Molecular and biochemical analysis of serine acetyltransferase and cysteine synthase towards sulfur metabolic engineering in plants

Summary. Serine acetyltransferase (SATase) and cysteine synthase (O-acetylserine (thiol)-lyase) (CSase) are committed in the final step of cysteine biosynthesis. Six cDNA clones encoding SATase have been isolated from several plants, e.g. watermelon, spinach, Chinese chive and Arabidopsis thaliana. Feedback-inhibition pattern and subcellular localization of plant SATases were evaluated. Two types of SATase that differ in their sensitivity to the feedback inhibition by l-cysteine were found in plants. In Arabidopsis, cytosolic SATase was inhibited by l-cysteine at a physiological concentration in an allosteric manner, but the plastidic and mitochondrial forms were not subjected to this feedback regulation. These results suggest that the regulation of cysteine biosynthesis through feedback inhibition may differ depending on the subcellular compartment. The allosteric domain responsible for l-cysteine inhibition was characterized, using several SATase mutants. The single change of amino acid residue, glycine-277 to cysteine, in the C-terminal region of watermelon SATase caused a significant decrease of the feedback-inhibition sensitivity of watermelon SATase. We made the transgenic Arabidopsis overexpressing point-mutated watermelon SATase gene whose product was not inhibited by l-cysteine. The contents of OAS, cysteine, and glutathione in transgenic Arabidopsis were significantly increased as compared to the wild-type Arabidopsis. Transgenic tobacco (Nicotiana tabacum) (F₁) plants with enhanced CSase activities both in the cytosol and in the chloroplasts were generated by cross-fertilization of two transgenic tobacco expressing either cytosolic CSase or chloroplastic CSase. Upon fumigation with 0.1 μL L⁻¹ sulfur dioxide, both the cysteine and glutathione contents in leaves of F₁ plants were increased significantly, but not in leaves of non-transformed control plants. These results indicated that both SATase and CSase play important roles in cysteine biosynthesis and its regulation in plants. Received November 27, 2001 Accepted December 21, 2001     

Conversion of feedback regulation in aspartate kinase by domain exchange

To elucidate the mechanism for the regulation of aspartate kinase (AK) via feedback inhibition, we constructed several chimeric enzymes between Bacillus subtilis AK II, a lysine-sensitive mesophilic enzyme, and Thermus flavus AK, a threonine-sensitive thermostable enzyme, each having the same alpha2beta2-type tetrameric structure. A chimeric AK, named BTT, composed of the chimeric alpha subunit that comprises of the N-terminal catalytic region from B. subtilis AK II and the C-terminal region from T. flavus, and the beta subunit from T. flavus, was inhibited only by threonine. Another chimeric enzyme, BT, which has a similar structure to that of BTT but lacks the beta subunit, having alpha2-type homo-dimeric structure, was also responsive only to threonine. However, the addition of threonine enhanced the activity of BT. These results indicate the regulatory function of C-terminal region and beta subunit in AK. BTT showed extremely high thermostability comparable to that of T. flavus, suggesting that the beta subunit also contributed to the stability of the AK.     

Carboxy-terminal deletion analysis of the major pneumococcal autolysin.

Autolysins are endogenous enzymes that specifically degrade the covalent bonds of the cell walls and eventually can induce bacterial lysis. One of the best-characterized autolysins, the major pneumococcal LytA amidase, has evolved by the fusion of two domains, the N-terminal catalytic domain and the C-terminal domain responsible for the binding to cell walls. The precise biochemical role played by the six repeat units that form the C-terminal domain of the LytA amidase has been investigated by producing serial deletions. Biochemical analyses of the truncated mutants revealed that the LytA amidase must contain at least four units to efficiently recognize the choline residues of pneumococcal cell walls. The loss of an additional unit dramatically reduces its hydrolytic activity as well as the binding affinity, suggesting that the catalytic efficiency of this enzyme can be considerably improved by keeping the protein attached to the cell wall substrate. Truncated proteins lacking one or two repeat units were more sensitive to the inhibition by free choline than the wild-type enzyme, whereas the N-terminal catalytic domain was insensitive to this inhibition. In addition, the truncated proteins were inhibited by deoxycholate (DOC), and the expression of a LytA amidase lacking the last 11 amino acids in Streptococcus pneumoniae M31, a strain having a deletion in the lytA gene, conferred to the cells an atypical phenotype (Lyt+ DOC-) (cells autolysed at the end of the stationary phase but were not sensitive to lysis induced by DOC), which has been previously observed in some clinical isolates of pneumococci. Our results are in agreement with the existence of several choline-binding sites and suggest that the stepwise acquisition of the repeat units and the tail could be considered an evolutionary advantage for the enzyme, since the presence of these motifs increases its hydrolytic activity.     

Engineering and characterization of the ribosomal L10.L12 stalk complex. A structural element responsible for high turnover of the elongation factor G-dependent GTPase

The ribosomal stalk protein L12 is essential for events dependent on the GTP-binding translation factors. It has been recently shown that ribosomes from Thermus thermophilus contain a heptameric complex L10.(L12)2.(L12)2.(L12)2, rather than the conventional pentameric complex L10.(L12)2.(L12)2. Here we describe the reconstitution of the heptameric complex from purified L10 and L12 and the characterization of its role in elongation factor G-dependent GTPase activity using a hybrid system with Escherichia coli ribosomes. The T. thermophilus heptameric complex resulted in a 2.5-fold higher activity than the E. coli pentameric complex. The structural element of the T. thermophilus complex responsible for the higher activity was investigated using a chimeric L10 protein (Ec-Tt-L10), in which the C-terminal L12-binding site in E. coli L10 was replaced with the same region from T. thermophilus, and two chimeric L12 proteins: Ec-Tt-L12, in which the E. coli N-terminal domain was fused with the T. thermophilus C-terminal domain, and Tt.Ec-L12, in which the T. thermophilus N-terminal domain was fused with the E. coli C-terminal domain. High GTPase turnover was observed with the pentameric chimeric complex formed from E. coli L10 and Ec-Tt-L12 but not with the heptameric complex formed from Ec-Tt-L10 and Tt.Ec-L12. This suggested that the C-terminal region of T. thermophilus L12, rather than the heptameric nature of the complex, was responsible for the high GTPase turnover. Further analyses with other chimeric L12 proteins identified helix alpha6 as the region most likely to contain the responsible element.     

Role of the basic C-terminal half of caldesmon in its regulation of F-actin: comparison between caldesmon and calponin

We previously reported that caldesmon (CaD), together with tropomyosin (TM), effectively protects actin filaments from gelsolin, an actin-severing protein. To elucidate the structure/function relationship of CaD, we dissected the functional domain of CaD required for the protection. The basic C-terminal half of rat nonmuscle CaD (D3) inhibits gelsolin activity to the same degree as intact CaD, although a smaller C-terminal region of D3 does not. This smaller C-terminal region contains the minimum regulatory domain responsible for the inhibition of actomyosin ATPase, and for the binding to actin, calmodulin and TM. These results suggest that the domain responsible for the inhibition of gelsolin activity lies outside the minimum regulatory domain, and that the positive charge possessed by the C-terminal half of CaD is important for its interaction with actin. Moreover, while the D3 fragment promotes the aggregation of F-actin into bundles as reported previously, this bundle formation is inhibited by the acidic N-terminal half of CaD, as well as by poly-l-glutamate. It seems likely that the acidic N-terminal half of CaD neutralizes the superfluous basic feature of the C-terminal half. A comparison between D3 and calponin, another actin-binding protein that is also basic and has similar actin-regulatory activities, is also discussed.     

Structural and functional characterization of α-isopropylmalate synthase and citramalate synthase, members of the LeuA dimer superfamily

The manipulation of modular regulatory domains from allosteric enzymes represents a possible mechanism to engineer allostery into non-allosteric systems. Currently, there is insufficient understanding of the structure/function relationships in modular regulatory domains to rationally implement this methodology. The LeuA dimer regulatory domain represents a well-conserved, novel fold responsible for the regulation of two enzymes involved in branched chain amino acid biosynthesis, α-isopropylmalate synthase and citramalate synthase. The LeuA dimer regulatory domain is responsible for the feedback inhibition of these enzymes by their respective downstream products. Both enzymes display multidomain architecture with a conserved N-terminal TIM barrel catalytic domain and a C-terminal (βββα)2 LeuA dimer domain joined by a flexible linker region. Due to the similarity of three-dimensional structure and catalytic mechanism combined with low sequence similarity, we propose these enzymes can be classified as members of the LeuA dimer superfamily. Despite their similarity, members of the LeuA dimer superfamily display diversity in their allosteric mechanisms. In this review, structural aspects of the LeuA dimer superfamily are discussed followed by three examples highlighting the diversity of allosteric mechanisms in the LeuA dimer superfamily.     

Pig liver carnitine palmitoyltransferase. Chimera studies show that both the N- and C-terminal regions of the enzyme are important for the unusual high malonyl-CoA sensitivity.

Pig and rat liver carnitine palmitoyltransferase I (L-CPTI) share common K(m) values for palmitoyl-CoA and carnitine. However, they differ widely in their sensitivity to malonyl-CoA inhibition. Thus, pig l-CPTI has an IC(50) for malonyl-CoA of 141 nm, while that of rat L-CPTI is 2 microm. Using chimeras between rat L-CPTI and pig L-CPTI, we show that the entire C-terminal region behaves as a single domain, which dictates the overall malonyl-CoA sensitivity of this enzyme. The degree of malonyl-CoA sensitivity is determined by the structure adopted by this domain. Using deletion mutation analysis, we show that malonyl-CoA sensitivity also depends on the interaction of this single domain with the first 18 N-terminal amino acid residues. We conclude that pig and rat L-CPTI have different malonyl-CoA sensitivity, because the first 18 N-terminal amino acid residues interact differently with the C-terminal domain. This is the first study that describes how interactions between the C- and N-terminal regions can determine the malonyl-CoA sensitivity of L-CPTI enzymes using active C-terminal chimeras.     

The N-terminal region of the starch-branching enzyme from Phaseolus vulgaris L. is essential for optimal catalysis and structural stability

Starch-branching enzymes (SBEs) play a pivotal role in determining the fine structure of starch by catalyzing the syntheses of alpha-1,6-branch points. They are the members of the alpha-amylase family and have four conserved regions in a central (beta/alpha)8 barrel, including the catalytic sites. Although the role of the catalytic barrel domain of an SBE is known, that of its N- and C-terminal regions remain unclear. We have previously shown that the C-terminal regions of the two SBE isozymes (designated as PvSBE1 and PvSBE2) from kidney bean (Phaseolus vulgaris L.) have different roles in branching enzyme activity. To understand the contribution of the N-terminal region to catalysis, six chimeric enzymes were constructed between PvSBE1 and PvSBE2. Only one enzyme (1Na/2Nb)-II, in which a portion of the N-terminal region of PvSBE2 was substituted by the corresponding region of PvSBE1, retained 6% of the PvSBE2 activity. The N-terminal truncated form (DeltaN46-PvSBE2), lacking 46 N-terminal residues of PvSBE2, lost enzyme activity and stability to proteolysis. To investigate the possible function of this region, three residues (Asp-15, His-24, and Arg-28) among these 46 residues were subjected to site-directed mutagenesis. The purified mutant enzymes showed nearly the same K(m) values as PvSBE2 but had lower V(max) values and heat stabilities than PvSBE2. These results suggest that the N-terminal region of the kidney bean SBE is essential for maximum enzyme activity and thermostability.     

Expression level of a chimeric kinase governs entry into sporulation in Bacillus subtilis

Upon starvation, Bacillus subtilis cells switch from growth to sporulation. It is believed that the N-terminal sensor domain of the cytoplasmic histidine kinase KinA is responsible for detection of the sporulation-specific signal(s) that appears to be produced only under starvation conditions. Following the sensing of the signal, KinA triggers autophosphorylation of the catalytic histidine residue in the C-terminal domain to transmit the phosphate moiety, via phosphorelay, to the master regulator for sporulation, Spo0A. However, there is no direct evidence to support the function of the sensor domain, because the specific signal(s) has never been found. To investigate the role of the N-terminal sensor domain, we replaced the endogenous three-PAS repeat in the N-terminal domain of KinA with a two-PAS repeat derived from Escherichia coli and examined the function of the resulting chimeric protein. Despite the introduction of a foreign domain, we found that the resulting chimeric protein, in a concentration-dependent manner, triggered sporulation by activating Spo0A through phosphorelay, irrespective of nutrient availability. Further, by using chemical cross-linking, we showed that the chimeric protein exists predominantly as a tetramer, mediated by the N-terminal domain, as was found for KinA. These results suggest that tetramer formation mediated by the N-terminal domain, regardless of the origin of the protein, is important and sufficient for the kinase activity catalyzed by the C-terminal domain. Taken together with our previous observations, we propose that the primary role of the N-terminal domain of KinA is to form a functional tetramer, but not for sensing an unknown signal.     

Functional analysis of chimeras derived from the Sinorhizobium meliloti and Mesorhizobium loti nodC genes identifies regions controlling chitin oligosaccharide chain length

The rhizobial nodulation gene nodC encodes an N-acetylglucosaminyltransferase that is responsible for the synthesis of chitin oligosaccharides. These oligosaccharides are precursors for the synthesis of the lipo-chitin oligosaccharides that induce cell division and differentiation during the development of nitrogen-fixing root nodules in leguminous plants. The NodC proteins of Mesorhizobium loti and Sinorizobium meliloti yield chitinpentaose and chitintetraose as their main products, respectively. In order to localize regions in these enzymes that are responsible for this difference in product chain length, a set of six chimeric enzymes, comprising different combinations of regions of the NodC proteins from these two bacteria, was expressed in Escherichia coli. The oligosaccharides produced were analyzed using thin-layer chromatography. The major conclusion from this work is that a central 91-amino acid segment does not play any obvious role in determining the difference in the chain length of the major product. Furthermore, the characteristically predominant synthesis of chitintetraose by S. meliloti NodC is mainly dependent on a C-terminal region of maximally 164 amino acids; exchange of only this C-terminal region is sufficient to completely convert the M. loti chitinpentaose synthase into an S. meliloti-like chitintetraose synthase. The N-terminal region of 170 amino acids also plays a role in restricting the length of the major product to a tetramer. However, the role of the C-terminal region is clearly dominant, since exchanging the N-terminal region has no effect on the relative amounts of chitintetraose and -pentaose produced when the C-terminal region of S. meliloti NodC is present. The length of a predicted beta-strand around residue 300 in the C-terminal region of various NodC proteins is the only structural element that seems to be related to the length of the chitin oligosaccharides produced by these enzymes; the higher the amount of chitintetraose relative to chitinpentaose, the shorter the predicted beta-strand. This element may therefore be important for the effect of the C-terminal 164 amino acids on chitin oligosaccharide chain length.     

Influence of C Terminus on Monoamine Oxidase A and B Catalytic Activity

Abstract: Monoamine oxidase (MAO) A and B play important roles in the metabolism of neurotransmitters and dietary amines. The domains important for enzyme specificities were studied by construction of chimeric MAOA/B enzymes. Exchange of the N-terminal 45 amino acids of MAOA with the N-terminal 36 residues of MAOB (chimeric enzymes B₃₆A and A₄₅B) resulted in the same substrate and inhibitor sensitivities as the wild-type MAOA or B. Thus, the N terminus may not be responsible for MAOA or B enzyme specificities. When MAOB C-terminal residues 393–520 were replaced with MAOA C-terminal residues 402–527 (chimeric B₃₉₃A) catalytic activity was not detectable. Chimeric B₃₉₃A consists of eight residues with different charges, three less proline residues (458, 476, and 490), and one additional proline at 518 compared with wild-type MAOB. These differences may have induced conformational changes and affected MAOB catalytic activity. Thus, the C terminus of MAOB is critical for maintaining MAOB in an active form. It is interesting that when the C terminus of MAOA was switched with MAOB (chimeric A₄₀₂B), little effect was observed on MAOA catalytic activity. This new information is valuable for further studies of the structure and function relationship of this important enzyme.     

Chimeric pneumococcal cell wall lytic enzymes reveal important physiological and evolutionary traits.

Two novel chimeric pneumococcal cell wall lytic enzymes, named LC7 and CL7, have been constructed by in vitro recombination of the lytA gene encoding the major autolysin (LYTA amidase) of Streptococcus pneumoniae, a choline-dependent enzyme, and the cpl7 gene encoding the CPL7 lysozyme of phage Cp-7, a choline-independent enzyme. In remarkable contrast with previous chimeric constructions, we fused here two genes that lack nucleotide homology. The CL7 enzyme, which contains the N-terminal domain of CPL7 and C-terminal domain of LYTA, exhibited a choline-dependent lysozyme activity. This experimental rearrangement of domains might mimic the process that have generated the choline-dependent CPL1 lysozyme of phage Cp-1 during evolution, providing additional support to the modular theory of protein evolution. The LC7 enzyme, built up by fusion of the N-terminal domain of LYTA and the C-terminal domain of CPL7, exhibited an amidase activity capable of degrading ethanolamine-containing cell walls. The chimeric amidase behaved as an autolytic enzyme when it was cloned and expressed in S. pneumoniae. The chimeric enzymes provided new insights on the mechanisms involved in regulation of the host pneumococcal autolysins and on the participation of these enzymes in the process of cell separation. Furthermore, our experimental approach confirmed the basic role of the C-terminal domains in substrate recognition and revealed the influence of these domains on the optimal pH for catalytic activity.     

Improving enzyme characteristics by gene shuffling; application to β-glucosidase

The genes of family 3 β-glucosidase enzymes consist of five distinct regions; the N-terminal residues, an N-terminal catalytic domain, a nonhomologous region, a C-terminal domain of unknown function and the C-terminal residues. The β-glucosidase genes derived from Cellvibrio gilvus (CG) and Agrobacterium tumefaciens (AT) have been subjected to gene deletion, truncation and shuffling. The folding information was found to be distributed unevenly across the different regions based on the gene manipulation results. Chimeric enzymes with improved enzyme characteristics were obtained only by gene shuffling at the C-terminal domain.     

Role of the structural domains in the functional properties of pancreatic lipase-related protein 2

Although structurally similar, classic pancreatic lipase (PL) and pancreatic lipase-related protein (PLRP)2, expressed in the pancreas of several species, differ in substrate specificity, sensitivity to bile salts and colipase dependence. In order to investigate the role of the two domains of PLRP2 in the function of the protein, two chimeric proteins were designed by swapping the N and C structural domains between the horse PL (Nc and Cc domains) and the horse PLRP2 (N2 and C2 domains). NcC2 and N2Cc proteins were expressed in insect cells, purified by one-step chromatography, and characterized. NcC2 displays the same specific activity as PL, whereas N2Cc has the same as that PLRP2. In contrast to N2Cc, NcC2 is highly sensitive to interfacial denaturation. The lipolytic activity of both chimeric proteins is inhibited by bile salts and is not restored by colipase. Only N2Cc is found to be a strong inhibitor of PL activity, due to competition for colipase binding. Active site-directed inhibition experiments demonstrate that activation of N2Cc occurs in the presence of bile salt and does not require colipase, as does PLRP2. The inability of PLRP2 to form a high-affinity complex with colipase is only due to the C-terminal domain. Indeed, the N-terminal domain can interact with the colipase. PLRP2 properties such as substrate selectivity, specific activity, bile salt-dependent activation and interfacial stability depend on the nature of the N-terminal domain.     

A comparative approach to structure-function studies of mammalian aromatases

To date, structure–function studies of aromatase cytochrome P450 (P450arom) have been advanced by point mutation analyses utilizing almost exclusively the human enzyme, in conjunction with computer-generated models of the three-dimensional form of the enzyme based on prokaryotic cytochromes P450. Recent studies have identified duplicated isozymes of porcine P450arom, the gonadal and placental forms of which appear to differ substantially in substrate utilization and inhibitor sensitivity. We present a comparative approach to define regions of P450arom responsible for specific functional characteristics using complimentary DNAs encoding the porcine isozymes. Constructs encoding the native and chimeric porcine and human P450arom enzymes were transiently expressed and activity was assessed using the tritiated water assay. Sensitivity to inhibition by the imidazole etomidate was investigated, and P450arom expression was assessed by immunoblot analysis. All constructs yielded active P450arom, suggesting that exchanging entire structural elements does not preclude catalytic function. The activity of the gonadal isozyme was shown to be inhibited by etomidate at concentrations 185 and 300-fold lower than those required to induce a similar inhibition of the placental and human enzymes, respectively. In contrast, there was only a two-fold difference in the sensitivity of the gonadal and placental isozymes to inhibition by CGS16949A. Analysis of chimeric constructs indicated that the sensitivity to etomidate was associated with residues in the B, B′ and C helices of the gonadal P450arom encompassing only one of six putative substrate recognition sites. Additionally, sensitivity to etomidate was not correlated with enzyme activity among the chimeric enzymes. Therefore, it appears that residues of the porcine gonadal P450arom that are responsible for etomidate binding may be distinct from those involved in substrate recognition and metabolism. These data support the notion that a comparative approach employing the use of chimeric enzymes provides a useful tool in directing point mutational analysis to determine residues important in inhibitor and perhaps substrate recognition of P450 enzymes such as P450arom. These studies are currently in progress.     

Studies into factors contributing to substrate specificity of membrane-bound 3-ketoacyl-CoA synthases.

We are interested in constructing a model for the substrate-binding site of fatty acid elongase-1 3-ketoacyl CoA synthase (FAE1 KCS), the enzyme responsible for production of very long chain fatty acids of plant seed oils. Arabidopsis thaliana and Brassica napus FAE1 KCS enzymes are highly homologous but the seed oil content of these plants suggests that their substrate specificities differ with respect to acyl chain length. We used in vivo and in vitro assays of Saccharomyces cerevisiae-expressed FAE1 KCSs to demonstrate that the B. napus FAE1 KCS enzyme favors longer chain acyl substrates than the A. thaliana enzyme. Domains/residues responsible for substrate specificity were investigated by determining catalytic activity and substrate specificity of chimeric enzymes of A. thaliana and B. napus FAE1 KCS. The N-terminal region, excluding the transmembrane domain, was shown to be involved in substrate specificity. One chimeric enzyme that included A. thaliana sequence from the N terminus to residue 114 and B. napus sequence from residue 115 to the C terminus had substrate specificity similar to that of A. thaliana FAE1 KCS. However, a K92R substitution in this chimeric enzyme changed the specificity to that of the B. napus enzyme without loss of catalytic activity. Thus, this study was successful in identifying a domain involved in determining substrate specificity in FAE1 KCS and in engineering an enzyme with novel activity.     

Molecular basis of the functional specificities of phototropin 1 and 2

A blue-light photoreceptor in plants, phototropin, mediates phototropism, chloroplast relocation, stomatal opening, and leaf-flattening responses. Phototropin is divided into two functional moieties, the N-terminal photosensory and the C-terminal signaling moieties. Phototropin perceives light stimuli by the light, oxygen or voltage (LOV) domain in the N-terminus; the signal is then transduced intramolecularly to the C-terminal kinase domain. Two phototropins, phot1 and phot2, which have overlapping and distinct functions, exist in Arabidopsis thaliana. Phot1 mediates responses with higher sensitivity than phot2. Phot2 mediates specific responses, such as the chloroplast avoidance response and chloroplast dark positioning. To elucidate the molecular basis for the functional specificities of phot1 and phot2, we exchanged the N- and C-terminal moieties of phot1 and phot2, fused them to GFP and expressed them under the PHOT2 promoter in the phot1 phot2 mutant background. With respect to phototropism and other responses, the chimeric phototropin consisting of phot1 N-terminal and phot2 C-terminal moieties (P1n/2cG) was almost as sensitive as phot1; whereas the reverse combination (P2n/1cG) functioned with lower sensitivity. Hence, the N-terminal moiety mainly determined the sensitivity of the phototropins. Unexpectedly, both P1n/2cG and P2n/1cG mediated the chloroplast avoidance response, which is specific to phot2. Hence, chloroplast avoidance activity appeared to be suppressed specifically in the combination of N- and C-terminal moieties of phot1. Unlike the chloroplast avoidance response, chloroplast dark positioning was observed for P2G and P2n/1cG but not for P1G or P1n/2cG, suggesting that a specific structure in the N-terminal moiety of phot2 is required for this activity.     

Characterization and expression analysis of a serine acetyltransferase gene family involved in a key step of the sulfur assimilation pathway in Arabidopsis

Ser acetyltransferase (SATase; EC 2.3.1.30) catalyzes the formation of O-acetyl-Ser from L-Ser and acetyl-CoA, leading to synthesis of Cys. According to its position at the decisive junction of the pathways of sulfur assimilation and amino acid metabolism, SATases are subject to regulatory mechanisms to control the flux of Cys synthesis. In Arabidopsis (Arabidopsis thaliana) there are five genes encoding SATase-like proteins. Two isoforms, Serat3;1 and Serat3;2, were characterized with respect to their enzymatic properties, feedback inhibition by L-Cys, and subcellular localization. Functional identity of Serat3;1 and Serat3;2 was established by complementation of a SATase-deficient mutant of Escherichia coli. Cytosolic localization of Serat3;1 and Serat3;2 was confirmed by using fusion construct with the green fluorescent protein. Recombinant Serat3;1 was not inhibited by L-Cys, while Serat3;2 was a strongly feedback-inhibited isoform. Quantification of expression patterns indicated that Serat2;1 is the dominant form expressed in most tissues examined, followed by Serat1;1 and Serat2;2. Although Serat3;1 and Serat3;2 were expressed weakly in most tissues, Serat3;2 expression was significantly induced under sulfur deficiency and cadmium stress as well as during generative developmental stages, implying that Serat3;1 and Serat3;2 have specific roles when plants are subjected to distinct conditions. Transgenic Arabidopsis plants expressing the green fluorescent protein under the control of the five promoters indicated that, in all Serat genes, the expression was predominantly localized in the vascular system, notably in the phloem. These results demonstrate that Arabidopsis employs a complex array of compartment-specific SATase isoforms with distinct enzymatic properties and expression patterns to ensure the provision of Cys in response to developmental and environmental changes.