gene expression in Arabidopsis could be induced by copper and to a lesser degree by zinc and cadmium. MT2a was the most responsive to copper, and MT protein expression is also induced under these conditions (Zhou and Goldsbrough, 1995; Murphy et al., 1997). Expression of MT2a RNA was also elevated in a copper-sensitive mutant of Arabidopsis that accumulated excess copper (van Vliet et al., 1995). Metal-induced expression of MT genes has also been reported in other plant species (see (Rauser, 1999)). However, some studies have shown no effect of particular thiol-binding metals on expression of particular MT genes.

The expression of some MT genes in plants is increased during senescence. This increase would allow those metal ions released after degradation of proteins to be chelated, reducing their reactivity and allowing the controlled mobilization of elemental nutrients to continue before the senescing tissue dies (Buchanan-Wollaston, 1994). The activity of Arabidopsis MT1 promoters in vascular tissues of senescing leaves may indicate that MTs are also involved in metal ion transport during this physiological process (P. Goldsbrough, personal communication).

Possible functions of MTs in heavy metal tolerance:

There is relatively little evidence indicating whether or not MTs in plants, or more particularly in Arabidopsis, have a role in heavy metal detoxification. In early studies non-plant MT genes expressed in plants generally resulted in an increase in heavy metal tolerance. However, there are few studies implicating endogenous MT gene expression in heavy metal tolerance. In a recent study, Cu-tolerance in Silene vulgaris has been associated with an amplification of a MT Type 2a gene (van Hoof et al., 2001). In a comparative study of different Arabidopsis genotypes (Murphy and Taiz, 1995), variation in copper tolerance among genotypes, measured by root growth inhibition, was highly correlated with the expression of MT2a RNA in copper-treated plants.

There have been no published studies on the phenotypes of MT gene mutants in Arabidopsis. It is possible that, in view of the number of MT genes expressed, there may be some degree of redundancy and a single gene mutant may not have an apparent phenotype. However, anti-sense studies in Arabidopsis have shown that transgenic plants expressing anti-sense MT1a and MT2a have a slight Cu-sensitive phenotype, and an anti-sense MT1a transgene expressed in a PC-deficient mutant background confers a slight increase in Cd-sensitivity (P. Goldsbrough and C. Cobbett, unpublished). Together these observations indicate that MTs in Arabidopsis are able to play a role in heavy metal tolerance. However, their primary role may be in heavy metal homeostasis. This question is likely to be resolved only when the complete series of MT knock-out mutants has been identified and characterized individually and in combination.

Phytochelatins:

Phytochelatins (PCs) are a family of peptides with the structure (g-GluCys)_n-Gly (where n > 1). The Glu and Cys residues in PCs are linked through a g-carboxylamide bond demonstrating that, unlike the Class I and II MTs, these peptides are not encoded directly by genes, but are the products of a biosynthetic pathway, shown in Figure 2A. These peptides have been referred to variously as cadystins (from the yeast S. pombe), poly-(g-EC)nG peptides, Cd-binding peptides, and phytochelatins, and they are broadly classified as Class III MTs. PCs have been identified throughout the plant kingdom including monocots, dicots, and gymnosperms and the algal progenitors of higher plants. In addition, PCs are found in various fungal species including Schizosaccharomyces pombe, Candida glabrata, and Neurospora crassa and in marine diatoms. A number of structural variants of PCs, for example (g-GluCys)_n-b-Ala, (g-GluCys)_n-Ser and (g-GluCys)_n-Glu, have been identified in some plant species.

PCs are derived from glutathione, (GSH, g-GluCysGly) or, in some cases, a related compound with other small amino acids substituting for the Gly residue (e.g., Ser). Genetic studies, in particular, have confirmed that GSH is the substrate for PC biosynthesis (Figure 2A). Mutants of both S. pombe and A. thaliana which have defects in genes encoding GSH biosynthetic enzymes are deficient in PCs and are hypersensitive to metal ions. GSH is synthesized from its constituent amino acids in a two-step pathway catalyzed successively by the ATP-dependent enzymes g-glutamylcysteine synthetase (GCS) and glutathione synthetase (GS). In plants, the GSH biosynthetic enzymes have been purified and some studies have suggested the existence of different isozymes of each in chloroplastic and cytosolic fractions. This genetic duplication has been interpreted as indicating the existence of separate GSH biosynthetic pathways in these two cellular compartments. Genes encoding these enzyme activities have been cloned from various plant species including Arabidopsis and Brassica juncea. For each GCS and GS in Arabidopsis, only a single gene has been identified and both enzymes appear to have chloroplast targeting sequences (May and Leaver, 1994; Skipsey et al., 2000). However, it is not known if these enzymes are exclusively targeted to chloroplasts. In B. juncea, in contrast, there appear to be multiple isozymes, possibly localized to different subcellular compartments (Schafer et al., 1998). Alternate genes encoding isovariants lacking the targeting sequence were not detected in searches of the Arabidopsis database (see Table 1).

GSH-deficient mutants have been identified in Arabidopsis. The cad2-1 mutant was identified as Cd-sensitive and partially deficient in PCs. It contains only 20–40% of wild-type levels of GSH. The CAD2 gene encodes GCS and the cad2-1 allele contains a 6 bp deletion within an exon (Cobbett et al., 1998). The mutant has only 30–40% of wild-type GCS levels of activity. A mutant with a more extreme phenotype, rml1 (rootmeristemless1), which has less than 2% wild-type levels of GSH, has a mutation in the same gene (Vernoux et al., 2000). These data suggest that the cad2-1 mutant gene product is partially active and, furthermore, that there is unlikely to be a second gene encoding GCS activity. The rml1 mutant fails to develop a root following germination due to an apparent lack of cell division (Cheng et al., 1995). This defect implies an unexpected role for GSH in controlling root growth and has led to speculation that GSH may be an important signal by which root growth is modulated in response to environmental stress.

Phytochelatin synthase genes and proteins:

PC synthase, or g-GluCys dipeptididyl transpeptidase (EC 2.3.2.15), catalyses the transpeptidation of the g-GluCys moiety of GSH onto a second GSH molecule to form PC₂, or onto a PC_n molecule to form PC_n+1. The enzyme was first identified by Grill et al. (1989). In vitro, the enzyme is active only in the presence of metal ions, including Cd, Ag, Bi, Pb, Zn, Cu, Hg, and Au cations; a profile of metal ions similar to those observed to induce PC biosynthesis in vivo.

Mutants at the cad1 locus in Arabidopsis are, like the cad2-1 mutant, Cd-sensitive and deficient in PCs (Howden et al., 1995). However, in contrast to cad2-1, cad1 mutants have wild-type levels of GSH and lack PC synthase activity suggesting a defect in PC synthase. The CAD1 gene was isolated by positional cloning (Ha et al., 1999), and an Arabidopsis cDNA (AtPCS1) (Vatamaniuk et al., 1999) identical to CAD1, and a similar cDNA from wheat (TaPCS1) (Clemens et al., 1999), were identified in parallel studies through their ability to confer Cd-resistance when expressed in yeast. Arabidopsis PCS1, purified as epitope-tagged derivatives (Clemens et al., 1999; Vatamaniuk et al., 1999) or expressed in E. coli (Ha et al., 1999), was sufficient for GSH-dependent PC biosynthesis in vitro and was activated to varying extents by metal ions. Interestingly, there is a second PC synthase gene in Arabidopsis (Table 1). The absence of detectable PCs in a severe cad1 mutant suggests this second gene is expressed in a highly tissue-specific manner or in response to particular environmental signals. The function and expression of this gene is yet to be determined.

Despite the absence of reports of PCs themselves in animals, PC synthase sequences have also been identified in a number of animal species. Functional PC synthase genes have been identified in the nematode, Caenorhabditis elegans (Clemens, 2001; Vatamaniuk et al., 2001) and in the slime mold, Dictyostelium discoideum (C. Cobbett, unpublished data). Furthermore, the suppression of CePCS1 expression by using a double-stranded RNA interference technique resulted in Cd-sensitivity, thereby demonstrating an essential role for PCs in heavy metal detoxification (Vatamaniuk et al., 2001). The PC synthase gene in S. pombe has also been identified (Clemens et al., 1999; Ha et al., 1999).

The predicted plant PC synthase proteins range in size from 51 to 56 kD and can be aligned across their full lengths. In contrast the predicted S. pombe, C. elegans and D. discoideum proteins are 47, 42, and 71 kD in size, respectively. The N-terminal regions of the plant, yeast, and animal PC synthases are similar in amino acid sequence, while the C-terminal sequences show little apparent conservation. The C-terminal regions of all the predicted proteins contain multiple Cys residues, often as pairs. However, there is no apparent conservation of the positions of these Cys residues relative to each other.

Regulation of PC biosynthesis:

There is supporting evidence for the coordinated transcriptional regulation of genes involved in sulfur transport and assimilation and in GSH biosynthesis in response to Cd exposure in Brassica juncea (Schafer et al., 1998; Heiss et al., 1999; Lee and Leustek, 1999) and in Arabidopsis (Xiang and Oliver, 1998). In Arabidopsis, the signal molecule jasmonate also induced transcription of the GSH biosynthetic genes (Xiang and Oliver, 1998). However, it is not apparent whether the effect of Cd on gene expression is mediated via jasmonate. In Arabidopsis transcription of AtPCS1/CAD1 is not induced by exposure to a range of heavy metals (Ha et al., 1999; Vatamaniuk et al., 2000). In contrast, TaPCS1 expression in wheat roots is induced on exposure to Cd (Clemens et al., 1999).

The primary mechanism for regulating PC synthase activity appears to be the activation of the enzyme by metal ions. Originally it was assumed that activation was due to a direct interaction between metal ions and the enzyme, but this explanation raised the question of how the enzyme might be activated by such a wide range of metals. Using purified recombinant AtPCS1, Vatamaniuk et al (2000) demonstrated that metal binding to the enzyme per se is not responsible for catalytic activation. The kinetics of PC synthesis are consistent with a mechanism in which heavy metal glutathione thiolate (eg., Cd-GS₂) and free GSH act as g-Glu-Cys acceptor and donor. The observation that S-alkylglutathiones can participate in PC biosynthesis in the absence of heavy metals is consistent with a model in which blocked glutathione molecules (metal thiolates or alkyl substituted) are the substrates for PC biosynthesis. Thus, rather than interacting directly with the enzyme itself, metal ions play their role an integral part of the substrate.

Other aspects of PC function:

PC-Cd complexes are sequestered to the vacuole in both plants and S. pombe, as suggested in Figure 1D. Vacuolar sequestration is an integral aspect of PC function in vivo and is described in detail below. In addition, in both fungi and plants, labile sulfide can also be detected in PC-Cd complexes. Labile sulfide also appears to have a role in PC function because a number of Cd-sensitive mutants in S. pombe are affected in aspects of sulfide metabolism resulting in decreased accumulation of vacuolar PC-Cd complexes. Parallel mutants have not been identified in Arabidopsis and the importance of labile sulfide in PC function in plants has not been determined.

The significance of PCs in metal detoxification:

The role of PCs in heavy metal detoxification is clear from characterization of the PC synthase-deficient mutants of Arabidopsis and S. pombe. A comparison of the relative sensitivity of the Arabidopsis and S. pombe mutants to different heavy metals revealed a similar but not identical pattern (Ha et al., 1999). In both organisms, PCs appear to play an important role in Cd- and arsenate-detoxification and no apparent role in the detoxification of Zn, Ni, and selenite ions. Minor differences between the two organisms were observed with respect to Cu, Hg, and Ag. Thus, because PC biosynthesis in vivo and in vitro is induced by all of these metal ions, it is likely that PCs do not play a significant role in the detoxification of many of the metals that can induce them. This may be because for some PC-metal complexes, other aspects of the detoxification mechanism, for example transport or sequestration, are absent. Or, for these metals, there may be more effective detoxification mechanisms, such as those based on MTs or histidine. PCs may be an important mechanism through which the phytoremediation capacity of plants might be manipulated. However, the effect of manipulating PC expression on heavy metal tolerance and accumulation, let alone phytoremediation is yet to be assessed.

Plant transport of organic pollutants:

The uptake and sequestration of toxic organics in plant roots and/or their concentration in vacuoles are essential to the remediation process. A subclass of the ABC transporters, originally named the multi drug resistance proteins (MRPs) after their activity in animal systems, is perhaps the best-characterized family of plant proteins that move organics across membranes. MRPs are generally responsible for uptake, efflux, and sequestration of a wide variety of high-molecular weight, toxic byproducts of metabolism, and xenobiotics (Lu et al., 1998; Tommasini et al., 1998). Some MRPs recognize glutathione conjugates of toxic organics, metals, and metalloids. Plant MRPs are the likely transporters of most xenobiotics like herbicides and of most endogenous organic compounds into or out of cells or into vacuoles (Keppler et al., 1999). Determining the precise roles of individual transporters will be a challange, because 1) there are so many transporters functioning simultaneously in plants; 2) there are no doubt many differences in tissue and subcellular membrane localization of there activities; and 3) there is an astronomical number of classes of potential substrates to be tested. Sorting out what pumps act on which classes of organic and elemental-conjugate will take an heroic effort. However, this work is essential to any rational design and/or breeding of plants optimally suited to phytoremediation.

Several exciting experiments have begun to dissect the specific plant MRP activities and their sites of action. The powerful genetics of Arabidopsis has played an important role in these first critical studies. We now know for example, that Arabidopsis AtMRP2 can transport glutathione- (GS-) conjugates and chlorophyll catabolites, but not bile acids, into vacuoles (Lu et al., 1998; Liu et al., 2001). AtMRP1 can transport GS-conjugates of xenobiotics and endogenous toxic chemicals, including herbicides and anthocyanins (Lu et al., 1997). AtMRP3 can transport GS-conjugates of Cd and chlorophyll catabolites (Tommasini et al., 1998), although no subcellular sites of transport have been reported for either AtMRP1 or AtMRP3. These findings are very exciting, because the accumulation of toxic organics in plant roots and plant vacuoles should favor subsequent degradation, providing plants have the appropriate catabolic enzymes.

Potential degradation of some toxic organic pollutants by native plants:

Although our basic knowledge of the plant-based degradation of organic pollutants lags far behind that for animals or bacteria, plants appear to transform and/or mineralize a wide variety of complex organics. Interpretation of plant-encoded degradation pathways is complicated by the fact that most studies are performed on contaminated soil instead of axenic culture, making it difficult to determine the extent to which the bacterial rich rhizosphere is contributing to measured activities. In fact, some studies on the degradation of organic pollutants have demonstrated a requirement for the rhizosphere (Anderson et al., 1993; Siciliano and Germida, 1999; Heinonsalo et al., 2000). A few exciting examples demonstrate the potential of plant metabolic systems as remediators of toxic xenobiotics are discussed in the following paragraphs. In the future, the availability of plant gene sequences made possible through the Arabidopsis database will make more of these systems accessible (see section D).

Plants contain uncharacterized aliphatic dehalogenases capable of degrading compounds like trichloroethylene (TCE) (Gordon et al., 1998). This is important because, among a long list of industrial solvents that pose a threat to wildlife and humans, TCE and related chlorinated solvents are the most widely distributed environmental pollutants in ground water and soils. Hundreds of TCE-contaminated Superfund Sites are listed by the United States Environmental Protection Agency ( www.epa.gov/superfund/sites/index.htm). Halogenated compounds are among the most difficult to metabolize and are usually toxic and carcinogenic. Plants grown at polluted sites are known to extract TCE, efficiently transpire it, and enhance the degradation of TCE in the rhizosphere by feeding biodegrading bacteria with root exudates (Anderson et al., 1993; Anderson and Walton, 1996). However, it has only recently become clear that plant enzymes play a direct role in the degradation process. Careful mass balance and isotopic labeling experiments demonstrate that axenically-grown hybrid poplars (Populus sp.) actively take up TCE and degrade TCE to trichloroethanol or dichloacetate (Figure 2F) ( http://umbbd.ahc.umn.edu/tce/tce_map.html), and finally CO₂ (Gordon et al., 1998). In one experiment with axenic poplar tissue culture cells, greater than 10% of the TCE was mineralized to CO₂ in ten days. These data suggest an oxidative degradation pathway, quite different from the reductive one often found in bacteria (Wackett et al., 1994).

Nitroaromatic compounds are highly toxic and carcinogenic, and notoriously difficult to metabolize. however, wide variety of plant species from diverse plant families appear to partially degrade TNT (2, 4, 6-trinitrotoluene) (Best et al., 1997; Miskiniene et al., 1998). The explosive TNT and a large family of related nitro-substituted linear organic compounds (GTN, glycerol trinitrate or nitroglycerin; GDN, glycerol dintrate; GMN, glycerol mononitrate; RDX, hexahydro-1,3,5-trinitro-1,3,5-triazine) are the explosive compounds used in munitions. Some of the same oxygen reactive properties that make them explosive also make them extremely toxic. As a result of their manufacture, storage, and disposal, toxic nitro-substituted compounds contaminate thousands of acres of land and hundreds of miles of rivers. Nearly 100 defense-related and industry Superfund sites are listed as contaminated with toxic explosive compounds by the United States EPA.

The plant-based degradation of TNT (Figure 2) appears to proceed through multiple complex pathways. However, the products of reductive mechanisms such as triaminotoluene (TAT), predominate and the minimal release of substrate-derived nitrate and carbon dioxide suggests complete mineralization occurs inefficiently (Best et al., 1997; Miskiniene et al., 1998), shown in Figure 2G. For example, axenically grown Microphyllum aquaticum plants and Cartharanthus roseus hairy root cultures both partially degrade TNT and release degradation intermediates into their growth medium (Hughes et al., 1997). Hairy root cultures of Catharanthus roseus were capable of the degrading most of the 25 ppm of TNT added to culture medium in a few days (Bhadra et al., 1999). Axenic cell cultures of sugar beet degrade GTN to the expected toxic intermediates GDN and GMN (Goel et al., 1997). Once denitrified it is likely that other enzyme systems can degrade their aromatic and linear hydrocarbon backbone of these compounds. For example, plants accelerate the degradation of polyaromatic hydrocarbons (PAHs) such as the extremely carcinogenic pyrenes (Liste and Alexander, 2000).

Organophosphate pesticides like TOCP (tri-o-cresylphosphate), halogenated pesticides like DDT (dichlorodiphenyltrichloroethane) (Figure 2), and their metabolic products are potent neurotoxins and suspected carcinogens. These and many other related pesticides pose serious threats to the environment due to their slow degradation rates. A wide variety of such pesticides are taken up at high levels by food crops (Mattina et al., 2000). Until recently these compounds were widely used by agriculture and municipalities for insect control, and now, due to past production and handling practices, they contaminate thousands of square miles of land and many lakes and streams. Selected aquatic plants like parrot feather and elodea degrade both classes of pesticides in aqueous culture media (Gao et al., 2000a; Gao et al., 2000b). These studies were performed in axenic culture, certifying that the activities are endogenous to these plants and are not due to surrounding microorganisms. It is unlikely that these or related catabolic activities are universally expressed at useful levels in most plant species or pesticides would not be so refractile to mineralization in the environment.

Chloro-substituted aromatics are among the worst pollutants due to their toxicity, carcinogenicity, wide distribution, and slow biodegradation in the environment. Polychlorinated biphenyls (e.g., PCBs) were used as insulation in electrical transformers at thousands of sites, as flame retardants in cotton clothing, and as plasticizers. Dioxins (e.g., TCDD, tetrachlorodibenzo-p-dioxin) are contaminating products of pesticide manufacture. Axenic cultures of some plant species have been shown to efficiently degrade several classes of PCBs (Kas et al., 1997). For example, sterile cultures of Solanum nigrum degrade several PCB congeners relatively efficiently (Mackova et al., 1997). PCBs with the most highly chlorinated benzene rings appear to be the most difficult for both bacteria and plants to break down, because there is little chemical access to the aromatic ring. The metabolic basis for the degradation of PCBs by plants has not been well characterized or quantified.

Foreign genes greatly enhance the metabolism of organic pollutants:

Another strategy for phytoremediating organic pollutants focuses on using the enormous existing database of well-characterized enzymes involved in the mammalian and bacterial detoxification of chlorinated solvents. For example, the human cytochrome P-450 2E1 aids in the detoxification of TCE, ethylene dibromide, carbon tetrachloride, chloroform, and vinyl chloride (Doherty et al., 1996) as shown for TCE in Figure 2E. Transgenic tobacco plants expressing this enzyme enhanced their TCE metabolism more than 100-fold compared to control plants (Doty et al., 2000). These plants also showed an increased uptake and debromination of ethylene dibromide. It is clear that plants over-expressing this or related plant and bacterial enzymes will be very useful in the remediation of sites contaminated with halogenated aliphatics. The potential offered by plant-encoded dehalogenases and P-450 oxygenases is discussed below in the section on the phytoremediation proteome.

In a striking demonstration of the potential of transgenic plant technologies to enhance the natural capacities of plants, a bacterial NADPH-dependent nitroreductase (French et al., 1996) greatly increased GTN degrading activity when expressed in tobacco (French et al., 1999). One possible initial step of nitroreductase in the degradation of TNT is shown in Figure 2G. Transgenic tobacco seedlings were about 10-fold more resistant to the toxic effects of GTN and TNT than their wild-type progenitors. Preliminary evidence suggests that these plants were able to break down both GTN and its first degradation product, GDN, twice as fast as the wild-type controls. While it is not clear what fraction of these nitro-substituted compounds is completely mineralized, the high level of plant resistance suggests mineralization may have been complete in these transgenic plants. This significant enhancement in degradation efficiency increases the feasibility of applying phytoremediation to toxic nitro-substituted pollutants.

Plants appear to contain a significant endogenous degradative capacity of their own and this can be greatly enhanced with the addition of genes encoding one key catabolic enzyme. Using bacterial and animal genes known to enhance the catabolism of organic pollutants, engineered plant species will be able to remediate all classes of organic pollutants. However, the potential of native plant genes to carry out these processes is essentially unexplored. Plants extract nearly all their nutrients from soil and water by competing in the natural rhizosphere already containing almost every known class of toxin produced by soil microorganisms such as the streptomyces (Hopwood, 1999). The ability to resist or metabolize these diverse compounds requires a tremendous and highly evolved genetic capacity. This genetic capacity is part of what gives plants their great potential for phytoremediation. The Arabidopsis genome is a gold mine of information and genetic material for the field of phytoremediation.

Constructing a working list of Arabidopsis phytoremediation proteome:

Phytoremediation related proteins (PRP) and the phytoremediation proteome can be defined as the portion of plant proteins that can contribute to the remediation of environmental pollutants through a direct action of binding or sequestering, transforming, or transporting organic or elemental pollutants. This distinguishes PRPs from other proteins that build normal plant structure, control gene expression, or transform central metabolites that might be the final degradation products of toxic organics, but do not act directly on environmental contaminants. The phytoremediation proteome is the sum of all the plant PRPs. The amino acid (a.a.) sequences of proteins with known roles in remediation were used as query sequences to search for homologous and similar sequences in the Arabidopsis database. The number of Arabidopsis PRP sequences can be taken as a reasonable minimum estimate and no attempt will be made to account for protein modifications. Most of the query sequences came from well-characterized enzymes with known roles in remediation processes previously identified and characterized in bacteria, fungi, and animals. Query sequences from non-plant kingdoms were used because they gave a better overview of the plant gene family size and the number of highly conserved and ancient functional protein motifs. Using plant query sequences would have given too much statistical weight to the closest Arabidopsis homologues to the plant query. No attempt was made to make this list exhaustive, because the list of organic and elemental pollutants is so long, and we lack sufficient knowledge to predict more than a fraction of the plant proteins contributing directly to phytoremediation processes. Our data on potential Arabidopsis PRPs are summarized in Table 1. Most query sequences turned up significant sequence matches in the Arabidopsis genome, and those query sequences that did not detect significantly-related sequences are listed in the footnotes of the table. In total, approximately 700 potential Arabidopsis PRP sequences were identified and they were comprised in most cases of clearly homologous protein sequences. This working list will need to be explored further by experimental scientists. Undoubtedly, only a subset of these sequences and their homologues in other plants are currently functioning in some way to contribute to phytoremediation processes such as uptake, detoxification, and sequestration of pollutants. However, all of these genes have the potential to be modified or over-expressed in some distinct way in the future, and hence make their contribution to remediative processes.

The potential Arabidopsis PRPs described in this study were identified using BLAST searches. During BLAST searches for related sequences, a method was developed to approximate alignments that optimize local similarity. This measure of the statistical significance of alignments is given as an E-value (Altschul et al., 1990). Simply stated, E-values can be defined as the expectation of finding a sequence with this much similarity by chance. The lower the E-value, the lower the possibility that such a match could have been found at random. For example, an E-value of 0.0001 (e × 10⁻⁴, written as e-4 in Table 1) suggests that you should not find this level of sequence relatedness to the query by chance more than once in 10,000 searches of a database of this complexity. Most Arabidopsis PRPs included in Table 1 had expectation ratios (E-values) less than 0.001 (Gerstein, 1998; Pearson, 2000). In many cases E-values less than 0.001 will not be stringent enough to decide relationships of homology (see below), but this low ratio does identify statistically significant relatedness. However, in most cases, decisions about which potential PRPs to score as a likely homologue were not based on E-values alone. Because proteins and protein subdomains evolve at different rates, any single statistical value will be too inclusive for some complex sequences in “extended” gene families. Those query sequences that failed to detect significantly related sequences in the Arabidopsis database are listed in a footnote at the bottom of the appropriate table. Several classes of Arabidopsis sequences passed criteria we set for statistical significance (E < 0.001), but we feel particularly uncertain of a relationship of evolutionary homology to the query sequence. In these cases, the number of sequences identified in the search and their E-values are included in each table (forth column), but the number of genes (second column of each table) is listed as zero.

The list of PRP genes in Table 1 is meant to be used as a guide to future investigation. The majority of PRP gene sequences discussed herein are likely homologues of their named bacterial, animal, and fungal counterparts. However, perhaps as many as 5 to 20% of the sequences identified as potential PRPs will be artifacts resulting from an examination of such a large database. Many sequences are presented as separate groupings in the database, when they actually should be grouped with other sequences, are at the same genetic locus, and are not gene duplications. This leads to a significant overestimate of the number of independent genes in a family. Other sequences may be placed in the wrong class of PRPs due to accidents of conserved subdomains, sequence convergence, and/or neutral drift. Undoubtedly, even larger numbers of PRP genes have not been detected in our study due to rapid rates of plant sequence divergence from their counterparts in bacteria or other eukaryotes, or due to evolution of novel motifs and sequences within plants. In summary, it is not possible to guarantee the absolute accuracy of these data on such large numbers of genes without further exploring each group of sequences identified in more detail.

Families of PRPs:

Of the 700 potential PRPs presented in Table 1, approximately 450 were enzymes catalyzing chemical or electrochemical (redox) reactions; approximately 250 were transport proteins, some of which also carry out chemical transformations like ATP hydrolysis; and several were the metal binding proteins, metallothioneins. Twenty classes of PRP enzymes and proteins identified in Arabidopsis were in small (2 to 10 genes) to moderate (10 to 25 genes) sized gene families (Table 1). However, five classes of PRPs were encoded by very large gene families (26 to 200 genes). Because of their critical roles in remediation, several of the largest families of PRP genes are discussed briefly below.

Cytochrome P-450s:

Cytochrome P-450s are a diverse class of enzymes well characterized for their roles in toxic metabolite degradation in animals and bacterial remediation of pollutants. P-450 enzymes catalyze diverse reactions, acting as monoxygenases, epoxidases, hydroxylases, and demethylases, and they recognize a wide variety of substrates. For example, the oxidation steps in the degradation of benzene (Kalf, 1987), polycyclic aromatics like benzo(a)pyrene shown in Figure 2H (Oesch et al., 1985), and chlorinated solvents like TCE shown in Figures 2E & 2F (Doty et al., 2000) all are known to be catalyzed by cytochrome P-450 enzymes. As shown in Table 1, when a human P-450 defined as a benzo(a)pyrene hydrolase was used as a query, it detected 284 significantly-related Arabidopsis P-450 sequences in the database. These data are in close agreement with a recent review suggesting the Arabidopsis P-450 superfamily might contain as many as 286 members (Xu et al., 2001). Further scrutiny of the P-450 sequences identified, however, suggests that there are many duplicates that were misgrouped as separate distinct sequences (Marshburn, McKinney, and Meagher, unpublished data). Nevertheless, the actual number of P-450 sequences still may be greater than 200, an extraordinarily large family. To emphasize the significance of plants encoding such a large number of P-450s, the human sequence query detected only seven yeast sequences with similar levels of significance (E values = 2e-12:0.03, not shown). The yeast genome encodes five times fewer genes than Arabidopsis (Kowalczuk et al., 1999), thus the Arabidopsis genome would encode only 35 P-450-related sequences if the number of P-450s were simply proportional to total gene number. A similar large number of cytochrome P-450 sequences were found in Arabidopsis when a rat TCDD dioxin-induced P-450 was used as a query (Table 1).

Oxygenases, dioxygenases, and peroxidases:

Oxygenases (laccases), dioxygenases, and/or peroxidases are required to degrade aromatic ring-containing compounds (benzene, benzo(a)pyrene) to catechols, dicarboxylic acid containing compounds, PCBs, and quinones. An example of how a dioxygenase participates in the degradation of the highly stable PCB dibenzyl(p) dioxin is given in Figure 2J ( www.rrz.uni-hamburg.de/biologie/ialb/mbio/doxinep.htm). Searching the Arabidopsis genome with a bacterial dioxygenase, a fungal laccase, and a bacterial peroxidase, 22 dioxygenase-related sequences, 38 laccase-related sequences and 13 peroxidase-related sequences, respectively, were found with likely homology to the queries (Table 1). Searching the yeast genome with the same query sequences shown in Table 1, only three dioxygenase-related sequences (E = 2e-5: 3e-4), three laccase-related sequences (E = 3e-53:6e-17), and one peroxidase-related sequence (E = 1e-20) were found. Perhaps it is not surprising to find so many dioxygenases, laccases, and peroxidases encoded by a plant genome proportionally to yeast considering their role in complex carbohydrate and lignin metabolism in plants. However, the role of these enzymes in degrading soil-born toxins is essentially unknown.

Dehalogenases:

Dehalogenases (hydrolases, hydratases) are required to hydrolyze chlorine and fluorine from halogenated linear hydrocarbon compounds like TCE and halogenated aromatic compounds like DDT or PCBs. A bacterial chlorobenzoyl dehalogenase query sequence (Table 1) revealed the presence of 14 or more significantly related Arabidopsis gene sequences (E = 1e-17: 2e-4) with the potential to encode dehalogenase activities. Most are listed as hydratases or hydrolases or unknown sequences. In the entire eukaryotic sequence database, only two animal sequences were any more related to the bacterial dehalogenase query than the best matched Arabidopsis sequences. For comparison, only three distantly related sequences (E = 6e-11:2e-4) were found in yeast. Plant dehalogenases could be part of a catabolic pathway degrading chlorinated toxins produced by soil microorganisms (Debono et al., 1980; Trew et al., 2000). In spite of these exciting results, it is difficult to know with certainty if any of these plant sequences have significant dehalogenase activity or if they are just part of the larger hydratase / hydrolase family. Genetic and biochemical studies will be necessary to determine the activities of the plant enzymes.

Transporters and pumps:

Approximately 250 Arabidopsis genes encode proteins with homology to various pumps and transporters with potential roles in phytoremediation. In addition to the normal transport of ions, small molecules, and toxic waste products within and between cells and organs, plants extract the majority of their nutrients from soil. Thus, plants can be expected to have a large number of root-specific transporters. Some of these transporters may function to inadvertently eliminate environmental pollutants taken up by plants, while playing the role for which they originally evolved, such as actively eliminating toxic products of plant metabolism. Other PRPs may accidentally pump toxic ions and chemicals into plants in an attempt to scavenge for required nutrients. For example, toxic levels of heavy metal ions like Cd(II), Co(II), Mn(II), and arsenate can be transported into or out of cells by one of the following transporters: the Ca⁺⁺/H⁺ ATPase related pumps (Nedelkoska and Doran, 2000; Tynecka et al., 2001), Ca⁺⁺/H⁺ and Zn⁺⁺/H⁺ antiporters (Nies, 1995; Endo et al., 1998; Goncalves et al., 1999), iron IRT transporters (Cohen et al., 1998), phosphate transporters (Huang and Lee, 1996; Pickering et al., 2000), and zinc ZNT transporters (Lasat et al., 2000; Pence et al., 2000).

The multidrug resistance proteins (MRPs) are ABC class transporters that can eliminate or sequester glutathione conjugates of toxins and other bulky toxic organic anions as discussed above. ABC transporters have many other transport functions relevant to phytoremediation (Sanchez-Fernandez et al., 2001). Yeast YCF1 is one of the best-characterized MRP-type ABC transporters. When used as a query sequence, YCF1 detects 118 significantly related homologues of ABC transporters in the Arabidopsis genome (Table 1). This number is in reasonable agreement with the 129 potential ABC transporters recently reported to be encoded by the Arabidopsis genome (Sanchez-Fernandez et al., 2001). Besides YCF1, yeast contains about 26 other ABC transporters with similar levels of homology (E = 0:0.001), about the number expected based on relative gene number in the two organisms.

Arabidopsis appears to contain large gene families encoding other classes of transporters as well. For example, using a yeast Ca⁺⁺/H⁺ ATPase pump as a query sequence, more than 70 Arabidopsis transporter sequences of this type were identified, compared to 23 related sequences in yeast. The yeast IRT1 iron transporter and ZNT1 zinc transporter query sequences were significantly related to 11 and 15 Arabidopsis sequences, respectively. While the yeast phosphate transporter Pho84 identified 77 significantly related Arabidopsis sequences, most of the sequences identified were more related to sugar transporters of the same transporter class. Only 13 were likely functional homologues expected to be acting as phosphate transporters and thus with the potential to bring in, for example, the phosphate analogue arsenate.