Characterization of a Novel Glucosamine-6-Phosphate Deaminase from a Hyperthermophilic Archaeon

ABSTRACT A key step in amino sugar metabolism is the interconversion between fructose-6-phosphate (Fru6P) and glucosamine-6-phosphate (GlcN6P). This conversion is catalyzed in the catabolic and anabolic directions by GlcN6P deaminase and GlcN6P synthase, respectively, two enzymes that show no relationship with one another in terms of primary structure. In this study, we examined the catalytic properties and regulatory features of the glmD gene product (GlmDTk) present within a chitin degradation gene cluster in the hyperthermophilic archaeon Thermococcus kodakaraensis KOD1. Although the protein GlmDTk was predicted as a probable sugar isomerase related to the C-terminal sugar isomerase domain of GlcN6P synthase, the recombinant GlmDTk clearly exhibited GlcN6P deaminase activity, generating Fru6P and ammonia from GlcN6P. This enzyme also catalyzed the reverse reaction, the ammonia-dependent amination/isomerization of Fru6P to GlcN6P, whereas no GlcN6P synthase activity dependent on glutamine was observed. Kinetic analyses clarified the preference of this enzyme for the deaminase reaction rather than the reverse one, consistent with the catabolic function of GlmDTk. In T. kodakaraensis cells, glmDTk was polycistronically transcribed together with upstream genes encoding an ABC transporter and a downstream exo-β-glucosaminidase gene (glmATk) within the gene cluster, and their expression was induced by the chitin degradation intermediate, diacetylchitobiose. The results presented here indicate that GlmDTk is actually a GlcN6P deaminase functioning in the entry of chitin-derived monosaccharides to glycolysis in this hyperthermophile. This enzyme is the first example of an archaeal GlcN6P deaminase and is a structurally novel type distinct from any previously known GlcN6P deaminase.

Amino sugars, such as N-acetylglucosamine (GlcNAc), Nacetylgalactosamine (GalNAc), and N-acetylmuramic acid, are important building blocks for structural polysaccharides or sugar chains in several organisms. In the metabolism of these sugars, the conversion between fructose-6-phosphate (Fru6P) and glucosamine-6-phosphate (GlcN6P) is a key step in both anabolic and catabolic directions. The anabolic reaction is catalyzed by GlcN6P synthase (L-glutamine:D-fructose-6-phosphate amidotransferase), while catabolism is mediated by GlcN6P deaminase (Fig. 1A). GlcN6P synthase catalyzes the irreversible formation of GlcN6P and glutamate from Fru6P and glutamine and is classified in a glutamine-dependent amidotransferase family (18) comprised of an N-terminal glutamine amide transfer (GAT) domain joined to a C-terminal sugar isomerase domain (Fig. 1B). The former domain produces ammonia from glutamine, and the generated ammonia is utilized for amination of Fru6P accompanied by isomerization to GlcN6P in the latter domain. Unlike other glutamine-dependent amidotransferases displaying ammonia-dependent ac-tivity, GlcN6P synthase cannot utilize free ammonia as the nitrogen donor in place of glutamine (19).
On the other hand, GlcN6P deaminase catalyzes the deamination-isomerization reaction from GlcN6P to Fru6P and ammonia and can also catalyze its reverse reaction under the presence of high concentrations of ammonia (7,20). Although GlcN6P synthase and GlcN6P deaminase catalyze similar reactions, there is no relation between the primary structures of these two enzymes. There have been many studies on GlcN6P synthase (3,5,13,24) and GlcN6P deaminase (1,8,(15)(16)(17)20) from Eucarya and Bacteria due to the importance of these enzymes in the regulation of amino sugar metabolism. In contrast, corresponding enzymes from Archaea have not been reported so far. Most intriguingly, archaeal genomes do not harbor any genes homologous to known GlcN6P deaminases.
We have previously found that the hyperthermophilic archaeon Thermococcus kodakaraensis KOD1 (2) has an ability to degrade chitin, a ␤-1,4-linked linear homopolymer of GlcNAc, and successfully identified a novel chitin catabolic pathway. Namely, chitin is first degraded into the disaccharide GlcNAc 2 by a unique extracellular chitinase from T. kodakaraensis (ChiA Tk ) possessing endo-and exo-type catalytic domains (25,28). The GlcNAc 2 is probably translocated across the cell membrane by an ABC transport system and then deacetylated by a deacetylase (Dac Tk ) with nonreducing end specificity. The partially acetylated disaccharide GlcN-GlcNAc is hydrolyzed into GlcN and GlcNAc by an exo-␤-glucosaminidase (GlmA Tk ), and the generated GlcNAc is further deacetylated to GlcN by Dac Tk (26,27), resulting in the complete conversion of chitin into GlcN monomers. The genes for these enzymes are highly clustered on the T. kodakaraensis genome, whose features have recently been reported (accession no. AP006878) (12). Here, we focused on a gene within this cluster, encoding a probable sugar isomerase related to the isomerase domain of GlcN6P synthase, and clearly demonstrated that this probable sugar isomerase exhibited GlcN6P deaminase activity. This report identifies not only the first archaeal GlcN6P deaminase involved in chitin degradation but also a novel type of GlcN6P deaminase distinct from previously known enzymes.

MATERIALS AND METHODS
Bacterial strains, plasmids, and media. T. kodakaraensis KOD1 was grown anaerobically at 85°C in a screw-cap bottle with MA medium (4.8 g and 26.4 g of Marine Art SF agents A and B, respectively [Senju Seiyaku, Osaka, Japan], 5 g of yeast extract, and 5 g of tryptone in 1 liter of deionized water) supplemented with elemental sulfur (5 g/liter). Escherichia coli DH5␣ and BL21-Codon-Plus(DE3)-RIL (Stratagene, La Jolla, CA) were used as hosts for the expression plasmid derived from pET-21a (Novagen, Madison, WI) and were cultivated in LB medium at 37°C.
DNA manipulations and sequencing. DNA manipulations were carried out by standard methods, as described previously by Sambrook and Russell (23). Restriction enzymes and other modifying enzymes were purchased from Takara Bio (Otsu, Shiga, Japan) or Toyobo (Osaka, Japan). Small-scale preparation of plasmid DNA from E. coli cells was performed with the QIAGEN plasmid mini kit (QIAGEN, Hilden, Germany). DNA sequencing was performed with the BigDye Terminator cycle sequencing ready reaction kit, version 3.1, and the model 3100 capillary DNA sequencer (Applied Biosystems, Foster City, CA).
Construction of the expression plasmid. The expression plasmid for glmD Tk was constructed by PCR as described below. Two oligonucleotides (sense, 5Ј-G GTGAGCATATGCACGCAACGCTTAGAG-3Ј, and antisense, 5Ј-CCGGAT CCCATCACCACTTTACGAC-3Ј [Underlined sequences indicate an NdeI site in the sense primer and a BamHI site in the antisense primer.]) and T. kodakaraensis genomic DNA were used as the primers and the template for DNA amplification, respectively. The amplified DNA was digested with NdeI and BamHI and then ligated with the corresponding sites in the plasmid pET-21a. The absence of unintended mutations in the insert was confirmed by DNA sequencing. The resulting plasmid was designated pET-glmD.
Purification of recombinant GlmD Tk . E. coli BL21-CodonPlus(DE3)-RIL cells harboring pET-glmD were induced for overexpression with 0.1 mM isopropyl-␤-D-thiogalactopyranoside at the mid-exponential growth phase and incubated for a further 4 h at 37°C. The cells were harvested by centrifugation (5,000 ϫ g for 10 min at 4°C), resuspended in buffer A (50 mM Tris-HCl [pH 8.0]), and then disrupted by sonication. The supernatant after centrifugation (15,000 ϫ g for 10 min) was incubated at 85°C for 10 min and centrifuged (15,000 ϫ g for 10 min) to obtain a heat-stable protein solution. The solution was applied to an anionexchange Resource Q column (6 ml) (Amersham Biosciences, Piscataway, NJ) equilibrated with buffer A. The proteins were eluted with a linear gradient of 0 to 0.5 M NaCl, and the peak fractions eluted at 0.3 M NaCl were concentrated using an Ultrafree-4 centrifugal filter unit Biomax-10 (Millipore, Bedford, MA). This was applied to a gel filtration Superdex-200 HR 10/30 column (Amersham Biosciences) equilibrated with buffer A containing 0.15 M NaCl. Protein concentration was determined with the Bio-Rad protein assay (Bio-Rad, Hercules, CA), with bovine serum albumin as a standard.
Enzyme assays. A qualitative assay for GlmD Tk toward various monosaccharides was performed with silica gel thin-layer chromatography (TLC) as described previously (25), with a modification in the developer to methanol-chloroform-acetic acid-water (30:20:10:1 [vol/vol/vol/vol]). For detection of the products, aniline-diphenylamine reagent and ninhydrin reagent were applied.
The pH values were adjusted at room temperature, and the values at higher temperatures were calculated according to the temperature coefficients for the respective buffers (9). To investigate the thermostability of GlmD Tk , the enzyme solution (50 ng GlmD Tk in 20 l of 75 mM CHES-NaOH [pH 8.3 at 60°C]) was incubated at 80°C or 90°C from 5 to 100 min before the first reaction, and the resulting activity was determined by the method described above. The activity observed prior to incubation was 100%.
The reverse reaction of GlcN6P deaminase was determined by measuring the generated GlcN6P. The reaction mixture (100 l), containing 5 mM Fru6P, 10 mM NH 4 Cl, and 150 ng GlmD Tk in 50 mM CHES-NaOH (pH 8.3 at 60°C), was incubated for 1 min at 60°C. The reaction was terminated by cooling and followed by filtration with a Microcon YM-10 (Millipore) to remove the enzyme. The amount of GlcN6P in the filtrate was determined by a modified method of Morgan and Elson (13).
Nucleotide sequence accession number. The nucleotide sequence data of the glmD Tk gene reported here (TK1755) have been included within the T. kodakaraensis KOD1 genome (accession number AP006878) in the EMBL, GenBank, and DDBJ nucleotide sequence databases.

RESULTS
A probable sugar isomerase gene within the chitin degradation gene cluster in T. kodakaraensis. We have previously reported that the genes for chitin catabolic enzymes chitinase (ChiA Tk encoded by TK1765) (25,28), GlcNAc 2 deacetylase (Dac Tk encoded by TK1764) (27), and ␤-glucosaminidase (GlmA Tk encoded by TK1754) (26) are clustered at the 1,553to 1,569-kbp region of the T. kodakaraensis KOD1 genome (12), as shown in Fig. 2. In this cluster, one gene (TK1755) was identified between the glmA Tk and putative ABC transporter genes (TK1756 to TK1760) in the same orientation. As these seven genes were overlapping or separated by short interval regions (from Ϫ8 bp to 47 bp), they were supposed to be transcribed into a single mRNA. The TK1755 gene, designated glmD Tk , consisted of 978 bp encoding a protein of 326 amino acids with a predicted molecular mass of 36,749 Da. The deduced amino acid sequence displayed high overall homologies to the proteins from the closely related hyperthermophilic archaea Pyrococcus furiosus (PF0362), Pyrococcus abyssi (PAB1348), and Pyrococcus horikoshii (PH0510) (61 to 63% identical at the amino acid level). GlmD Tk also showed weak but notable similarities to the C-terminal isomerase domain of GlcN6P synthases from various sources (25 to 32% identities), although it lacks an N-terminal GAT domain (Fig. 1C). The GlmD Tk -related proteins and domains commonly contained a tandem repeat of two sugar isomerase subdomains (SIS domain, Pfam01380). Previous X-ray crystallographic analyses for GlcN6P synthase from Escherichia coli (GlmS Ec ) have identified catalytically important amino acid residues, Glu-488, His-504, and Lys-603, for the amination and isomerization of Fru6P (30,31). We found that the corresponding residues were also conserved in GlmD Tk as Glu-214, His-230, and Lys-322, suggesting a sugar isomerization activity in GlmD Tk with a catalytic mechanism similar to that of GlcN6P synthase. It should be noted that the T. kodakaraensis genome harbors a separate gene (designated glmS Tk ) encoding a protein entirely homologous to GlcN6P synthase (Fig. 1C), composed of GAT and isomerase domains, at a different locus (TK0809). The C-terminal isomerase domain showed 27% identity to Glm-D Tk .
As described above, T. kodakaraensis degrades chitin to GlcN by the functions of three catabolic enzymes, ChiA Tk , Dac Tk , and GlmA Tk , but the metabolic fate of the resulting GlcN was unclear. Bacterial and eucaryal pathways for GlcN metabolism include GlcN6P deaminase for the generation of the glycolytic intermediate Fru6P from GlcN6P. A representative is NagB from E. coli, which belongs to the GlcN6P isomerase/6-phosphogluconolactonase family (Pfam01182). However, no obvious ortholog for NagB and the related enzymes has been identified on the T. kodakaraensis genome or on other known archaeal genomes. These facts raised a possibility that glmD Tk within the gene cluster for chitin degradation might participate in catabolism of the amino sugar, so we investigated the catalytic properties and regulatory features of GlmD Tk , as described below.
Production and purification of recombinant GlmD Tk . To investigate the function of the protein product of glmD Tk , the gene was expressed in E. coli with the pET expression system. The recombinant protein was obtained in a soluble form and was purified to apparent homogeneity in SDS-PAGE by heat treatment and column chromatography (Fig. 3A). The molecular mass of the recombinant GlmD Tk was estimated to be 37.3 kDa by SDS-PAGE and 71.8 kDa by gel filtration chromatog- raphy, indicating formation of a homodimer. This dimeric structure was the same as that of the typical GlcN6P synthase but differed from the general homohexameric structure of GlcN6P deaminases. Some glutamine-dependent amidotransferases have been known to possess a GAT domain as a distinct subunit (18). We therefore examined the possibility of heteromeric association of GlmD Tk with other proteins in T. kodakaraensis cells by native PAGE Western blot analysis, as shown in Fig. 3B. GlmD Tk was actually expressed in the cells grown in the presence of GlcNAc 2 , and the mobility of native GlmD Tk coincided with that of the purified, recombinant protein. This result indicated that GlmD Tk was present in this archaeon as a homodimeric enzyme without any other subunits. Enzymatic property of GlmD Tk . We then investigated the catalytic properties of the purified recombinant GlmD Tk against GlcN6P or Fru6P by TLC. As shown in the left lane of Fig. 4, the incubation of GlmD Tk with GlcN6P led to the formation of Fru6P, and the release of ammonia was also confirmed with the ninhydrin reagent (data not shown). These facts indicated a GlcN6P deaminase activity of GlmD Tk . On the other hand, GlmD Tk showed no activity toward Fru6P with glutamine (Fig. 4, middle), which was consistent with the lack of a GAT domain in GlmD Tk . We also detected activity for the reverse reaction producing GlcN6P from Fru6P in the presence of ammonia (Fig. 4, right). This demonstrates the catalytic ability of GlmD Tk for reversible deamination/amination accompanied by isomerization between GlcN6P and Fru6P, as in bacterial and eucaryal GlcN6P deaminases. Nonphosphorylated compounds GlcN and Fru did not act as substrates for this enzyme, and Glc6P could not replace Fru6P as a substrate for the amination reaction in the presence of ammonia (data not shown).
GlcN6P deaminase activity levels of GlmD Tk were determined by quantifying the generated Fru6P with an enzymecoupled assay using phosphoglucose isomerase and Glc6P dehydrogenase. In this assay, no activity was detected in a control experiment without phosphoglucose isomerase. This fact verified the production of Fru6P, but not Glc6P, from GlcN6P by GlmD Tk , as shown by the TLC analysis where the discrimination of Fru6P and Glc6P was somewhat difficult due to similar mobilities. This also implied that the enzyme does not exhibit phosphoglucose isomerase activity. The optimal pH of Glm-D Tk was 8.0 to 8.5. This enzyme was thermostable as expected, the optimal temperature was 95°C to 100°C, and the half-lives at 80°C and 90°C were determined to be 103 and 19 min, respectively. We then performed kinetic analysis of GlmD Tk for both forward (GlcN6P deamination/isomerization) and reverse (Fru6P amination/isomerization) reactions. The reverse activity was determined by measuring the amount of produced GlcN6P with a modified Morgan and Elson method (13). As the optimal pH for the reverse reaction was determined to be the same as that for the forward one, assays were commonly performed at pH 8.3. The enzyme followed typical Michaelis-Menten kinetics for both directions, indicating no homotropic allosteric properties. The kinetic constants obtained are summarized in Table 1. The values for the deaminase reactions of GlmD Tk were comparable to those of the GlcN6P deaminase from E. coli (k cat , 158 to 160 s Ϫ1 ; K m , 0.55 to 0.75 mM; and k cat /K m , 2.1 to 2.9 ϫ 10 5 M Ϫ1 s Ϫ1 ) (6,15). The value of k cat in the forward reaction was two times higher than that in the reverse reaction, and the K m value for GlcN6P was one and two orders lower than those for Fru6P and ammonia, respectively. As a result, the relative k cat /K m values for Fru6P and ammonia in the amination reaction were much lower (14.0% and 0.801%, respectively) than that for GlcN6P in the deamination reaction. These results clearly indicated that GlmD Tk kinetically favored deamination/isomerization of GlcN6P rather than its reverse reaction. This preference was the same as those of classical GlcN6P deaminases and was consistent Expression profiles of GlmD Tk in T. kodakaraensis. As described above, the glmD Tk gene was assumed to be cotranscribed with the upstream ABC transporter genes and the downstream glmA Tk , which corresponds to a transcript size of approximately 9,200 bp. Northern blot analysis using a specific probe against glmD Tk resulted in smeared signals of lower sizes (1,500 to 2,500 bp), probably due to degradation of the long mRNA (data not shown). We therefore performed RT-PCR for three overlapping segments of the transcript (Fig. 2) and confirmed the amplification of all three fragments as shown in Fig. 5A. This result supported a single transcriptional unit for the genes of the ABC transporter, glmD Tk and glmA Tk . Then, RNA dot blot hybridization was performed using the glmD Tk probe. In this analysis, a probe for glmS Tk encoding putative GlcN6P synthase (TK0809) was also applied in order to investigate the regulation of the interconversion between GlcN6P and Fru6P. It has previously been revealed that the expression of glmA Tk was induced by GlcNAc 2 , an end product of chitin degradation by ChiA Tk (27). As shown in Fig. 5B, almost no transcription of glmD Tk was detected in the absence of GlcNAc 2 , whereas transcription was strongly induced in GlcNAc 2 -containing medium, as expected. In contrast, glmS Tk was constitutively transcribed in the three growth conditions (no sugar, or addition of GlcNAc 2 or maltose) as in the case of the control DNA ligase gene (lig Tk ).
The expression of GlmD Tk in T. kodakaraensis was also examined at the protein level by Western blot analysis. Cells grown in media containing various sugars were used for the analysis, and the result is shown in Fig. 5C. The expression of GlmD Tk was not observed under the basal culture condition (Fig. 5C, lane 1), while it was induced by the addition of GlcNAc 2 (lane 2). Chitobiose (GlcN 2 ) and maltose did not act as inducers for GlmD Tk at all (Fig. 5C, lanes 3 and 4). This profile coincided with that of GlmA Tk and reflected the same tendency as that seen in the transcription analysis mentioned above. Clearly, the chitin-catabolic operon consisting of glmA Tk , glmD Tk , and ABC transporters was specifically regulated at the transcriptional level by GlcNAc 2 .

DISCUSSION
In this study, we revealed that one gene, TK1755, in the gene cluster for chitin degradation on the T. kodakaraensis genome, encoded a GlcN6P deaminase structurally distinct from previously known deaminases. The protein product (GlmD Tk ) displayed similarity to the sugar isomerase domain of GlcN6P synthases. GlmD Tk kinetically preferred the deamination of GlcN6P rather than the reverse ammonia-dependent amination of Fru6P (Table 1), supporting the function of this enzyme in amino sugar catabolism in vivo. Based on these facts together with our previous results (25)(26)(27)(28), we can now clearly envision the chitin metabolism in this hyperthermophile in its near entirety, as summarized in Fig. 6. GlmD Tk was estimated to have a role in the final step to allow entrance of the chitinderived catabolites into glycolysis. The highly clustered genes for this pathway were transcriptionally induced by GlcNAc 2   (14). Previous studies have indicated that GlcN6P deaminase and the isomerase domain of GlcN6P synthase share some general similarities despite the lack of significant homology between their primary structures (21,31). Both proteins have related nucleotide-binding folds, although GlcN6P deaminase has a dehydrogenase-like six-stranded fold, whereas the fold of GlcN6P synthase is a five-stranded flavodoxin type. They commonly catalyze 2R aldose-ketose isomerization by abstracting the C1 pro-R hydrogen of a substrate to form cis-enolamine, followed by reprotonation of C2 at the same re face of the intermediate. Therefore, the catalytic property of GlmD Tk as a GlcN6P deaminase is feasible. However, an intriguing difference between GlmD Tk and the isomerase domain of GlcN6P synthase is the ability of the former to accept exogenous ammonia. GlcN6P synthase lacks ammonia-dependent activity unlike other amidotransferases as described above, and in addition, it has been reported that the isomerase domain alone, prepared by limited proteolysis, did not show ammonia-dependent GlcN6P synthesis activity (10). Recent tertiary structural analysis of GlmS Ec has proposed that the binding of glutamine to the GAT domain promotes a conformational change of the protein that results in the opening of an intramolecular channel, through which the ammonia derived from glutamine migrates to the isomerase active site (29). The unique structure of the intramolecular channel seems to be the reason why this enzyme cannot gain access to free ammonia. In contrast, the reversible catalytic property of GlmD Tk implies that free ammonia is easily accessible to the active site of this enzyme. However it should be noted that the C-terminal decapeptide of GlmS Ec , playing a central role in the migration of ammonia, is highly conserved in GlmD Tk as well as in the GlcN6P synthases (29,31). In GlmD Tk , this region might have a function distinct from those in the usual GlcN6P synthases, such as in the recognition and incorporation of exogenous ammonia.
Besides the isomerase domain of GlcN6P synthase, GlmD Tk shares overall homology against proteins existing in a subset of Archaea and Bacteria. These homologs are also comprised of tandem SIS domains and classified as COG2222 in the Cluster of Orthologous Group of proteins database. Among them, the Pyrococcus homologs not only showed high homology to glm-D Tk but also were commonly located within the completely conserved gene cluster from dac to glmA for GlcNAc 2 catabolism. Although separated into two open reading frames at a different locus, P. furiosus harbors chitinase genes similar to ChiA Tk . Therefore, the cluster in P. furiosus is likely to be involved in chitin degradation, as in the case of T. kodakaraensis. Interestingly, P. abyssi and P. horikoshii also harbor this gene cluster despite the absence of a chitinase ortholog. We have previously reported that Gly Tk , which displays broad specificity to various ␤-1,4-glycosides (11), is encoded in this cluster (Fig. 2). The enzyme does not seem to participate in GlcNAc 2 metabolism (26), but the induction behavior was the same as those of the other clustered genes for chitinolysis (27). Although the physiological role of this ␤-glycosidase will have to be clarified, the common presence of the gene cluster, containing gly in P. abyssi and P. horikoshii, suggests an additional role of this cluster among members of the Thermococcales order. The cluster may be involved in the degradation of other cellular ␤-1,4-linked heteropolysaccharides. The proteins classified into COG2222 include AgaS encoded within the GalNAc catabolic cluster in E. coli. It has been reported that the bacterial metabolism of GalNAc was similar to that of GlcNAc (4,22). In this pathway, galactosamine-6-phosphate (GalN6P) was predicted to be converted to tagatose-6-phosphate by the function of AgaI corresponding to the classical GlcN6P deaminase NagB. Although the catalytic property of AgaS has not been investigated yet, our results raise the possibility that AgaS might be functional as an additional GalN6P deaminase to achieve efficient processing of GalNAc.
In this study, we identified and characterized a new type of GlcN6P deaminase from the hyperthermophilic archaeon T. kodakaraensis and hereby provided an overview of the chitin catabolic pathway to glycolysis in this organism. As the enzymes in this pathway, ChiA Tk , GlmD Tk , Dac Tk , and GlmA Tk , were all novel enzymes with high thermostability, it is expected that these enzymes can be useful catalysts for future conversion of the unused biomass chitin.