Structure–function analyses of two plant meso-diaminopimelate decarboxylase isoforms reveal that active-site gating provides stereochemical control

meso-Diaminopimelate decarboxylase catalyzes the decarboxylation of meso-diaminopimelate, the final reaction in the diaminopimelate l-lysine biosynthetic pathway. It is the only known pyridoxal-5-phosphate–dependent decarboxylase that catalyzes the removal of a carboxyl group from a d-stereocenter. Currently, only prokaryotic orthologs have been kinetically and structurally characterized. Here, using complementation and kinetic analyses of enzymes recombinantly expressed in Escherichia coli, we have functionally tested two putative eukaryotic meso-diaminopimelate decarboxylase isoforms from the plant species Arabidopsis thaliana. We confirm they are both functional meso-diaminopimelate decarboxylases, although with lower activities than those previously reported for bacterial orthologs. We also report in-depth X-ray crystallographic structural analyses of each isoform at 1.9 and 2.4 Å resolution. We have captured the enzyme structure of one isoform in an asymmetric configuration, with one ligand-bound monomer and the other in an apo-form. Analytical ultracentrifugation and small-angle X-ray scattering solution studies reveal that A. thaliana meso-diaminopimelate decarboxylase adopts a homodimeric assembly. On the basis of our structural analyses, we suggest a mechanism whereby molecular interactions within the active site transduce conformational changes to the active-site loop. These conformational differences are likely to influence catalytic activity in a way that could allow for d-stereocenter selectivity of the substrate meso-diaminopimelate to facilitate the synthesis of l-lysine. In summary, the A. thaliana gene loci At3g14390 and At5g11880 encode functional. meso-diaminopimelate decarboxylase enzymes whose structures provide clues to the stereochemical control of the decarboxylation reaction catalyzed by these eukaryotic proteins

meso-diaminopimelate decarboxylase adopts a homodimeric assembly. On the basis of our structural analyses, we suggest a mechanism whereby molecular interactions within the active site transduce conformational changes to the active-site loop. These conformational differences are likely to influence catalytic activity in a way that could allow for D-stereocenter selectivity of the substrate meso-diaminopimelate to facilitate the synthesis of L-lysine. In summary, the A. thaliana gene loci At3g14390 and At5g11880 encode functional. meso-diaminopimelate decarboxylase enzymes whose structures provide clues to the stereochemical control of the decarboxylation reaction catalyzed by these eukaryotic proteins Stereochemical control of catalysis is a hallmark feature of enzymes. By restricting substrate accessibility and guiding substrate-binding orientation, enzymes precisely dictate substrate specificity and control stereochemistry. For example, in the bacterial type I polyketide synthases, the stereochemical outcome of the native substrate can be predicted simply by the presence of particular sequence motifs within the catalytic domain (1). Similarly, mutation studies have shown that the formation of different stereoisomers by heteroyohimbine synthase appears to be determined solely by a single histidine residue positioned in an active-site loop (2).
In organisms that use the diaminopimelate pathways for the synthesis of lysine, the final step is the decarboxylation of meso-diaminopimelate to produce L-lysine and carbon dioxide (3,4). This reaction is catalyzed by the enzyme meso-diaminopimelate decarboxylase (DAPDC). 5 The substrate of the reaction, meso-diaminopimelate, features two carboxylate groups The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S3 and supporting Refs. 1 located at two different stereocenters: one is an L-stereocenter, and the other is a D-stereocenter. Decarboxylation of mesodiaminopimelate specifically at the D-stereocenter results in the production of L-lysine (Fig. 1). The specificity toward the D-stereocenter of the substrate is a feature unique to DAPDC, as all other pyridoxal-5-phosphate (PLP)-dependent decarboxylases characterized thus far act on L-stereocenters (3). Crystal structures of bacterial DAPDC enzymes reveal that these proteins have an optimal substratebinding pocket to accommodate the substrate, meso-diaminopimelate, and present the D-stereocenter for decarboxylation (5)(6)(7)(8)(9). The mechanism by which this stereospecificity is achieved in eukaryotic orthologs remains uncertain.
DAPDC is a member of the alanine racemase family of foldtype III PLP-dependent enzymes. This family also includes the ornithine and arginine decarboxylases (10). Within the active site of these enzymes, the aldehyde moiety of the PLP cofactor is covalently bound to a conserved lysine residue forming a Schiff base (also referred to as an internal aldimine). Upon binding of the substrate, the lysine residue is exchanged for the amino group of the substrate forming an external aldimine. In the reaction catalyzed by DAPDC, the bond between the ␣-carbon of the external aldimine and the carboxylate group is broken. PLP facilitates the reaction by withdrawing electron density from the substrate through a series of double bonds conjugated to the cationic pyridinium ring, stabilizing the negative charge that develops on the substrate during the reaction (10).
DAPDC is an attractive and putative target for the rational design of herbicides given the fact that the enzyme catalyzes a reaction that is essential for the anabolism of one of the 20 ubiquitous proteogenic amino acids. In addition, it is plausible that herbicides that specifically inhibit DAPDC will not be toxic to humans and animals in general because the DAPDC and lysine anabolic pathways are absent in the animal kingdom.
Our current understanding of the structure and function of DAPDC is largely based on work conducted on archaeal and bacterial orthologs from Methanococcus jannaschii (9), Mycobacterium tuberculosis (7,8), Helicobacter pylori (5), and Corynebacterium glutamicum (6). Previous findings suggest that DAPDC is a dimer in solution, although a tetrameric assembly has been observed in the asymmetric unit of crystal structures of M. tuberculosis, and small-angle X-ray scattering data suggest a weak dimer-tetramer equilibrium (8). More recently, solution studies using analytical ultracentrifugation and small-angle X-ray scattering coupled with enzyme kinetics have concluded that DAPDC enzymes from a range of bacterial species form dimers in solution (11). It appears that dimerization is required for catalysis due to the contribution of residues to the active site from both subunits. Consequently, disrupting self-association may provide a means to inhibit these enzymes. Thus far, the only studies that have been reported for a eukary-otic DAPDC are of the two Arabidopsis thaliana orthologs that were initially identified and verified through a functional complementation assay (12). Preliminary structural studies have been carried out on these orthologs, but the quaternary association has not yet been investigated (13).
Here, we report the X-ray crystallographic structures of two DAPDC isoforms from the plant species A. thaliana, the first in-depth structural characterization of eukaryotic orthologs. Functional complementation studies and kinetic analyses reveal that these isoforms are both authentic DAPDC enzymes, whereas analytical ultracentrifugation and small-angle X-ray scattering solution studies reveal that the protein adopts a homodimeric assembly. We successfully solved the structure of one of the A. thaliana isoforms in an asymmetric arrangement, with one active-site loop in the open position and the other closed. These novel observations provide new insight into the catalytic mechanism, giving clues as to how this enzyme exerts stereochemical control of the decarboxylation reaction.

Complementation analysis of an Escherichia coli lysine auxotrophic mutant demonstrates that At-DAPDC1 and At-DAPDC2 can rescue endogenous lysine biosynthesis
To assess whether the identified genes annotated by the locus tags At3g14390 and At5g11880 are functional DAPDC enzymes, complementation analyses were conducted. The E. coli CGSC strain 10193 harbors a mutation in the gene that encodes for DAPDC (lysA). This mutation results in a lysine auxotrophic phenotype where the mutant is able to grow only in the presence of exogenous lysine. As such, we can test the functionality of At-DAPDC enzymes by whether or not they can functionally complement the lysA mutant. It should be noted that previous work by our group shows that both At-DAPDC isoforms were able to complement another lysA mutant in a study that surveyed the lysine anabolic genes from A. thaliana (12). However, the previous analysis used fulllength cDNAs for both proteins, although the analysis reported here was completed without the sequences that encode the predicted transit peptide that denotes localization to the chloroplast (13).
To ascertain whether the genes of interest are able to complement the E. coli lysA mutation, mutants were transformed with an empty vector (pBAD33) or with pBAD33 expressing At-DAPDC1 or At-DAPDC2. The plasmid pBAD33 possesses a promoter that is inducible in the presence of arabinose and repressible in the presence of glucose (14). Surprisingly, both At-DAPDC enzymes were only able to rescue the mutant when their expression was repressed by glucose (Fig. 2). It is believed that under this condition there is still enough basal expression to allow for the synthesis of adequate quantities of lysine for protein synthesis to facilitate growth.
Conversely, in cells in which the expression of At-DAPDC is enhanced (in the presence of arabinose), no growth was seen (Fig. 2). Rather than this being due to a lack of lysine production to facilitate protein synthesis, we propose that this result is due to the depletion of meso-diaminopimelate, the substrate of DAPDC. In Gram-negative bacteria, such as E. coli, meso-di-

Stereochemical control of meso-diaminopimelate decarboxylase
aminopimelate is used as the cross-linking amino acid in the peptide stem of peptidoglycan, a process facilitated by mesodiaminopimelate ligase (MurE) (15). Therefore, the induced expression of At-DAPDC enzymes in lysine-free media leads to cell death because the native MurE cannot compete with the overexpressed DAPDC, and thus, although the organism is able to synthesize L-lysine, it has impaired peptidoglycan synthesis. Nonetheless, these results indicate that the two genes encode authentic DAPDC enzymes and are able to function in E. coli without the predicted chloroplast transit peptide sequences.

Recombinant At-DAPDC1 and At-DAPDC2 are both enzymatically active
Although both enzymes are active, their rate of turnover is surprisingly low. The catalytic activities of recombinant At-DAPDC1 and At-DAPDC2 were examined using the DAPDC-SDH-coupled assay (16). We first determined that the rate of catalysis was linear across an enzyme concentration range of 0 -800 nM (Fig. 3A). We therefore chose to collect the initial rate data at an enzyme concentration of 550 nM. The initial rate data for both enzymes fitted well to the Michaelis-Menten equation (Fig. 3, B and C). The purified At-DAPDC enzymes are both catalytically active with apparent K m values of 5.1 and 4.8 mM, V max values of 85 and 229 nmol min Ϫ1 , and catalytic rate constants (k cat ) of 0.005 and 0.014 s Ϫ1 , for At-DAPDC1 and At-DAPDC2, respectively ( Table 1). The K m values are consistent with other previously characterized DAPDC enzymes (0.68 mM (Bacillus anthracis), 0.97 mM (E. coli), 1.6 mM (M. tuberculosis), and 1.9 mM (Vibrio cholerae) (11)). In contrast, the catalytic rate constants for the A. thaliana DAPDC orthologs are much lower than their bacterial counterparts (58 s Ϫ1 (B. anthracis), 55 s Ϫ1 (E. coli), 28 s Ϫ1 (M. tuberculosis), and 22 s Ϫ1 (V. cholerae) (11)). This may be a reflection of the lower temperatures used here (30°C, compared with 37°C). Peverelli et al. (11) found that dimerization of DAPDC was required for catalysis, given that residues are required from each subunit to form the active site. Therefore, we sought to investigate the quaternary structure of the recombinant At-DAPDC enzymes in solution.

At-DAPDC enzymes exist in a monomer-dimer equilibrium
Although monomeric (5,17), dimeric (6,7,9), and tetrameric (8) assemblies have been observed in crystal structures of prokaryotic orthologs of DAPDC, the quaternary structure has not yet been investigated for any plant orthologs. To address this, the quaternary structure of the two At-DAPDC enzymes were characterized by analytical ultracentrifugation. Sedimentation velocity data were fitted to continuous size distribution models yielding c(s) distributions (Fig. 4). For both proteins (at concentrations between 5 and 20 M), two species with sedimentation coefficients of around 3.5 and 5 S are evident in solution. When fitted to a continuous mass distribution model, the estimated masses of each species are ϳ50 and ϳ100 kDa (based on frictional ratios of 1.51 for At-DAPDC1 and 1.43 for At-DAPDC2, calculated during the c(s) analyses); therefore, these peaks correspond to a monomer and dimer (given the theoretical monomeric molecular masses of 53,807 Da for At-DAPDC1 and 54,240 Da for At-DAPDC2).
The plots in Fig. 4, A and B, have been normalized to the height of the peak corresponding to the dimer, revealing that with increasing protein concentrations the proportion of protein found in the dimeric form increases. This indicates that the protein is in a concentration-dependent monomer-dimer selfassociation. The sedimentation velocity data were further analyzed using the advanced van-Holde/Weischet method. This method is model-independent and directly assesses the shape of the sedimenting boundary to provide an estimate of the proportion of each species in solution (Fig. 4, C and D). The half-parabolic shape of the van-Holde/Weischet plots are characteristic of mass action and support the conclusion that this protein exists in equilibrium between its monomeric and dimeric forms.
The predominant species at these concentrations is the dimer, suggesting that the dimer dissociation constants (K D 2-1 ) for these enzymes are below the concentrations examined here. Sedimentation velocity data at the lowest concentration (5 M) were fitted to a monomer-dimer model. These analyses estimate the K D 2-1 values for both At-DAPDC enzymes to be less than 1 M ( Table 2). The enzyme assays were conducted at a concentration of 550 nM; thus, there is likely a mixture of monomers and dimer in solution at this concentration. However, because the rate versus enzyme concentration (Fig. 3A) is linear from 200 to 800 nM, which spans the estimated K D 2-1 for both enzymes (ϳ650 nM), either both the monomer and dimer are similarly active or the presence of substrates lowers the dissociation constant, resulting in a predominantly dimeric species at the concentration used to collect the initial rate data. Given the results of Peverelli et al. (11), where mutants that disrupt the dimer interface of V. cholerae DAPDC have attenuated activity, we suggest that the latter is a more reasonable explanation of the data.

Stereochemical control of meso-diaminopimelate decarboxylase Structural characterization of At-DAPDC enzymes
At-DAPDC1 and At-DAPDC2 were crystallized as reported previously (13). All relevant data-collection and processing parameters are given in Table 3. Globally, the structures of the two At-DAPDC enzymes show considerable similarity to each other and to previously characterized bacterial DAPDC enzymes (Fig. S1) (5)(6)(7)(8)(9). For both structures, the asymmetric unit is a dimer (Fig. 5A), with each monomer composed of two domains: an alanine racemase-like fold and an (␣/␤) 8 barrel domain (Fig. 5B). The alanine racemase-like domain (shown in yellow, Fig. 5B) is composed of interlocking ␤-strands from both the N and C termini (residues 1-50 and 301-440, amino acid numbering corresponding to that of At-DAPDC1). The (␣/␤) 8 barrel domain (residues 51-300, shown in red in Fig. 5B) contains the active site at its core and features an extended active-site loop between the fifth ␤-sheet (␤5) and sixth ␣-helix (␣6) of the barrel (residues 165-194). The monomers associate in a head-to-tail arrangement to form a dimer, with the two active sites located at the dimer interface.
To validate that the dimeric assembly observed in the crystal structures is an accurate representation of the protein structures in solution, small-angle X-ray scattering experiments were conducted. The structural parameters obtained from these analyses are summarized in Table 4. The maximum dimensions of the particles (105 and 113 Å for At-DAPDC1 and At-DAPDC2) and estimated molecular masses (95 and 94 kDa) are most consistent with that of a dimer (given the theoretical dimeric molecular masses of 107 and 108 kDa for At-DAPDC1 and At-DAPDC2). Furthermore, theoretical scattering curves were generated based on the crystal structures of At-DAPDC for both a monomer and a dimer. The experimental scattering obtained for samples of At-DAPDC in solution are a much closer fit to the scattering calculated for a dimer than for a monomer ( 2 ϭ 0.41 and 0.71, compared with 3.26 and 8.09), suggesting that the dimeric quaternary arrangements observed in the crystal structures are correct (Fig. 5, C and D).

Active sites of At-DAPDC enzymes are similar to prokaryotic orthologs of DAPDC
In the structure of At-DAPDC1, electron density consistent with the structure of a PLP molecule can be observed in both active sites (Fig. 6A). The density is continuous with a free lysine molecule, probably resulting from the supplementation of free lysine in the crystallization conditions. This external aldimine is likely to closely reflect the catalytic intermediate formed between meso-diaminopimelate and PLP during the decarboxylation reaction (5). PLP is positioned near a conserved lysine residue (Lys-77) that likely forms the internal aldimine with PLP when no substrate is present. In contrast, in the structure of At-DAPDC2, electron density attributable to PLP and a molecule of free lysine is visible in one active site only (Fig. 6B). No such electron density is observed in the other active site.
The same interactions are present in the PLP-bound active sites of At-DAPDC1 and At-DAPDC2 (see LigPlot in Fig. S2). For the following analyses, residue numbering according to At-DAPDC2 is used, to allow for comparisons to be made between the ligand-bound and apo monomers. PLP forms hydrogen bonds directly with Ser 223 , Arg 299 , Tyr 396 , Gly 298 , Gly 260 , and Glu 296 and via water molecules with Leu 261 , Arg 299 , Ser 300 , Arg 169 , and Cys 367* (the asterisk denotes that this residue is contributed from the opposing monomer) (Fig. 6C). His 220 is positioned parallel to PLP, ϳ3.5 Å above the aromatic ring of the cofactor, appropriately positioned to facilitate a -stacking interaction (Fig. 6C). The supplemented lysine molecule forms hydrogen bonds with Arg 299 , Arg 335 , Tyr 339 , and Glu 368* (Fig. 6D).
The active-site architecture described here for the A. thaliana orthologs of DAPDC is similar to that observed in previously characterized bacterial orthologs (5-9), suggesting that eukaryotic DAPDC enzymes employ the same methods to direct stereochemical control. This includes anchoring of the L-stereocenter of meso-diaminopimelate presenting the D-stereocenter for catalysis, facilitated by the conserved residues Arg 299 , Arg 335 , Tyr 339 , and Glu 368* (see Fig. S3 for sequence alignment). This also includes the contribution of residues from both subunits toward the active site, suggesting that, as in bacterial orthologs (11), the dimer is the catalytically active form.  (19). A, rate of reaction relative to enzyme concentration shows the expected linear response between 200 and 800 nM for both At-DAPDC1 and At-DAPDC2. Given this plot, the concentration of enzyme used for the initial rate experiments (B and C) was 550 nM, because this was in the linear range. B and C, data (circles) represent the activity of At-DAPDC as a function of substrate (meso-diaminopimelate) concentration, fitted to the Michaelis-Menten equation (solid line). The R 2 for the fit to At-DAPDC1 initial rate data was 97%, and the fit to the At-DAPDC2 initial rate data was 98%. The green dashed lines indicate K m values, and the red dashed lines indicate V max .

Stereochemical control of meso-diaminopimelate decarboxylase Conformational differences between the apo and PLP-bound monomers of At-DAPDC2 provide insight into stereochemical control
The PLP-bound monomer of At-DAPDC2 shows distinct conformational differences to the apo monomer (without PLP and lysine). When superimposed, the r.m.s.d. value is 0.456, compared with 0.128 when aligning the two monomers of At-DAPDC1 (both PLP-bound). This is the first time an asymmetric dimer has been observed in a structure of DAPDC and may give further insight into the stereochemical control of the decarboxylation reaction mediated by DAPDC.
One such conformational difference between the apo and PLP-bound forms is the position of the active-site loop. The active-site loop is the region between Arg 169 and Lys 197 (numbering according to At-DAPDC2) that extends past the entrance to the active site (Fig. 7). The electron density of this region is often poorly defined in structures of DAPDC (5) (Fig. 7B). Conversely, in the apo monomer (blue loop), Phe 191 is shifted by 19 Å and is buried in the interface between the ␣7 and ␣8 helices of the (␣/␤) 8 barrel (Fig. 7B). Interestingly, judging by the position of this phenylalanine residue, all structures of DAPDC orthologs currently deposited in the PDB appear to have loops in the closed conformation.
In the active site of the apo monomer of At-DAPDC2, several differences can be observed compared with the PLP-bound active site. Most of these involve reorientations of side chains that are no longer bound by hydrogen bonds to either PLP or lysine. However, some residues undergo more significant conformational change, including Arg 169 , His 218 , and His 220  Stereochemical control of meso-diaminopimelate decarboxylase (Fig. 7C). In the apo-form, Arg 169 exists in a twisted conformation, forming a hydrogen bond with Asp 145 . His 218 and His 220 are located in close proximity facing toward Arg 169 . In the PLPbound form, His 220 is reoriented to form a -stacking interaction with PLP, whereas His 218 fills the space vacated by His 220 . Arg 169 adopts an extended conformation running parallel to His 220 and His 218 that is stabilized through a hydrogen bond with the backbone carbonyl of His 220 . This constitutes a 2.5 Å movement of the C␣ atom of Arg 169 . Arg 169 is the first residue of the active-site loop, and thus it is conceivable that this conformational rearrangement may be associated with repositioning of the active-site loop into the closed configuration. Closure of the active-site loop has been shown to enhance the rate of catalysis in DAPDC from H. pylori (5). As well as stabilizing the catalytic intermediate, closure of the active-site loop may be important in excluding other molecules from entering the active site, preventing side reactions, or excluding water from the active site during the reaction; in so doing, this may enhance catalysis as production of the nonpolar product, CO 2 , would be expected to occur faster in a nonpolar environment than surrounded by water.
When meso-diaminopimelate is modeled in the active site of At-DAPDC2 in the same configuration as L-lysine (shown in yellow in Fig. 7D), the D-stereocenter carboxyl can be suitably accommodated within the active site, whereas steric hindrance (with Cys 367 and Glu 368 ) would prevent binding in the reverse orientation. In this configuration, we can also see a possible hydrogen bond between the substrate and His 220 (within 3 Å of each other). It is possible, therefore, that approach of the substrate is able to stabilize the position of His 220 over the pyridoxal ring and subsequently encourage the closure of the active-site loop, but this would occur only when the D-stereo-center is oriented for decarboxylation, not the L-stereocenter. If active-site loop closure is required to enhance catalysis in At-DAPDC, as it does in H. pylori (5), this may provide an extra means by which DAPDC exerts stereochemical control over the decarboxylation of meso-diaminopimelate.

Conclusion
We have established that the A. thaliana gene loci, At3g14390 and At5g11880, encode functional DAPDC enzymes. The recombinantly expressed eukaryotic enzymes are catalytically active but exhibit much lower activity than previously characterized bacterial orthologs. To determine whether this is a feature of plant DAPDC enzymes will require further characterization of eukaryotic DAPDC orthologs. Solution studies reveal that both At-DAPDC enzymes exist in a monomer-dimer equilibrium in solution; however, the dimer is likely the catalytically active form due to the requirement of residues from each subunit to form the active site. This is in line with previously characterized bacterial orthologs of DAPDC.
The crystal structures of the At-DAPDC enzymes reveal that the eukaryotic orthologs characterized here for the first time are very similar structurally, including the active-site architecture, to the prokaryotic forms of DAPDC previously characterized. The fortuitous capture of At-DAPDC2 in an asymmetric dimer, with one monomer in the apo-form and the other in a ligand-bound form, provides valuable insight into the conformational transition that occurs upon PLP binding. Furthermore, this is the first DAPDC structure that has the active-site loop in an open conformation.
Based on our structural data, we propose that the following conformational changes are important for catalysis and stereochemical selection of the substrate. First, binding of PLP initiates rearrangements within the active site, specifically the reorientation of Arg 169 , His 218 , and His 220 . These conformational changes are propagated to the active-site loop, through Arg 169 which is a part of this loop; the position of the active-site loop is known to influence the catalytic activity of DAPDC enzymes (5). Second, we have shown that the substrate, meso-diaminopimelate, is positioned through interactions with conserved active-site residues to form an additional hydrogen bond to His 220 . This further stabilizes the closed active-site loop conformation, mediated through Arg 169 , and can only occur when the D-stereocenter is oriented for decarboxylation. These findings provide further clues into how DAPDC targets the D-stereocenter of the substrate, meso-diaminopimelate, for decarboxylation during catalysis-knowledge that is necessary to assist in the design of effective inhibitors that may act as herbicides or antibiotics.

Cloning of the cDNAs annotated by the locus tags At3g14390 and At5g11880
The cDNAs At3g14390 (At-DAPDC1) and At5g11880 (At-DAPDC2), encoding the two isoforms of DAPDC from A. thaliana, were cloned as described previously (13). The forward and reverse primers used to amplify the At3g14390 cDNA were 5Ј-CCCCCGGATCCATGGCCGCCGTCGTTTCTCAAAA-CTCGTCC-3Ј and 5Ј-CCCCCGTCGACTCATAGACCTTC-

Stereochemical control of meso-diaminopimelate decarboxylase
AAAGAAACGCAAATGGTC-3Ј. The forward and reverse primers used to amplify the At5g11880 cDNA were 5Ј-CCCC-CGGATCCATGGCCGCCGTTTCCCAAAACTCCTACCA-AA-3Ј and 5Ј-CCCCCGTCGACTCATAGTCCTTCAAAGA-AACGTAAATGGTC-3Ј. The underlined nucleotides represent the restriction enzyme sites used in the cloning of the cDNAs, and the bold italicized nucleotides represent initiation and termination codons. The PCR protocols were designed to exclude the first 144 and 147 nucleotides from the At3g14390 and At5g11880 cDNAs, respectively, as these nucleotides encode a predicted transit peptide that allows localization to the chloroplast. The truncated cDNAs were cloned between the BamHI and SalI restriction sites of pET30A to produce the plasmids pET30a::At3g14390 and pET30a::At5g11880. The constructs have the following amino acid residues (MHHHHHHSSGLVPRGSGMKETAAAKFDRQHM-DSPDLGTDDDDKAMADI) at the N-terminal end of each recombinant enzyme, yielding polypeptides of 53,807 Da (At-DAPDC1) and 54,240 Da (At-DAPDC2).

Functional complementation
The plasmids used for functional complementation were produced by subcloning the XbaI and SalI fragment from pET30a::At3g14390 and pET30a::At5g11880 into pBAD33 (14) to produce the plasmids pBAD33::At3g14390 and pBAD33:: At5g11880. The proteins produced from the pBAD33 constructs are identical to the proteins produced from the pET30a constructs.

Protein expression and purification of At-DAPDC for structural characterization
The expression vectors pET30a::At3g14390 and pET30a:: At5g11880 were transformed into E. coli Rosetta (DE3) cells. The transformed cells were grown in LB medium (with 50 g ml Ϫ1 kanamycin and 34 g ml Ϫ1 chloramphenicol) at 37°C until the A 600 reached ϳ0.5. Expression at 26°C overnight was  Table 4 Small-angle X-ray scattering structural parameters.

Stereochemical control of meso-diaminopimelate decarboxylase
induced by addition of 0.5 mM isopropyl D-1-thiogalactopyranoside (IPTG). The cell pellet was resuspended in buffer (50 mM sodium phosphate, 300 mM sodium chloride, 10 mM imidazole, pH 8.0) and lysed by sonication using a cell disruptor (Vibra-Cell VC 750; Sonics). The lysate was clarified by centrifugation (12,000 ϫ g, 4°C, 10 min) and loaded onto a His-Trap FF Crude 5-ml column (GE Healthcare). The bound His-tagged protein was eluted with increasing imidazole concentrations up to 250 mM. Fractions containing protein were pooled and concentrated to 3 ml using a 30-kDa molecular mass cutoff spin concentrator (Vivaspin centrifugal concentrator, Sartorius). The concentrated solution was further purified on a HiLoad 10/30 Superdex 200 column (GE Healthcare) pre-equilibrated with buffer (20 mM Tris, 100 mM sodium chloride, pH 8.0). All purification steps were carried out at 4°C. Peak fractions were concentrated to 10 mg ml Ϫ1 using a 30-kDa molecular mass cutoff spin concentrator. Protein concentrations were determined spectrophotometrically using the extinction coefficient at 280 nm of 33,350 M Ϫ1 cm Ϫ1 , calculated from amino acid sequences using ExPASy ProtParam (19), and Beer's law (20).

Protein expression and purification of At-DAPDC and Sc-SDH for enzymatic assays
pET30a::At3g14390, pET30a::At5g11880 and pET16b::ScLys1 (expressing saccharopine dehydrogenase from Saccharomyces cerevisiae) were transformed into E. coli BL21-CodonPlus-RIPL strain (Agilent Technologies). Cells were grown in LB media containing 50 g ml Ϫ1 kanamycin and 34 g ml Ϫ1 chloramphenicol (At-DAPDC) or 50 g ml Ϫ1 ampicillin and 34 g ml Ϫ1 chloramphenicol (Sc-SDH) at 37°C, to an A 600 of 0.5. Protein expression was induced with 1.0 mM IPTG for 24 h at 18°C. The cells were lysed by sonication in a solution of 50 mM sodium phosphate, pH 8.0, and 300 mM sodium chloride. The extract was incubated with Talon Metal Affinity Resin (Clontech) for 30 min at 4°C. The resin was washed five times with sonication buffer containing 10 mM imidazole, pH 8.0, and the enzyme was eluted with buffer containing 250 mM imidazole. The pure protein was concentrated in an Amicon Ultra 10,000 molecular weight cutoff spin concentrator replacing the elution buffer with 100 mM Tris containing 1 mM DTT and 2 mM EDTA, pH 8.0. Protein concentration was determined by the Bradford assay (21).

DAPDC kinetic assay
A coupled assay, using Sc-SDH as the coupling enzyme, was used to measure the activity of At-DAPDC1 and At-DAPDC2 (16). L-Lysine produced from meso-diaminopimelate by DAPDC is used as the substrate of Sc-SDH along with ␣-ketoglutarate and NADH. The oxidation of NADH to NAD ϩ can be monitored spectrophotometrically as a decrease in the absorbance at 340 nm over time.
Titration assays varying the concentration of substrate, meso-diaminopimelate, were buffered with 100 mM Tris, pH 8.0, using a final assay concentration of 200 M NADH, 25 mM ␣-ketoglutarate, and an excess of Sc-SDH (40 g). With a final reaction volume of 500 l, meso-diaminopimelate was added to various final concentrations ranging from 0 to 15 mM. Reactions were initiated with the addition of either At-DAPDC1 or At-DAPDC2 to a final concentration of 550 nM. All coupled assays were performed at 30°C, and absorbance at 340 nm was measured as a function of time. Data were analyzed using a Michaelis-Menten kinetic model to determine the V max , K m , and k cat values.

Analytical ultracentrifugation
Sedimentation velocity experiments were performed in a Beckman Coulter XL-I analytical ultracentrifuge. Sedimentation was monitored using the absorbance optical system. At-DAPDC1 and At-DAPDC2 were analyzed at three protein concentrations: 5, 10, and 20 M in 20 mM Tris, 150 mM sodium chloride at pH 8.0. Sample solutions were loaded into 12-mm double-sector quartz cells and centrifuged at 50,000 rpm at 20°C. Intensity data were recorded over a radial position range of 6.0 to 7.25 cm, with a step size of 0.003 cm. Different wavelengths were monitored depending on the protein concentration; 281 nm was used for 20 M samples, and lower concentrations were analyzed at 236 nm.
The sedimentation data were fitted to a continuous sizedistribution model in SEDFIT (22) using the partial specific volume of At-DAPDC and the solvent viscosity and density computed in SEDNTERP (23). Two-dimensional spectrum analyses with 50 Monte Carlo iterations and enhanced van-Holde/Weischet analyses were conducted in UltraScan III (24 -26). The data were fitted to a reversible monomerdimer model with genetic algorithm optimization (27). 96 Monte Carlo iterations were performed in each genetic algorithm fit to determine the 95% confidence intervals for each parameter. During fitting, the monomer molecular mass was fixed at the molecular mass derived from the protein sequence (53.8 kDa for At-DAPDC1 and 54.2 kDa for At-DAPDC2).

X-ray crystallography
Purified At-DAPDC1 and At-DAPDC2 were crystallized and X-ray diffraction data collected as reported previously (13). The crystal structures were refined using REFMAC (28) with iterative rounds of model building performed in COOT (29). The structures were deposited in the PDB with accession numbers 6N2A and 6N2F.

Small-angle X-ray scattering
Data were collected at the SAXS/WAXS beamline at the Australian Synchrotron Facility using a 1M Pilatus detector. The two protein samples were injected at 10 mg ml Ϫ1 into a Superdex 200 PC 3.2/30 gel-filtration column (GE Healthcare) pre-incubated with buffer (20 mM Tris, 100 mM sodium chloride, pH 8.0), before being exposed to the X-ray beam through a 1.5-mm glass capillary at a continuous flow, with 2 s of X-ray exposure per image collected.
Data reduction was performed using CHROMIXS (30), whereas data analysis was performed using the ATSAS software suite (31). Guinier and pairwise distribution analyses were performed using PrimusQT (32), and CRYSOL (33) was used to fit Stereochemical control of meso-diaminopimelate decarboxylase