Identification, functional characterization, and structural analysis of an atypical l-threonate 3-dehydrogenase

TL;DR

Scientists identified and characterized a new enzyme that helps bacteria break down sugar acids in a unique way.

Contribution

The discovery of GL300_RS07945 as an NADP+-preferring l-threonate 3-dehydrogenase with unique structural features.

Findings

GL300_RS07945 is a novel NADP+-preferring Ltn3D enzyme from Paracoccus litorisediminis.

The enzyme catalyzes a bypass route for converting l-threonate to 3-oxo-l-threonate.

Ltn3D efficiently oxidizes homologous sugar acids, including d-gluconate to d-ribulose 5-phosphate.

Abstract

Diverse bacteria possess unusual gene clusters containing cryptic genes of unknown function, which are often related to the metabolism of sugars and sugar acids. In 1964, Aspen and Jakoby first isolated and characterized an NAD+-dependent l-threonate 3-dehydrogenase (Ltn3D; Enzyme Commission 1.1.1.129) from Pseudomonas sp. (J Biol Chem 239, 710–713), the molecular identity of which has remained unknown for over 60 years. Here, we have utilized bacterial genome context, together with biochemical and structural characterization, to reveal that GL300_RS07945 in Paracoccus litorisediminis encodes a representative NADP+-preferring Ltn3D. The crystal structure of the Michaelis ternary complex indicated that this enzyme is a member of the short-chain dehydrogenases/reductase superfamily, yet it differed in the recognition of the 2′-phosphate group of NADP+ between two adjacent arginine residues (Arg33 and Arg34). The C-3 atom of the competitive inhibitor tartronate was rationally positioned in close proximity to the nicotinamide ring for the catalysis. The reaction catalyzed by Ltn3D constitutes a distinct bypass route for the direct conversion of l-threonate to 3-oxo-l-threonate, which differs from the known sequential steps involving a dehydrogenase (l-threonate 2-dehydrogenase) and an isomerase (OtnI). In contrast to l-threonate 2-dehydrogenase, Ltn3D efficiently oxidized the 3-OH of homologous five- and six-carbon sugar acids, in addition to l-threonate. Among them, d-gluconate, potentially produced by GL300_RS07940 as a bifunctional 2-keto-d-gluconate/2-keto-l-gluconate reductase, could be converted to d-ribulose 5-phosphate by Ltn3D, followed by the action of a kinase (3OtnK) and a decarboxylase (3OtnC) in vitro. Altogether, our data suggest that Ltn3D constitutes a unique evolutionary innovation for the catabolism of four- to six-carbon sugar acids.

Figures

Full text

Advances in genome sequencing technology provide an opportunity to identify “classically reported enzyme activity” in cell-free extracts from organisms by using the modern molecular biological methods, and the corresponding gene may be contained among the number of hypothetical genes (proteins) within public databases. Although their functions are initially elucidated by comparing them with previously annotated proteins, this process frequently leads to misleading or erroneous annotations because of low sequence homology (1, 2, 3). In this regard, genome context and screening of a library of potential substrates can identify and/or correct the enzyme function and physiological meaning of a hypothetical (cryptic) protein. Here, we focus on the sugar (carbohydrate) metabolism by microorganisms, one of the most historical areas of biochemistry.

There are three main glycolytic routes in bacteria: the Embden–Meyerhof–Parnas, Entner–Doudoroff, and oxidative pentose-phosphate pathways (4). In the former two, six-carbon (C6) metabolic intermediates of fructose-l,6-bisphosphate and d-2-keto-3-deoxygluconate 6-phosphate (D-KDGlu 6P) are immediately converted to d-glyceraldehyde 3-phosphate (GAP) and dihydroxyacetone phosphate (DHAP) and to pyruvate and GAP, respectively, by different aldol-cleavage reactions. On the other hand, several hyperthermophilic and/or halophilic archaea possess a so-called nonphosphorylating Entner–Doudoroff pathway that ultimately yields pyruvate and d-glyceraldehyde (5). In the “semiphosphorylating” modification, the D-KDGlu intermediate is alternatively phosphorylated and aldol cleaved to yield the same products as the Entner–Doudoroff pathway. The former is promiscuous and represents an equivalent route for d-glucose and d-galactose metabolism in Sulfolobus solfataricus (6), and Haloferax volcanii metabolizes d-galactose through the DeLey–Doudoroff pathway, which is schematically homologous to the latter (7, 8).

On the other hand, bacterial catabolism of pentoses, including l-arabinose, d-xylose, and d-arabinose, may occur through four known pathways. The well-known phosphorylating pathway consists of isomerases, kinases, and epimerases and ultimately provides d-xylulose 5P, which is further metabolized in the pentose-phosphate pathway (9). The remaining three routes are partially homologous to the nonphosphorylating Entner–Doudoroff pathway from archaea, in which pentose is commonly converted to d- or l-2-keto-3-deoxypentonate through sequential reactions by aldose 1-dehydrogenase, sugar lactone hydrolase, and sugar acid dehydratase (AraC, XylC, and UxaA). The metabolic fates of the d- or l-2-keto-3-deoxypentonate intermediate are as follows (Fig. 1A): in the Dahms pathway (10), pyruvate and glycolaldehyde by unidentified aldolase; in the Weimberg pathway (11), α-ketoglutarate by dehydratase (AraD and XylD) and dehydrogenase (AraE and XylE) via α-ketoglutaric semialdehyde; in the Watanabe pathway (12), pyruvate and glycolate via 2,4-diketo-3-deoxypentonate by dehydrogenase (AraF and XylF) and hydrolase (AraG and XylG). Regarding “metabolic crosstalk,” the conversion from a monosaccharide to α-ketoglutaric semialdehyde and 2,4-diketo-3-deoxypentonate is catalyzed by different metabolic enzymes for each pentose sugar. On the other hand, chirality at C2, C3, and C4 is completely lost at these intermediates, and downstream steps are metabolized by a common aldehyde dehydrogenase (AraE/XylE) and hydrolase (AraG/XylG), respectively. Metabolic genes related to these nonphosphorylating pentose pathways often cluster together with the putative sugar transporter genes and transcriptional regulator gene on the bacterial genomes.Figure 1**Microbial metabolism of sugar acids.**A, metabolic pathways of C4 (l-threonate and d-erythronate) (purple), C5 (l-arabinonate, d-xylonate, d-arabinonate, and l-xylonate) (yellow green), and C6 sugar acids (d-galactonate and d-gluconate) (yellow and green, respectively). There are three metabolic fates of KDP intermediates through the Dahms, Weimberg, or Watanabe pathways. In the last half of the Entner–Doudoroff and DeLey–Doudoroff pathways, d-KDGlu and d-KDGal intermediates are commonly converted to pyruvate and GAP. In the four-step route of the l-threonate pathway, a focus of the present study, l-threonate 3-dehydrogenase (Ltn3D) catalyzes the oxidization of l-threonate to 3-oxo-l-threonate (red), which differs from Ltn2D (and OtnI) (right panel) and contributes to cometabolizing other C5/C6 sugar acids (red broken line). A bifunctional 2-keto-d-gluconate/2-keto-l-gluconate reductase (2KDR; pink) is significantly related to d-gluconate metabolism. Asterisks indicate sugar acids related to vitamin C degradation. B, gene clusters related to C4/C5/C6 sugar acid metabolism from Paracoccus litorisediminis and Roseovarius sp. CAU 1744. Gene components are indicated in the same color and correspond to A. Genes in gray are putative transcriptional regulators and/or transporters. Genes from P. litorisediminis with stars were functionally characterized in this study. C, growth curves of P. litorisediminis on monosaccharides (left) and sugar acids (right) as sole carbon sources (20 mM).

Although the utilization of a C4 sugar acid (tetronate), including d- and l-threonate and erythronate, has long been overlooked, the degradation of these carbon sources from bacteria was recently reported at the molecular level (Fig. 1A). In the C4 sugar acid pathways, the focus of the present study, l-threonate and d-erythronate are commonly converted to 2-oxo-tetronate by l-threonate 2-dehydrogenase (Ltn2D) and d-erythronate 2-dehydrogenase, respectively. In the first and recently discovered routes involving “4-steps” (13) and “3-steps” (14), the 2-oxo-tetronate intermediate is metabolized to DHAP through three and two sequential steps by 2-oxo-tetronate isomerase (OtnI), 3-oxo-tetronate kinase (3OtnK), and 3-oxo-tetronate 4P decarboxylase (3OtnC) and by 2-oxo-tetronate kinase (2OtnK) and 2-oxo-tetronate 4P decarboxylase (2OtnC), respectively. Although these metabolic genes are also clustered on the bacterial genome, the genetic locus is considered to be separate from the C5/C6 sugar and/or sugar acid gene cluster(s), and there is no metabolic crosstalk between these carbon sources. On the other hand, in 1964, Aspen and Jakoby (15) found NAD^+^-dependent oxidization activity of l-threonate in Pseudomonas sp. cells grown on l-threonate as a sole carbon source. The corresponding enzyme was partially purified and given Enzyme Commission [EC] number, EC 1.1.1.129, as an l-threonate 3-dehydrogenase (Ltn3D), whereas its molecular identity, biochemical properties, and physiological role have remained unknown for over 60 years.

In the present study, we detected a hypothetical protein from the marine bacterium Paracoccus litorisediminis, locus GL300_RS07945, in an unusual gene cluster related to C4/C5/C6 sugar acid metabolism, which functions as an NAD(P)^+^-dependent Ltn3D that directly converts l-threonate to 3-oxo-tetronate (3-oxo-l-threonate). Biochemical and crystallographic analyses revealed a unique mechanism for recognizing coenzymes and substrates in the short-chain dehydrogenase/reductase (SDR) superfamily. The promiscuous substrate specificity and characterization of the neighborhood gene (protein) suggest that this enzyme significantly contributes to the metabolism of other C5/C6 sugar acids, particularly d-gluconate, through the C4 sugar acid pathway.

Results

Unique gene cluster related to C4/C5/C6 sugar acid metabolism

We performed a homology search against bacterial genome sequences using metabolic genes involved in several nonphosphorylating sugar (and sugar acid) pathways as a probe. As a result, we focused on GL300_RS07850-08000 genes from the marine bacterium P. litorisediminis and homologous gene clusters from other bacteria, including Roseovarius sp. CAU 1744, in which several genes for C4/C5/C6 sugar acid metabolism were present (Figs. 1B and S1, and Supplementary discussion). Briefly, among two known C4 sugar acid pathways, only the four-step route for l-threonate and d-erythronate was found (13). No corresponding gene to aldose 1-dehydrogenase was mostly present, and C5 and C6 sugar acids might alternatively enter into the Weimberg/Watanabe and Entner–Doudoroff/DeLey–Doudoroff pathways, respectively.

When P. litorisediminis was cultivated in artificial seawater, significant growth was observed on not only monosaccharides (d-glucose, d-galactose, l-arabinose, and d-xylose) but also sugar acids (d-gluconate, d-galactonate, l-arabinonate, d-galactonate, l-threonate, and d-erythronate but not d-xylonate) as a sole carbon source (20 mM) (Fig. 1C), conforming to the hypothetical pathways expected from the GL300_RS07850 to 08000 gene cluster. Collectively, this type of gene cluster may play a role as a “metabolic funnel” for sugar acids (but not monosaccharides) (Fig. 1A); however, it was unclear why metabolic genes for the C4 sugar acid pathway clustered with those for the C6/C5 sugar acid pathways.

Functional characterization of GL300_RS07945

It is important to note that the homologous gene to GL300_RS07945 was mostly present in the C6/C5/C4 sugar acid gene cluster (red in Fig. S1). This protein belongs to the SDR superfamily, in which several metabolic enzymes related to nonphosphorylating sugar (and sugar acid) pathways are contained (yellow green in Fig. 2) (16, 17, 18, 19, 20, 21, 22, 23). However, there is no functional member (including them) with less than 30% sequence similarity to it. Based on the genome context, as described above, four compounds categorized into monosaccharides, sugar acids, 2-keto-3-deoxysugar acids, and aldehydes were screened as a potential substrate(s) for the recombinant GL300_RS07945 protein (Fig. S2A). Among them, dual oxidization activity for NAD^+^ and NADP^+^ at optimum pH 10 (using a glycine–NaOH buffer) was observed for several sugar acids with the configurations of the C2(R) and C3(S) hydroxyl groups, namely, l-threonate (C4), l-arabinonate (C5), d-xylonate (C5), d-gluconate (C6), d-galactonate (C6), and d-fuconate (C6); l-altronate (C6) and l-idonate (C6) were not tested (Fig. 3A). No activity was observed for l-tartrate, which corresponds to l-threonate with the C4 hydroxyl methyl group replaced by a carboxyl group. The ratio of NADP^+^ to NAD^+^ in the kcat/Km value for l-threonate was 3.1 (Fig. 4A and Table 1). The kcat/Km value for d-glycerate (C3) with only the C2(R) hydroxyl group was markedly reduced by two orders of magnitude from C4/C5/C6 sugar acid substrates, which was attributed to the extremely low kcat value. On the other hand, GL300_RS07915, a homolog to Ltn2D (Fig. 1B), exhibited strict specificities for l-threonate and NAD^+^ as well as the known enzyme (Table 1) (13); there are two different dehydrogenases for l-threonate within a single gene cluster for l-threonate metabolism. The enzyme function of Ltn2D (and 3OtnK and 3OtnC) was identified by both the coupled-enzyme spectrophotometric assay in vitro (see below) and the phenotype of the gene disruptant in vivo (13).Figure 2Phylogenetic analysis of Ltn3D. All SDR members have been functionally and structurally characterized. The numbers at the nodes are bootstrapping values according to neighbor joining (1000 replicates, shown value: %) for reliability of the different groups. The scale bar shows amino acid substitutions per site. In the Ltn3D subfamily (red), yellow branches correspond to the species from the Paracoccus genus. Numbers in circles correspond to list nos. 1 to 24 in Figure 6B, and colored members are discussed in the text. Despite their similar structural folding, the overall phylogenetic tree of the SDR superfamily consists of long branch lengths because of the low sequence similarity. Ltn3D, l-threonate 3-dehydrogenase; SDR, short-chain dehydrogenase/reductase.Figure 3**Configurations of substrates.A, substrates of GL300_RS07945. The gray-shaded configuration is identical to that of l-threonate. No activity was observed for l-tartrate. B, potential reaction products through sequential reactions by Ltn3D (red), 3OtnK (blue), and 3OtnC (green). The lower panel indicates a schematic conversion of l-threonate to DHAP; the decarboxylated carbon is colored in yellow. Gray-colored routes are not operative in Paracoccus litorisediminis. C, comparisons of two potential oxidization products from l-threonate with tartronate and hydroxypyruvate. The gray-shaded configuration is identical to the latter. Hydroxypyruvate is also identical with the C2-oxidized form of d-glycerate, whereas NADPH-dependent reduction activity was observed in d-glyceraldehyde, which corresponds to the C3-oxidized form. D, stereoselective reduction at the (gray-shaded) C2 positions of 2-keto-d-gluconate and 2-keto-l-gulonate. Only the 2(R) hydroxyl group is yielded by GL300_RS07940 (2KGR) (black line). DHAP, dihydroxyacetone phosphate; 2KGR, 2-keto-l-gulonate reductase; Ltn3D, l-threonate 3-dehydrogenase; 3OtnC, 3-oxo-tetronate 4P decarboxylase; 3OtnK, 3-oxo-tetronate kinase.Figure 4Biochemical characterization of GL300_RS07945 as an Ltn3D.**A, kinetics for l-threonate under NAD(P)^+^. The drawn curve is a fit of the data points to the Michaelis–Menten equation. The determined Km, kcat, and Vmax values were shown in the box. Values are the means ± SD, n = 3. B, Michaelis–Menten (left) and Lineweaver–Burk (right) competitive inhibition by tartronate. l-threonate concentrations were changed with fixed concentrations of tartronate, which are indicated. Right, the inset is Km values versus [tartronate]. C, HPLC analysis of sequential reactions by Ltn3D and 3OtnK from l-threonate using NAD^+^(left) or NADP^+^ (right) as a coenzyme. The elution profile of each component is shown in Fig. S3A. The star and asterisk correspond to peaks derived from 3-oxo-l-threonate 4P and glycine (buffer), respectively. D, HPLC analysis using l-arabinonate, d-galactonate, and d-gluconate as substrates. Ltn3D, l-threonate 3-dehydrogenase; 3OtnK, 3-oxo-tetronate kinase.Table 1. Kinetic parameters of GL300_RS07945 and GL300_RS07915SubstrateCoenzymeKm (mM)kcat (min^−1^)kcat/Km (min^−1^ mM^−1^)GL300_RS07945 (Ltn3D)l-ThreonateNADP^+^6.50 ± 0.29a198 ± 430.9 ± 0.3NAD^+^10.0 ± 1.3a98.9 ± 7.410.1 ± 0.3d-FuconateNADP^+^6.09 ± 0.85a489 ± 2980.9 ± 6.4l-ArabinonateNADP^+^9.98 ± 0.89a614 ± 3461.6 ± 2.0g-GluconateNADP^+^0.682 ± 0.059b49.4 ± 1.772.6 ± 3.8d-XylonateNADP^+^1.96 ± 0.06a104 ± 152.8 ± 1.2d-GalactonateNADP^+^50.8 ± 1.0a883 ± 1817.4 ± 0.1d-GlycerateNADP^+^6.01 ± 0.05a7.52 ± 0.261.25 ± 0.03d-GlyceraldehydeNADPH1.75 ± 0.05a0.359 ± 0.0070.206 ± 0.002GL300_RS07915 (Ltn2D)l-ThreonateNAD^+^0.232 ± 0.029b2060 ± 788940 ± 778NADP^+^10.8 ± 1.2a214 ± 1419.8 ± 0.8Values are the means ± SD, n = 3.aEight different substrate concentrations between 1 and 10 mM were used.bEight different substrate concentrations between 0.1 and 1 mM were used.

To elucidate the reaction product from l-threonate by GL300_RS07945, we attempted the following four approaches. Among active substrates of GL300_RS07945, the C2- and C3-oxidized forms of d-glycerate, hydroxypyruvate, and d-glyceraldehyde, respectively, are chemically stable (Fig. 3C); 2-keto-d-gluconate corresponds to the former of d-gluconate (Fig. 3D). Among them, only NADPH-dependent reduction activity toward d-glyceraldehyde was detected (Table 1). Furthermore, among tartronate and hydroxypyruvate, which are structural analogs of 3-oxo- and 2-oxo-l-threonate, respectively (Fig. 3C), only the former competitively inhibited activity, and the Ki value (55.5 μM) was markedly lower than the Km value for l-threonate (6.50 mM in Table 1 and Fig. 4B). Coupled-enzyme assays as the third and fourth approaches were performed by both HPLC and spectrophotometric analyses. In the HPLC analysis, when l-threonate was incubated with GL300_RS07945 and 3OtnK (GL300_RS02600) in the presence of NAD(P)^+^ and ATP, a novel peak appeared in a time-dependent manner, concomitant with decreases in NADP^+^ and ATP (Fig. 4C). Its retention time (∼12 min) was close to those of several phosphorylated compounds, such as d-KDGlu 6P, d-KDGal 6P, and GAP (Fig. S3A); the further addition of 3OtnC (GL300_RS02605) had no effect on the elution profile (Fig. S3B). In the spectrophotometric assay (13), the oxidization of l-threonate was monitored following the addition of the Ltn3D enzyme in the presence of ATP and NAD^+^ (Fig. S4A). The addition of 3OtnK and NADH-dependent glycerol 3P dehydrogenase (G3PDH) did not affect the increase observed in absorbance at 340 nm because of the reduction of NAD^+^ to NADH. The final addition of 3OtnC significantly decreased absorbance, which was caused by the oxidation of NADH by G3PDH (the reduction of DHAP to produce glycerol 3P). This phenomenon was not observed when d-galactonate was used as a substrate instead of l-threonate (its product was d-xylulose 5P) (Fig. S4B).

Collectively, these results strongly suggest that DHAP was produced from l-threonate through sequential reactions by GL300_RS07945, 3OtnK, and 3OtnC in the absence of OtnI, in which GL300_RS07945 functioned as Ltn3D, in contrast to the known Ltn2D. In this alternative three-step route, only 3-oxo-l-threonate was produced as a 3-oxo-tetronate (right panel in Fig. 1A). As described in the “Introduction” section, Ltn3D (EC 1.1.1.129) from Pseudomonas sp., reported by Aspen and Jakoby (15), exhibits strict specificities for NAD^+^ and l-threonate, whereas its sharp pH optimum (pH 10.8) is similar, compared with Ltn3D of P. litorisediminis (PlLtn3D).

Dihydroxyacetone as a potential reaction product by Ltn3D

Oxidization steps by 6-phosphogluconate dehydrogenase and l-gulonate dehydrogenase are homologous to Ltn3D, and their reaction products are spontaneously decarboxylated to d-ribulose 5P and l-xylulose, respectively (possibly via a 1,2-enediol intermediate) (Fig. S5, A and B) (24, 25). Aspen and Jakoby (15) identified the reaction product from l-threonate by Ltn3D as dihydroxyacetone (DHA). Furthermore, Luo and Huang (26) reported that d-ribonate was metabolized through a homologous pathway to the four-step route of the C4 sugar acid pathway, whereas the related gene cluster(s) commonly contained no (putative) decarboxylase gene. In consideration of these findings, one possibility is that Ltn3D catalyzed the decarboxylating oxidization reaction of l-threonate (Fig. S5C).

In the so-called oxidative glycerol pathway from bacteria (27), DHA is produced from glycerol by a dehydrogenase and is then converted to DHAP by a kinase (DhaK). P. litorisediminis was found to possess a homolog of the DhaK gene (GL300_RS19155; Fig. 1B), and biochemical characterization revealed that this protein significantly phosphorylated DHA but not 3OtnK (Table S1). When GL300_RS19155 was used in the coupled-enzyme spectrophotometric assay, the addition of G6PDH (slightly) decreased absorbance at 340 nm (Fig. S4C). Based on these results, we assumed that a trace amount of DHA appeared during the reaction of Ltn3D. A loss of growth on l-threonate in the 3OtnK gene disruptant of Ralstonia eutropha H16 indicates that the metabolic fate of 3-oxo-l-threonate was only phosphorylation to 3-oxo-l-threonate 4P by 3OtnK (13), whereas this bacterium possessed no homolog to the DhaK gene (data not shown). Therefore, a hypothetical alternative pathway by Ltn3D and DhaK may be specific for P. litorisediminis (Fig. 1A).

Overall structures of apo- and holo-forms of PlLtn3D

To elucidate the reaction mechanism of PlLtn3D, the crystal structure of the apo-form was elucidated by the molecular replacement method using the structure predicted by AlphaFold2 (28) as a search model and then refined with sufficient stereochemical quality to a resolution of 1.90 Å in the P1 space group with two molecules in the asymmetric unit (Protein Data Bank [PDB] ID: 9X6I) (Table 2). A tight homotetramer is described as a dimer of dimers with 222-point group symmetry mediated by three perpendicular twofold axes that are conventionally termed the P, Q, and R-axes (Fig. 5A), which correspond to a tetrameric species with ∼110 kDa, calculated by a gel-filtration chromatography analysis (theoretical monomer molecular weight of the [His]6-tagged protein: 27,334 Da) (Fig. S2B), as commonly reported for other SDR members (29, 30, 31). Outliers of the Ramachandran plot (0.43%; Table 2) are due to Tyr240, whose main-chain conformation is distorted by the preceding Pro139; these residues are not close to active sites (Fig. S6). However, its electron density is clearly visible, including surrounding residues, and a valid structural model has been constructed.Table 2. Data collection and crystallographic refinement statistics of PlLtn3DLigand-free formNADP^+^- and tartronate-bound formPDB code9X6I9XAXData collection Space groupP1P1 a, b, c (Å)69.763, 73.454, 141.28869.359, 72.552, 140.251 α, β, γ (°)83.15, 80.763, 83.81485.279, 82.573, 84.517 Wavelength (Å)1.000001.00000 Resolution range (Å)47.75–1.90 (1.93–1.90)48.29–2.08 (2.12–2.08) R_merge_0.101 (1.087)0.200 (0.538) R_meas_0.109 (1.180)0.226 (0.610) CC_1/2_0.999 (0.716)0.859 (0.314) I/σ I19.3 (2.2)16.3 (9.1) Completeness (%)97.9 (96.6)97.8 (96.6) Redundancy7.3 (6.7)5.8 (5.6)Refinement R/R_free_0.1687/0.20000.1687/0.2002 No. of atoms Protein20,73020,892 Ligands196672 Water18261599 B-factors (Å^2^) Protein30.1513.27 Ligands34.0313.46 Water35.0719.17 RMSD Bond lengths (Å)0.0080.004 Bond angles (°)0.950.75 Ramachandran plot Favored (%)98.5097.63 Allowed (%)1.071.94 Outliers (%)0.430.43Values in parentheses are for the highest-resolution shell.Figure 5**Overall crystal structure of PlLtn3D.**A, ribbon representation of the tetrameric assembly. Two subunits, A and D in the asymmetric unit are colored in green and blue, and the other two subunits, B and C, related by crystallographic twofold symmetry, are shown in yellow and green, respectively. Three perpendicular twofold axes (P-, Q-, and R-axes) that generate 222 point-group symmetry are also indicated. The inset photo is a crystal of PlLtn3D. B, ribbon diagram of subunits. α-Helices and β-strands are colored in green and yellow, respectively. C, superimposition of the apo-form (green) to the holo-form (cyan). A loop region containing α9 (Leu183-Lys192; red) in the small domain was close to the active site pocket. Bound NADP^+^ and tartronate are represented as a stick model. D, electron density of the Fo–Fc omit map contoured at 1.0 σ and models of NADP^+^ and tartronate. PlLtn3D, Ltn3D of Paracoccus litorisediminis.

PlLtn3D adopted the classical SDR fold with a highly conserved active site architecture comprising the Asn106–Ser136–Tyr149–Lys153 catalytic tetrad (Fig. 5B). Each subunit consisted of an N-terminal Rossmann fold with a central, parallel β-sheet (β7-β6-β5-β4-β1-β2-β3) sandwiched by α-helices on either side (α1-α2-α10 and α3-α5-α7-α8). A small C-terminal domain containing short α4 and α9 was slightly separated from the main body of the subunit. Comparisons with other structures in the PDB using a pairwise distance matrix-alignment (DALI) analysis showed that there were many close structural homologs from the SDR superfamily with diverse substrate specificities, whereas their sequence similarities were commonly less than 30%, as follows (green in Fig. 2): 2,4-dienoyl-CoA reductase from human (1W73; root mean square deviation from PlLtn3D of 1.9 Å over 284 Cα atoms; sequence identity of 21%) (32); hydroxysteroid 17-beta dehydrogenase 13 from dog (8G84; 2.6 Å over 222 Cα atoms; 25%) (33); biphenyl dehydrogenase from Pandoraea pnomenusa (3ZV3; 2.4 Å over 225 Cα atoms; 24%) (34); and carveol dehydrogenase from Mycobacterium avium (3PXX; 2.4 Å over 221 Cα atoms; 25%) (35). Furthermore, the root branch of the Ltn3D subfamily has a bootstrap value of 99.9%, indicating the independent subclass to other functional SDR members (Fig. 2).

To gain further insights into the reaction mechanism, crystals in the apo-form of PlLtn3D were soaked in a solution containing an inhibitor and coenzyme for a few minutes, and the structure was refined at a resolution of 2.08 Å (PDB ID: 9XAX) (Table 2) (Fig. 5C). Clear density in the Fo–Fc omit map in all chains allowed for the modeling of tartronate and NADP^+^ bound at the active site (Fig. 5D). In the same experimental procedure, attempts to trap the complex with a substrate or NAD^+^ were unsuccessful. In comparisons with the ligand-free form, a loop region between Leu183 and Lys192 in the small domain was close to the active site pocket (red region in Fig. 5C), which corresponded to the so-called “closed conformation,” often caused by a coenzyme and/or substrate binding in other SDR superfamily enzymes (21, 22, 23). In the case of PlLtn3D, there was no amino acid residue that directly interacted with a substrate in this region, as described below.

NADP+ binding mode of PlLtn3D

Dinucleotide binding involved interactions with the Rossmann fold domain, including a hydrophobic pocket to accommodate the adenosine ribosyl group and a polar ribose binding residue (Asp54), similar to other SDR members (Fig. 6A) (30). The region of Gly8-Gly9-Gly10-Ser11-Gly12-Val13-Gly14, located between the end of β1 and the start of α1, corresponded to the extended consensus sequence for coordinating the nucleotide diphosphate group: Gly-X_3_-Gly-X-Gly (violet-colored region). The side chains and main-chain nitrogen atoms of Arg33 and Arg34 formed salt bridges with the 2′-phosphate group of NADP^+^, whereas that of Ser14 formed a hydrogen bond with the 3′-hydroxyl group. These residues were tightly conserved among (putative) Ltn3D enzymes.Figure 6**Ligand binding mode of PlLtn3D.**A, recognition of the 2′-phosphate and 3′-hydroxyl groups of NADP^+^. Hydrogen bonds (with distance) are shown as a black broken line. The purple-colored region corresponds to the characteristic sequence, [Gly/Ala]-X_3_-Gly-X-Gly, in a typical dinucleotide binding Rossmann fold motif, corresponding to B. Numbers in circles (also in C and E) correspond to sites 1 to 5 in B. B, partial multiple sequence alignment of PlLtn3D (star) in the SDR superfamily. The GenBank accession number is shown in Figure 2. Values in parentheses indicate sequential homology with PlLtn3D. In the left panel, sites 1 to 4 are generally important for discriminating between NAD^+^ and NADP^+^ in SDR superfamily enzymes, and an arginine residue at site 5 (red) is specific for a few (RR) members. In the right panel, sites 6, 9, and 10 correspond to a motif of the catalytic triad, as well as other SDR superfamily enzymes, and each alanine mutant of sites 6 to 9 and 11 to 14 was constructed in a site-directed mutagenic study. Binding sites of tartronate (C) and binding models of d-gluconate and d-xylonate (D) of PlLtn3D. In D, tartronate is shown in a translucent pink model. Among the substrates tested, the interaction with the O7∗ atom of NADP^+^ (in box) is specific for d-gluconate and d-xylonate (Fig. S7). E, comparisons of the (putative) active sites and structural formulae of substrates among SDR members, including PlLtn3D (cyan), 6ZZQ (green), and 3O03 (gold). 6ZZQ, (R)-3-hydroxybutyrate dehydrogenase in complex with NAD^+^ and acetoacetate; 3O03, gluconate 5-dehydrogenase in complex with NADP^+^ and d-gluconate. Abstraction of the proton occurs at the yellow-shaded carbon atom. PlLtn3D, Ltn3D of Paracoccus litorisediminis; SDR, short-chain dehydrogenase/reductase.

Amino acid residues at equivalent positions to 10, 11, 32, and 33 in PlLtn3D (sites 1, 2, 3, and 4 in Figure 6B, respectively) are important for the coenzyme specificities of SDR members (typical members) (30). Hydrophilic residues, including mostly serine at site 1 or 2, often interact with the 3′-hydroxyl group of a coenzyme as well as PlLtn3D (yellow). NADP^+^-dependent enzymes possess an arginine residue at site 2 or 4 (cyan) but not an aspartate residue at site 3 (yellow green). Although the neighboring arginine residues at sites 4 and 5 were detected in a few SDR enzymes (nos. 12–16 in Figs. 2A [blue] and 6B; RR members), their coenzyme specificities show strict NADP^+^ dependence, which differs from PlLtn3D (36, 37, 38, 39, 40). Based on these insights, we constructed two serine mutants of PlLtn3D, R33S and R34S. In comparisons with the WT enzyme, no significant difference was observed in kinetic parameters for l-threonate (Table S2). On the other hand, their kcat/Km values for NADP^+^ decreased by 783- and 34-fold, respectively, because of the marked increase in Km values, by which the ratio of NADP^+^ to NAD^+^ changed from 280 to 2.2 and 36, respectively (Table 3). Furthermore, a large-scale phylogenetic analysis using the Protein-BLAST program revealed several (putative) Ltn3D enzymes with the pair of Arg-X at sites 4 and 5 (nos. 6 and 8–11 in Figs. 2A [cyan] and 6B). These results suggest that Arg33 is more important for the NADP^+^ preference (and catalysis) than Arg34.Table 3. Kinetic parameters of WT and mutants of PlLtn3D for NAD^+^ or NADP^+^EnzymeCoenzymeKm (mM)kcat (min^−1^)kcat/Km (min^−1^ mM^−1^)NADP^+^/NAD^+^ (fold)WTNADP^+^0.00334 ± 0.00010a196 ± 357,400 ± 1270280NAD^+^0.459 ± 0.054b98.6 ± 1.3205 ± 11R33SNADP^+^0.857 ± 0.031b63.4 ± 1.073.3 ± 0.72.2NAD^+^2.60 ± 0.25b85.9 ± 6.032.6 ± 0.4R34SNADP^+^0.0902 ± 0.0016b154 ± 41670 ± 3436NAD^+^1.41 ± 0.08b65.5 ± 2.146.0 ± 0.5Values are the means ± SD, n = 3.aSix different substrate concentrations between 0.00250 and 0.1 mM were used.bSix different substrate concentrations between 0.1 and 1 mM were used.

Tartronate binding mode of PlLtn3D

Tartronate is a prochiral molecule with two carboxyl groups (Fig. 3C). The distances between C3 of tartronate (but not C2) and C4∗ of the nicotinamide ring of NADP^+^ and the angle of N1∗–C4∗–C3 (3.1 Å and 109°, respectively) were consistent with those obtained from various structural analyses of the hydride-transferring enzyme in complex with NAD(P)^+^ (Fig. 6C). The side chains of Arg143 and Ser138 formed hydrogen bonds with the C1 carboxyl group, which corresponded to that of l-threonate. There was a salt bridge between Arg143 and Glu204. The C2 hydroxyl group formed hydrogen bonds with the side-chain and main-chain nitrogen atoms of Asn181, and the latter further interacted with the side chain of Arg143. The C3 carboxyl group of tartronate formed hydrogen bonds with the side chains of Ser136 and Tyr149, which were present in Ser136–Tyr149–Lys153, corresponding to a motif of the catalytic triad, as well as other SDR superfamily enzymes (sites 6, 9, and 10 in Fig. 6B) (29). Among two (hydrophobic) methionine residues (Met186 and Met238), the side chain of the latter was oriented toward the C1 carboxyl group; a distance of ∼3.8 Å.

Based on these results, we constructed alanine mutants for Ser136, Ser138, Arg143, Tyr149, Asn181, Met186, Glu204, and Met238 (sites 6–9 and 11–14, respectively). Among them, the R143A, Y149A, N181A, and E204A mutants were inactive; the S136A mutant exhibited markedly reduced activity, and Km value of the M238A mutant markedly increased, conforming to the structural insights obtained (Table 4). Despite the direct interaction with the C1 carboxyl group of tartronate (also the substrate), the S138A mutant was significantly active. A few (putative) Ltn3D enzymes from the other species in the Paracoccus genus natively possessed the homologous alanine residue (nos. 3 and 4 [and 9] in Figures 2B [violet] and 6B), suggesting the greater importance of Arg143 for recognizing the C1 carboxyl group of a substrate than Ser138.Table 4. Kinetic parameters for l-threonate by mutants of PlLtn3DEnzymeKm (mM)kcat (min^−1^)kcat/Km (min^−1^ mM^−1^)WT6.50 ± 0.29a198 ± 430.9 ± 0.3S136A6.16 ± 0.09a0.150 ± 0.0020.0243 ± 0.0001S138A6.36 ± 0.25a200 ± 531.1 ± 0.2R143ANDbNDNDY149ANDbNDNDN181ANDbNDNDM186A12.3 ± 0.4a269 ± 821.8 ± 0.1E204ANDbNDNDM238A125 ± 0.5c513 ± 34.10 ± 0.01ND, not determined.Values are the means ± SD, n = 3.aEight different substrate concentrations between 1 and 10 mM were used.bNot assessed because of extremely weak activity.cEight different substrate concentrations between 10 and 100 mM were used.

To elucidate the structural basis of substrate specificity in more detail, we constructed a model structure in complex with a substrate (l-threonine, l-arabinonate, d-xylonate, d-galactonate, or d-gluconate) (Figs. 6D and S7). The C2 hydroxyl group of tartronate without a chiral carbon was completely overlaid to the C2(R) hydroxyl group of the substrate. Among the substrates tested, the C6 and C5 hydroxyl groups of d-gluconate and d-xylonate, respectively, with low Km values (Table 1), interacted with the O7∗ atom of NADP^+^, and the distance to the former was closer than that to the latter: 2.1 and 2.5 Å, respectively. On the other hand, the C1 carboxyl group of d-galactonate (and l-arabinonate) separated from the side chain of Arg143, which was essential for catalysis (Table 4). Therefore, the structure of the holo-form significantly compensated to functionally identify GL300_RS07945 as an Ltn3D and may explain why d-gluconate and d-galactonate were the optimal and worst substrates, respectively (Table 1).

Comparison with other SDR enzymes

l-gulonate 3-dehydrogenase catalyzed the homologous oxidization reaction of the 3(S)-hydroxyl group with Ltn3D (Fig. S5B), although they have no evolutionary relationship (41). In the SDR superfamily, two enzymes utilize a hydroxyl acid as a substrate (Fig. 2). 3(R)-Hydroxybutyrate dehydrogenase (HBDH; orange) has been shown to catalyze the enantioselective reduction of 3-oxo to 3(R)-hydroxyl carboxylates (42). In the case of d-gluconate 5-dehydrogenase (IdnO; pink), its oxidization of d-gluconate occurs at the C5 position, but not at the C3 position (17). The superposition of the crystal structure of PlLtn3D in complex with NADP^+^ and tartronate to their holo-forms revealed that despite the structural similarity of the substrate, the C1 carboxyl group was not recognized by any homologous amino acid residues between Ltn3D, 3(R)-hydroxybutyrate dehydrogenase, and IdnO, and two catalytic serine and tyrosine residues were completely overlaid (Fig. 6E). Therefore, in the putative catalytic mechanism of Ltn3D, the C3 hydroxyl group of l-threonate was deprotonated by Tyr149 as a general base catalyst, with the concurrent transfer of the hydride ion to NADP^+^, and the pKa value of its hydroxyl group was reduced by Lysl53, leading to the stabilization of the tyrosinate anion at physiological pH (Fig. S8).

Physiological role of PlLtn3D

In a zymogram staining analysis using a crude extract prepared from P. litorisediminis cells grown on a marine broth medium supplemented with l-threonate, there were two active bands derived from dehydrogenase isozymes I and II for l-threonate (Fig. S9). Isozyme I clearly showed a preference for NAD^+^, whereas NADP^+^-dependent isozyme II with lower activity appeared in an upper range. The native molecular weight of Ltn2D using the purified recombinant enzyme was estimated to be ∼50 kDa by gel filtration (dimeric structure), which was markedly smaller than that of Ltn3D (Fig. S2B). These results indicate that isozymes I and II corresponded to Ltn2D and Ltn3D, and both enzymes were physiologically operative in l-threonate metabolism by P. litorisediminis.

Previous studies investigated metabolic crosstalk between l-threonate and other sugar acid(s). Gerlt et al. reported that the Δ3OtnK gene mutant of Pectobacterium atrosepticum strain SCRI 1043 was unable to grow with l-threonate, whereas growth on d-gluconate was similar to the WT strain (13). On the other hand, the same 3OtnK gene was found to be necessary for d-gluconate utilization as a carbon source in P. atrosepticum strain WPP14 (43). Furthermore, Zhang et al. (44) demonstrated that Ltn2D from Agrobacterium radiobacter K84 (Arad_9439) utilized not only l-threonate but also l-idonate, and its product, 2-oxo-l-idonate, was subsequently isomerized and phosphorylated by OtnI and 3OtnK in vitro; 3OtnC may further decarboxylate 3-oxo-l-idonate 4P to l-xylulose 5P, which subsequently enters into the pentose-phosphate pathway as a d-xylulose 5P via l-ribulose 5P (Fig. 3B) (45). Regarding the genome context, gene clusters from P. atrosepticum SCRI 1043 and A. radiobacter K84 are not linked to the metabolism of other sugar acids, which differs from the present results.

The HPLC analysis revealed that d-gluconate (as well as d-galactonate and l-arabinonate) was converted to 3-oxo-d-gluconate by Ltn3D and 3OtnK (Fig. 4D). Since the G3PDH-coupled spectrophotometric assay was only used for l-threonate (DHAP), we newly prepared NADPH-dependent d-ribulose 5P reductase (R5PR) from Staphylococcus aureus (46), which catalyzes the reduction of d-ribulose 5P to d-ribitol 5P. When 3OtnC was added during the NADP^+^-linked Ltn3D/3OtnK-catalyzed reaction for d-gluconate in the copresence of R5PR, we observed the immediate oxidation of NADPH (the reduction of d-ribulose 5P) (Fig. S4D). Collectively, the genome context and the results of the biochemical analysis indicate that d-gluconate physiologically entered into the four-step route of the l-threonate pathway as d-ribulose 5P (Fig. 3B), conforming to its highest affinity of PlLtn3D among the C5/C6 sugar acid substrates (Table 1).

In the so-called tetritol (C4 polyol) catabolic pathway(s) from Mycobacterium smegmatis (47), l-erythrulose 4P is converted into d-erythrose 4P via d-erythrulose 4P by l-erythrulose 4P epimerase (LerI) and d-erythrulose 4P isomerase (DerI2) to enter into the pentose-phosphate pathway (Fig. 3B). P. litorisediminis possesses no homolog of the LerI and DerI genes, conforming to no growth on d-xylonate (Fig. 1C); l-arabinonate is alternatively metabolized through the Weimberg pathway and/or the Watanabe pathway (Fig. 1A).

Characterization of GL300_RS07940

A homologous gene to GL300_RS07940 was mostly contained within the C4/C5/C6 sugar acid gene cluster as well as Ltn3D (pink in Fig. S1) and showed ∼45% sequence identity with NAD(P)^+^-dependent l-2-hydroxyacid dehydrogenase from Ketogulonicigenium vulgare Y25 (48). This bacterium is useful for the bioconversion of 2-keto-l-gulonate from l-sorbose in the manufacture of vitamin C, and 2-keto-l-gulonate is unfavorably converted to l-idonate, which may be attributed to the function of l-2-hydroxyacid dehydrogenase as a 2-keto-l-gulonate reductase (Kv2KGR) (Fig. 3D). To elucidate its physiological role in the C4/C5/C6 sugar acid pathway(s), we biochemically characterized GL300_RS07940 from P. litorisediminis using the purified recombinant protein (Fig. S2A).

This protein exhibited NADPH-dependent reduction activity for 2-keto-l-gulonate, and no significant difference was observed in kcat/Km values between 2-keto-l-gulonate and 2-keto-d-gluconate: 376 and 654 min^−1^ mM^−1^, respectively (Table 5). On the other hand, the coenzyme specificity of Pl2KGR strictly exhibited a preference for NADPH, which differed from Kv2KGR (48). A pair of Asp175–Ile176 in Kv2KGR for recognizing NAD^+^ was substituted to Ser173–Arg174, which favorably interacted with the 2′-phosphate group of NADP^+^ in Pl2KGR. In the reverse oxidization reaction (Fig. 3D), the kcat/Km value for d-gluconate was 31-fold higher than that for d-mannonate (and also l-gulonate). Furthermore, d-galactonate (C5) and d-fuconate (C6) with the configurations of C2 and C3 as d-gluconate were also substrates but not l-threonate (C4).Table 5. Kinetic parameters of GL300_RS07940SubstrateCoenzymeKm (mM)kcat (min^−1^)kcat/Km (min^−1^ mM^−1^)2-Keto-d-gluconateNADPH13.9 ± 0.8a9020 ± 357654 ± 6NADH15.2 ± 0.6a170 ± 511.2 ± 0.1d-GluconateNADP^+^1.80 ± 0.20b177 ± 14100 ± 2d-MannonateNADP^+^21.6 ± 1.5a73.1 ± 3.73.41 ± 0.032-Keto-l-gulonateNADPH19.9 ± 1.5a7400 ± 407376 ± 4l-IdonateNADP^+^NDcNDNDl-GulonateNADP^+^4.26 ± 0.16a1.47 ± 0.020.350 ± 0.004l-ThreonateNADP^+^NDdNDNDd-GalactonateNADP^+^4.16 ± 0.10a60.9 ± 1.314.6 ± 0.1d-FuconateNADP^+^3.47 ± 0.24a94.4 ± 3.526.8 ± 0.5ND, not determined.Values are the means ± SD, n = 3.aEight different substrate concentrations between 1 and 10 mM were used.bEight different substrate concentrations between 0.1 and 1 mM were used.cNot assessed because of commercial unavailability.dNot assessed because of extremely weak activity.

The C4/C5/C6 sugar acid gene clusters often contained not only 2KGR but also l-idonate 5-dehydogenase (IdnD) and 5-keto-d-gluconate reductase (IdnO) (49) (Fig. S1). Their reversible oxidoreductive reactions must be enabled to further diversify the metabolism of not only d-gluconate but also 2-keto-l-gulonate, 2-keto-d-gluconate, 5-keto-d-gluconate, and l-idonate (Fig. 1A). Among them, P. litorisediminis showed significant growth on 2-keto-d-gluconate as the sole carbon source (Fig. 1C), which is potentially produced from 2,5-diketo-d-gluconate by GL300_RS19150 (Fig. 1B). Collectively, these results further strengthen the hypothesis of a metabolic link between l-threonate and d-gluconate.

Discussion

The SDR superfamily is one of the largest protein groups, consisting of ∼170,000 primary structures and ∼550 three-dimensional structures (29, 30, 31). Among them, the crystal structure of a hypothetical protein from Xanthobacter autotrophicus, deposited in 2012, shows ∼57% sequence identity with GL300_RS07945 (4DYV; brown in Fig. 2). However, the corresponding gene is not associated with any notable gene cluster. Therefore, PlLtn3D is a typical example of the advantage of the genome context for correctly estimating the potential substrate of a hypothetical protein.

The known C4 sugar acid pathways via a 2-oxo-tetranate intermediate enable the metabolism of both l-threonate and d-erythronate but not any other sugar acid (Fig. 1A). Therefore, as expected from the genome context, the substrate promiscuity of Ltn3D must first be demonstrated by combining gene clusters between the C4 and C5/C6 sugar acid pathways (Fig. S1). The degradation of C5/C6 sugar acids by Ltn3D, 3OtnK, and 3OtnC resulted in their direct entry into the pentose-phosphate pathway but not glycolysis and/or the tricarboxylic acid cycle (Fig. 3B).

One of the origins of d-gluconate and ketogluconates in nature is the incomplete oxidation of d-glucose by membrane-bound dehydrogenases from microorganisms (50). On the other hand, among them, l-idonate and 5-keto-d-gluconate are also intermediates in biosynthesis from vitamin C to l-tartrate in higher plants (51), and l-threonate and l-xylonate are also found in the (chemical) degradation of vitamin C (stars in Fig. 1A) (52). Therefore, these common origins of the C6, C5, and C4 sugar acids may be of significant evolutionary merit for the formation of a single gene cluster.

The remaining question is why there is only one bypass route of l-threonate by Ltn3D. ApnO–OiaK–OiaX genes, related to the degradation of d-apionate (53), are located near the C6/C5/C4 sugar acid gene clusters on the genomes of a few bacteria, including Roseovarius sp. CAU 1744 (Figs. 1B and S1). d-apionate is enzymatically produced from d-apiose, a uniquely branched pentose in the cell walls of higher plants. Among these metabolic enzymes, ApnO (d-apionate oxidoisomerase; EC 1.1.1.421) catalyzes the oxidization of the C2 hydroxyl group of d-apionate and the migration of a hydroxymethyl group to yield 3-oxo-d-apionate. It is important to note that this reaction for d-erythronate was effectively identical to that by (hypothetical) d-erythronate 3-dehydrogenase (Fig. S10). A biochemical analysis and elucidation of the crystal structure of ApnO are currently in progress.

Experimental procedures

General procedures

P. litorisediminis NBRC 112902 was purchased from the National Institute of Technology and Evaluation. Bacterial genomic DNA was prepared using a DNeasy Tissue Kit (Qiagen). PCR was performed using GeneAmp PCR System 2700 (Applied Biosystems) for 30 cycles in 50 μl of a reaction mixture containing 1 unit of KOD One DNA polymerase (Toyobo), appropriate primers (15 pmol), and template DNA under the following conditions: denaturation at 98 °C for 10 s, annealing at 55 °C for 5 s, and extension at 68 °C for the periods calculated as an extension rate of 1 kbp/s. All sugar acids, except for l-threonate (Ca^2+^ salt; Tokyo Chemical Industry), d-erythronate (K^+^ salt; Combi-Blocks), and l-gulonate (Li^2+^ salt; Sigma–Aldrich), were prepared by hypoiodite-in-methanol oxidization from the corresponding sugars as K^+^ or Ba^2+^ salts (54) and were purified using the column of AG 1-X8 Resin (200–400 mesh, formate form) (Bio-Rad). 2-Keto-3-deoxysugar acids were enzymatically synthesized from appropriate sugar acids (12, 22).

Plasmid construction

The primer sequences used in the present study are shown in Table S3. The GL300_RS07945 (Ltn3D), GL300_RS07915 (Ltn2D), GL300_RS02600 (3OtnK), GL300_RS02605 (3OtnC), GL300_RS07940 (2KGR), and GL300_RS19155 (DhaK) genes were amplified by PCR using primers containing appropriate restriction enzyme sites at the 5′- and 3′-ends and the genomic DNA of P. litorisediminis as a template. Each amplified DNA fragment was introduced into the BamHI–HindIII or BamHI–PstI site in pQE-80L (Qiagen), by which 11 additional residues (MRGSHHHHHHG) were attached at the N terminus. A site-directed mutation was introduced into the GL300_RS07945 gene by a single round of PCR with sense and antisense primers and the cloned pQE-80L plasmid as a template. Mutant proteins were expressed and purified by the same procedures as the WT enzyme.

Protein overexpression and purification

Recombinant Escherichia coli DH5α cells harboring the constructed plasmids were grown at 37 °C in LB medium containing ampicillin (50 mg/l). When an absorbance at 600 nm reached approximately 0.6, isopropyl-β-d-thiogalactopyranoside was added at a final concentration of 1 mM and cultured at 25 °C for 18 h to induce the expression of the respective (His)6-tagged protein. Cells were harvested, resuspended in buffer A (50 mM sodium phosphate buffer [pH 8.0] containing 300 mM NaCl and 10 mM imidazole), disrupted by sonication on ice using an Ultra Sonic Disruptor UD-211 (Tomy Seiko), and centrifuged to pellet the insoluble debris. The supernatant was loaded onto a nickel–nitrilotriacetic acid Superflow column (Qiagen) equilibrated with buffer A. The column was washed with buffer B (50 mM sodium phosphate buffer [pH 8.0] containing 300 mM NaCl, 10% [v/v] glycerol, and 25 mM imidazole). The enzymes were then eluted with buffer C (pH 8.0, buffer B containing 250 mM instead of 25 mM imidazole). Regarding the crystallization of GL300_RS07945, the proteins were further purified through size-exclusion chromatography using a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) with buffer D (20 mM Tris–HCl [pH 8.0] containing 150 mM NaCl). The main single-peak fractions were collected and concentrated by ultrafiltration with Amicon Ultra-15 (Millipore).

The native molecular mass of recombinant proteins was estimated by gel filtration, which was conducted using an HPLC system at a flow rate of 1 ml/min. The purified enzyme was loaded onto a HiLoad 16/600 Superdex 200 pg column equilibrated with buffer D. A high–molecular-weight gel filtration calibration kit (GE Healthcare) was used with the following molecular markers: thyroglobulin, 669 kDa; ferritin, 440 kDa; aldolase, 158 kDa; conalbumin, 75 kDa; and ovalbumin, 44 kDa.

Crystallization and X-ray crystallography

Crystallization of the purified PlLtn3D protein was initially performed with Index and Crystal Screen HT (Hampton Research) using the sitting-drop vapor diffusion method at 20 °C. In this method, drops (0.5 μl) of ∼20 mg/ml protein in buffer D were mixed with equal amounts of the reservoir solution (as follows) and equilibrated against 70 μl of the same reservoir solution by vapor diffusion: 0.2 M Li_2_SO_4_·H_2_O, 0.1 M Hepes–NaOH (pH 7.5), and 25% (w/v) PEG 3350. Regarding the crystallization of PlLtn3D in complex with the coenzyme and inhibitor, crystals of the apo-form were soaked in the reservoir solution containing 1.5 mM NADP^+^ and 1 mM tartronate (Thermo Scientific). Crystals were soaked in the reservoir solution supplemented with 30% (w/v) PEG 3350, mounted on a nylon loop, flash-cooled, and kept in a stream of nitrogen gas at 100 K during data collection.

Diffraction data were collected with the PILATUS 6M detector of BL45XU at SPring-8 as well as the processed ZOO system and XDS (55, 56, 57). The structure was elucidated using the molecular-replacement pipeline program BALBES (58) with the structure predicted by AlphaFold2 (28) as the search model. Further model building for structures was manually performed with COOT (59) and crystallographic refinement with PHENIX (60). Detailed data collection and processing statistics are shown in Table 2.

Molecular docking

Molecular docking was conducted using the OEDocking suite (version 4.3.2.1) (https://docs.eyesopen.com/applications/oedocking/index.html). The crystal structure of PlLtn3D in complex with NADP^+^ and tartronate was processed with the Make_Receptor module to generate the receptor input. The three-dimensional structures of substrates were retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov) and subjected to conformer generation using OMEGA (version 6.0.0.1) in the “pose” mode (https://www.eyesopen.com/omega) (61). The resulting conformers were then docked into the active site using the POSIT module (62).

Enzyme assay

Enzyme assays were performed with a Shimadzu UV-1800 spectrophotometer (Shimadzu GLC Ltd). The consumption or formation of NAD(P)H was monitored as the decrease or increase in absorbance at 340 nm with an extinction coefficient (ε) of 6220 M^−1^ cm^−1^. Km and kcat (=Vmax/[E]t) values were calculated by directly fitting the Michaelis–Menten equation, vo/[E]t = kcat[S]/(Km + [S]), to the primary data plotted as [S] versus v0/[E]t (Fig. S11) using GraphPad Prism software (version 10.6.1).

GL300_RS07945: The activity of Ltn3D was routinely assayed in 50 mM glycine–NaOH buffer (pH 10) containing 10 mM l-threonate and 1.5 mM NAD(P)^+^ with a final reaction volume of 1 ml.

To estimate the potential enzymatic product, a spectrophotometric investigation of the reaction sequence (13) was conducted. The initial solution (1 ml) contained 50 mM glycine–NaOH buffer (pH 10), 1.5 mM l-threonate, 1.5 mM ATP, 1 mM MgCl_2_, and 3OtnK (GL300_RS02600; 10 μg). The reaction was started by the addition of 15 mM NAD^+^ (100 μl), and absorption increased with the production of NADH. After the reaction reached equilibrium, NAD(H)-dependent G3PDH (Oriental Yeast; 1 unit) and 3OtnC (GL300_RS02605; 10 μg) were added. To detect the product from d-gluconate, NADP^+^ and R5PR were used in the reaction solution containing 1 mM ZnSO_4_ instead of NAD^+^ and G3PDH, respectively. The R5PR gene (TarJ) was amplified by PCR using the genomic DNA of S. aureus and synthetic DNA primers and was then introduced into the BamHI–PstI site in pQE-80L. The (His)6-tagged recombinant 3OtnK, 3OtnC, and R5PR (and Ltn2D, DhaK, and 2KGR) proteins were expressed in E. coli DH5α cells and purified through the same procedures as PlLtn3D.

Alternatively, the continuous reaction(s) through GL300_RS07945, 3OtnK, and/or 3OtnC was analyzed at 35 °C on an Aminex HPX-87H Organic Analysis column (300 × 7.8 mm; Bio-Rad) linked to an RID-8020 refractive index detector (Tosoh) and eluted with 5 mM H_2_SO_4_ at a flow rate of 0.6 ml/min. The reaction mixture consisted of 50 mM glycine–NaOH buffer (pH 10), 1 mM l-threonate, 2 mM ATP, 1 mM MgCl_2_, and 2.5 mM NAD(P)^+^.

GL300_RS07915: The activity of Ltn2D was routinely assayed in 50 mM Tris–HCl buffer (pH 8) containing 10 mM l-threonate and 1.5 mM NAD^+^.

GL300_RS19155 and GL300_RS02600: The activity of DhaK was coupled with pyruvate kinase and lactate dehydrogenase reactions using F-Kit glycerol (Roche) in accordance with the manufacturer’s instructions.

GL300_RS07940: The activity of 2KGR was assayed in 50 mM Tris–HCl buffer (pH 8) containing 10 mM l-threonate and 1.5 mM NADP^+^. In the reverse direction, the enzyme was measured in 50 mM Hepes–NaOH buffer (pH 7.2) containing 10 mM 2-keto-l-gulonate (Santa Cruz Biotechnology) and 0.15 mM NADPH.

Bacterial strain, culture conditions, and preparation of cell-free extracts

P. litorisediminis was cultured aerobically with vigorous shaking at 30 °C in marine broth or artificial seawater supplemented with 20 mM of a carbon source. In the zymogram staining analysis, P. litorisediminis was alternatively cultured in marine broth medium containing l-threonate (20 mM). Grown cells were harvested by centrifugation and washed with buffer D. Washed cells were resuspended in buffer D, disrupted by sonication, and then centrifuged. The cell-free extract obtained was separated on a 5% to 20% gradient native PAGE gel (ATTO) at 4 °C. The gels were then soaked in 10 ml of staining solution consisting of 50 mM glycine–NaOH (pH 10), 50 mM l-threonate, 0.25 mM p-iodonitrotetrazolium violet, 0.06 mM phenazine methosulfate, and 1.5 mM NAD(P)^+^ at room temperature. Dehydrogenase activity appeared as a dark red band.

Phylogenetic analysis

Protein sequences were analyzed using the Protein-BLAST and Clustal W programs distributed by the Kyoto Encyclopedia of Genes and Genomes of Japan (https://www.genome.jp/kegg/) (63). An unrooted fast maximum likelihood–based phylogenetic tree of SDR superfamily including Ltn3D was displayed using TreeView X.

Data availability

Atomic coordinates and structural factor files have been deposited in the PDB under accession codes 9X6I (the ligand-free form) and 9XAX (the NADP^+^- and tartronate-bound form).

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.