Structure and Mechanism of D‑Glucosaminate-6-phosphate Ammonia-lyase: A Novel Octameric Assembly for a Pyridoxal 5′-Phosphate-Dependent Enzyme, and Unprecedented Stereochemical Inversion in the Elimination Reaction of a D‑Amino Acid

Abstract

D-Glucosaminate-6-phosphate ammonia-lyase (DGL) is a pyridoxal 5′-phosphate (PLP)-dependent enzyme that produces 2-keto-3-deoxygluconate 6-phosphate (KDG-6-P) in the metabolism of D-glucosaminic acid by Salmonella enterica serovar typhimurium. We have determined the crystal structure of DGL by SAD phasing with selenomethionine to a resolution of 2.58 Å. The sequence has very low identity with most other members of the aminotransferase (AT) superfamily. The structure forms an octameric assembly as a tetramer of dimers that has not been observed previously in the AT superfamily. PLP is covalently bound as a Schiff base to Lys-213 in the catalytic dimer at the interface of two monomers. The structure lacks the conserved arginine that binds the α-carboxylate of the substrate in most members of the AT superfamily. However, there is a cluster of arginines in the small domain that likely serves as a binding site for the phosphate of the substrate. The deamination reaction performed in D₂O gives a KDG-6-P product stereospecifically deuterated at C3; thus, the mechanism must involve an enamine intermediate that is protonated by the enzyme before product release. Nuclear magnetic resonance (NMR) analysis demonstrates that the deuterium is located in the pro-R position in the product, showing that the elimination of water takes place with inversion of configuration at C3, which is unprecedented for a PLP-dependent dehydratase/deaminase. On the basis of the crystal structure and the NMR data, a reaction mechanism for DGL is proposed.

Graphical Abstract

The metabolism of D-glucosaminic acid in Salmonella enterica servar typhimurium was shown recently to proceed by phosphorylation at C-6 to give D-glucosaminate 6-phosphate (DGA-6-P).¹ This is followed by elimination of ammonia by the product of the DgaE gene, D-glucosaminate-6-phosphate ammonia-lyase (DGL, EC 4.3.1.29), as shown in eq 1, to produce 2-keto-3-deoxygluconate 6-phosphate (KDG-6-P), a key intermediate in the Entner–Doudoroff pentose phosphate pathway.^1,2

(1)

GL is a pyridoxal 5′-phosphate (PLP)-dependent enzyme that can be assigned to the aminotransferase (AT) superfamily (fold I) on the basis of sequence homology. However, DGL has a homology (<20% identity) quite remote from those of most members of the AT superfamily, and DGL, together with bacterial selenocysteine synthase (SelA), represents a novel subfamily within the AT superfamily.^3,4 Furthermore, the substrate of DGL is a D-amino acid, whereas the other approximately 200 members of the AT superfamily, with the exception of isoleucine epimerase⁵ and CAP-15,⁶ act exclusively on L-amino acids. Nevertheless, DGL orthologues are widely distributed among α-proteobacteria, and in some firmicutes,¹ and the Dga operon has been linked to pathogenicity in S. enterica, Escherichia coli, Enterobacter faecalis, Clostridium difficile, and Lactobacillus casei.⁷ Thus, the three-dimensional structure of DGL is of considerable interest. We have now determined the three-dimensional structure of DGL by X-ray crystallography to a resolution of 2.58 Å, and we have discovered that it has a novel octameric assembly heretofore unobserved in the AT superfamily. Furthermore, nuclear magnetic resonance (NMR) analysis demonstrates that the elimination reaction takes place with unprecedented inversion of stereochemistry at C-3. A reaction mechanism for DGL is proposed on the basis of the crystal structure of DGL and NMR analysis of the reaction product.

MATERIALS AND METHODS

Materials.

DGA-6-P was prepared from D-glucosamine-HCl (Sigma-Aldrich), and DgaF was prepared as described previously.² L-Selenomethionine and NADH were obtained from Sigma-Aldrich. Lactate dehydrogenase was a product of USBiochemicals. Buffers, salts, and solvents were obtained from Fisher Scientific.

Enzyme.

Salmonella typhimurium DGL was overexpressed and purified on a Ni-metal chelate column as described previously.^1,2 It should be noted that native DGL binds selectively to a Ni-metal chelate column without addition of a hexa-His tag. SeMet-substituted DGL was expressed in 0.5 L of Studier PASM-5052 autoinduction medium⁸ containing 125 μg/mL L-selenomethionine in a 2 L Erlenmeyer flask at 37 °C and 250 rpm using the previously reported BL21(DE3) cells containing the DGL expression plasmid.² The yield of purified selenomethionine-substituted protein was 100 mg. The purified protein contained 81% replacement of methionines by selenomethionine as determined by electrospray ionization mass spectrometry (Figure S1). DGL was assayed in 0.1 M triethanolamine-HCl (pH 8.0) containing 4 μL of crude DgaF (2-keto-3-deoxygluconate-6-phosphate aldolase), 10 μg/mL lactate dehydrogenase, and 0.2 mM NADH, at 37 °C, following the decrease in absorbance at 340 nm (Δε = −6220 M⁻¹ cm⁻¹) as described previously.^1,2

Rapid-Scanning Stopped-Flow Measurements.

DGL (2.9 mg/mL) in 0.05 M Tris-HCl (pH 8.0) was mixed with 20 mM DGA-6-P in 0.05 M Tris-HCl (pH 8.0) at room temperature in an OLIS RSM-1000 rapid-scanning stopped-flow spectrophotometer. Data were collected between 240 and 800 nm at 1 kHz for 2 s, or at 10 Hz for 150 s. The data were analyzed with OLIS Global Works.⁹

NMR of the DGL Reaction Product.

D-Glucosaminate 6-phosphate (2.75 mg, 0.01 mmol) was dissolved in 600 μL of D₂O, and then 1.4 μL of 14.2 M NaOD (0.02 mmol) was added, followed by 1.4 μL of DGL (61 mg/mL) to give a pH of ~6. The reaction mixture was incubated for 30 min, and then the NMR spectrum was recorded with water suppression by presaturation in a Bruker 400 MHz spectrometer. The unlabeled product was obtained by performing the enzymatic reaction in water, followed by lyophilization and redissolving in D₂O. To prepare D-ribonolactone 5-phosphate, 5.3 mg of DGA-6-P was dissolved in 600 μL of D₂O, followed by 3.3 μL of 14.2 M NaOD and 2 μL of DGL. After the reaction was complete as determined by NMR, 5 μL of 30% H₂O₂ was added, and the reaction mixture was incubated for 2 h. The solution was then frozen and lyophilized. The dry residue was dissolved in 1 mL of trifluoroacetic acid-D for NMR analysis of the lactone. NMR spectra for KDG and D-ribonolactone were calculated with ORCA,¹⁰ with B3LYP using the DEF2-TZVPP basis set with the DEF2/JK auxiliary to provide GIAOs. The spectra were referenced by calculating the NMR spectrum of acetone with the same parameters and assigning it to δ 2.0.

Crystallization and Data Collection.

DGL was crystallized in hanging drops under two conditions. The protein, 2 μL of a 12 mg/mL solution in 0.1 M Tris-HCl (pH 8.0), was mixed in a 1:1 ratio with a solution from a well containing 1 mL of 0.1 M Tris-HCl (pH 8.5), 0.1–0.4 M CaCl₂, 6–10% glycerol, and 10–12% PEG 4000 (condition 1) and suspended over the well at ambient temperature (~20 °C). The other condition used 2 μL of a 12 mg/mL solution in 0.1 M Tris-HCl (pH 8.0) mixed with 0.1 M MES-Na (pH 6.5), 0.2–0.4 M Mg(OAc)₂, and 12–16% PEG 8000 (condition 2) in the well. Under both conditions, crystals of DGL appeared as yellow prisms as large as 200 μm within 1 week. The SeMet-substituted enzyme gave crystals only under condition 2.

The native protein crystals were transferred to cryosolvent solutions containing 0.1 M Tris-HCl (pH 8.0) and 0.4 M CaCl₂ with 20% PEG 4000 and a 20% ethylene glycol/dimethyl sulfoxide/glycerol (1:1:1) mixture, and flash-cooled in liquid nitrogen. The SeMet-substituted protein crystals were transferred to cryosolvent containing 0.1 M Mes-Na (pH 6.5), 0.4 M Mg(OAc)₂, 20% PEG 8000, and 10% DMSO, and flash-cooled in liquid nitrogen. Data were collected on beamline ID-22 at SER-CAT at Argonne National Laboratory for native protein on August 15, 2019, and for the SeMet-substituted enzyme on September 26 and October 15, 2020. For the native protein, data were collected at 1.0000 Å with 0.25° oscillations and a 0.25 s exposure for 180° at 100 K. The data for the SeMet-substituted enzyme were collected at 0.979 Å with 0.25° oscillations and a 0.25 s exposure for 360° at 100 K.

Data Processing.

The data were integrated with XDS¹¹ and scaled with AIMLESS¹² and POINTLESS¹³ using AutoPROC.¹⁴ Data collected at 0.979 Å from two crystals of SeMet-substituted DGL obtained under condition 2 were found to show moderate anomalous scattering, extending to ~3 Å, and were used for SAD phasing.¹⁵ The crystals were in space group P2₁ and had the following dimensions: a = 53.8 Å, b = 206.3 Å, c = 162.5 Å, and β = 94.9°. The self-rotation function showed two 2-fold rotations and one 4-fold rotation. Because the asymmetric unit was predicted to contain 7–10 monomers on the basis of the solvent content, we concluded that the asymmetric unit was likely to be an octamer. The data, with a resolution limit of 2.974 Å, were analyzed by HySS,¹⁶ and a substructure with 91 heavy atoms was found, compared to a predicted 88 seleniums for the octamer based on the amino acid composition, with a CC of 0.21. This substructure was used as input for AutoSOL,¹⁷ using PHASER¹⁸ for map creation and RESOLVE¹⁹ for density modification. The resulting map had a FOM of 0.26, an R of 0.3915, and an RMSD of 0.61. However, attempted model building using BUCCANEER²⁰ with this map gave only 8% completion, with an R_free of 0.54. The data were moderately anisotropic, and the ellipsoidally truncated data output from STARANISO²¹ with a resolution limit of 2.60 Å was then used for experimental phasing. Ellipsoidal diffraction limits for this data were as folows: a* = 3.02 Å, b* = 2.60 Å, and c* = 2.61 Å. HySS was run with these data and found a substructure of 105 heavy atom sites at a resolution of 3.1 Å, with a CC of 0.36. Running AutoSOL with this substructure resulted in a density-modified map, with a FOM of 0.33, an R of 0.3074, and an RMSD of 0.77. Model building with this map using BUCCANEER in CCP4i2²² successfully resulted in a model with 98% completion, including five complete chains of 369 residues, with an R_work of 0.22 and an R_free of 0.29. The structure was completed by model building with COOT²³ using NCS. The resulting model was refined with PHENIX.refine²⁴ using NCS and TLS. TLS regions were determined with PHENIX.-find_tls_groups. This structure [Protein Data Bank (PDB) entry 7LC0] was refined to a resolution of 2.60 Å.

The native data, from crystals collected under condition 1, were used for molecular replacement with PHASER with the SeMet-substituted enzyme structure as the search model. These crystals were in space group P222₁ and had the following dimensions: a = 60.4 Å, b = 171.5 Å, c = 195.0 Å, and α = β = γ = 90°. STARANISO was also used to process the moderately anisotropic data: a* = 2.58 Å, b* = 2.79 Å, and c* = 3.01 Å. The self-rotation function showed three 2-fold rotation axes, and the asymmetric unit was predicted to contain 4–6 monomers, on the basis of solvent content. Hence, two tetrameric search models were built by splitting the octamer in half, either on the catalytic dimers or on the noncatalytic dimers. The tetrameric model with the octamer split at the noncatalytic dimers failed to give a solution with PHASER. The model split at the catalytic dimers was used for molecular replacement with PHASER and gave a single solution with an LLG of 12256 and a TFZ of 51. This model was then rebuilt with BUCCANEER to replace the SeMet in the search model with Met. NCS-averaged maps were prepared with phenix.ncs.average and used for model building in COOT. Refinement of this structure (PDB entry 7LCE) was performed with PHENIX.refine to a resolution of 2.58 Å.

Gel Filtration of DGL.

DGL [0.5 mL, 10 mg/mL in Tris-HCl (pH 8.0)] was injected into a column of Sephacryl S-200HR (1 cm × 50 cm) equilbrated with 0.1 M Tris-HCl (pH 8.0) and 0.1 mM PLP and eluted at a flow rate of 0.5 mL/min using a Bio-Rad LP system. The column was calibrated by injection of 0.5 mL of bovine serum albumin [10 mg/mL in 0.1 M Tris-HCl (pH 8.0)]. The proteins were detected by absorbance at 280 nm.

RESULTS AND DISCUSSION

Bioinformatics of DGL.

DGL is a member of the AT superfamily of PLP-dependent enzymes, which contains ~200 different enzyme activities. However, there is very low sequence identity of DGL with most other members of the family, such as aspartate aminotransferase (AAT), kynureninase (KynU), and tyrosine phenol-lyase (TPL), other than in the essential PLP-binding aspartate and lysine residues (Figure 1A).²⁵ The conserved lysine binds the PLP through formation of a Schiff base linkage in the internal aldimine. The conserved aspartate forms a hydrogen bond with the pyridine nitrogen of the PLP, ensuring that the nitrogen remains protonated to increase the electrophilicity of the cofactor. There is also a partially conserved serine that donates a hydrogen bond to the phosphate of PLP, three residues prior to the PLP-binding lysine. The highest sequence homology of DGL is observed with SelA (PDB entry 3W1H) from Aquifex aeolicus,³ which has only 21% identity for the full length protein, and Cap15 from Streptomyces griseus, a novel PLP-dependent decarboxylating oxygenase,⁶ which has 41% identity with DGL (Figure 1B) but does not have a crystal structure. Because DGL and SelA make up a newly recognized subfamily (IPR006337) of the AT superfamily,⁴ CAP15 must also be a member of this subfamily. Furthermore, due to the high sequence identity, the structure of CAP15 must be very similar to that of DGL. A proposed selenocysteine synthase from Methanocaldococcus janaschii based on sequence homology (PDB entry 2AEU) was shown to have no selenocysteine synthase activity²⁶ and has only 14% sequence identity with DGL.

Figure 1.

(A) Sequence alignment of the active site residues of S. typhimurium DGL, E. coli aspartate aminotransferase (AAT), Pseudomonas fluorescens kynureninase (KynU), Aquifex aeolicus selenocysteine synthase (SelA), and Citrobacter freundii tyrosine phenol-lyase (TPL). The PLP-binding aspartate and lysine residues are colored red with yellow highlights, and the partially conserved serines are colored blue with yellow highlights. (B) Sequence alignment of S. typhimurium DGL with A. aeolicus SelA and Streptomyces griseus CAP15. Conserved residues are colored magenta. Residues involved in catalytic dimer contacts in DGL are highlighted in yellow, and those involved in noncatalytic dimer contacts are highlighted in green. The residues involved in PLP binding in DGL are highlighted in cyan.

Phasing of DGL.

Numerous attempts to phase crystals of native DGL by molecular replacement with PHASER¹⁸ using the structure of SelA in either the full length form (PDB entry 3W1H) or the truncated dimeric mutant form (PDB entry 3WCO) failed to find a solution that could be used for model building, with either monomers or dimers as the search model. Other AT superfamily PLP-dependent enzymes showing even less identity (11–20%) also did not give usable solutions with PHASER using molecular replacement. Phasing was finally accomplished by SAD^15,27 using SeMet-substituted DGL, with crystals obtained under different conditions, which provided a map suitable for building of an initial model of the octameric assembly. This model was then used for phasing by molecular replacement with PHASER of a tetrameric substructure on a native data set. The resulting map had sufficient electron density to build all 369 residues in all four chains in the asymmetric unit. The final model of the tetrameric asymmetric unit of native DGL was refined to a resolution of 2.58 Å with R_work and R_free values of 0.193 and 0.229, respectively, and the selenoMet-substituted octamer was refined to 2.60 Å resolution with R_work and R_free values of 0.193 and 0.226, respectively. The statistics for the final structures of SeMet-substituted and native DGL are listed in Table 1.

Table 1.

Data Collection and Refinement Statistics^a

	7LCE, DGL	7LC0, DGL-SeMet
wavelength (Å)	1.000	0.9790
ellipsoidal resolution range (Å)	64.41−2.58 (3.02−2.58)	53.87−2.60 (2.77−2.60)
ellipsoidal resolution limits (Å)	a* = 2.58, b* = 2.79, c* = 3.01	a* = 3.02, b* = 2.60, c* = 2.61
spherical resolution limits (Å)	64.41−2.73 (3.02−2.73)	52.59−2.66 (2.71−2.66)
space group	P2221	P1211
unit cell	60.374, 171.569, 195.054, 90, 90, 90	54.089, 205.572, 164.026, 90, 95.187, 90
total no. of reflections	715473 (120126)	621088 (32823)
no. of unique reflections	48940 (8157)	87468 (4373)
multiplicity	14.6 (14.7)	7.2 (7.3)
anomalous multiplicity	−	3.6 (3.8)
completeness (%)	94.6 (75.0)	93.3 (60.3)
anomalous completeness (%)	−	92.6 (60.2)
mean I/σ(I)	18.1 (1.7)	10.6 (1.4)
Wilson B-factor (Å2)	79.15	55.91
R meas	0.137 (1.983)	0.147 (1.529)
CC(1/2)	0.999 (0.629)	0.997 (0.522)
CC*	1 (0.525)	0.999 (0.643)
no. of reflections used in refinement	48914 (8157)	87468 (4373)
no. of reflections used for Rfree	2000 (397)	1985 (24)
R work	0.1926 (0.3014)	0.1930 (0.2975)
R free	0.2287 (0.3256)	0.2275 (0.3819)
CC(work)	0.963 (0.779)	0.934 (0.702)
CC(free)	0.950 (0.870)	0.913 (0.759)
no. of non-hydrogen atoms	11388	22560
macromolecules	11164	22192
ligands	40	115
solvent	184	253
no. of protein residues	1476	2952
RMSD for bonds (Å)	0.002	0.002
RMSD for angles (deg)	0.55	0.59
Ramachandran favored (%)	97.73	97.15
Ramachandran allowed (%)	2.27	2.82
Ramachandran outliers (%)	0	0.45
Rotamer outliers (%)	0.81	0.68
Clashscore	1.97	2.87
average B-factor (Å2)	87.59	68.63
macromolecules	90.73	68.74
ligands	111.97	72.45
solvent	75.70	57.98
no. of TLS groups	11	44

^aStatistics for the highest-resolution shell are shown in parentheses.

Monomer of DGL.

The basic unit of the AT superfamily of PLP-dependent enzymes is a monomer with a large rigid domain, which binds PLP, and a small mobile domain, which can rotate to open and close the active site when substrates or ligands bind. The secondary structure of DGL is shown in Figure S2. The monomer of DGL has a large domain composed of residues 32–262, with a core composed of seven β-strands, five parallel and two antiparallel, flanked by α-helices on each side (Figure 2A, red, magenta, and cyan), that forms the PLP-binding site. There are nine extended helices and two short helical turns in the large domain. This core structure of β-strands and helices is conserved in the other members of the AT superfamily, such as SelA (Figure 2B), aspartate aminotransferase (Figure 2C), tyrosine phenol-lyase (Figure 2D), and kynureninase (Figure 2E), despite the low identity (<20%) in the amino acid sequences. However, the small domains of these enzymes are extremely diverse in both sequence and structure, as one can see in Figure 2. The small domain of DGL has three extended helices, two short helical turns, and five antiparallel β-strands and is made from residues 1–16 and 269–369 (Figure 2A, red, green, and yellow). These two domains are connected by linkers, a loop composed of residues 17–31, and a short strand of residues 263–269 (Figure 2A, orange). The small domain is more disordered than the large domain, with much higher B-factors (Figure S3), consistent with the high mobility expected for the small domain in the absence of a ligand in the active site.

Figure 2.

Comparison of the structures of the monomers of DGL, selenocysteine synthase, aspartate aminotransferase, tyrosine phenol-lyase, and kynureninase. (A) Monomer of DGL. (B) Monomer of selenocysteine synthase (PDB entry 3WCO). (C) Monomer of E. coli aspartate aminotransferase (PDB entry 1ASN). (D) Monomer of C. freundii tyrosine phenol-lyase (PDB entry 2TPL). (E) Monomer of P. fluorescens kynureninase (PDB entry 1QZ9). The large domain is on the bottom of the structures.

Dimer of DGL.

The minimum functional unit of AT superfamily PLP-dependent enzymes is a dimer, because the active site is formed by residues contributed by both monomers at the monomer–monomer interface. The catalytic dimeric assembly of DGL in Figure 3A is formed by hydrophobic effects and extensive hydrogen bonding contacts. On one side, the hydrogen bonding network extends across the width of the monomer–monomer interface (Figure 3B). On the other side, the hydrogen bonding contacts are clustered around a 2-fold axis (Figure 3C). The total contact surface in the catalytic dimer is 3117 Ų. This dimeric assembly is similar to that observed in the other members of the AT superfamily.

Figure 3.

Catalytic dimer contacts of DGL. (A) Hydrogen bonding contacts of the catalytic dimer. (B) Close-up crossed-eye stereoview of the hydrogen bonding contacts in the top of the catalytic dimer. (C) Crossed-eye stereoview of the hydrogen bonding contacts on the bottom of the catalytic dimer.

There is a hydrophobic cluster composed of the side chains of Leu-126, Ile-240, Ala-243, and Met-244 and another hydrophobic cluster with the side chains of Ile-5, Tyr-6, Val-45, Ile-47, Leu-50, and Val-252 on the contact surface. Hydrogen bonds are seen among Pro-3, Tyr-6, and Gln-8 in helix 1, and Tyr-330 in the small domain, and Ile-23, Leu-24, Val-29, Arg-39, Ala-40, Ala-41, Ser-42, Val-45, Glu-46, Ile-47, Ala-94, Met-96, Ser-122, and Arg-125 in helix 7, and Thr-219, Glu-249, and Asn-250 in helix 14, in the large domain of chains A and B (Figures 1B and 3B,C). The Met-96 O, connected to cis-Pro-97, accepts a hydrogen bond from NH of Arg-125, which in turn donates a hydrogen bond to the O of Ala94. In the active site, NH1 of Arg-242 donates a hydrogen bond to OP3 of the LLP of the other chain, and NH2 hydrogen bonds to the O of Tyr-116.

Two monomers of DGL can also interact to form a noncatalytic dimer, which does not create an active site, as shown in Figure 4. The noncatalytic dimer contacts have few hydrophobic interactions but show hydrogen bonding contacts between Arg-110, Asn-113, and Asp-115 of one chain and Glu-133, Ser-136, Ser-137, and Asn-138 of the other chain (Figures 1B and 4). These contacts are less extensive than those in the catalytic dimer, with a surface area of only 542 Ų.

Figure 4.

Crossed-eye stereoview of the hydrogen bonding contacts in the noncatalytic dimer.

Octamer of DGL.

The combined catalytic and noncatalytic dimer–dimer contacts in DGL generate a complete assembly that is an octamer, as a tetramer of dimers, with a shape reminiscent of a “templar cross” (Figure 5A,B). To the best of our knowledge, this is an assembly that has not been observed heretofore in the AT superfamily, although octamers are well-known for other proteins. A search of the PDB with the keywords “pyridoxal-5′-phosphate” and “octamer” did not reveal any octameric assemblies in the AT superfamily. These enzymes have been found previously to form dimers (e.g., AAT, PDB entry 1ASN²⁸), tetramers (e.g., tryptophan indole-lyase, PDB entry 1AX4²⁹), hexamers (e.g., E. coli glutamate decarboxylase, PDB entry 1PMM³⁰), decamers (e.g., SelA, PDB entry 3W1H³), and dodecamers (e.g., Pseudomonas dacunhae aspartate β-decarboxylase, PDB entry 3FDD³¹). In the complete DGL octamer, there are other dimer–dimer contacts, in addition to the catalytic and noncatalytic dimer contacts. There are π-cation interactions between the guanidinium group of Arg-90 on one dimer and the indole ring of Trp-143 on another dimer (Figure S4).

Figure 5.

Octamer of DGL. (A) Crossed-eye stereoview of the octamer of DGL, looking down the 4-fold axis. The PLP is shown as a space filling model. (B) Crossed-eye stereoview of the octamer of DGL, looking down the 2-fold axis in the catalytic dimer. (C) Crossed-eye stereoview of the octamer of DGL, looking down the 2-fold axis in the noncatalytic dimer.

Bacterial SelA forms a decameric assembly as a pentamer of dimers.³ The main contacts between the catalytic dimers of SelA differ from those in DGL in that an antiparallel β-sheet forms between the SelA dimers.³ Mutations that disrupt the dimer–dimer contacts in SelA result in a stable dimer that is inactive to convert Ser-tRNA^Sec to Sec-tRNA^Sec.³² The decamer of SelA was shown to be capable of binding 10 tRNAs, with the tRNA-binding site formed between the dimers in the decamer.³ Thus, the intact decamer is essential for the catalytic activity of SelA. For DGL, it is not known at present if the intact octameric assembly is necessary for catalytic activity. Previously, we had determined a molecular weight of ~80 kDa for DGL, corresponding to a dimer, using gel filtration in a relatively dilute solution in the absence of PLP.² We repeated the gel filtration at a higher concentration in the presence of 0.1 mM PLP and found that the major peak corresponds to an apparent molecular weight of 260 kDa, compared to a predicted value of 312 kDa for the octamer, and a shoulder at 80 kDa, consistent with the dimer (Figure S5). Furthermore, PISA analysis³³ predicts that the octamer is the most stable assembly in solution (Figure S6). This suggests that there is an octamer–dimer equilibrium in solution that may be dependent on temperature, pH, and protein and PLP concentrations. The surface contacts of the catalytic dimers are more extensive than those of the noncatalytic dimers, so it is likely that the dissociation of DGL leads to catalytic rather than noncatalytic dimers. Glutamate decarboxylase from Arabidopsis thaliana shows a similar hexamer–dimer equilibrium, responsive to protein concentration, pH, ionic strength, and Ca²⁺/calmodulin, that is proposed to play a role in the regulation of activity in vivo.³⁴

Active Site of DGL.

PLP is bound to DGL as a Schiff base linkage with Lys-213 (Figures 1A and 6A), as expected for a member of the AT superfamily. The phosphate of PLP contacts DGL by hydrogen bonds donated by N (2.7 Å) of Ala-73, N (3.1 Å) and OG (2.9 Å) of Ser-74, OG (3.1 Å) of Ser-210, NH1 (2.9 Å) of Arg-242*, and NZ of Lys-245* (3.7 Å) from the other monomer in the catalytic dimer (Figure S7). In addition, the phosphate caps the N-terminus of helix H4 (Figure S2), which contains Ala-73 and Ser-74, as is normal in the AT superfamily. The pyridinium NH of PLP donates a hydrogen bond to OD2 (2.8 Å) of Asp-189 to maintain the protonated state, as one can see from the ultraviolet–visible absorption peak of the bound cofactor at 424 nm. Different from most other members of the AT superfamily, there is no close hydrogen bonding contact of PLP O3 with the protein, although the OG of Ser-161 is 4.2 Å from and NZ of LLP-213 3.1 Å from O3, which could donate a hydrogen bond. Generally, the AT superfamily has an aromatic residue side chain forming a π–π stacking interaction with the PLP, and DGL has Tyr-116 fulfilling this role (Figure 6A). The OH of Tyr-116 also donates a hydrogen bond to OG of Ser-74. These contacts of PLP are very similar to those in the active site of SelA, and all of these PLP contact residues are conserved in SelA, except for Ala-73 and Ser-74, which are Asn and Ala, respectively (Figure 1B), but still donate hydrogen bonds to the phosphate from the peptide backbone NHs, and Lys-245*, which is replaced by an Arg. The PLP-binding residues of DGL are also conserved in the sequence of CAP15, except for Ser-74, which is substituted by Ala (Figure 1B).

Figure 6.

(A) Crossed-eye stereoview of the PLP-binding site in DGL. The simulated omit mF_o – DF_c map is contoured at 3σ. (B) Crossed-eye stereoview of DGA-6-P manually docked into the active site showing the proposed phosphate-binding arginine cluster in the small domain.

Most members of the AT superfamily have a conserved arginine residue in the small domain in the active site that forms a salt bridge with the α-carboxylate of the bound substrate that is critical for substrate binding.²⁵ The presence of this arginine orients the α-carboxylate in the external aldimine on the same side as the 3′-O of the PLP. Unsurprisingly, this arginine residue is absent from SelA, because its substrate is the tRNA ester of L-serine, which lacks a free α-carboxylate. DGL also lacks the conserved α-carboxylate-binding arginine residue in the active site, despite having an α-carboxylate in the substrate. However, a cluster of arginine residues (Arg-301, Arg-332, and Arg-346) is located nearby in the small domain, which could serve as a binding site for the phosphate of DGA-6-P (Figure 6B). Manual docking of the substrate shows that binding of the phosphate to this arginine cluster would position the amino group to attack the re-face of the Schiff base, and the α-carboxylate would be positioned near Arg-242*, which could donate a hydrogen bond (Figure 6B). This positions the α-CH on the same face as Lys-213, allowing it to perform deprotonation of the external aldimine to form the quinonoid intermediate. Thus, there is a reversal of the orientation of the side chain and α-carboxylate of the substrate in the active site of DGL compared to other members of the AT superfamily, which allows for the reaction of a D-amino acid substrate with Lys-213, in accord with Dunathan’s hypothesis.³⁵ Previously, we suggested that His-162 may be a catalytic base to allow for reaction of a D-amino acid, based on a homology model.² However, the crystal structure suggests that His-163 is closer to the substrate-binding site (Figure 6A,B). This histidine is also conserved in the sequence of CAP15, but not in the sequence of SelA (Figure 1B). Both of these histidines are strictly conserved in the sequences of DGL, even those from firmicutes, which have only ~45% overall sequence identity with the enzyme from S. typhimurium. This exposed pair of vicinal histidines may explain the observed binding of DGL to a Ni²⁺-metal chelate column, even without the addition of a His tag.²

Catalytic Properties of Selenomethionine-Substituted DGL.

SeMet-substituted DGL contained 81% selenomethionine in place of methionine, based on electrospray mass spectrometry (Figure S1). SeMet-substituted DGL was found to have activity slightly reduced from that of the native enzyme (Table 2). Suprisingly, there have been very few cases in which the activity of SeMet-substituted enzymes has been quantitatively compared with that of the native enzyme. Phosphomannose isomerase was found to have its K_m increased 4-fold for the SeMet-substituted enzyme, without an effect on V_max.³⁶ This was explained as being due to the steric effect of SeMet residues located in the active site. In our case, we find a doubling of K_m as well as a 40% reduction in k_cat, for an overall 3.2-fold decrease in k_cat/K_m (Table 2). There is a methionine, Met-21, in possible van der Waals contact with the substrate (Figure 6B), and as with phosphomannose isomerase, this may explain the effects observed in the kinetics of SeMet-substituted DGL (Table 2).

Table 2.

Catalytic Properties of Native and SeMetSubstituted DGL

enzyme	Km (mM)	kcat (s−1)	kcat/Km (M−1 s−1)
native	0.139 ± 0.026	27.4 ± 1.4	(1.97 ± 0.31) × 105
SeMet-substituted	0.283 ± 0.025	17.5 ± 3.5	(6.15 ± 0.03) × 104

Mechanism of DGL.

Mixing of DGL with DGA-6-P in the rapid-scanning stopped-flow spectrophotometer results in changes in the absorption spectrum of the PLP cofactor (Figure 7). Within the mixing time (~2 ms) of the instrument, the 424 nm absorption peak of the internal aldimine (Figure 7, dashed line) decreases and a quinonoid intermediate, as a shoulder with a λ_max of 478 nm, is formed (Figure 7, black line). This intermediate decays with a k_obs of 318 ± 25 s⁻¹ to give an intermediate with absorption around 340 nm that is likely to be a substituted enamine resulting from dehydration of the substrate (Figure 7, red line). Similar aminoacrylate intermediates have been observed with other members of the AT superfamily, and they exhibit broad absorptions with peaks at 340–350 nm.^37,38 This absorbance peak remains constant during the steady state and then decays back to the initial aldimine spectrum after the substrate is exhausted (Figure 7, green line and inset).

Figure 7.

Visible spectra of DGL during the reaction with DGA-6-P: dashed line, free enzyme; black line, 0.002 s; red line, 1.00 s; blue line, 90.00 s; green line, 150.00 s.

When the reaction of DGL is performed in D₂O, only one of the C3 protons in the product is replaced by deuterium (Figure S8A), which shows that the enamine intermediate is stereospecifically protonated by the enzyme before product release. Surprisingly, the stereochemical assignment of the C3 diasterotopic hydrogens of KDG-6-P has not been reported previously. KDG-6-P forms an equilibrium mixture of α- and β-furanoses in aqueous solution.³⁹ NMR analysis of furanoses to determine stereochemistry by coupling constants and NOEs is complicated by conformational mobility as well as rapid exchange between the anomers and the open chain.³⁹ The upfield protons at C3 are coupled to the downfield resonances at 4.2–4.4 ppm from C4, as one can see in the COSY spectrum (Figure S8B). Irradiation of the C4 protons did not give an interpretable NOE of the C3 protons. However, NMR analysis of the pyranose of KDG produced by D-glucosaminate dehydratase showed that the most upfield signal of the C3 diastereotopic hydrogens is from the pro-R-H, and the downfield signal is from the pro-S-H.⁴⁰ The upfield peaks at 1.85 ppm (α-anomer) and 2.15 ppm (β-anomer) of DGA-6-P are the ones substituted by deuterium (Figure S8A), and on the basis of the previous studies,⁴⁰ we assign this resonance to the pro-R proton of C3. In support of this conclusion, density functional theory (DFT) calculations of the pyranose and furanose forms of KDG show that the pro-R hydrogen is always upfield of the pro-S hydrogen (Table S1). To confirm this assignment, we oxidized the KDG-6-P product of DGL in D₂O with H₂O₂ and performed NMR analysis of the ribonic acid 5-phosphate product cyclized in trifluoroacetic acid-d. The upfield peak of the diastereotopic C2 hydrogens of the D-ribonolactone 5-phosphate was found to be substituted by deuterium (Figure S8C). DFT calculations of the NMR spectrum of ribonolactone show that the upfield peak is the pro-R hydrogen (Table S1), providing additional support for our stereochemical assignment.

The deamination of L- or D-serine to give pyruvate, and L- or D-threonine to give α-ketobutyrate, is catalyzed by a number of PLP-dependent enzymes in the AT superfamily and the β-family. Deamination of serine results in pyruvate, which is not intrinsically chiral. By substituting one of the diastereotopic methylene hydrogens of serine with a hydrogen isotope and performing the elimination in water with a different hydrogen isotope, one can obtain pyruvate with a chiral methyl group.⁴¹ Using this method, tyrosine phenol-lyase was shown to catalyze deamination of L-serine with retention of configuration in the pyruvate.⁴² Similarly, deamination of L-serine by tryptophan indole-lyase proceeds with retention of stereochemistry.⁴³ The isolated β₂ subunit of tryptophan synthase, in the β-family, catalyzes the deamination of L-serine with retention of configuration.⁴⁴ D-Serine dehydratase from E. coli, also in the β-family, was also shown to give pyruvate with retention of configuration.⁴⁵

In contrast to serine, deamination of threonine in D₂O results directly in a chiral product. Deamination of both L-threonine and D-threonine by different enzymes in Serratia marcescens was found to produce deuterated α-ketobutyrates with retention of configuration.⁴⁶ D-Glucosaminate dehydratase, which catalyzes a reaction similar to DGL, also incorporates deuterium in the KDG product with retention of configuration,⁴⁰ because the pro-S hydrogen is replaced by deuterium. The retention of configuration in these eliminations is due to a single lysine residue serving as both the base to deprotonate the C_α atom of the substrate and the acid to protonate the water leaving group. Thus, with a single base, both reactions must take place on one face of the substrate complex. In contrast to these other examples, the deamination of DGA-6-P to KDG-6-P takes place with inversion of configuration at C3, giving (3R)-3-deutero-KDG-6-P.

A mechanism consistent with the crystal structure and the kinetic and NMR data is shown in Scheme 1. The binding of the substrate to the PLP to form the external aldimine and subsequent deprotonation of the Cα-H to give the quinonoid intermediate are very fast, because they occur within the dead time of the stopped-flow instrument (~2 ms) as shown in Figure 7. The quinonoid intermediate decays rapidly, by elimination of water, assisted by the transfer of a proton from Lys-213 as a general acid, giving the enamine intermediate, which is protonated by His-163 on the si-face of C3 to give a ketimine intermediate. Subsequent transaldimination of the ketimine intermediate occurs via a gem-diamine to release the imine product, which undergoes non-enzymatic hydrolysis to provide the observed products, KDG-6-P and ammonia. In D₂O, His-163 transfers a deuteron to the si-face of the intermediate to give the observed (3R)-KDG-6-P.

Scheme 1. Mechanism of DGL.

CONCLUSIONS

DGL is a PLP-dependent enzyme that forms an octameric assembly, as a tetramer of dimers, not previously observed in the AT superfamily. DGL catalyzes the deamination of DGA-6-P with inversion of stereochemistry unprecedented for a PLP-dependent dehydratase/deaminase. This inversion results from protonation of the dehydrated enamine intermediate by His-163 on the opposite face of the substrate from the catalytic base, Lys-213.