Characterization of l-2-keto-3-deoxyfuconate aldolases in a nonphosphorylating l-fucose metabolism pathway in anaerobic bacteria

Abstract

The genetic context in bacterial genomes and screening for potential substrates can help identify the biochemical functions of bacterial enzymes. The Gram-negative, strictly anaerobic bacterium Veillonella ratti possesses a gene cluster that appears to be related to l-fucose metabolism and contains a putative dihydrodipicolinate synthase/N-acetylneuraminate lyase protein (FucH). Here, screening of a library of 2-keto-3-deoxysugar acids with this protein and biochemical characterization of neighboring genes revealed that this gene cluster encodes enzymes in a previously unknown “route I” nonphosphorylating l-fucose pathway. Previous studies of other aldolases in the dihydrodipicolinate synthase/N-acetylneuraminate lyase protein superfamily used only limited numbers of compounds, and the approach reported here enabled elucidation of the substrate specificities and stereochemical selectivities of these aldolases and comparison of them with those of FucH. According to the aldol cleavage reaction, the aldolases were specific for (R)- and (S)-stereospecific groups at the C4 position of 2-keto-3-deoxysugar acid but had no structural specificity or preference of methyl groups at the C5 and C6 positions, respectively. This categorization corresponded to the (Re)- or (Si)-facial selectivity of the pyruvate enamine on the (glycer)aldehyde carbonyl in the aldol-condensation reaction. These properties are commonly determined by whether a serine or threonine residue is positioned at the equivalent position close to the active site(s), and site-directed mutagenesis markedly modified C4-OH preference and selective formation of a diastereomer. I propose that substrate specificity of 2-keto-3-deoxysugar acid aldolases was convergently acquired during evolution and report the discovery of another l-2-keto-3-deoxyfuconate aldolase involved in the same nonphosphorylating l-fucose pathway in Campylobacter jejuni.

Introduction

Microorganisms, including bacteria, archaea, yeasts, and fungi, utilize not only hexoses but also pentoses (d-xylose, l-arabinose, and d-arabinose) and 6-deoxyhexose sugars (l-rhamnose and l-fucose) as their sole carbon source. The metabolism of these sugars is classified into two pathways with or without phosphorylated intermediates. The former pathways consisting of isomerases (EC 5.3.1.-), kinases (EC 2.7.1.-), epimerases (EC 5.1.3.-), and/or aldolase (EC 4.1.2.-) have been extensively examined (such as l-fucose; Fig. 1A) (1). The latter pathways are (partially) analogous to the nonphosphorylative Entner–Doudoroff pathway from archaea (2) and have been classified into three pathways, in which the sugar is commonly converted into a 2-keto-3-deoxysugar acid intermediate through the participation of aldose 1-dehydrogenase (EC 1.1.1.-), sugar lactonase (EC 3.1.1.-), and sugar acid dehydratase (EC 4.2.1.-). In the “route I” pathway (such as l-rhamnose; Fig. 1D), the 2-keto-3-deoxysugar acid intermediate is cleaved through an aldolase reaction into the appropriate aldehyde and pyruvate, which is completely homologous to the nonphosphorylative Entner–Doudoroff pathway from archaea. The “route II” pathway corresponds to an alternative pathway of pentoses, in which the 2-keto-3-deoxypentonate (KDP)² intermediate is converted to α-ketoglutarate via α-ketoglutaric semialdehyde by KDP dehydratase (EC 4.2.1.141(43)) and α-ketoglutaric semialdehyde dehydrogenase (EC 1.2.1.26) (3, 4). On the other hand, in the “route III” pathway (such as l-fucose; Fig. 1C), 2-keto-3-deoxysugar acid intermediates are converted to pyruvate and α-keto acid by the sequential actions of dehydrogenase and hydrolase (5–7). Metabolic genes related to these pathways often cluster together with putative sugar transporter genes and a transcriptional regulator gene in the genomes (Fig. 1G), and the encoding enzymes are classified into limited numbers of the known protein superfamilies: cluster of orthologous groups of proteins (COG). These insights facilitate estimations of potential substrates and/or metabolic pathways.

Figure 1.

Sugar metabolism by microorganisms. A–F, schematic representation of the phosphorylative l-fucose (A), nonphosphorylative l-fucose (B and C), l-rhamnose (D), d-xylonate (E), and l-arabinose pathways (F). Homologous genes (proteins) are indicated in the same color and correspond to G. In the DHDPS/NAL protein superfamily (COG0329), 2-keto-3-deoxyaldolases, examined in the present study, are phylogenetically classified into three groups. In B–F, sugars are commonly converted into their corresponding 2-keto-3-deoxysugar acid, the subsequent metabolic fate of which is aldol cleavage (route I), dehydration (route II), or dehydrogenation (route III). Genes in boxes were expressed, and the recombinant proteins were purified and characterized in the present study. The pathway of Fig. 1B was discovered in the present study. G, schematic gene clusters corresponding to the metabolic pathways shown in A–F. Genes in light and dark gray are putative transcriptional regulators and sugar transporters, respectively. The dashed line indicates the pseudogene.

Among these metabolic enzymes, 2-keto-3-deoxysugar acid aldolases are classified into two classes. Class I aldolases are characterized by a covalent intermediate, which is a protonated Schiff base formed between a lysine residue and the carbonyl carbon of the substrate. The DHDPS/NAL protein superfamily (COG0329) contains the enzymes for d-2-keto-3-deoxygluconate (d-KDGlu) (8), l-2-keto-3-deoxyrhamnonate (l-KDR) (9), l-2-keto-3-deoxygalactonate (l-KDGal) (10), and d-2-keto-3-deoxypentonate (d-KDP) (11) from archaea, fungi, and/or bacteria together with the archetypical DHDPS (EC 4.2.1.52) and NAL (EC 4.1.3.3) (Fig. 2A). Only l-KDR aldolase from bacteria has been reported in class II aldolases (COG3836; HpcH) with the absolute requirement of a divalent metal cofactor (9). On the other hand, three genetically unidentified bacterial aldolases for l-KDP (12), d-KDP (13), and d-2-keto-3-deoxyfuconate (d-KDF) (14) may also belong to this type because of their similar metal dependence. Physiological route I pathways are generally only considered to be found in d-glucose, d-galacturonate, l-rhamnose, d-fucose, d-xylose, and l-arabinose metabolism.

Figure 2.

2-Keto-3-deoxysugar acid aldolases in DHDPS/NAL protein superfamily. A, partial multiple sequence alignment of the deduced amino acid sequences containing l-KDF aldolase. Color of active site residues correspond to B and C. B, schematic aldol reaction. These enzymes reversibly catalyze cleavage and condensation between 2-keto-3-deoxysugar acid (green), pyruvate (aqua), and aldehyde (blue). In the latter reaction, pyruvate is bound via a Schiff base linkage to a lysine residue (active site 4 in A), and this facilitates proton abstraction from the methyl group to generate the enzyme bound enamine (upper panel). The attack of this enamine on the (Re)-face of the aldehyde carbonyl group (red line) yields the C4-(R)-OH product, whereas an attack on the (Si)-face of the aldehyde (orange line) yields the C4-(S)-OH product (lower panel). In the present study, a library consisting of nine 2-keto-3-deoxysugar acids, except for d-KDR, was prepared. The enzyme in the box plays a physiological role. C, schematic diagram showing the interactions of d-KDGlu (Protein Data Bank code 1W3N) and d-KDGal (Protein Data Bank 1W3T) and nearby residues in KdgA from S. solfataricus (16). The hydrogen bonds are shown as dashed lines with the bond lengths (Å). W indicates a water molecule.

(Reverse) aldol condensation is more useful for the synthesis of biologically, pharmaceutically, and/or agrochemically significant compounds than (forward) physiological aldol cleavage (15). This reaction enables the formation of a carbon–carbon bond between two carbonyl compounds, during which two new stereogenic centers are made. Among several reported class I aldolase enzymes, d-KDGlu aldolase (KdgA) from the hyperthermophilic Archaeon Sulfolobus solfataricus (as described above) has been well-studied biochemically and structurally, and the catalytic mechanism containing eight active-site residues at least is elucidated significantly (8, 16) (Fig. 2A). In the aldol condensation reaction, the ω-amino functionality of Lys¹⁵⁵ (site 4) forms a Schiff base with the carbonyl C2 of pyruvate, which tautomerizes to afford an enamine intermediate that may attack the (Re)- or (Si)-face of the carbonyl of d-glyceraldehyde (Fig. 2B). This reaction is mediated by the phenolic substituent of Tyr¹³⁰ (active site 2), which plays a key catalytic role in shuttling protons between reactive enzyme-bound intermediates to yield two diastereomers: d-KDGlu and d-KDGal. In the aldol cleavage reaction, their C4-OH groups are also in a position to interact with this tyrosyl oxygen, which mediates proton extraction by the substrate, and the C5- and C6-OH groups make bridging water-mediated hydrogen-bonding interactions with (the main and/or side chains of) active sites 3, 6, 7, and 8 (Fig. 2C). The carboxylate group is recognized by the main and side chains of Thr⁴³ and Thr⁴⁴ in a characteristic Gly-Xaa-Xaa-Gly-Glu motif (active site 1).

In the present study, I focused on the EO124_RS00050-RS00025 gene cluster from Veillonella ratti ATCC 17746, a strictly anaerobic bacterium. Although gene components are partially homologous with those of known l-fucose pathways, a putative DHDPS/NAL-like protein (EO124_RS00030; FucH) was also present. The library screening of the 2-keto-3-deoxysugar acids of this protein (and biochemical characterization of neighboring components) revealed that the gene cluster was responsible for (unidentified) route I of the nonphosphorylative l-fucose pathway. This approach also enabled more detailed comparisons of the substrate specificities and stereochemical selectivities of not only FucH as a l-KDF aldolase but also other 2-keto-3-deoxysugar acid aldolases in the DHDPS/NAL protein superfamily.

Results and discussion

Gene cluster from V. ratti

I previously characterized gene clusters related to nonphosphorylative sugar metabolism from Herbaspirillum huttiense and Acidovorax avenae, in which an altronate dehydratase-like gene (protein) (COG2721; UxaA) functions as a dehydratase for d-arabinonate, l-xylonate, d-altronate, d-idonate, l-gluconate, and/or l-fuconate (7, 17). On the other hand, the homologous gene from V. ratti ATCC 17746 (EO124_RS00050; FucC) clustered with several genes related to putative l-fucose metabolism: EO124_RS00040, l-fucose/H⁺ symporter (TC 2.A.1.7; FucP); the C-terminal half of EO124_RS00045, l-fucose mutarotase (EC 5.1.3.29; FucM); EO124_RS00045, l-fucose 1-dehydrogenase (EC 1.1.1.122; FucA); the N-terminal half of EO124_RS00045, l-fuconolactonase (FucB; EC 3.1.1.-); EO124_RS00025, l-lactaldehyde reductase (EC 1.1.1.77; FucO); and EO124_RS00050; function unknown DHDPS/NAL-like gene (FucH) (Fig. 1G).

To date, the metabolic pathways of l-fucose in microorganisms (bacteria) have been classified into two routes. The final products of the “phosphorylative route” are dihydroxyacetone phosphate and l-lactaldehyde, with the latter being aerobically oxidized to l-lactate (by l-lactaldehyde dehydrogenase (EC 1.2.1.22); AldA) or anaerobically reduced to (S)-1,2-propanediol (by FucO) (Fig. 1A) (18). The nonphosphorylative pathway corresponds to route III, and the metabolic fate of the l-KDF intermediate is pyruvate and l-lactate via l-2,4-diketo-3-deoxyfuconate (Fig. 1C) (5, 7). FucH was not present within gene clusters related to these l-fucose pathways (Fig. 1G). Therefore, to elucidate the physiological role(s) of this type of gene cluster, I initially characterized FucH from V. ratti (referred to as VrFucH).

Potential enzyme activity of VrFucH

Limited numbers of 2-keto-3-deoxysugar acids are generally used for functional characterization because of the difficulties associated with their preparation (not commercially available) (8–14). Alternatively, I enzymatically prepared nine 2-keto-3-deoxysugar acid(s) from sugar acids, as described under “Experimental procedures.” The potential enzyme activity of VrFucH was initially assayed by HPLC (Fig. S1). When each 2-keto-3-deoxysugar acid was incubated with the purified recombinant protein, a novel peak with a later retention time appeared in a time-dependent manner and was identical to lactaldehyde (∼12.6 min for l-KDF, d-KDF, and l-KDR), glycolaldehyde (∼12.4 min for d-KDP and l-KDP), and glyceraldehyde (∼11.3 min for d-KDGlu, l-KDGlu, d-KDGal, and l-KDGal). When catalyzed with KDGlu(s) and KDGal(s), novel peaks that differed from glyceraldehyde were commonly observed and found to be identical to pyruvate (∼9.8 min) and the alternative diastereomer. l-KDF and d-KDP were the best substrates (Fig. 3A). These results indicated that VrFucH catalyzes the reversible retro-aldolase reaction for 2-keto-3-deoxysugar acid(s) to afford pyruvate and aldehyde.

Figure 3.

Characterization of the aldol cleavage reaction. WT and/or mutant enzymes of VrFucH (A and F), PsLRA4 (B and G), HjLGA1 (C and H), EcYagE (D and I), and CjFucH (E and J) were assayed to directly estimate the concentration of released lactaldehyde (for KDF and KDR), glycolaldehyde (for KDP), or glyceraldehyde (for KDGlu and KDGal) by HPLC (A–E) (Fig. S1) or to continuously monitor released pyruvate by a spectrophotometer using l-lactate dehydrogenase as a coupling enzyme (F–J). Kinetic parameters assessed by the latter method are summarized in Tables 1 and 2 and Tables S1–S3, from which only k_cat/K_m values are plotted on a graph. Arrows indicate the physiological substrate of each enzyme.

Substrate specificity of VrFucH

To investigate substrate specificity in more detail, the activity of the aldol cleavage reaction was spectrophotometrically assessed using l-lactate dehydrogenase as a coupling enzyme. Kinetic parameters from the Lineweaver–Burk plot are shown in Table 1 and Fig. 3F. The k_cat/K_m values for l-KDF and d-KDP were similar (6570 and 6340 min⁻¹ mm⁻¹, respectively), whereas those for the other substrates were reduced by ∼1–3 orders of magnitude and caused by very small k_cat values. Among the substrates tested, clear structural preferences were detected at C4-(R)-OH and C6-CH₃ but not at C5-OH; for example, the ratios of l-KDF to l-KDR (C4) and l-KDF to l-KDGal (C6) were 29 and 14, respectively (red bars in Fig. 4A).

Table 1.

Kinetic parameters of EO124_RS00030 from (VrFucH)

The values are the means ± S.D. (n = 3). NP indicates no preparation.

Enzymea	Substrateb	C4-OH	Km	kcat	kcat/Km	R/Sc
			mm	min−1	min−1 mm−1	fold
WT	l-KDF	R	0.327 ± 0.004	2150 ± 8	6570 ± 65	29
	l-KDR	S	1.28 ± 0.07	289 ± 12	226 ± 3
	d-KDR	R	NP	NP	NP	NP
	d-KDF	S	0.535 ± 0.133	4.39 ± 0.68	8.37 ± 0.71
	d-KDP	R	0.430 ± 0.075	2690 ± 234	6340 ± 500	4.3
	l-KDP	S	0.923 ± 0.183	1340 ± 210	1470 ± 56
	d-KDGlu	R	0.692 ± 0.223	107 ± 21	161 ± 18	47
	d-KDGal	S	1.85 ± 0.38	6.21 ± 0.95	3.39 ± 0.23
	l-KDGal	R	1.47 ± 0.43	208 ± 48	144 ± 9	9.5
	l-KDGlu	S	1.16 ± 0.04	17.5 ± 0.4	15.2 ± 0.2
T164S	l-KDF	R	0.583 ± 0.074	1260 ± 97	2180 ± 120	11
	l-KDR	S	2.54 ± 0.21	503 ± 35	198 ± 2
	d-KDR	R	NP	NP	NP	NP
	d-KDF	S	1.08 ± 0.01	163 ± 1	151 ± 0
	d-KDP	R	0.345 ± 0.023	293 ± 8	853 ± 33	2.7
	l-KDP	S	1.88 ± 0.13	599 ± 39	319 ± 7
	d-KDGlu	R	0.570 ± 0.051	15.0 ± 1.0	26.4 ± 0.6	1.3
	d-KDGal	S	0.900 ± 0.116	17.9 ± 1.3	20.1 ± 1.1
	l-KDGal	R	3.10 ± 0.50	158 ± 23	51.1 ± 0.8	2.5
	l-KDGlu	S	2.11 ± 0.16	42.7 ± 2.6	20.3 ± 0.3

^a A His₆-tagged recombinant enzyme was used.

^b Eight different substrate concentrations between 0.1 and 1 mm were used.

^c The ratio of k_cat/K_m.

Figure 4.

Stereospecificity of the aldol cleavage reaction at C4, C5, and C6 positions of 2-keto-3-deoxysugar acids. A–E, WT and/or mutant enzymes of VrFucH (A), PsLRA4 (B), HjLGA1 (C), EcYagE (D), and CjFucH (E). The ratios in k_cat/K_m values were calculated from Tables 1 and 2 and Tables S1–S3.

In the aldol condensation reaction, the HPLC analysis showed that l-lactaldehyde, d-lactaldehyde, and glycolaldehyde were rapidly consumed in a time-dependent manner, and the reaction reached equilibrium within 6 h (Fig. 5A and Fig. S2). However, glyceraldehyde(s) were suitable for estimating selectivity because of similar retention between KDF, KDR, KDP, and pyruvate. During their reactions, both peaks for glyceraldehyde and pyruvate decreased in parallel, and two product peaks corresponding to KDGlu and KDGal were clearly visible throughout the time course of the reaction, affording 96:4 and 95:5 mixtures of d-KDGlu:d-KDGal and l-KDGal:l-KDGlu, respectively (Fig. 5B), suggesting strict (Re)-facial selectivity (Fig. 2B). Aliphatic aldehydes, including acetaldehyde and propionaldehyde, were not acceptable for VrFucH.

Figure 5.

Characterization of the aldol condensation reaction. The reaction by WT and/or mutant enzymes of VrFucH (A–C), PsLRA4 (D and E), HjLGA1 (F and G), EcYagE (H–J), and CjFucH (K) were monitored by HPLC (Fig. S2). The values indicate the diastereomer rates.

Characterization of native aldolases for l-KDR, l-KDGal, and d-KDP

As described in the introduction, the DHDPS/NAL protein superfamily contains three 2-keto-3-deoxysugar acid aldolases. l-KDR aldolase (LRA4) converts l-KDR (a C4-diastereomer of l-KDF) into pyruvate and l-lactaldehyde (the same products as VrFucH) (Fig. 2B), and the encoding gene is located within a unique fungal gene cluster related to route I of the nonphosphorylative l-rhamnose pathway (Fig. 1, D and G) (9). l-KDGal aldolase (EC 4.1.2.54; LGA1) is involved in fungal d-galacturonate metabolism as the fourth enzyme, in which l-KDGal is converted into pyruvate and l-glyceraldehyde (10). On the other hand, YagE is more closely related phylogenetically to VrFucH (31% sequence identity) than any DHDPS/NAL members (∼25%) (Fig. 6) and functions as d-KDP aldolase in d-xylonate metabolism (Fig. 1, E and G) (11, 19). Therefore, LRA4, LGA1, and YagE from Pichia stipites (also named Scheffersomyces stipitis), Hypocrea jecorina (also named Trichoderma reesei), and Escherichia coli were enzymatically characterized in detail using a library of 2-keto-3-deoxysugar acids (referred to as PsLRA4, HjLGA1, and EcYagE, respectively).

Figure 6.

Phylogenetic tree of the DHDPS/NAL protein superfamily. The number on each branch indicates the bootstrap value. Asterisks indicate proteins from anaerobic bacteria. S and T in parentheses indicate serine and threonine residues at active site 5 in Fig. 2A.

PsLRA4

In the aldol cleavage reaction, l-KDR (the physiological substrate) and l-KDP were the best substrates, and the k_cat/K_m values of the other substrates were reduced by 1–3 orders of magnitude because of very small k_cat values (Fig. 3, B and G, and Table S1). The C4-(S)-OH preference was completely opposite to that of VrFucH (for example, the ratio of l-KDR to l-KDF in k_cat/K_m was 48), whereas the C6-CH₃ preference was similar (red bars in Fig. 4B). In the aldol condensation reaction, PsLRA4 catalyzed the selective formation of d-KDGal and l-KDGlu in 83 and 98% diastereomer rates (DRs), respectively (Fig. 5D). Therefore, PsLRA4 showed (Se)-facial selectivity (Fig. 2B).

HjLGA1

l-KDF was the best substrate in the aldol cleavage reaction, whereas the k_cat/K_m value for l-KDGal (the physiological substrate) was 21-fold lower than that for l-KDF because of a high K_m value; 69.0 and 1430 min⁻¹ mm⁻¹, respectively (Fig. 3, C and H, and Table S2). The C4-(S)-OH and C6-CH₃ preferences were similar to those of VrFucH; for example, the ratios of l-KDR to l-KDF (C4) and l-KDF to l-KDGal (C6) in k_cat/K_m were 24 and 21 (red bars in Fig. 4C). In the aldol condensation reaction, l-glyceraldehyde was the only active substrate, affording a 83:17 mixture of l-KDGal:l-KDGlu (Fig. 5F). The reaction with d-glyceraldehyde using 10-fold amounts of the enzyme gave a 83:17 mixture of d-KDGlu and d-KDGal. Therefore, HjLGA1 showed (Re)-facial selectivity (Fig. 2B).

EcYagE

d-KDP was the best substrate in the aldol cleavage reaction, whereas the k_cat/K_m value was 4.8-fold higher than that for l-KDF (Fig. 3, D and I, and Table S3). In the aldol condensation reaction using d-glyceraldehyde, EcYagE catalyzed the selective formation of d-KDGlu in 88% DR (Fig. 5H). On the other hand, when catalyzing with l-glyceraldehyde, this enzyme exhibited no diastereocontrol, which gave a 53:47 mixture of l-KDGal and l-KDGlu. This property is consistent that there is no C4-(R)-OH preference between the two diastereomers (red bars in Fig. 4D). Therefore, the structural preference of C4-OH and facial selectivity in the reversible aldol reaction were clearly linked, suggesting a common amino acid residue(s) responsible for these stereochemical specificities.

Identification of amino acid residue(s) responsible for stereoselectivity

The catalytic mechanism of VrFucH as a l-KDF aldolase (and PsLRA4, HjLGA1, and EcYagE) may be typical of DHDPS/NAL members, particularly KdgA from S. solfataricus (SsKdgA), as described in the introduction; Lys¹⁶² and Tyr¹³³ correspond to two lysine and tyrosine residues at active sites 4 and 2, respectively (Fig. 2B) (16). On the other hand, SsKdgA utilizes not only d-glyceraldehyde (a natural substrate) but also l-glyceraldehyde as acceptable aldehydes and exhibits poor diastereocontrol, which gives a 50:50 mixture of d-KDGlu:d-KDGal or l-KDGal:l-KDGlu (8, 20). In the crystal structure(s) (16), the C4-OH group of bound d-KDGlu and d-KDGal is commonly recognized by Thr¹⁵⁷ at active site 5, whereas Gly¹⁷⁸, Asp¹⁸¹, and Ala¹⁹⁸ at active sites 7 and 8 contribute more for the hydrogen-bond network to d-KDGlu than d-KDGal via several water molecules (Fig. 2C). Based on the saturation mutagenesis of Thr¹⁵⁷ and combination with mutations at Tyr¹³², Ala¹⁹⁸, or Asp¹⁸¹, the best double mutants that provided better stereocontrol were T157C/Y132V and T157F/Y132V for d-KDGlu in ∼92% DR and T157V/A198L/D181Q for d-KDGal in 88% DR, respectively, whereas their k_cat/K_m values decreased by 3 orders of magnitude (21). Alternatively, in the present study, we noted that VrFucH, HjLGA1, and EcYagE with (Re)-facial selectivity possess a threonine at active site 5, which is substituted to a serine in PsLRA4 with (Si)-facial selectivity, respectively, and the highly conserved tyrosine residue at active site 3 is substituted to a phenylalanine only in EcYagE (Fig. 2A). VrFucH_T164S, VrFucH_Y135F, HjLGA1_T169S, EcYagE_T169S, EcYagE_F140Y, and PsLRA4_S173T were designed based on this phylogenetic information.

Mutation at active site 5

In the aldol condensation reaction using d-glyceraldehyde, the selective formation of d-KDGal by FucH_T164S, HjLGA1_T169S, and EcYagE_T169S was markedly improved in 42% (4%), 32% (17%), and 38% (12%) DRs, respectively (values for each WT are in parentheses) (Fig. 5, C, G, and I). PsLRA4_S173T catalyzed the diastereoselective formation of d-KDGlu with a markedly improved 85% (17%) DR (Fig. 5E). When catalyzing with l-glyceraldehyde, these mutants only resulted in small improvements in diastereocontrol: for example, l-KDGal with 89% (95%) DR for VrFucH_T164S and l-KDGlu with 82% (92%) DR for PsLRA4_S173T. Concomitantly, in the aldol cleavage reaction, these mutations relieved structural preference only at the C4 position (but not C5 or C6); d-KDGlu was a better substrate for PsLRA4_S173T than d-KDGal (blue bars in Fig. 4, A–D; Table 1; and Tables S1–S3).

Mutation at active site 3

VrFucHY135F was completely inactive for any substrate. In the case of EcYagEF140Y, substrate and stereochemical specificities were similar to YagE_WT (green bars in Fig. 4D, and Fig. 5J), whereas the k_cat/K_m value for d-KDP was only 12% (Fig. 3I and Table S3). In the phylogenetic tree, PsLRA4 and HjLGA1 form “distinct groups” in the “same branch” (Fig. 6). Furthermore, HjLGA1 could be considered to be l-KDF aldolase in vitro because it is more active against l-KDF than l-KDGal (Fig. 3, C and H). Therefore, each opposite stereoselectivity may have been subjected to the acquisition of a serine or threonine residue at active site 5 in the evolutionary stage. Because the branch containing VrFucH also contains several enzymes with an equivalent serine residue (Fig. 6), their physiological substrate(s) do not appear to be l-KDF or d-KDP.

Another type of l-KDF aldolase from Campylobacter jejuni

C. jejuni subsp. jejuni NCTC 11168 possesses a gene cluster consisting of cj0480c–cj0490, the transcription of which is up-regulated by l-fucose (22); Cj0486, Cj0488, Cj0485, Cj0487, Cj0482–0483 (pseudogene), Cj0481, and Cj0489 (pseudogene) corresponding to FucP, FucM, FucA, FucB, FucC, FucH, and AldA, respectively (Fig. 1G). Among them, the putative amino acid sequence of Cj0481 was only slightly similar to that of VrFucH (only 23% identity) (Fig. 6). Therefore, to estimate the physiological role of this gene cluster in more detail, Cj0481 was enzymatically characterized by the recombinant protein (referred to as CjFucH).

Among the 2-keto-3-deoxysugar acids tested, l-KDF and d-KDP were significant active substrates for CjFucH, and the k_cat/K_m values of the other substrates were reduced by 1 order of magnitude (Fig. 3, E and J, and Table 2). The ratio of l-KDF to l-KDR in k_cat/K_m was 11, indicating the C4-(R)-OH preference (Fig. 4E). The aldol condensation reaction resulted in the selective formation of d-KDGlu and l-KDGal in both 83% DRs, namely, (Re)-facial selectivity (Fig. 5K). Despite having the same stereospecificity as VrFucH (and HjLGA1 and EcYagE), CjFucH possessed a serine residue (Ser¹⁶⁴) at active site 5, similar to PsLRA4 (Fig. 2A). In the phylogenetic tree of the DHDPS/NAL protein superfamily, CjFucH was closer to the NAL subclass than any 2-keto-3-deoxysugar acid aldolases (Fig. 6). All NAL enzymes possess a threonine residue at active site 5 and are not strictly specific with regards to stereochemistry at the C4 position, and the site-directed mutation leads to the preparation of selective diastereoisomeric sialic acid analogues (23). These results indicate that this active-site position may be universal for playing a key role in influencing the stereoselectivity of its aldol reactions. We are now conducting crystallographic analyses of VrFucH, PsLRA4, and CjFucH, the findings of which will contribute to the more detailed elucidation of their unique properties.

Table 2.

Kinetic parameters of Cj0481 from C. jejuni (CjFucH)

The values are the means ± S.D. (n = 3). A His₆-tagged recombinant enzyme was used. NP indicates no preparation.

Substratea	C4-OH	Km	kcat	kcat/Km	R/Sb
		mm	min−1	min−1 mm−1	fold
l-KDF	R	0.569 ± 0.057	747 ± 49	1310 ± 44	11
l-KDR	S	1.47 ± 0.08	172 ± 9	118 ± 0
d-KDR	R	NP	NP	NP	NP
d-KDF	S	0.205 ± 0.012	12.1 ± 0.2	59.2 ± 2.3
d-KDP	R	1.33 ± 0.14	832 ± 71	628 ± 13	8.7
l-KDP	S	1.31 ± 0.28	93.6 ± 14.2	72.2 ± 3.9
d-KDGlu	R	1.90 ± 0.15	83.8 ± 4.8	44.2 ± 0.9	2.7
d-KDGal	S	0.179 ± 0.021	2.93 ± 0.13	16.5 ± 1.2
l-KDGal	R	2.42 ± 0.27	160 ± 16	66.3 ± 0.6	2.4
l-KDGlu	S	1.35 ± 0.05	37.2 ± 0.9	27.6 ± 0.4

^a Eight different substrate concentrations between 0.1 and 1 mm were used.

^b The ratio of k_cat/K_m.

Novel nonphosphorylative pathway of l-fucose metabolism

Although the gene cluster analysis and enzymatic characterization suggested that the VrFucH gene (protein) plays a role in l-fucose metabolism, V. ratti is a strictly anaerobic coccus, and neither minimal medium for growth nor gene expression analyses have been reported to date. Alternatively, biochemical characterization of the neighbor EO124_RS00025 (57% sequential identity with FucO from E. coli) (24) was performed (referred to as VrFucO) (Fig. 1G). k_cat/K_m values to l-lactaldehyde in the presence of NAD(P)H were ∼2 orders of magnitude higher than those to l-glyceraldehyde, glycolaldehyde, and propionaldehyde (neither d-lactaldehyde nor d-glyceraldehyde was active) and were attributed to low K_m values (Table 3). Glu³⁹ in NADH-dependent FucO from E. coli favorably interacts with the 2′- and 3′-hydroxyl groups of the ribose of the adenine moiety (24) and is substituted to a glycine residue (Gly³⁸) in VrFucO, by which dual coenzyme specificity may be explained.

Table 3.

Kinetic parameters of EO124_RS00025 from V. ratti (VrFucO) and Cj0489–0490 from C. jejuni (CjAldA)

NA indicates not assessed because of trace activity.

Enzymesa	Substrate	Product	Coenzyme	Km	kcat	kcat/Km
				mm	min−1	min−1 mm−1
VrFucO	l-Lactaldehydeb	l-1,2-Propanediol	NADH	0.150 ± 0.005	1960 ± 12	13,100 ± 371
			NADPH	0.106 ± 0.001	656 ± 1	6220 ± 32
	d-Lactaldehydeb	d-1,2-Propanediol	NADH	NA	NA	NA
	l-Glyceraldehydeb	Glycerol		1.16 ± 0.11	671 ± 45	580 ± 16
	d-Glyceraldehydeb			NA	NA	NA
	Glycolaldehydeb	Ethylene glycol		3.62 ± 0.38	1180 ± 125	327 ± 5
	Propionaldehydeb	Propanol		4.15 ± 0.89	380 ± 72	92.3 ± 3.3
CjAldA	l-Lactaldehydec	l-Lactate	NADP+	0.142 ± 0.008	948 ± 51	6690 ± 15
			NAD+	0.205 ± 0.011	153 ± 6	751 ± 10
	d-Lactaldehydeb	d-Lactate	NADP+	2.93 ± 0.31	173 ± 17	59.3 ± 1.3
	l-Glyceraldehydec	l-Glycerate		0.221 ± 0.005	470 ± 6	2130 ± 26
	d-Glyceraldehydeb	d-Glycerate		3.62 ± 0.24	167 ± 12	46.1 ± 0.3
	Glycolaldehydeb	Glycolate		2.60 ± 0.42	1220 ± 179	468 ± 6

The values are the means ± S.D. (n = 3).

^a A His₆-tagged recombinant enzyme was used.

^b Eight different substrate concentrations between 0.1 and 1 mm were used.

^c Eight different substrate concentrations between 0.01 and 0.1 mm were used.

Among AldA-like genes from several Campylobacter and Helicobacter species, clustered with fucABCHM genes, only cj0489 from C. jejuni subsp. jejuni NCTC 11168 was identified as a pseudogene (Fig. 1G). On the other hand, the putative amino acid sequence of the gene fused into a single ORF (as described under “Experimental procedures”) showed 61% identity with that of AldA from E. coli (25) (referred to as CjAldA). The recombinant CjAldA protein showed 7.19 units/mg protein of specific activity for l-lactaldehyde in the presence of NADP⁺, whereas AldA from E. coli strictly depended on NAD⁺. The k_cat/K_m value for l-lactaldehyde (6690 min⁻¹ mm⁻¹) was 3.1-, 14-, 112-, and 145-fold higher than those of l-glyceraldehyde, glycolaldehyde, d-lactaldehyde, and d-glyceraldehyde, respectively (Table 3).

Although a genetic engineering approach is not currently available for V. ratti, the mutation of cj0481 encoding CjFucH was previously shown to result in the loss of the ability to grow on l-fucose (22). Collectively, the present genetic and biochemical results strongly suggest that FucH(s), FucO, and AldA physiologically function as l-KDF aldolase, l-lactaldehyde reductase, and l-lactaldehyde dehydrogenase involved in (unidentified) route I of the nonphosphorylative l-fucose pathway from anaerobic bacteria.

Schomburg and coworkers (26) suggested that SsKdgA is involved in route I of the nonphosphorylative pathways of not only d-glucose, d-galactose, d-xylose, and l-arabinose (8, 27) but also l-fucose metabolism. However, in the aldol condensation reaction, the k_cat/K_m value for l-lactaldehyde was ∼3000-fold lower than that for d-glyceraldehyde, and the encoding gene was neither transcriptionally nor translationally responsible for l-fucose; this hypothesis warrants further study. d-KDP is a significant substrate for VrFucH and CjFucH, and the FucC protein is a bifunctional dehydratase for l-fuconate and d-arabinonate (7, 17), whereas glycolaldehyde utilizes neither VrFucO nor CjAldA (Table 3), indicating no “metabolic promiscuity” in route I of the nonphosphorylative l-fucose pathway in V. ratti and C. jejuni, in contrast to S. solfataricus. Furthermore, although E. coli K-12 utilizes functions of FucO and AldA during aerobic and anaerobic growth, respectively (18), these distinct functions may not apply to anaerobic species such as Veillonella, Campylobacter, and Helicobacter species.

Conclusion

Based on enzyme activities in cell-free extracts from experiments performed several decades ago, many (hypothetical) route I nonphosphorylative pathways for l-arabinose (12), d-xylose (13), and d-fucose metabolism (14), but not l-fucose metabolism, have been described; therefore, the present study is unique. Because FucH homologues are only found in strictly anaerobic (and difficulty culturable) bacteria (Fig. 6), the gene context in bacterial genomes and the library screening of potential substrates will be beneficial for identifying biochemical and physiological functions as a l-KDF aldolase. Library screening with originally synthesized 2-keto-3-deoxysugar acids revealed the substrate specificities and facial selectivities of not only l-KDF aldolase but also three aldolases for l-KDR, l-KDGal, and d-KDP previously registered in the DHDPS/NAL protein superfamily in more detail (9–11). These properties are commonly determined by whether a serine or threonine residue is positioned at the equivalent position close to the active site(s), which may be extended to the evolution of other useful biocatalysts for the stereoselective synthesis of biologically active molecules. Substrate specificity for 2-keto-3-deoxysugar acid aldolase may have been convergently acquired in the evolutionary stage, in contrast to other DHDPS/NAL members. If this hypothesis is true, another type of l-KDF aldolase may be physiologically involved in the same nonphosphorylative l-fucose pathway (22).

Experimental procedures

Bacterial strains

V. ratti ATCC 17746, H. jecorina QM 6a, and a plasmid harboring the EcYagE gene were purchased from the RIKEN BioResource Center (Saitama, Japan), the National Institute of Technology and Evaluation (Chiba, Japan), and the National BioResource Project (Shizuoka, Japan), respectively. Genomic DNA was prepared using a DNeasy tissue kit (Qiagen). Genomic DNA from C. jejuni subsp. jejuni NCTC 11168 was kindly gifted from Dr. Takaaki Shimohata (Tokushima University).

Gene cloning and protein overexpression and purification

The primer sequences used in the present study are shown in Table S4. Target genes were amplified by PCR from genomic DNA using KOD FX DNA polymerase (Toyobo) and introduced into the BamHI–HindIII, BamHI–PstI, or BamHI–SalI site in pQE-80L (Qiagen), a plasmid vector that encodes an N-terminal His₆ tag added to the expressed protein, to obtain each pQE-based expression plasmid. A site-directed mutation was introduced by sequential steps using PCR with sense and antisense primers (Table S4) and each pQE-based expression plasmid as a template. A functional CjAldA gene was constructed by substituting G455678 with A, and inserting a single A between T455680 and A455681 in the original genome sequence of C. jejuni subsp. jejuni NCTC 11168 by sequential steps of PCR using sense and antisense primers (Table S4).

E. coli strain DH5α harboring pQE-based expression plasmids was grown at 37 °C to a turbidity of 0.6 at 600 nm in LB medium containing ampicillin (50 mg/liter). After the addition of 1 mm isopropyl-β-d-thiogalactopyranoside, the culture was grown for 6 h to induce the expression of the respective His₆-tagged proteins. The cells were harvested and resuspended in buffer A (50 mm sodium phosphate buffer (pH 8.0) containing 300 mm NaCl and 10 mm imidazole). The cells were then disrupted by sonication, and the solution was centrifuged. The supernatant was loaded onto a nickel–nitrilotriacetic acid Superflow column (Qiagen). 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 imidazole instead of 25 mm imidazole).

Synthesis of 2-keto-3-deoxysugar acid substrates

All sugar acids were prepared by hypoiodite-in-methanol oxidization from the corresponding sugars as K⁺ or Ba²⁺ salts (28). K⁺ salts were dissolved in an approximate amount of water, and pH was adjusted to ∼10.0 with NaOH. Ba²⁺ salts were dissolved by the addition of appropriate H₂SO₄, and BaSO₄ produced was removed by centrifugation. The pH of the filtrate was adjusted as described above. The solution containing sugar acids was applied to the column of an AG® 1-X8 resin (200–400 mesh, formate form) (Bio-Rad). The column was washed thoroughly with water and developed with a gradient of 0–1 m formic acid. In the detection of sugar acids, HPLC was performed using a multistation LC-8020 model II system (TOSOH). Samples were applied at 35 °C to 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 H₂SO₄ at a flow rate of 0.6 ml/min. Fractions containing sugar acids were combined and lyophilized to yield the corresponding lactone-sugars. Sugar acids were obtained by the base hydrolysis of the lactone sugar according to the method of Yew et al. (5).

Nine 2-keto-3-deoxysugar acids, d-KDP, l-KDP, d-KDGlu, l-KDGlu, d-KDGal, l-KDGal, l-KDF, d-KDF, and l-KDR, were synthesized from d-xylonate, l-arabinonate, d-gluconate, l-mannonate, d-galactonate, l-galactonate, l-fuconate, d-fuconate, and l-rhamnonate, respectively, using l-arabinonate dehydratase from Azospirillum brasilense (3) (for d-KDP, l-KDP, d-KDGlu, d-KDGal, and d-KDF), d-arabinonate dehydratase from H. huttiense (7) (for l-KDGal and l-KDF), or l-rhamnonate dehydratase from Azotobacter vinelandii (9) (for l-KDGlu and l-KDR). The reaction mixture (100 ml) consisted of 50 mm HEPES–NaOH buffer (pH 7.2), 10 mm sugar acid, and 1 mm MgCl₂. After the addition of ∼50 mg of purified sugar acid dehydratase, the mixture was left at 30 °C overnight. The solution was filtered and then applied to an AG® 1-X8 resin column followed by a 0–1 m gradient of formic acid, and fractions containing 2-keto-3-deoxysugar acid (detected using the semicarbazide method) (29) were combined, lyophilized, and dissolved. A stock solution (100 mm) was prepared by neutralization (pH ∼7.0) with 1 n NaOH.

Enzyme assay

All enzyme assays were performed at 30 °C. l-KDF aldolase–Aldolase activity in the forward direction (aldol-cleavage reaction) was continuously assayed at 340 nm in 50 mm HEPES–NaOH buffer (pH 7.2) containing 10 mm 2-keto-3-deoxysugar acid and 1 unit of l-lactate dehydrogenase (Oriental Yeast Co., Ltd.). The reaction was initiated by the addition of 1 mm NADH solution (100 μl) with a final reaction volume of 1 ml. In the reverse direction (aldol-condensation reaction), the enzyme was measured in 50 mm HEPES–NaOH buffer (pH 7.2) containing 10 mm pyruvate and 10 mm l-lactaldehyde. The production of 2-keto-3-deoxysugar acid was estimated by HPLC using an Aminex HPX-87H organic analysis column, as described above. l- and d-Lactaldehydes were chemically synthesized from l- and d-threonines, respectively, according to the method of Huff and Rudney.

l-Lactaldehyde reductase

Reductase activity was assayed spectrophotometrically at 340 nm in 50 mm potassium phosphate (pH 7.0) containing 10 mm l-lactaldehyde and 0.15 mm NADH.

l-Lactaldehyde dehydrogenase

Dehydrogenase activity was assayed spectrophotometrically at 340 nm in 50 mm potassium phosphate (pH 8.5) containing 10 mm l-lactaldehyde and 1.5 mm NADP⁺.

Sequence comparison

Protein sequences were analyzed using the Protein-BLAST and Clustal W programs distributed by the DNA Data Bank of Japan.

Author contributions

S. W. conceptualization; S. W. resources; S. W. data curation; S. W. investigation; S. W. methodology; S. W. writing-original draft; S. W. project administration; S. W. writing-review and editing.