d-Xylose Degradation Pathway in the Halophilic Archaeon Haloferax volcanii

The pathway of d-xylose degradation in archaea is unknown. In a previous study we identified in Haloarcula marismortui the first enzyme of xylose degradation, an inducible xylose dehydrogenase (Johnsen, U., and Schönheit, P. (2004) J. Bacteriol. 186, 6198–6207). Here we report a comprehensive study of the complete d-xylose degradation pathway in the halophilic archaeon Haloferax volcanii. The analyses include the following: (i) identification of the degradation pathway in vivo following 13C-labeling patterns of proteinogenic amino acids after growth on [13C]xylose; (ii) identification of xylose-induced genes by DNA microarray experiments; (iii) characterization of enzymes; and (iv) construction of in-frame deletion mutants and their functional analyses in growth experiments. Together, the data indicate that d-xylose is oxidized exclusively to the tricarboxylic acid cycle intermediate α-ketoglutarate, involving d-xylose dehydrogenase (HVO_B0028), a novel xylonate dehydratase (HVO_B0038A), 2-keto-3-deoxyxylonate dehydratase (HVO_B0027), and α-ketoglutarate semialdehyde dehydrogenase (HVO_B0039). The functional involvement of these enzymes in xylose degradation was proven by growth studies of the corresponding in-frame deletion mutants, which all lost the ability to grow on d-xylose, but growth on glucose was not significantly affected. This is the first report of an archaeal d-xylose degradation pathway that differs from the classical d-xylose pathway in most bacteria involving the formation of xylulose 5-phosphate as an intermediate. However, the pathway shows similarities to proposed oxidative pentose degradation pathways to α-ketoglutarate in few bacteria, e.g. Azospirillum brasilense and Caulobacter crescentus, and in the archaeon Sulfolobus solfataricus.

D-Xylose, a constituent of the polymer xylan, is the major component of the hemicellulose plant cell wall material and thus one of the most abundant carbohydrates in nature. The utilization of D-xylose by microorganisms has been described in detail in bacteria and fungi, for which two different catabolic pathways have been reported. In many bacteria, such as Escherichia coli, Bacillus, and Lactobacillus species, xylose is converted by the activities of xylose isomerase and xylulose kinase to xylulose 5-phosphate as an intermediate, which is further degraded mainly by the pentose phosphate cycle or phosphoketolase pathway. Most fungi convert xylose to xylulose 5-phosphate via xylose reductase, xylitol dehydrogenase, and xylulose kinase. Xylulose 5-phosphate is also an intermediate of the most common L-arabinose degradation pathway in bacteria, e.g. of E. coli, via activities of isomerase, kinase, and epimerase (1).
Recently, by genetic evidence, a third pathway of xylose degradation was proposed for the bacterium Caulobacter crescentus, in analogy to an alternative catabolic pathway of L-arabinose, reported for some bacteria, including species of Azospirillum, Pseudomonas, Rhizobium, Burkholderia, and Herbasprillum (2,3). In these organisms L-arabinose is oxidatively degraded to ␣-ketoglutarate, an intermediate of the tricarboxylic acid cycle, via the activities of L-arabinose dehydrogenase, L-arabinolactonase, and two successive dehydration reactions forming 2-keto-3-deoxy-L-arabinoate and ␣-ketoglutarate semialdehyde; the latter compound is further oxidized to ␣-ketoglutarate via NADP ϩ -specific ␣-ketoglutarate semialdehyde dehydrogenase (KGSADH). 2 In a few Pseudomonas and Rhizobium species, a variant of this L-arabinose pathway was described involving aldolase cleavage of the intermediate 2-keto-3-deoxy-L-arabinoate to pyruvate and glycolaldehyde, rather than its dehydration and oxidation to ␣-ketoglutarate (4). Because of the presence of some analogous enzyme activities in xylose-grown cells of Azosprillum and Rhizobium, the oxidative pathway and its variant was also proposed as a catabolic pathway for D-xylose. Recent genetic analysis of Caulobacter crecentus indicates the presence of an oxidative pathway for D-xylose degradation to ␣-ketoglutarate. All genes encoding xylose dehydrogenase and putative lactonase, xylonate dehydratase, 2-keto-3-deoxylonate dehydratase, and KGSADH were found to be located on a xylose-inducible operon (5). With exception of xylose dehydrogenase, which has been partially characterized, the other postulated enzymes of the pathway have not been biochemically analyzed.
The pathway of D-xylose degradation in the domain of archaea has not been studied so far. First analyses with the halophilic archaeon Haloarcula marismortui indicate that the initial step of D-xylose degradation involves a xylose-inducible xylose dehydrogenase (6) suggesting an oxidative pathway of xylose degradation to ␣-ketoglutarate, or to pyruvate and glycolaldehyde, in analogy to the proposed oxidative bacterial pentose degradation pathways. Recently, a detailed study of D-arabinose catabolism in the thermoacidophilic crenarchaeon Sulfolobus solfataricus was reported. D-Arabinose was found to be oxidized to ␣-ketoglutarate involving D-arabinose dehydrogenase, D-arabinoate dehydratase, 2-keto-3-deoxy-D-arabinoate dehydratase, and ␣-ketoglutarate semialdehyde dehydrogenase (3).
In this study, we present a comprehensive analysis of the complete D-xylose degradation pathway in the halophilic archaeon Haloferax volcanii. This halophilic archaeon was chosen because it exerts several suitable properties for the analyses. For example, it can be cultivated on synthetic media with sugars, e.g. xylose, an advantage for in vivo labeling studies in growing cultures. Furthermore, a shotgun DNA microarray of H. volcanii is available (7) allowing the identification of xyloseinducible genes, and H. volcanii is one of the few archaea for which an efficient protocol was recently described to generate in-frame deletion mutants.
Accordingly, the D-xylose degradation pathway was elucidated following in vivo labeling experiments with [ 13 C]xylose, DNA microarray analyses, and the characterization of enzymes involved and their encoding genes. The functional involvement of genes and enzymes was proven by constructing corresponding in-frame deletion mutants and their analysis by selective growth experiments on xylose versus glucose. The data show that D-xylose was exclusively degraded to ␣-ketoglutarate involving xylose dehydrogenase, a novel xylonate dehydratase, 2-keto-3-deoxyxylonate dehydratase, and ␣-ketoglutarate semialdehyde dehydrogenase.

Growth of H. volcanii
DS70 strain H26 (⌬pyrE2), obtained from Thorsten Allers (University of Nottingham, United Kingdom), contained a uracil auxotrophy required for the construction of gene deletion mutants. The organism was grown aerobically at 42°C on synthetic medium (7) containing 25 mM either D-xylose or D-glucose and uracil (50 g/ml). Growth was followed by an increase in absorbance at 600 nm; consumption of xylose and glucose during growth was determined as described (6). For gene deletion experiments involving the so-called pop-in/ pop out strategy (see below) and for growth of deletion mutants (⌬HVO_B0027, ⌬HVO_B0028, ⌬HVO_B0029, ⌬HVO_B0038A, and ⌬HVO_B0039), a complex medium (8) containing 1% casamino acids was used. For pop-out selection, the medium was supplemented with 50 g/ml each 5-fluoroorotic acid and uracil.

In Vivo [ 13 C]Xylose Labeling Experiments
H. volcanii was grown aerobically at 42°C in 100-ml Erlenmeyer flasks (shaken at 150 rpm) filled with 20 ml of synthetic medium, containing [1-13 C]xylose or [2-13 C]xylose (each 25 mM). Cells were harvested during exponential growth phase (A 600 , 1.4) by centrifuging 20 ml of culture broth at 48,000 ϫ g and 4°C for 30 min. The biomass pellet was washed twice with 1 ml of 0.9% NaCl, hydrolyzed in 1.5 ml of 6 M HCl for 24 h at 110°C in sealed 2-ml Eppendorf tubes, and desiccated overnight in a heating block at 85°C under a constant air stream. The hydrolysate was dissolved in 50 l of 99.8% dimethyl formamide and transferred into a new Eppendorf cup within a few seconds. For derivatization, 30 l of N-methyl-N-(tert-butyldimethylsilyl)-trifluoroacetamide was added, which readily silylates hydroxyl groups, thiols, primary amines, amides, and carboxyl groups (9), and the mixture was incubated at 550 rpm and 85°C for 60 min. 1 l of the derivatized sample was injected into a Series 8000 gas chromatograph, combined with an MD 800 mass spectrometer (Fisons Instruments), and analyzed as described earlier (10). Gas chromatograph temperature profile was 150°C for 2 min, increased to 280°C at 10°C per min, and held at 280°C for 3 min. Injector temperature was set to 280°C, split ratio to 1:20, flow rate to 2 ml per min, and carrier gas was helium in a SPB-1 column (30 m ϫ 0.32 mm ϫ 0.25 m; Agilent Technologies). Electron impact spectra were obtained at Ϫ70 eV with a full scan ranging from 70 to 560 m/z and a solvent delay of 4 min.
Mass spectra of the derivatized amino acids alanine, aspartate, glutamate, proline, and threonine were corrected for the natural abundance of all stable isotopes and unlabeled biomass from inoculum. Glycine, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, serine, tyrosine, and valine were not used in this study, whereas arginine, asparagine, cysteine, glutamine, and tryptophan were not detectable. The labeling patterns of the detected amino acids were direct and quantitative evidence for metabolic pathways leading from carbon substrate to the respective precursors.

DNA Microarray Analysis
H. volcanii was grown in synthetic medium as described (7), with either 0.25% (w/v) glucose or 0.25% xylose (w/v) as sole carbon and energy sources. Cells were harvested during midexponential growth phase (3-5 ϫ 10 8 cells/ml). RNA isolation, cDNA synthesis and labeling, DNA microarray hybridization, scanning, and data analysis were performed as described (7). Four biological replicates were performed. A dye-swap was included into the experimental design. All signals were visually inspected, and artifacts were removed. The results were exported from the scanning software (Genepix Pro 3.0), and further analysis was performed with Excel. All signals were removed that had a low or moderate signal intensity in the "xylose cDNA" (Ͻ1000 arbitrary units) to exclude all genes with low or moderate transcript levels during growth on xylose. Average values were calculated, and the results were sorted according to the induction level in xylose medium. Initially, only spots were included in the analysis that represent highly induced genes and were quantified in at least two of the four replicates. However, as first blast searches against the contigs of the provisional H. volcanii genome revealed that most of these clones were concentrated on two genomic regions, all clones representing these regions with at least 6-fold induction levels were included, even if they were quantified only in a single experiment. It should be noted that singletons can be regarded as "replicates" when they overlap or indicate induction of several genes of one operon. A table with clone identifiers and genomic localizations was prepared (Table 2), and the two genomic regions represented by 25 clones were annotated.

Annotation of Genome Regions
The regions from 80,000 to 110,000 and 328,000 to 340,000 of contig 275 of the provisional H. volcanii genome sequence (March, 2006) were annotated. In a first step, all open reading frames (ORFs) with at least 100 codons were identified in all six reading frames. Haloarchaea have a high guanine-cytosine content and a very low fraction of stop codons, leading to an "overprediction problem" (11). Two approaches were used to separate real genes from artificial ORFs. First, a codon preference analysis of the two genomic regions was performed with the program "codon preference" of the HUSAR program package, which had been shown earlier to aid in the identification of genes as well as the correct regions of translation initiation (12). In addition, blast searches of the predicted proteins against protein sequence libraries (TrEMBL, Swiss Prot) were performed. Furthermore, multiple sequence analyses were performed with ClustalW to enhance the quality of start codon selection. In summary, this led to an annotation, which was virtually identical to the currently available annotation of the H. volcanii genome. One gene (between HVO_B0038 and HVO_B0039) coding for a putative protein of the enolase superfamily was not included in the official annotation. During this study, this gene, designated HVO_B0038A, was identified as a functional xylonate dehydratase.

Preparation of Crude Extracts and Enzyme Assays
Crude extracts were prepared from late log phase xylosegrown cells and stored at Ϫ20°C in 0.1 M Tris-HCl, pH 7.5, containing 3.4 M NaCl (buffer A). Cells were disrupted by sonication, and after centrifugation, enzyme activities were determined in the supernatant. Protein was determined by the Bradford method. All enzyme assays were performed at 42°C. Dehydrogenase activities were assayed by measuring following the rate of reduction of NADP ϩ or NAD ϩ at 340 nm. The standard assay mixture for XDH activity contained 0.1 M Tris-HCl, pH 8.

Conversion of D-Xylonate to ␣-Ketoglutarate or to Pyruvate in Crude Extracts
The NAD(P) ϩ -dependent conversion of D-xylonate to ␣-ketoglutarate was assayed in crude extracts of D-xylose-and D-glucose-grown cells of H. volcanii. The assay mixture contained 0.1 M Tris-HCl, pH 7.5, 2.5 M KCl, 50 mM MgCl 2 , 4 mM NADP ϩ or NAD ϩ , 20 mM D-xylonate, and crude extract. Samples were incubated at 42°C up to 120 min; the reaction was stopped on ice, and the amount of ␣-ketoglutarate formed was quantified by measuring the reductive amination of ␣-ketoglutarate to glutamate at 365 nm. The reaction mixture (1 ml) contained 0.1 M Tris-HCl, pH 7.5, 0.3 mM NADH, 40 mM ammonium chloride, 0.94 units of glutamate dehydrogenase (Sigma) and 40 l of the samples.
The xylonate-dependent formation of pyruvate was tested with crude extracts from D-xylose-grown cells of H. volcanii. The assay mixture contained 0.1 M Tris-HCl, pH 7.5, 2.5 M KCl, 50 mM MgCl 2 , 20 mM D-xylonate, and crude extract. After incubation at 42°C up to 120 min, samples were cooled on ice, and the pyruvate formed was measured by reduction to lactate with lactate dehydrogenase at 365 nm and 20°C. The reaction mixture (1 ml) contained 0.1 mM potassium phosphate buffer, pH 7.2, 0.3 mM NADH, 3 units of lactate dehydrogenase, and 40 l of the samples.

Purification of Enzymes from H. volcanii
For purification of XAD and KGSADH, cells were suspended in 0.1 M Tris-HCl, pH 8.0, containing 2 M ammonium sulfate (buffer B); for XAD, the buffer was supplemented with 50 mM MgCl 2 . Cells were disrupted by passing through a French pressure cell followed by centrifugation for 90 min at 100,000 ϫ g at 4°C. Further purification steps were as follows.
XAD-The 100,000 ϫ g supernatant was applied to a phenyl-Sepharose column (60 ml) equilibrated with buffer B containing 50 mM MgCl 2 . After washing the column, protein was eluted with a linear gradient of decreasing ammonium sulfate concentration with 0.1 M Tris-HCl, pH 8.0, containing 50 mM MgCl 2 . Fractions containing the highest enzyme activity were pooled and concentrated to 1 ml by ultrafiltration (cutoff 10 kDa). This solution was applied to a Superdex 200 HiLoad gel filtration column (1.6 ϫ 60 cm) that had been equilibrated with 50 mM Tris-HCl, pH 7.5, containing 2 M KCl. The purity of fractions containing activity was checked by SDS-PAGE, and the 52-kDa protein band was analyzed by N-terminal amino acid sequencing (14). A further purification step involves chromatography on a Sepharose CL4B column (5 ml). Protein samples obtained from Superdex column were dialyzed against a 1000-fold volume of buffer B containing 50 mM MgCl 2 and then applied to CL4B column equilibrated with the same buffer. Protein was eluted with 0.1 M Tris-HCl, pH 8.0, containing 50 mM MgCl 2 . After this purification step, XAD was purified about 100-fold to a specific activity of 1.1 units/mg with xylonate.
KGSADH-The 100,000 ϫ g supernatant was applied to a phenyl-Sepharose column (60 ml) equilibrated with buffer B. After washing the column with buffer B, the protein was eluted with a linear gradient of decreasing ammonium sulfate concentration with 0.1 M Tris-HCl, pH 8.0, containing 10% glycerol (buffer C). Fraction containing the highest enzyme activity was pooled and concentrated to 1 ml by ultrafiltration (cutoff 10 kDa). The concentrated protein solution was applied to a Superdex 200 HiLoad gel filtration column (1.6 ϫ 60 cm) that had been equilibrated with 50 mM Tris-HCl, pH 7.5, containing 2 M NaCl and 10% glycerol (buffer D). Fractions containing enzyme activity were diluted with buffer containing ammonium sulfate to a final concentration of 2 M ammonium sulfate in 0.1 M Tris-HCl, pH 8.0, and were applied to a Sepharose CL4B (5 ml) column. After washing the column with buffer B, protein was eluted with a linear gradient of decreasing ammonium sulfate concentration with buffer C. Eluted fractions containing enzyme activity indicated purified enzyme and were stored at 4°C. The purity of the preparation was checked by SDS-PAGE, and the N-terminal amino acid sequence of the purified enzyme was determined.

Cloning and Heterologous Expression of Genes
HVO_B0028, HVO_B0029, HVO_B0038A, and HVO_B0039 coding for XDH, putative glucose-fructose oxidoreductase (GFOR), XAD, and KGSADH, respectively, were amplified by PCR from genomic DNA using ProofStart polymerase (Qiagen) or Pwo polymerase (peqlab) (supplemental Table 1S). The PCR products were each cloned into pET19b. The resulting plasmid were harvested from E. coli JM109, sequenced, and transformed into E. coli Rosetta(DE3)-pLysS expression strain (Stratagene). For expression, cells were grown in Luria-Bertani medium at 37°C. The expression was started by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside. After 4 h of further growth, cells were harvested by centrifugation.

Purification and Reactivation of Recombinant Enzymes
Recombinant XDH and putative GFOR, XAD, and KGSADH were purified from transformed E. coli cells, which were disrupted by passages through French pressure cells and centrifuged. The first purification step used for all enzymes was Ni-NTA chromatography. The proteins, which were specifically eluted with imidazole, do not show catalytic activity due to misfolding after expression in the nonhalophilic E. coli. To generate catalytic active recombinant enzymes, they were reactivated following unfolding by urea and refolding in the presence of salts and substrates by a modified protocol according to Ref. 15. Specific conditions for further purification steps were as follows.
Putative GFOR-After Ni-NTA chromatography, selected fractions were concentrated with a 10-kDa cutoff filter and unfolded in the presence of urea. Refolding was initiated by dialysis in 50 mM Tris-HCl, pH 7.5, containing 10% glycerol, 3 M KCl, 0.5 mM D-xylose, 0.1 mM NAD ϩ , 2 mM reduced glutathione, and 0.2 mM oxidized glutathione overnight. Protein was concentrated and applied to a Superdex 200 HiLoad 16/60 gel filtration column and eluted with an isocratic flow in buffer D. The refolded enzyme showed xylose dehydrogenase activity measured as xylose-dependent reduction of NADP ϩ .
XAD-After Ni-NTA chromatography selected fractions were concentrated with a 30-kDa cutoff filter and unfolded by urea treatment in the absence of EDTA. Refolding was initiated by dialysis in 50 mM Tris-HCl, pH 7.5, containing 2 M KCl, 50 mM MgCl 2 , and 0.25 mM D-xylonate. However, no XAD activity could be detected after 24 or 48 h.
KGSADH-After Ni-NTA chromatography selected fractions were concentrated with a 30-kDa cutoff filter in unfolding buffer without EDTA. Refolding was initiated by dialysis in 50 mM Tris-HCl, pH 8.0, containing 2.5 M KCl, 10% glycerol, 0.05 mM NADP ϩ , 1 mM glutaraldehyde, 1.5 mM glutathione (reduced), and 0.15 mM glutathione (oxidized). Protein was concentrated and applied to a Superdex 200 HiLoad 16/60 gel filtration column, and was eluted with an isocratic flow in buffer D. After this procedure, catalytic active enzyme was obtained measured as glutaraldehyde-dependent reduction of NADP ϩ .

RT-PCR Experiments
RNA from xylose-grown cells of H. volcanii was isolated in the exponential growth phase at A 600 of 1.6. Cells were cooled to 4°C, harvested by centrifugation at 10,000 ϫ g for 15 min, and disrupted by freezing and crushing under liquid nitrogen by passing them through a Qiashredder column (Qiagen, Hilden, Germany). RNA was extracted by using the RNeasy mini kit and RNase-free DNase set (Qiagen) as specified by the manufacturer. RT-PCR was carried out by the Qiagen OneStep RT-PCR kit using 5-40 ng of total RNA (supplemental Table  2S).

Construction of In-frame Deletion Mutants of H. volcanii
The construction of in-frame deletion mutants (⌬HVO_ B0028, ⌬HVO_B0029, ⌬HVO_B0038A, ⌬HVO_B0027, or ⌬HVO_B0039) was performed using an optimized version of the so-called pop-in/pop-out method (16) as described recently (17). In each case two PCR fragments of about 500 bp (supplemental Table 3S) were amplified. PCR fragments include the up-and downstream regions of the corresponding gene. The two fragments were fused by PCR, generating an in-frame deletion product. The products were ligated into pTA131 (16), sequenced, and transformed in H. volcanii H26 ⌬pyrE2. Growth in uracil-free medium was used to select for clones that had integrated the vector into their genome via homologous recombination at the indicated locus (pop-in). Clones that had experienced a second homologous recombination event (popout) were selected by growth on complex medium containing D-Xylose Degradation Pathway in Haloferax volcanii 5-fluororotic acid, which is toxic for cells containing the pyrE gene. Clones obtained after pop-out steps were either wild type or the in-frame deletion variant of the indicated open reading frame. The mutated clones were identified by PCR and additionally verified by Southern blot analysis. For Southern blot analysis the isolation of genomic DNA was performed by using Qiagen DNeasy kit (Qiagen) followed by digestion with restriction endonucleases (supplemental Table 3S). Hybridization with a probe was performed at 50°C. Hybridization probe was labeled with a digoxigenin DNA labeling kit (Roche Diagnostics). Chemiluminescent detection of the labeled fragments was performed as described by the manufacturer (Roche Diagnostics).

RESULTS
The degradation pathway of D-xylose in H. volcanii was analyzed as follows: (i) in vivo 13 C-labeling experiments following growth on [ 13 C]xylose; (ii) enzyme measurements in cell extracts; (iii) DNA microarray analyses to identify xylose-inducible genes; (iv) purification and characterization of enzymes involved in xylose degradation; and (v) proof of functional involvement of genes and enzymes in xylose degradation by analyzing the corresponding in-frame deletion mutants.

In Vivo D-Xylose Degradation Pathway in H. volcanii
H. volcanii was grown on a synthetic medium on D-xylose or D-glucose as the sole carbon and energy source. The cells grew exponentially with a doubling time of 11 h on xylose and 6.5 h on glucose, respectively, up to a maximal cell density of ⌬A 600 of 1.6 for both sugars. During exponential growth, 14 mM D-xylose and 11 mM D-glucose were consumed (data not shown).
To identify the xylose degradation pathway in vivo, we performed labeling experiments with [ 13 C]xylose in growing cultures of H. volcanii. Two possible oxidative pathways for D-xylose degradation can be considered. D-Xylose might be directly converted to ␣-ketoglutarate in analogy to arabinose oxidation pathways in Azospirillum brasilense and Sulfolobus solfataricus, or it might be the converted to pyruvate and glycolaldehyde as proposed for few Pseudomonas and Azospirillum species (3,4). To determine which pathway of xylose degradation is in vivo operative in H. volcanii, the organism was grown on [1-13 C]and [2-13 C]xylose, and the labeling patterns of proteinogenic amino acids, isolated from exponentially grown cells, were analyzed by gas chromatography-mass spectrometry. The labeling patterns of glutamate and proline with 74 -86% label at position 1 (Table 1) clearly indicate an almost exclusive direct degradation of D-xylose toward ␣-ketoglutarate, the precursor of these two amino acids. The remaining unlabeled fraction can be explained by tricarboxylic acid cycle activity, whereas aspartate and threonine, which are derived from oxaloacetate, are almost completely unlabeled because position 1 of ␣-ketoglutarate is lost as CO 2 during conversion to succinate. Direct degradation of xylose to pyruvate yielding label at position 1 can be excluded, because significant label, if any, was found exclusively at position 2 or 3 of alanine i.e. 1-3 and 2-3 fragments are equally labeled within the resolution of the analysis. Using [2-13 C]xylose, the labeled carbon was not lost during conversion of ␣-ketoglutarate to succinate and led to oxaloacetate labeled equally at positions 1 and 4 ( Table 1). The unlabeled fraction of the glutamate fragment can again be explained by tricarboxylic acid cycle activity. The gluconeogenic reaction from malate or oxaloacetate toward pyruvate can now be tracked and led to alanine labeled at position 1, whereas position 4 of oxaloacetate or malate was lost as CO 2 ( Table 1). The [ 13 C]xylose labeling experiments clearly indicate that D-xylose in H. volcanii was exclusively oxidized to ␣-ketoglutarate in vivo and thus excludes a significant direct degradation to pyruvate, as a possible alternative route involving aldol cleavage.

Xylose-inducible Enzyme Activities in Cell Extracts of H. volcanii
The 13 C-labeling data were supported by measuring xyloseinducible enzyme activities in crude extracts. Extracts of xylose-grown cells catalyzed the NADP ϩ -dependent oxidation of xylose at a 3-fold higher specific activity (0.23 Ϯ 0.009 units/ mg) as compared with glucose-grown cells (0.06 Ϯ 0.006 units/ mg) indicating the presence of an xylose-inducible D-xylose dehydrogenase. Furthermore, extracts of xylose-grown cells catalyzed the NADP ϩ -and NAD ϩ -dependent conversion of D-xylonate to ␣-ketoglutarate, at 15-fold higher activities (0.025 Ϯ 0.003 units/mg) as compared with glucose-grown cells. This inducible multistep conversion involves two dehydratase steps, one with xylonate and a second with 2-keto-3deoxyxylonate as substrates, and the NAD(P) ϩ -dependent oxidation of ␣-ketoglutarate semialdehyde to ␣-ketoglutarate. Therefore, in accordance with the inducible overall conversion of xylonate to ␣-ketoglutarate, a 2-fold induced xylonate dehydratase activity (0.013 Ϯ 0.002 unit/mg) and a highly (Ͼ100fold) induced NADP ϩ -dependent KGSADH activity (0.7 Ϯ 0.022 unit/mg, measured with glutaraldehyde as substrate) were detected in xylose-grown cells. Furthermore, in accordance with the in vivo labeling experiments, which exclude the formation of pyruvate from D-xylose, cell extracts of xylose-grown cells did not catalyze the conversion of D-xylonate to pyruvate. To identify xylose-inducible genes and their encoded enzymes, DNA microarray experiments were performed followed by enzyme characterization.

DNA Microarray Analyses
A shotgun DNA microarray with a 1-fold coverage of the H. volcanii genome was constructed a few years ago (7). It consists of genomic regions with an average length of 1.5 kbp, which typically represent one or two genes. It has been successfully applied for the analysis of metabolic regulation, translational control, and regulatory mutants as well as the identification of a regulated promoter for conditional gene expression (Ref. 18 and references therein). The DNA microarray is comprised of 3000 clones, which have been sequenced from both ends, allowing their localization in the genome of H. volcanii.
For the identification of xylose-inducible genes, H. volcanii cultures were grown in synthetic medium with either xylose or glucose as sole carbon and energy source. The cultures were harvested at mid-exponential growth phase (about 4 ϫ 10 8 cells/ml), and the transcriptomes were compared using the DNA microarray as described (7). Four biological replicates were performed. It was expected that the transcript levels of genes encoding xylose degradation proteins should vary considerably in the two media. Therefore, the analysis was concentrated on clones that showed a high induction level during growth on xylose to exclude genes that moderately induced transcript levels, which are caused by indirect effects, e.g. different growth rates during growth on xylose and glucose. Inspection of the sorted results (Table 2) revealed that no transcripts existed with induction levels between 2.5-and 6-fold. Therefore, a 6-fold induction was chosen as a threshold, and all clones with higher average induction levels that could be quantified in at least two of the four replicates were tabulated. This led to the identification of 20 clones with induction levels between 6-and 30-fold.
At the time of the analysis, the genome sequencing project of H. volcanii was underway, and a provisional sequence comprised of several hundred contigs was available at the TIGR website. The 20 clones were compared with the provisional  OCTOBER 2, 2009 • VOLUME 284 • NUMBER 40 genome sequence using blast searches, and it turned out that 16 of them represented two regions of the genome. As the complete genome sequence was released, it was found that both regions are localized on plasmid pHV3. The four additional clones represented four different genomic regions and were quantified only in two replicates. Therefore, they were regarded as false positives and not further investigated. Table 2 summarizes the 20 clones, induction levels, genomic localizations, and represented genes. All clones were added to Table 2, which represent one of the two regions, show an induction level of at least 6-fold, and were quantified in only one of the four replicates (nine clones). It should be noted that these singletons can be informative and regarded as replicates when they overlap or when they indicate induction of several genes of one operon. For example, the first six clones of Table 2 are all singletons, but all six of them show the induction of genes for a single ABC transporter. The two genomic regions are schematically shown in Fig. 1, and the localizations of the 25 microarray clones are indicated. The analysis revealed 21 genes to be highly induced during growth on xylose. Several of these genes overlap or have intergenic regions that are too small to harbor a promoter. Thus, they probably form operons and are transcribed into polycistronic mRNAs, e.g. HVO_B0021 to HVO_B0024, HVO_B0027 to HVO_B0029, HVO_B0034 to HVO_B0038, and HVO_B0228 to HVO_B0231. Cotranscription of HVO_B0027, HVO_B0028, and HVO_B0029 was proven by RT-PCR experiments performed for three overlapping segments of the transcript. Amplification of all three fragments was confirmed (Fig.  2). This result supports a single transcriptional unit for HVO_B0027, HVO_B0028, and HVO_B0029 and suggests that these genes are regulated by a single promoter situated upstream of HVO_B0027.

D-Xylose Degradation Pathway in Haloferax volcanii
The annotations of the xyloseinduced genes (Table 3) include putative ABC transporters and enzymes likely to be involved in xylose (sugar) degradation. Surprisingly, three different clusters of ABC transporters (HVO_B0021-24, HVO_B0034-38, and HVO_B0227-230)  Table  2. a, first genomic region; b, second genomic region. encoded by 13 genes are highly induced in xylose medium. In addition, transcripts of eight further genes encoding enzymes were induced, including a gene encoding a putative xylose dehydrogenase (HVO_B0028). However, most genes had annotations that were rather vague, e.g. sugar epimerase and aldehyde dehydrogenase, or likely to be wrong, e.g. glucose-fructose oxidoreductase.
Taken together, the microarray analysis led to the identification of a very small number of genes with highly induced transcript levels during growth on xylose, including genes for ABC transporters and genes probably encoding cytoplasmic enzymes. To characterize the functional roles of the latter gene products, a careful analysis was performed, including enzyme isolation and characterization, construction of in-frame deletion mutants, and analyses of their phenotypes in growth experiments, for example.

Enzymes and Their Encoding Genes Functionally Involved in Xylose Degradation
XDH-Xylose-grown cells contained inducible xylose dehydrogenase activity. Both HVO_B0028 and HVO_B0029, which were highly induced in a single transcriptional unit, are likely candidates coding for functional xylose dehydrogenase. Both HVO_B0028, annotated as putative D-xylose dehydrogenase, and HVO_B0029, annotated as putative GFOR, showed high sequence identity to characterized D-xylose dehydrogenase from H. marismortui (6).
The catalytic properties of the encoded proteins were determined after cloning and expression in E. coli of both ORFs as histidine-tagged fusion proteins. The recombinant halophilic proteins were purified by Ni-NTA chromatography as catalytically inactive proteins, and catalytically active enzymes were obtained after unfolding by urea and refolding in the presence of salts and the substrates xylose and NADP ϩ or NAD ϩ .
The reactivated HVO_B0028 gene product showed NADP ϩdependent xylose dehydrogenase activity of 24.1 Ϯ 0.6 units/ mg. The apparent molecular mass of native enzyme was 165 Ϯ 7 kDa. SDS-PAGE revealed a 55.5 Ϯ 3-kDa subunit, which is significantly higher than the calculated molecular mass of 42.3 kDa. This overestimation of apparent molecular masses of halophilic proteins on SDS is a common feature reported for several halophilic proteins (6,19). The data suggest an ␣ 4 oligomeric structure of xylose dehydrogenase from H. volcanii. The purified enzyme catalyzed the oxidation of D-xylose with both NADP ϩ and NAD ϩ with a 12-fold lower apparent K m for NADP ϩ compared with NAD ϩ indicating NADP ϩ to be the physiological electron acceptor. The enzyme was highly specific for D-xylose; D-glucose was used at significantly lower catalytic efficiency, 130-fold (Table 4). L-Arabinose, D-arabinose, D-ribose, D-mannose, and L-mannose were not oxidized at significant rates.
The purified reactivated HVO_B0029 gene product catalyzed NADP ϩ -dependent oxidation of D-xylose at 10 Ϯ 1 units/mg (at 25 mM xylose). From Lineweaver-Burk plots an apparent K m for D-xylose of 89 Ϯ 1 mM was extrapolated. 20% of activity was observed with 25 mM glucose. The enzyme was specific for NADP ϩ with an apparent K m of 0.75 Ϯ 0.06 mM, and no activity was found with NAD ϩ as cofactor.
The gene products of both HVO_B0028 and HVO_B0029 showed D-xylose dehydrogenase activity. To identify the enzyme functionally involved in D-xylose catabolism, in-frame deletion mutants of both ORFs were constructed using the recently developed so-called pop-in/pop-out method. As an example for this method, the genomic organizations of the HVO_B0028 in the wild type and the deletion mutant are shown in Fig. 3a. Deletion was confirmed by Southern blot analysis (Fig. 3b). The phenotypes of wild type and mutant were analyzed by growth studies on synthetic media containing xylose or glucose as sole energy and carbon substrates. ⌬HVO_B0028 could not grow on xylose but grew unaffected on glucose as the wild type (Fig. 4, a and b). However, ⌬HVO_B0029 grew equally well on both xylose and glucose (Fig. 4c). These data clearly indicate that HVO_B0028 rather than HVO_B0029 is essential for xylose degradation; thus, HVO_B0028 encodes the physiologically relevant XDH.
XAD-XAD was purified from D-xylose-grown cells by chromatography on phenyl-Sepharose, gel filtration, and Sepharose CL4B column, yielding a 100-fold enriched protein with a specific activity of 1.1 units/mg. After gel filtration step on Superdex, the 52 Ϯ 3-kDa protein band was analyzed by N-terminal sequencing. The sequence, VEQAKLNDPNAEYTMRDPL, matched exactly with the deduced N-terminal amino acid sequence of a protein, encoded by an ORF located between HVO_B0038 and HVO_B0039, which was not annotated so far. This ORF, which we designated as HVO_B0038A, showed high xylose-specific induction of transcription by microarray analyses. To prove the functional involvement of

annotation of xylose induced genes in H. volcanii, including annotation of genes functionally involved in xylose degradation (this study, marked in boldface)
HVO_B0038A in xylose degradation, an in-frame deletion mutant was constructed. The mutant had lost its ability to grow on xylose, but it showed the same phenotype as the wild type during growth on glucose (Fig. 4d). Thus, HVO_B0038A represents the gene coding for functional D-xylonate dehydratase XAD in xylose degradation and was annotated as xad. HVO_B0038A codes for a protein of 412 amino acids with a calculated molecular mass of 45.5 kDa. By gel filtration the molecular mass of the native enzymes was determined as 340 Ϯ 25 kDa suggesting a homo-octameric structure. XAD catalyzed the dehydration of D-xylonate at a similar catalytic efficiency as compared with D-gluconate (Table 4). D-Galactonate was not utilized.
2-Keto-3-deoxyxylonate Dehydratase (KDXD)-DNA microarray analyses indicate a high degree of up-regulation of transcription of HVO_B0027, as part of the transcription unit together with HVO_0028 encoding XDH. To prove the functional involvement of HVO_B0027 in xylose catabolism of H. volcanii, an in-frame deletion mutant was constructed and its phenotype analyzed. The mutant could not grow on xylose, but growth on glucose was unaffected (Fig. 4e), indicating HVO_B0027 to be essential for xylose catabolism.
KGSADH-KGSADH was purified from xylose grown cells from H. volcanii by three chromatographic steps to a specific activity of 7.14 Ϯ 0.72 units/mg, measured as NADP ϩ reduction with glutaraldehyde. The enzyme showed a high preference for NADP ϩ (K m 0.03 Ϯ 0.002 mM) over NAD ϩ (K m 2.6 Ϯ 0.55 mM) as cofactor. The apparent molecular mass of native enzyme was 233 Ϯ 15 kDa; SDS-PAGE revealed a 75 Ϯ 3-kDa band (Table 4). Based on N-terminal amino acid sequence, MTDMSKNYVNGEXVXSEXG, of the purified protein, HVO_B0039, which was annotated as aldehyde dehydrogenase, was identified as a gene encoding KGSADH. In accordance Northern blot analysis revealed that HVO_0039 is highly expressed during growth on xylose, whereas its transcript could not be detected in glucose-grown cells (data not shown).
HVO_B0039 was expressed as 77 Ϯ 3-kDa histidine-tagged fusion protein in E. coli. The recombinant protein was purified by affinity chromatography unfolded in urea and refolded in the presence of KCl and substrates and a gel filtration step. This procedure yielded a catalytically active enzyme, catalyzing the NADP ϩ -specific dehydrogenation of the substrates ␣-ketoglutarate semialdehyde at 0.12 unit/mg (single measurements and of glutaraldehyde at 0.56 Ϯ 0.28 unit/mg) (Table 4).
Furthermore, an in-frame deletion mutant of KGSADH gene was constructed; the mutant did not show substantial growth on xylose (Fig. 4f) but grew on glucose as the wild type proving HVO_B0039 to be essential for xylose catabolism by encoding a functional KGSADH.

DISCUSSION
The elucidation of a xylose degradation pathway in the halophile H. volcanii reported in this study required a multidisciplinary approach involving in vivo labeling experiments, DNA microarray analyses, and characterization of native and recombinant enzymes. Furthermore, the recently established method to construct in-frame deletion mutants of candidates genes was used to prove their functional involvement in D-xylose catabolism. The analyses led to the identification of the first D-xylose degradation pathway in archaea, in which xylose is converted to ␣-ketoglutarate, an intermediate of the tricarboxylic acid cycle. The pathway involves XDH, a novel XAD, KDXD, and KGSADH (Fig. 5). A putative lactonase, proposed by genome context analysis, was included in Fig. 5. This pathway of D-xylose degradation differs from the "classical" xylose degradation pathway in  in archaea, and homologs of the encoding genes xylA and xylB were not found in available archaeal genomes. In the following, the enzymes involved in the oxidative xylose degradation pathway in H. volcanii will be discussed in compar-  OCTOBER 2, 2009 • VOLUME 284 • NUMBER 40 ison with characterized and putative homologs in archaea and bacteria.

D-Xylose Degradation Pathway in Haloferax volcanii
XDH from H. volcanii showed 59% sequence identity to XDH (rrnAC3034) from H. marismortui, which we have previously characterized as a new type of dehydrogenase (6) that belongs to the GFO/IDH/MocA family (Pfam02894). XDHs from both halophiles show similar molecular properties being homotetramers and show similar preference for NADP ϩ as cofactor. However, H. marismortui XDH effectively utilized D-ribose, which was not converted by the Haloferax enzyme. An XDH ortholog, Hlac_2241, with high sequence identity (56%) was also found in sugar utilizing Halorubrum lacusprofundi. XDH from H. volcanii was not related (11% sequence identity) to the XDH, CC0821, from the bacterium C. crescentus. This bacterial XDH, which belongs to the short chain dehydrogenase/reductase family, was partially characterized; it is composed of 30-kDa subunits; the oligomeric state of the native enzyme was not reported (5). The enzyme was highly specific for D-xylose and NAD ϩ , and other pentose or hexoses as well as NADP ϩ as cofactor were not used. Thus, the bacterial XDH has a smaller subunit size, 30 kDa versus 42.3 kDa, and a different cofactor specificity, NAD ϩ versus NADP ϩ , as compared with archaeal XDH from H. volcanii.
For putative D-xylonolactonase in H. volcanii, HVO_B0030 was annotated as a member of senescence marker protein-30(SMP) superfamily, which includes various lactonases. HVO_B0030 showed significant sequence identity (24%) to characterized L-arabinolactonase from A. brasilense (4) functionally involved in oxidative L-arabinose degradation pathway, and to putative D-xylonolactonase (CC0820, 28% identity) as part of the xylose cluster from C. crescentus (5). The putative H. volcanii D-xylonolactonase contains the same secondary struc-tural elements and several conserved features of the active site of a sugar acid lactonase as concluded from the recent first crystal structure of a bacterial gluconolactonase from Xanthomonas campestris (20). These conserved features are also present in the A. brasilense and C. crescentus enzymes and include two highly conserved regions and amino acids, involved e.g. in calcium binding, being crucial for lactone hydrolysis. Furthermore, because HVO_B0030 is located in close vicinity to XDH encoding gene HVO_B0028, it is likely that HVO_B0030 might encode a functional D-xylonolactonase in H. volcanii. However, other sugar-utilizing halophilic archaea, xylose utilizing H. marismortui and H. lacusprofundi, do not contain an ortholog of HVO_B0030 suggesting other enzymes to be potential lactonases.
XAD from H. volcanii represents the first identified and characterized xylonate dehydratase in archaea. Orthologs of high sequence identity, 75 and 90%, were found in the sugarutilizing halophiles, H. marismortui, rrnAC0575, and H. lacusprofundi, Hlac_2242, respectively, suggesting functional XAD in these organisms. The enzymes belongs to the mandelate racemase/muconate lactonizing enzyme family (MR/MLE) of the enolase-like superfamily, which also includes the archaeal D-arabinoate dehydratase, araD (SSO3124), from S. solfataricus, as well as characterized and putative archaeal gluconate dehydratases from Sulfolobus and Thermoplasma sp., from Thermoproteus tenax and Picrophilus torridus, involved in modified Entner-Doudoroff pathways of these organisms (21). A multiple sequence alignment of XAD from H. volcanii and its putative halophilic orthologs, araD from S. solfataricus, and archaeal gluconate dehydratases (GADs) is shown in Fig. 6. The typical sequence signatures 1 and 2, indicative for MR/MLE family, as well as metal ion ligands and amino acids involved in general acid/base catalysis, are indicated. The KXK motif, typical for gluconate dehydratases (22,23), is not present in xylonate dehydratase (Fig. 6). The halophilic XAD showed moderate sequence identity to araD (20%), to the archaeal GADs (24 -29%), and to enolase group (20 -25%) of the enolase superfamily. This is reflected by the phylogenetic relationship of these enzymes, each forming distinct clusters in the enolase superfamily (Fig. 7).

D-Xylose Degradation Pathway in Haloferax volcanii
different enzyme families; the XDH of H. volcanii is a member of the glucose-fructose oxidoreductase/inositol dehydrogenase MocA family, and the Caulobacter enzyme belongs to the short chain dehydrogenase/reductase family. Furthermore, the first identified archaeal xylonate dehydratase of H. volcanii is a member of the MR/MLE family of the enolase superfamily, whereas the putative bacterial counterpart belongs to the ILVD/EDD family. The last two enzymes of the oxidative pathway in H. volcanii and C. crescentus, 2-keto-3-deoxyxylonate dehydratase and ␣-ketoglutarate semialdehyde dehydrogenase, both belong to the same protein families, to the fumaroylacetoacetate hydrolase superfamily and aldehyde dehy-

D-Xylose Degradation Pathway in Haloferax volcanii
drogenase superfamily, respectively. These data suggest convergent evolution of oxidative D-xylose degradation pathway in archaea and bacteria.