Chemical genetic identification of glutamine phosphoribosylpyrophosphate amidotransferase as the target for a novel bleaching herbicide in Arabidopsis

A novel phenyltriazole acetic acid compound (DAS734) produced bleaching of new growth on a variety of dicotyledonous weeds and was a potent inhibitor of Arabidopsis ( Arabidopsis thaliana ) seedling growth. The phytotoxic effects of DAS734 on Arabidopsis were completely alleviated by addition of adenine to the growth media. A screen of ethylmethanesulfonate-mutagenized Arabidopsis seedlings recovered seven lines with resistance levels to DAS734 ranging from five to 125-fold. Genetic tests determined that all the resistance mutations were dominant and allelic. One mutation was mapped to an interval on chromosome 4 containing At4g34740, which encodes an isoform of glutamine phosphoribosylamidotransferase (AtGPRAT2), the first enzyme of the purine biosynthetic pathway. Sequencing of At4g34740 from the resistant lines showed that all seven contained mutations producing changes in the encoded polypeptide sequence. Two lines with the highest level of resistance (125-fold) contained the mutation Arg264Lys. The wild-type and mutant AtGPRAT2 enzymes were cloned and functionally overexpressed in Escherichia coli . Assays of the recombinant enzyme showed that DAS734 was a potent slow-binding inhibitor of the wild-type enzyme (I 50 ~0.2 µM) whereas the mutant enzyme R264K was not significantly inhibited by 200 µM DAS734. Another GPRAT isoform in Arabidopsis, AtGPRAT3, was also inhibited by DAS734. This combination of chemical, genetic and biochemical evidence indicates that the phytotoxicity of DAS734 arises from direct inhibition of GPRAT and establishes its utility as a new and specific chemical genetic probe of plant purine biosynthesis. The effects of this novel GPRAT inhibitor are compared to the phenotypes of known AtGPRAT genetic mutants.

The developing field of chemical genetics relies on the ability of certain small molecules to mimic the effect of genetic mutations by blocking, or otherwise modulating, specific cellular processes (Blackwell and Zhao, 2003, Mayer, 2003, Stockwell, 2000. The use of small molecules to perturb plant biological processes can have several convenient advantages over genetic methods. Compounds can be applied and removed at specific times and locations to rapidly produce their effects and they have the potential to be used on a variety of different species including those that are not genetically tractable. Recent chemical genetic studies using plant systems have identified several new chemical effectors of a variety of plant cellular functions (Armstrong et al., 2004, Asami et al., 2003, Surpin et al., 2005, Zhao et al., 2003, Zheng et al., 2006. Compounds that intervene in primary metabolism can be of special interest for chemical genetic studies modulating plant biosynthetic pathways and their downstream events. Many herbicides have potent and specific action within these pathways and so are instructive chemical probes of plant biosynthetic processes. For example, commercial herbicide targets include enzymes in branched chain and aromatic amino acid, fatty acid, cellulose and plastoquinone biosynthesis (Wakabayashi and Boger, 2002). Much has been learned about the individual target enzymes, the pathways and their role in plant metabolism via detailed understanding of the effects of these herbicides. This suite of chemical probes has been expanded by many experimental herbicidal compounds that (for a variety of reasons) have not achieved commercialization but can still serve as informative inhibitors of additional pathways. Examples are inhibitors of histidine biosynthesis (Cox et al., metabolism (Smith and Atkins, 2002). Thus small molecule inhibitors of purine synthesis may be useful for modulating a wide variety of biologically important plant processes, as well as being potentially useful leads for herbicide design. The glutamine antagonists azaserine, acivicin and 6-diazo-5-oxo-L-norleucine have often been used as purine biosynthesis inhibitors but these compounds interfere with many amidotransferases and so lack specificity (Lyons et al., 1990). The natural product phytotoxins hydantocidin and ribofuranosyl triazolonone have been shown to exert their toxic effect by bioconversion via phosphorylation into inhibitors of adenylosuccinate synthetase (AdSS), the enzyme catalyzing the penultimate step in AMP biosynthesis (Cseke et al., 1996, Heim et al., 1995, Schmitzer et al., 2000, Siehl et al., 1996. Another natural product, hadacidin, inhibits plant AdSS (Hatch, 1967). Purine biosynthesis is also a well known target for many antineoplastic and antiviral agents designed for potential pharmaceutical use (Christopherson et al., 2002).
We found that treatment of plants with a simple but novel phenyltriazole acetic acid compound results in a phytotoxic phenotype of interest in several dicotyledonous plant species by producing distinctive bleaching of developing leaves. The compound is especially potent on Arabidopsis thaliana. A better understanding of the observed biological effects (phenotypes) of chemical genetic probes can be gained by knowledge of their specific target site(s). However, determination of the precise biochemical targets of effective compounds is often a difficult and rate-limiting step in the exploitation of new compounds for chemical genetic studies (Burdine andKodadek, 2004, Tochtrop andKing, 2004). We have used a combination of genetic, chemical and biochemical techniques to elucidate the molecular target of the phenyltriazole acetate compound as glutamine phosphoribosylpyrophosphate amidotransferase (GPRAT), the first committed enzyme of de novo purine biosynthesis. The effects of the compound can be compared to the phenotypes of known AtGPRAT genetic mutants (Hung et al., 2004, van der Graaff et al., 2004. We have functionally expressed Arabidopsis thaliana GPRAT (AtGPRAT) in Escherichia coli at high levels and have shown that DAS734 is a slow tight-binding inhibitor of the enzyme. Thus we have identified a new and specific chemical probe of the first step in plant purine biosynthesis and characterized its mode of interaction with its cognate target.

Phytotoxic activity of DAS734
The phenyltriazole acetic acid [5-(4-chlorophenyl)-1-isopropyl-1H-[1,2,4]triazol-3yl]-acetic acid (DAS734, Fig. 1B) exerted modest herbicidal activity on the seedlings of several dicotyledonous weed species in greenhouse postemergent applications. It was most active on morningglory (Ipomoea hederacea), field pansy (Viola arvensis), redroot pigweed (Amaranthus retroflexus) and buckwheat (Polygonum convolvulus). The phytotoxic symptoms were characterized by extensive bleaching of new growth such that newly emerging leaves turned white (Fig. 1C). Although greenhouse activity against commercially relevant weeds was modest, we found that DAS734 was very phytotoxic to Arabidopsis seedlings grown on agarose media containing the compound (Fig. 1D). The concentration of DAS734 required to inhibit root growth by 50% (RI 50 ) was 200 nM. The ethyl ester of DAS734 was even more potent with an RI 50 of 30 nM. It is likely that the neutral ethyl ester may have improved uptake relative to the free acid form. The high level of activity in Arabidopsis seedling assays approaches that of some potent commercial herbicides such as acetolactate synthase inhibitors.
The phytotoxic effects of many inhibitors of primary metabolism can often be reversed by addition of the end product(s) of the affected pathway (Subramanian et al., 1999). A variety of metabolites were added individually or in pools to the Arabidopsis seedling growth medium in the presence of DAS734. Of these, only the addition of 185 µM adenine was found to completely alleviate all of the phytotoxic effects of DAS734 including bleaching and root inhibition ( Fig. 2A) whereas amino acids, pyrimidines and various other products of primary metabolism had no effect. Alleviation of phytotoxicity was found to a lesser extent with addition of adenosine and adenine nucleotides but other purines such as hypoxanthine, guanine and inosine had no effect (data not shown). This pattern of symptom alleviation restricted to adenine and its derivatives is similar to that seen with hydantocidin and hadacidin (Heim et al., 1995, Siehl et al., 1996. However DAS734 had no effect in enzyme assays of AdSS, the target of hydantocidin and hadacidin, or the subsequent enzyme in purine biosynthesis, adenylosuccinate lyase (data 8 not shown). This suggested that DAS734 may act at some novel site involved in purine biosynthesis. As there are ten additional enzymes in de novo purine biosynthesis, many of which are difficult to directly assay, we elected to take a genetic approach to dissect the precise site of action. Because DAS734 had no inhibitory activity on E. coli, cyanobacteria, green algae or yeast but was a potent inhibitor of Arabidopsis growth, Arabidopsis was chosen as the genetic model organism of choice for these studies.

Screen for mutants resistant to DAS734
To identify mutants resistant to DAS734, ethylmethanesulfonate (EMS)-mutagenized M 2 Arabidopsis seeds were germinated and grown in medium containing sub-lethal concentrations of the herbicide. Several pilot screens of ~42,000 seedlings were performed at DAS734 concentrations between 0.18 and 1.1 µM and a concentration of 0.54 µM DAS734 was selected to perform a large-scale screen of 480,000 seedlings. At this concentration, seedlings formed purple cotyledons, did not produce true leaves and had significantly inhibited root growth. Resistant mutants were identified as seedlings with green, fully expanded cotyledons, emerging true leaves and normal root length.
Sixteen mutants were recovered and confirmed as showing some level of resistance after rescreening M3 seed on media containing 0.54 µM DAS734. When assessed on a lower concentration of DAS734 (0.054 µM), nine mutants produced green cotyledons but had some inhibition of growth of roots and true leaf formation. Seven mutants (R01, R03, R04, R06, R07, R08, and R22) showed complete resistance with growth equivalent to control seedlings. These mutants were used in further analyses.
To quantify the levels of resistance, the seven mutants were grown on media containing increasing concentrations of DAS734 and RI 50 values were determined for each mutant (Figs. 2B and C). The mutants could be readily classified into three distinct groups showing strong (>110-fold), moderate (~60-fold) and weak (~5-fold) resistance (Table 1) relative to wild-type Col-0 seedlings. All seven mutants were phenotypically normal, healthy and fertile under our standard growth conditions on agarose media and in the greenhouse. The adult Arabidopsis mutant plants also showed significant resistance to DAS734 in greenhouse postemergent spray applications of the herbicide. The application rate of the compound giving a 50% reduction in growth relative to an untreated control was 80 g/ha for wild-type plants, whereas treatments of up to 2000 g/ha had little or no effect on the strongly resistant R06 mutant line (Fig. 2D). This corresponds to >25-fold resistance for the adult plants.

Resistance to other herbicidal chemistries
To ensure that the observed herbicide resistance was specific to DAS734, the seven mutant lines were tested for resistance to three other herbicidal chemistries (structures shown in Fig. 1b). DAS073 is a close structural analog of DAS734 and shares the same phytotoxic phenotype that can be alleviated by addition of adenine (data not shown). All of the DAS734-resistant mutants tested were cross-resistant to DAS073, consistent with a shared mode of action. DAS309 is an isomer of DAS734 differing only in the attachment position of the isopropyl group on the triazole ring. However, it produces a different phytotoxic phenotype (no bleaching) that is not alleviated by adenine. All of the DAS734-resistant mutants were sensitive to DAS309 suggesting that the resistance mechanism is specific to the mode of action rather than via recognition of similar chemical structures. This provided some reassurance that resistance could be due to mutations at the target site and not from serendipitous chemical recognition by mechanisms involving metabolism or pumping to another cell compartment. Hadacidin is structurally unrelated to DAS734 but is known to inhibit purine biosynthesis via inhibition of AdSS. Its phytotoxic symptoms can be alleviated by adenine (Heim et al., 1995). All of the mutants were sensitive to hadacidin suggesting that DAS734 has a site of action affecting purine biosynthesis that is different from that of the hadacidin target, AdSS. The chemical specificity of the resistance of the mutants to different herbicidal chemistries encouraged us to further genetically characterize the mutants.

Genetic characterization of the DAS734-resistant mutants
In four of the seven DAS734-resistant mutants tested (R01, R06, R07 and R08), M 3 progeny segregated for resistance. The ratio of resistant to sensitive progeny was approximately 3:1 suggesting that these lines contained a dominant mutation and the M 2 plants were heterozygous for the mutation. The progeny from the remaining three mutants were 100% resistant to DAS734 indicating that the mutations in these lines were homozygous for the mutation. Further test crosses with wild-type plants showed that these DAS734 resistant mutants also contained dominant mutations. To determine if the mutations in the DAS734-resistant mutants were allelic, M 3 lines identified as homozygous for the resistance mutation were crossed with each other. In every case, all of the F 2 seed was resistant to DAS734. This indicates that the DAS734 resistant phenotypes in all seven lines were caused by mutations in the same or closely linked genes. Because all the mutants appeared to have a mutation at the same locus, only one mutant line was used to map the resistance gene.

Genetic mapping and identification of the resistance mutations
To identify the gene responsible for the mutant phenotype, the mutation in the strongly resistant line R03 was used for genetic mapping experiments. The wild-type allele was mapped to the lower arm of chromosome 4 between 1.26 x 10 7 bp and 1.58 x 10 7 bp using single nucleotide polymorphism (SNP) markers (Supplemental Fig. S1).
Genes in this region were inspected to identify if any were involved in purine biosynthesis or metabolism. One gene (At4g34740) was found with annotations matching these criteria and encodes GPRAT, the first committed enzyme in the purine biosynthetic pathway. There are three isoforms of GPRAT in the Arabidopsis genome and At4g34740 encodes AtGPRAT2. To determine if mutant line R03 contained a mutation in At4g34740, the gene was PCR amplified from both the R03 mutant and wild-type genomic DNA and the PCR products were sequenced. A single transition mutation of G to A was detected at nucleotide 791 relative to the start codon. This mutation causes Lys264 to be changed to Arg in the encoded polypeptide (Fig. 3). Because the mutations in the other DAS734-resistant mutants appeared to be allelic with the R03 mutation, the At4g34740 gene was PCR amplified and sequenced from the other mutant lines. All seven DAS734-resistant lines were found to contain mutations affecting the coding sequence of this gene (Table 1, Fig. 3) thus these mutations are likely to be responsible for conferring resistance to DAS734. R06 (145-fold resistance to DAS734) contained the same mutation in AtGPRAT2 as R03, an Arg-to Lys substitution at codon 264. R01, R07 and R08 (~50-fold resistance to DAS734) all contain a Pro to Ser mutation at residue 476. These mutants do not appear to be siblings as R01 and R07 each contained another different mutation in AtGPRAT2.
R01 contained a Thr-to-Ala mutation at residue 130 and R07 contained a silent A-to-G mutation at nucleotide 416 (relative to the start codon). The T130A mutation in R01 probably does not contribute to resistance as the two other phenotypically indistinguishable mutants that contain only the P476S mutation (R07 and R08) do not contain this mutation. The two mutants displaying the lowest level of resistance to DAS734, R04 and R22, contained unique mutations. R04 contained two mutations, a Proto-Ser mutation at amino acid 265 and a Tyr-to-Phe at amino acid 494 whereas R22 contained a single Gly-to-Ser mutation at amino acid 371 (Table 1, Fig. 3). The P265S mutation in R04 is perhaps a more likely candidate to confer resistance than Y494F as it is adjacent to the R264K mutation that gives high levels of resistance in R03 and R06.

Heterologous Expression of AtGPRAT2 in E. coli
GPRATs are highly regulated enzymes and they exert significant control over de novo purine biosynthesis via feedback inhibition by purine nucleotides (Chen et al., 1997). At this point in our investigations it was unclear if DAS734 directly inhibited GPRAT and mutations in the enzyme conferred resistance or if the mutations altered regulation of the enzyme allowing the mutant plant to overproduce purines and therefore alleviate the effects of the herbicide, as addition of exogenous adenine does. We therefore cloned and overexpressed AtGPRAT2 in E. coli to ascertain the direct effects of DAS734 on the enzyme.
All eukaryotic and many microbial GPRATs possess a short N-terminal propeptide that is autocatalytically cleaved to yield a conserved N-terminal cysteine. The γsulfhydryl group of this cysteine residue acts as the nucleophile in the glutamate transaminase portion of the two-step reaction catalyzed by the enzyme, thus GPRATs are members of the self-processing Ntn (N-terminal nucleophile)-hydrolase class of enzymes (Smith, 2004). In addition, AtGPRAT2 is predicted to contain a chloroplast transit peptide (Hung et al., 2004) (Fig. 3). For expression in E. coli, we truncated the AtGPRAT2 gene to eliminate the chloroplast targeting sequence leaving eleven residues before the putative catalytic Cys-87 (Fig. 3). This protein was expressed in E. coli and a novel polypeptide of ~54 kD was detected by SDS-PAGE analysis of cell pellet extracts ( Fig. 4A). The extracts were assayed for GPRAT activity by measuring the PRPPdependent production of glutamate from glutamine. Extracts from E. coli containing the plasmid encoding the AtGPRAT2 gene contained high levels of GPRAT activity whereas extracts produced from E. coli containing a control plasmid had negligible activity (<5% of the activity in extracts from cells expressing AtGPRAT2). The Km for glutamine of the recombinant enzyme was 1.34 mM (Supplemental Fig. S2). This is considerably lower than that determined for soybean GPRAT purified from root nodules (18 mM), but similar to that of recombinant GPRATs from microbial sources (Reynolds et al., 1984).

Inhibition of AtGPRAT2 by DAS734
The enzyme activity of AtGPRAT2 was potently inhibited by the addition of DAS734 to the assays (Fig. 4B). Initial Michaelis-Menten kinetic characterization showed that DAS734 behaved like a noncompetitive inhibitor with respect to glutamine (Supplemental Fig. S2). However this type of kinetic behavior can be manifested by slow tight-binding inhibitors. To evaluate whether DAS734 exhibited slow tight-binding characteristics, the enzyme reaction progress was monitored over time after simultaneous addition of substrates and inhibitor. The time course was clearly curved indicating that inhibition of the enzyme increased with time ( Fig. 4C). I 50 curves generated by preincubating the inhibitor with enzyme over increasing time intervals also showed increasing potency (data not shown). Maximal inhibition could be achieved by a ten minute pre-incubation of DAS734 with AtGPRAT2 without any other substrate additions being required, yielding an average I 50 value under our assay conditions of 0.2 µM. This suggests that DAS734 binds to GPRAT without requirement for glutamine or PRPP binding. Enzyme that was fully inhibited by DAS734 was passed over a desalting column to remove unbound ligand. Forty seven per cent of the enzyme activity could be recovered relative to an untreated control, indicating that the dissociation rate of DAS734 from the enzyme is relatively slow but reversible and accounts for the slow tight-binding behavior. showing >500-fold increase in I 50 over that of wild-type (Fig. 4B). Also, DAS734 had no effect on the reaction time course catalyzed by the mutant enzyme ( Fig.   4C), in marked contrast to the wild-type form. Although the mutant enzyme was uninhibited by DAS734 at 100 µM in biochemical assays of GPRAT activity, root growth of the Arabidopsis lines harboring this mutated GPRAT (R03 and R06) was somewhat inhibited at DAS734 concentrations above 20 µM (Fig. 2B). This could be due to additional non-specific effects on root growth at these high levels of compound.

Inhibition of other AtGPRAT isoforms by DAS734
There are three expressed GPRAT homologs in the Arabidopsis genome that are differentially expressed in various plant tissues (Boldt and Zrenner, 2003, Hung et al., 2004, Ito et al., 1994. AtGPRAT1 shares 93% amino acid identity with AtGPRAT2 over the length of the mature polypeptide whereas AtGPRAT3 has 72% identity.  Studies on other GPRATs from microbes and vertebrates show that the enzyme is allosterically inhibited by end products of the purine biosynthesis pathway (Chen et al., 1997). We wished to ascertain if the highly resistant mutant AtGPRAT2 R264K was altered in its ability to respond to feedback regulators. This could affect its biological function and possibly contribute to the resistance of intact mutant plants to DAS734 treatment. The precise combination and concentration of purine nucleotides that exert maximal effect on GPRATs appears to vary according to the source organism. Little information is available for plant enzymes although soybean GPRAT purified from root nodules was found to be inhibited by AMP, IMP and GMP (Reynolds et al., 1984). We examined the effect of purine nucleotides on wild-type AtGPRAT2 and found that maximal inhibition of the enzyme was produced by combinations of adenine nucleotides ( Fig. 4D) whereas minimal inhibition was seen with IMP or GMP under our assay conditions. The mutant enzyme AtGPRAT2 R264K had the same response to adenine nucleotides as the wild-type enzyme indicating that the R264K mutation conferring DAS734 resistance did not affect allosteric inhibition by purine nucleotides (Fig. 4E).

Treatment with DAS734 phenocopies atd2
The Arabidopsis genetic mutant atd2 has a T-DNA insertion within the AtGPRAT2 gene and was identified by its variegated bleached seedling phenotype and strong growth retardation. The wild-type phenotype could be restored by addition of 5 mM IMP (van der Graaff et al., 2004). We assessed if low concentrations of adenine could also restore normal growth to this mutant. We found that bleaching of atd2 was completely alleviated and normal growth restored by addition of 185 µM adenine to the medium (Fig. 6). In contrast, addition of 185 µM hypoxanthine had no effect. Thus the phenotype of atd2 showing bleaching and reduced growth (van der Graaff et al., 2004) is similar to that found by treatment of wild-type seedlings with DAS734.

Reversal of phytotoxic effects of DAS734 by adenine
Through a combination of chemical and genetic techniques, we have determined the precise biochemical site of action of DAS734 as AtGPRAT, the first committed enzyme in de novo purine biosynthesis. This identification was facilitated by the observation that the purine adenine completely alleviated the phytotoxic symptoms from germination to formation of the first true leaves. The linear purine biosynthetic pathway splits after the formation of IMP to produce either AMP or GMP via different enzymes (Fig. 1A). Thus it might be expected that alleviation of inhibition of purine biosynthesis prior to IMP formation may require addition of guanine or xanthine in addition to adenine. Indeed this logic was used to initially define the site of action of hydantocidin as AdSS (Heim et al., 1995). With our discovery of a potent and specific inhibitor of GPRAT, we observed that addition of a relatively low concentration (185 µM) of exogenous adenine is sufficient to restore normal seedling growth. This is likely due to the presence of efficient purine recycling mechanisms in plants such that adenine is readily taken up and converted to AMP by adenine phosphoribosyltransferase (APRT) (Ashihara et al., 2000, Katahira and Ashihara, 2006, Lee and Moffatt, 1994. AMP can then be converted to IMP by AMP deaminase (ADA) to feed the GMP biosynthesis branch of the pathway (Fig. 1A). Unlike vertebrates and microbes, plants do not seem to contain high levels of hypoxanthine/guanine phosphoribosyltransferase activity to efficiently recover exogenous hypoxanthine or guanine to form IMP or GMP (Ashihara et al., 2001, Katahira and Ashihara, 2006). Thus adenine (rather than adenosine, inosine, hypoxanthine or guanine) is the most efficient exogenous reagent to reverse the toxic effects of DAS734.

GPRAT redundancy in Arabidopsis
There are three functional GPRAT genes in the Arabidopsis genome (Boldt and Zrenner, 2003, Hung et al., 2004, Ito et al., 1994, van der Graaff et al., 2004. This is in contrast to most of the other genes encoding enzymes of purine biosynthesis that are present in single copies (Boldt and Zrenner, 2003). However, all of the DAS734resistance mutations that we identified occur in only one GPRAT isoform, AtGPRAT2.
Thus restoration of normal growth can be achieved by resistance mutations in this single GPRAT isoform even though we showed that AtGPRAT3 was also susceptible to DAS734 inhibition. (We were unable to express AtGPRAT1 in E. coli to enable inhibitor testing.) This is consistent with the observation that AtGPRAT2 is the major GPRAT expressed in roots, leaves and flowers (Hung et al., 2004). AtGPRAT3 is expressed at much lower levels in these tissues whereas AtGPRAT1 is expressed at low levels in roots and flowers only (Hung et al., 2004, Ito et al., 1994. Mutants lacking AtGPRAT2 (atd2, cia1) are severely stunted with bleached leaves but are viable and fertile under low light conditions. The bleached seedling phenotype of cia1 and atd2 can be alleviated by the addition of 5 mM AMP or IMP to the medium (Hung et al., 2004, van der Graaff et al., 2004. We found that 185 µM adenine was sufficient to abolish bleaching of atd2, similar to the result found with DAS734 inhibition. The viability of atd2 and cia1 indicates that Arabidopsis can survive but does not thrive via the function of AtGPRATs 1 and/or 3.
The AtGPRAT1 knockout mutant is indistinguishable from wild type. Double mutants lacking both AtGPRAT1 and AtGPRAT2 have the same phenotype as those lacking AtGPRAT2, again suggesting that AtGPRAT2 is the primary GPRAT (Hung et al., 2004). However, treatment of Arabidopsis seedlings with >5 µM DAS734 is lethal. This may be explained by our observation that AtGPRAT3 is also strongly inhibited by DAS734 and hence the compound can shut down additional GPRAT activity in the plant.
Although the genetic and chemical data are consistent with AtGPRAT2 being the predominant isoform supplying nucleotides for growth and development of Arabidopsis, AtGPRAT3 may be required for more specialized but important secondary functions.
Thus treatment by DAS734 may be equivalent to double or triple knockout mutants lacking AtGPRATs 2 and 3 or all GPRAT activity.

Bleaching effect from GPRAT inhibition
The phenotype produced by DAS734 is distinctive in comparison with many other importing proteins in in vitro assays. As ATP and GTP are required for transport of proteins into plastids (Soll and Schleiff, 2004), this defect may be because of lower availability of purine nucleotides in the AtGPRAT2 mutant chloroplasts. However, poor import efficiency could also arise as a secondary effect of defective chloroplast development. Lack of chloroplast development and characteristic bleaching of developing leaves is also produced by the phytotoxin tagetitoxin, an inhibitor of RNA polymerase (Lukens andDurbin, 1985, Mathews andDurbin, 1990). RNA synthesis could be reduced by a block in purine nucleotide availability and so this may offer another mechanism whereby loss of GPRAT function leads to the observed physiological symptoms. Van der Graaff et al. (2004) hypothesized that the bleached appearance of AtGPRAT2 mutants could be due to photooxidative damage but it is unclear why loss of GPRAT function should lead to this phenomenon. Treatment of plants with hydantocidin, an herbicidal inhibitor that blocks the penultimate step in AMP biosynthesis after the bifurcation of the purine biosynthesis pathway at IMP, does not produce the same bleaching effect as DAS734 on Arabidopsis or other plant species (Heim et al., 1995).
Bleaching symptoms from loss of GPRAT function (induced either genetically or chemically) may therefore be a consequence of a specific depletion of guanine nucleotides, or a combined deficiency of both adenine and guanine nucleotide pools.

Addition of exogenous adenine can supplement both AMP and GMP pools (via APRT
and ADA) to recover the normal phenotype.
Other herbicidal compounds that produce similar distinctive bleached-white phenotypes via photooxidative mechanisms are metabolic inhibitors of pathways that are either directly or indirectly involved in producing photoprotectant carotenoids. These include phytoene desaturase inhibitors e.g., norflurazon (Sandmann and Boger, 1997), 4hydroxyphenylpyruvate dioxygenase inhibitors e.g., mesotrione (Mitchell et al., 2001 and inhibitors of the non-mevalonate isoprenoid pathway e.g., fosmidomycin and clomazone (Lange et al., 2001, Mueller et al., 2000. It will be of interest to determine if these pathways are particularly sensitive to depletion of purines or intermediates of the purine pathway, either directly or indirectly. In addition to having a bleaching effect on emerging leaves, DAS734 is a potent inhibitor of root growth. This is in contrast to the above compounds that act exclusively by photooxidative mechanisms and so have little direct effect on non-photosynthetic tissues (T. Walsh, P. Schmitzer, R. Neal; unpublished results). Thus DAS734 has additional effects beyond bleaching of new growth, as would be expected from interruption of purine biosynthesis. DAS734 will be a useful chemical probe to ascertain the full scope of these metabolic consequences.

Facile heterologous overexpression of AtGPRAT
Many previous biochemical studies of GPRATs have been hindered by the extreme oxygen sensitivity of the enzymes (Bernlohr and Switzer, 1981, Itakura and Holmes, 1979). This arises from the lability of the structural 4Fe-4S cluster present in some microbial and all eukaryotic GPRATs. The 4Fe-4S cluster is presumably present in plant GPRATs as the essential ligand residues for the iron-sulfur center are conserved (Fig. 3,   Supplemental Fig. S3). We were successful in functionally overexpressing AtGPRAT2 at high levels in E. coli without the chloroplast transit peptide but retaining an eleven residue propeptide preceding the predicted active site Cys residue. Mass spectrometric analysis of purified recombinant AtGPRAT2-His expressed in E. coli indicated that the propeptide was removed to expose the N-terminal catalytic cysteine residue. Work with the B. subtilis enzyme has shown that this processing can occur autocatalytically (Brannigan et al., 1995). We took no special precautions to protect AtGPRAT2 from oxygen exposure and found the recombinant plant enzyme to be relatively stable to daily handling. No significant loss in activity was apparent after several weeks of storage at -70° C. The extraordinary stability of recombinant AtGPRAT2 to oxygen relative to other GPRATs from non-photosynthetic organisms could perhaps be an adaptive consequence of its location in the oxygen-rich environment of the chloroplast.

Sites of the DAS734-resistance mutations in GPRAT
Detailed studies of the Bacillus subtilis and E. coli GPRATs have shown that they are highly functionalized enzymes composed of two domains in tight allosteric communication with each other (Bera et al., 2000, Chen et al., 1997, Smith, 1998. One domain contains the glutaminase site where glutamine is hydrolyzed to yield ammonia.
The ammonia traverses an intramolecular tunnel, formed during the catalytic cycle, to the adjacent domain where it is condensed with PRPP to form the unstable product phosphoribosylamine (Krahn et al., 1997a).  (Bera et al., 1999, Smith, 1998

DAS734 as a novel chemical probe of plant purine biosynthesis
The combination of chemical and genetic approaches used in this study has provided a novel and unique small molecule ligand for probing GPRAT function and de novo purine biosynthesis in plants. Unlike many other non-specific inhibitors of purine biosynthesis (Lyons et al., 1990), DAS734 appears to be unusually specific for GPRAT from certain dicotyledonous plant species, in particular Arabidopsis. Its utility as a biochemical probe and an herbicidal lead is enhanced by the fact that it does not require bioactivation to a phosphorylated form as do inhibitors of other enzymes in the purine biosynthesis pathway such as hydantocidin (Cseke et al., 1996, Siehl et al., 1996.
DAS734 is a good example of one of the potential advantages of chemical genetics because it inhibits at least two of the three functional isoforms of GPRAT in Arabidopsis and so can overcome GPRAT genetic redundancy. Thus the compound will be of further utility in dissecting how chemical or genetic disruption of GPRAT activity leads to impaired chloroplast development and function and leaf bleaching.

Materials
Seeds of atd2 were obtained from the Arabidopsis Biological Resource Center.
DAS734 was synthesized as described in Supplemental Methods S1. All other reagents were obtained from Sigma Chemical Company.

Mutant Screening and Root Growth Assays
EMS-mutagenized M 2 Col-0 seed (Lehle Seeds, Round Rock, TX) were screened for resistance as described in (Walsh et al., 2006). DAS734-resistant seedlings were identified as plants with green fully expanded cotyledons and developing true leaves.
Arabidopsis seedling assays and metabolite reversal studies were also conducted as

Greenhouse Tests
Seeds of test species were planted in Metromix 360 (Brehob, Indianapolis, IN) 8-12 days prior to spraying with compound and grown in a greenhouse with a 16 hr photoperiod at 24-29°C. Compounds were dissolved in acetone-DMSO (97:3, v/v) and dilutions made with water-isopropanol-crop oil concentrate (78:20:2, v/v/v) containing 0.02% Triton X-155. Compounds were applied using a DeVilbiss compressed air sprayer.

Genetic Analyses
To resistant lines, M 4 seed resulting from self-fertilization of each parent used in the crosses was harvested and tested for resistance to DAS734. M 3 homozygotes produced M 4 seed that were 100% resistant to DAS734. F 1 progeny from crosses between the identified homozygous parents resistant to DAS734 were grown and allowed to self-fertilize. At least 300 F 2 seedlings from each cross were tested for resistance to DAS734. Lines containing independently segregating resistance genes should produce 93% resistant F2 seedlings whereas as allelic lines should produce 100% resistant seedlings. In all cases where the parent was homozygous for the resistance gene, F1 progeny from the test cross were 100% resistant to DAS734. When the herbicide-resistant parent was heterozygous for the resistance gene, approximately 50% of the F 1 progeny showed herbicide resistance.

Mapping of the R03 mutation
To generate the mapping population, a resistant homozygous M 3 line of R03 (from the Col-0 accession) was crossed with a wild-type plant of the Landsberg erecta (Ler) accession. The resulting F 1 plants were allowed to self-fertilize and produce F 2 seed.
Because the resistance phenotype is caused by a dominant mutation, the wild-type allele from the Ler background was mapped. The F 2 seed was germinated in modified MS medium containing 0.54 µM DAS734. Plants sensitive to the herbicide were identified, rescued onto medium without herbicide and transplanted to soil after seven days. When plants were at the rosette stage, one leaf was removed for genomic DNA isolation.
Mapping was performed using single nucleotide polymorphism markers as described in Walsh et al. (2006).
For sequencing, the At4g34740 (AtGPRAT2) gene was PCR amplified by Pfu Turbo DNA polymerase (Stratagene, LaJolla, CA) using the primers TCAACTTTTCATAA-TTGGTTTGTGTGTATTT and AATTTCGTAGAACTCGCCACAAGCA. The PCR products were sequenced using Applied Biosystems (Foster City, CA) BigDye Terminator v3.1 cycle sequencing kit and run on an ABI 3100.

Expression of AtGPRAT in E. coli
The gene encoding AtGPRAT2 was PCR amplified from an Arabidopsis thaliana cDNA library using the forward DNA: 5'-CATATGGATGATTATGACGAGAAGCC-TCGGGAAGAGTGTGGAG-3' and the reverse primer: 3'-TAGGTGTTGTAACT-TCCTCCAACCCATGCCATCATTCCTAGG-5'. To facilitate gene cloning, the forward and reverse primers encoded an NdeI and BamHI site respectively (underlined). In B.
subtilis, propeptide autoprocessing to reveal an N-terminal cysteine is required to generate an active GPRAT enzyme. The 5' primer was therefore designed to encode a comparable 11 amino acid propeptide (in bold) and deletion of the chloroplast transit peptide of AtGPRAT2. The amplification product was gel-purified and cloned using were grown for 24 hours at 28°C. Cells were harvested by centrifugation and the resulting cell pellets were frozen in dry ice and stored at -80°C.

GPRAT enzyme activity
For enzyme extraction, pelleted E. coli cells expressing AtGPRAT were resuspended in 0.1 M Tris, pH 7.4 containing 1 mg/mL lysozyme (5 mL/cells from 250 mL culture) at room temperature. After ~15 min, the suspension was frozen in liquid N 2 then thawed.
DNase was added to 0.02 mg/mL final concentration and MgCl 2 to 1 mM. After the extract was no longer viscous, dithiothreitol was added to 10 mM and the extract was

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Map-based cloning of DAS734-resistance mutation in resistant line R03 Supplemental Figure S2. Enzyme kinetic characterization of E. coli-expressed

AtGPRAT2 and inhibition by DAS734
Supplemental Figure S3. Comparison of the amino acid sequences of Arabidopsis and microbial GPRATs Supplemental Methods S1. Chemical synthesis of phenyltriazole acetate DAS734.

ACKNOWLEDGMENTS
We are grateful to several Dow AgroSciences scientists for invaluable assistance.    subtilis GPRAT crystal structure, PDB ID 1AO0 (Chen et al., 1997) and within 10 Å of the 6-diazo-5-oxonorleucine-derivatized active site cysteine in the E. coli GPRAT structure, PDB ID 1ECC1 (Krahn et al., 1997a) are denoted with a red dot below the sequence. These residues are therefore at or close to the glutaminase site of GPRAT.
Residues within 10 Å of the bridging oxygen of the ribose pyrophosphate in GMP in the B. subtilis 1AO0 structure and in the carbocyclic PRPP analog in the E. coli 1ECC1 structure) are denoted with a blue dot below the sequence. These residues are therefore at or close to the PRPP catalytic site. The four cysteine residues that are equivalent to the ligands for the 4Fe-S cofactor in B. subtilis GPRAT are marked.   hypoxanthine (C). The atd2 bleached mutant phenotype is reversed to a normal green color by addition of adenine whereas addition of hypoxanthine has no effect.