Structure and Function of Cyanobacterial DHDPS and DHDPR

Lysine biosynthesis in bacteria and plants commences with a condensation reaction catalysed by dihydrodipicolinate synthase (DHDPS) followed by a reduction reaction catalysed by dihydrodipicolinate reductase (DHDPR). Interestingly, both DHDPS and DHDPR exist as different oligomeric forms in bacteria and plants. DHDPS is primarily a homotetramer in all species, but the architecture of the tetramer differs across kingdoms. DHDPR also exists as a tetramer in bacteria, but has recently been reported to be dimeric in plants. This study aimed to characterise for the first time the structure and function of DHDPS and DHDPR from cyanobacteria, which is an evolutionary important phylum that evolved at the divergence point between bacteria and plants. We cloned, expressed and purified DHDPS and DHDPR from the cyanobacterium Anabaena variabilis. The recombinant enzymes were shown to be folded by circular dichroism spectroscopy, enzymatically active employing the quantitative DHDPS-DHDPR coupled assay, and form tetramers in solution using analytical ultracentrifugation. Crystal structures of DHDPS and DHDPR from A. variabilis were determined at 1.92 Å and 2.83 Å, respectively, and show that both enzymes adopt the canonical bacterial tetrameric architecture. These studies indicate that the quaternary structure of bacterial and plant DHDPS and DHDPR diverged after cyanobacteria evolved.

However, the structure of a plant DHDPR has not yet been determined, but a recent study employing small angle X-ray scattering suggests that the enzyme from Arabidopsis thaliana adopts a novel dimeric structure (Fig. 1C) 28 . Accordingly, there appears to be structural diversity between bacterial and plant orthologues of both DHDPS and DHDPR.
In this study, we aimed to determine the structure and function of the first DHDPS and DHDPR enzymes from a cyanobacterial species. Given that endosymbiotic theory suggests that the chloroplasts of plants were derived from the symbiosis of separate single bacterial cells 38 , we were interested in characterising the structure and function of DHDPS and DHDPR from the model cyanobacterial species, Anabaena variabilis (Av) 39 . Here, we present an in-depth characterisation of the structure and function of Av-DHDPS and Av-DHDPR in both solution and crystal states. We show that Av-DHDPS and Av-DHDPR both adopt the canonical bacterial structures, suggesting that the point of quaternary structural divergence between the bacterial and plants enzymes occurred after cyanobacteria evolved.

Results and Discussion
Purified recombinant Av-DHDPS and Av-DHDPR are active and folded. Av-DHDPS and Av-DHDPR were expressed in E. coli as His-tagged constructs and purified to > 98% homogeneity using immobilised metal affinity chromatography (IMAC) (Fig. 2). The specific activity of purified Av-DHDPS and Av-DHDPR were determined to be 8.81 and 66.7 U/mg, respectively (Table 1), which correlate well to previous studies of recombinant orthologs 10,40 . MS/MS sequencing following trypsin digestion demonstrates that both recombinant A. variabilis enzymes are comprised of the correct primary structure (Table 1). CD spectroscopy was subsequently employed to demonstrate that recombinant Av-DHDPS (Fig. 3, open symbols) and Av-DHDPR (Fig. 3, solid symbols) contain 45-51% α /β structure, which is consistent with previous studies of bacterial and plant orthologues 7,9,14,16,19,22,25,27,41 . Enzyme kinetic properties. Having determined that the recombinant cyanobacterial enzymes were homogenous, folded and enzymatically-active, we next set out to quantify their enzyme kinetic properties. Firstly, we characterised Av-DHDPS. Plots of initial rate as a function of varying pyruvate and ASA concentrations reveal typical Michaelis-Menten hyperbolic relationships (Fig. 4A). These data were globally fitted to yield a best fit to a bi-bi ping-pong mechanism without substrate inhibition (R 2 = 0.99), providing the kinetic parameters summarised in Table 2. The resulting kinetic parameters agree well with previous studies of bacterial orthologues 9,14,16,20,22,24 . To establish whether recombinant Av-DHDPS is sensitive to allosteric feedback inhibition by   Raw data for Av-DHDPS (○ ) and Av-DHDPR (• ) were fitted by nonlinear regression using the CDPro software and the CONTINLL algorithm ( -), resulting in 33% α -helix, 18% β -structure, 14% β -turn and 35% unordered structure for Av-DHDPS with a RMSD of 0.070, and 18% α -helix, 27% β -structure, 13% β -turn and 42% unordered structure for Av-DHDPR with a RMSD of 0.037.
(S)-lysine, which is the end product of the diaminopimelate pathway (Fig. 1A), enzyme assays were also performed with increasing (S)-lysine concentrations [1][2][3]18,19,29 . DHDPS from E. coli (Ec-DHDPS) and V. vinifera (Vv-DHDPS) were used as controls, given that previous studies show these orthologues are allosterically inhibited by (S)-lysine 18,29 . The dose-response curves for Av-DHDPS, Ec-DHDPS and Vv-DHDPS are shown in Fig. 4B with the nonlinear best fits to a four-parameter logistic function yielding an IC 50 LYS = 0.068 mM (R 2 = 0.99) for Av-DHDPS (Table 2), which is closer to the value obtained for Vv-DHDPS [IC 50 LYS = 0.030 mM (R 2 = 0.98)] than for Ec-DHDPS [IC 50 LYS = 0.210 mM (R 2 = 0.98)]. Interestingly, a recent study revealed that the amino acid at position 56 (E. coli numbering) determines whether DHDPS enzymes will be inhibited by (S)-lysine 19 . Moreover, a His or Glu at this position is a marker of allosteric inhibition, whereas DHDPS sequences that contain Lys or Arg at position 56 are insensitive to lysine-mediated allosteric inhibition. For Av-DHDPS, there is a Glu at position 58 (equivalent to position 56 in Ec-DHDPS), which is consistent with the recently established determinants of allostery for DHDPS enzymes 19 .
For Av-DHDPR, the enzyme kinetic parameters were determined employing E. coli DHDPS as the coupling enzyme using increasing DHDP and NADH concentrations (Fig. 5A). The nonlinear least squares global fit was obtained to a ternary complex model (R 2 = 0.98), yielding the kinetic values reported in Table 2. A comparison of NADH and NADPH showed that Av-DHDPR is inhibited by its substrate, DHDP, when NADPH is employed as the cofactor (Fig. 5B). Subsequent thermodynamic measurements using microscale thermophoresis 19,42 revealed that the cyanobacterial enzyme binds the substrate analogue 2,6-pyridinedicarboxylate (2,6-PDC) 34 only when NADP + , and not NAD + , is present in the titration (Fig. 5C). This suggests that Av-DHDPR is inhibited by DHDP in the presence of the oxidised phosphorylated cofactor, which is consistent with S. aureus DHDPR 11 .
Scientific RepoRts | 6:37111 | DOI: 10.1038/srep37111 diffracted to 2.83 Å (Fig. 7E,F). The diffraction data were subsequently used to determine the three-dimensional structure of Av-DHDPS and Av-DHDPR by molecular replacement. Consistent with the analytical ultracentrifugation (AUC) studies in solution, the crystal structure of Av-DHDPS reveals a homotetramer that resembles the 'head-to-head' dimer-of-dimers canonical to bacterial orthologues (Fig. 8A). Each monomer is comprised of 297 residues that folds to form a N-terminal (β /α ) 8 -barrel domain and a C-terminal domain consisting of 3 α -helices. Close inspection of the active site shows that it is comprised of the key conserved residues known to be important for catalysis, namely Lys164, Thr46, Tyr109, Tyr136  and Arg141, which are equivalent to Lys161, Thr44, Tyr107, Tyr133 and Arg138 in E. coli DHDPS (Fig. 8B) 17,18 . Likewise, inspection of the allosteric site confirms the presence of Glu at position 58 (His56 in Ec-DHDPS), but also reveals that Trp occupies position 55 (His53 in Ec-DHDPS), which is common in plants but not bacteria (Fig. 8C) [26][27][28][29] . The presence of a Trp at this positon is likely to explain the plant-like lysine IC 50 for Av-DHDPS (Fig. 4B, Table 2).
The Av-DHDPR crystal structure also reveals a homotetramer (Fig. 9A), which is consistent with the in-solution AUC studies, and agrees well with previously characterised bacterial DHDPR structures [31][32][33][34][35][36] . Each monomeric unit consists of 287 residues with an N-terminal nucleotide and a C-terminal tetramerisation domain connected via a hinge region. Although residue variation is observed at the nucleotide binding site, the physicochemical properties of the residues are still conserved (Fig. 9B) as observed for other bacterial DHDPR species [31][32][33][34][35][36] . By contrast, the substrate binding cleft of Av-DHDPR is predominantly conserved (Fig. 9C). However, Av-DHDPR has a unique prolonged solvent-exposed loop located between β -sheets B4 and B5, consisting of residues Val107 to Gly116 (Fig. 9D). Overall, the loop residues have a neutral, slightly hydrophobic nature, and the function of the loop is unclear. This prolonged loop is absent in all other published DHDPR structures suggesting this is a unique feature of cyanobacterial DHDPR 31-36 . Bioinformatics analysis. This study reveals that both Av-DHDPS and Av-DHDPR adopt the canonical bacterial tetrameric architecture. This was unexpected given the endosymbiosis theory 38 and the shared aminotransferase pathway found in both cyanobacteria and plants. Consequently, bioinformatics sequence analyses of DHDPS and DHDPR from several bacterial and plant species were performed to predict when the plant structures first evolved. For DHDPS, the dataset employed consisted of sequences from 150 bacteria, 85 cyanobacteria, 84 plants, 12 green algae and 2 red algae from the NCBI protein database (www.ncbi.nlm.nih.gov/protein, 2016). A representative subset of these sequences are aligned in Fig. 10A. It was noted that the motifs Arg43 to Asp45, Arg108 to Gln116 and Gly325 to Tyr/His327 (V. vinifera numbering) are conserved in plants and form an interacting network at the tetramerisation interface in plant structures [26][27][28][29] . These motifs were also found in green algae but not in bacteria, cyanobacteria or red algae species (Fig. 10A). This finding suggests that the point of divergence of the DHDPS quaternary structures occurred between red and green algae. This remains to be verified experimentally.
To examine whether the same pattern is observed for DHDPR, sequences from 25 bacteria, 57 cyanobacteria, 38 plants, 12 green algae and 2 red algae from the NCBI protein database were obtained (July 2016). Figure 10B shows a representative multiple sequence alignment of a subset of these species. The length of the loop motif in cyanobacteria (Gly183-Ser203, E. coli DHDPR numbering) is similar in length to other bacterial species (Fig. 10B). However, this loop is significantly longer in plant, red algae and green algae sequences (Fig. 10B). This suggests that DHDPR from red and green algae may adopt a similar dimeric quaternary architecture to plant orthologues. This also remains to be verified experimentally.

Conclusions
In this study, we determined for the first time the enzyme kinetic parameters, solution properties and three-dimensional structures of DHDPS and DHDPR from a cyanobacterial species. We show both enzymes exist as homotetramers in solution and in the crystal state, and that they adopt the canonical bacterial quaternary architecture. Our results suggest that the point of structural divergence differentiating bacterial and plant DHDPS and DHDPR enzymes occurred between cyanobacteria and lower plants.

Cloning, expression and purification of Av-DHDPS and Av-DHDPR. Synthetic codon-optimised
Av-DHDPS (i.e. dapA) and Av-DHDPR (i.e. dapB) genes cloned into pRSET-A expression vectors were purchased from GeneArt. The plasmids were subsequently transformed into E. coli BL21-DE3 pLysS cells for the overexpression of the recombinant enzymes. Recombinant protein was produced by treating E. coli BL21-DE3 pLys cells with 1.0 mM IPTG at 25 °C for 8 h. Cells were harvested by centrifugation (5000 × g) and resuspended in 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 20 mM imidazole, 5% (v/v) glycerol, which included 10 mM pyruvate for Av-DHDPS, given that pyruvate is known to stabilise DHDPS enzymes 16,22 . The cell suspension was lysed on ice by sonication using a Vibra Cell VC40 (Sonics & Materials) at 40 micron using 6 cycles of 10 sec on followed by 2 min off. Recombinant His-tagged enzymes were isolated from the cell lysate using IMAC employing a 5 ml His-Trap column (GE Healthcare) and a 0-500 mM imidazole linear gradient over 17 column volumes. Av-DHDPS and Av-DHDPR eluted at 195 mM and 140 mM imidazole, respectively. The purified protein was dialysed overnight against 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5% (v/v) glycerol, which included 10 mM pyruvate for Av-DHDPS.
Tandem mass spectrometry. Purified recombinant Av-DHDPS and Av-DHDPR were subjected to trypsin digestion and tandem mass spectrometry (MS/MS) sequencing using a Thermo Scientific LTQ Orbitrap Elite ETD Mass Spectrometer as previously reported 44,45 . Circular dichroism spectroscopy. Circular dichroism (CD) spectra of Av-DHDPS and Av-DHDPR were obtained using an Aviv Model 420 CD spectrometer using similar methods reported previously 14,16,22,27,41,46 . Briefly, wavelength scans were performed between 195 and 240 nm with a 4.0 sec averaging time in 20 mM Tris, pH 8.0, 150 mM NaCl (also containing 1 mM pyruvate for Av-DHDPS) in a 1.0 mm quartz cuvette. Data were analysed using the CDPro software package incorporating the SP22X database 47,48 . DHDPS-DHDPR coupled enzyme kinetic assay. Kinetic analyses of Av-DHDPS and Av-DHDPR were performed employing the DHDPS-DHDPR coupled assay as previously described 7,9,11,14,16,19,22,24,25,[27][28][29] . Briefly, assays were performed in triplicate at 30 °C in a 1 cm acrylic cuvette using E. coli DHDPR and E. coli DHDPS as the coupling enzymes for Av-DHDPS and Av-DHDPR, respectively. Mixtures were allowed to equilibrate in a temperate-controlled Cary 4000 UV-Vis spectrophotometer for 12 min before initiating the reaction with ASA. The initial reaction rate data were analysed using the ENZFITTER software (Biosoft). Data were fitted to various models, including the bi-bi ping-pong and ternary complex models with and without substrate inhibition, with best fits determined from the highest R 2 value.
Microscale thermophoresis. Affinity measurements using microscale thermophoresis (MST) were carried out with a Monolith NT. LabelFree instrument (NanoTemper Technologies) 19 49,50 . 400 μ l of buffer and 380 μ l of sample at an initial concentration ranging from 0.1 μ M to 7.0 μ M were loaded into the reference and sample sectors of the cells, respectively. The rotor was accelerated to 40,000 rpm and data were collected continuously at 230 nm using a step size of 0.003 cm without averaging. Initial scans were carried out at 3,000 rpm to determine the optimal wavelength and radial positions for the high speed experiment. Solvent density, solvent viscosity, and estimates of the partial specific volume of Av-DHDPS (0.736 ml/g) and Av-DHDPR (0.736 ml/g) at 20 °C were calculated using SEDNTERP 51 . Data were fitted using the SEDFIT software (www. analyticalultracentrifugation.com) to a continuous size-distribution model 52-55 . Crystallisation and X-ray diffraction. Av-DHDPS and Av-DHDPR were crystallised using the hanging-drop vapour diffusion method as described previously 10,30,40,56 . For X-ray data collection, crystals were transferred to reservoir solution containing 20% (v/v) glycerol and 12.5 mM MnCl 2 , and directly flash frozen in liquid nitrogen. Intensity data were collected at the Australian Synchrotron using the MX2 beamline. Diffraction data were processed using MOSFLM 57 and scaled using SCALA 58 . Molecular replacement was performed using the MR protocol of Auto-Rickshaw 59 with B. anthracis DHDPS (PDB ID: 1XKY) as the search model for Av-DHDPS and M. tuberculosis DHDPR (PDB ID: 1YL5) as the search model for Av-DHDPR. Structural refinement was performed using REFMAC5 60 with iterative model building using COOT 61 . The refinement statistics are provided in Table 4. For Av-DHDPS, Ramachandran statistics showed 91.3% in the preferred region, 8.3% in the additionally allowed region and 0.4% in the disallowed region consistent with previous structural reports 15,21,62 . For Av-DHDPR, Ramachandran statistics showed 89.5% in the preferred region, 8.8% in the additionally allowed region and 1.3% in the generously allowed region consistent with previous studies [31][32][33][34][35][36] . However, 0.4% of residues (i.e. Ser89 and Gln112) were in the disallowed region due to poor electron density.