Bifunctional Phosphoglucose/Phosphomannose Isomerases from the Archaea Aeropyrum pernix and Thermoplasma acidophilum Constitute a Novel Enzyme Family within the Phosphoglucose Isomerase Superfamily*

The hyperthermophilic crenarchaeon Aeropyrum pernix contains phosphoglucose isomerase (PGI) activity. However, obvious homologs with significant identity to known PGIs could not be identified in the sequenced genome of this organism. The PGI activity from A. pernix was purified and characterized. Kinetic analysis revealed that, unlike all known PGIs, the enzyme catalyzed reversible isomerization not only of glucose 6-phosphate but also of epimeric mannose 6-phosphate at similar catalytic efficiency, thus defining the protein as bifunctional phosphoglucose/phosphomannose isomerase (PGI/PMI). The gene pgi/pmi encoding PGI/PMI (open reading frame APE0768) was identified by matrix-assisted laser desorption ionization time-of-flight analyses; the gene was overexpressed in Escherichia coli as functional PGI/PMI. Putative PGI/PMI homologs were identified in several (hyper)thermophilic archaea and two bacteria. The homolog from Thermoplasma acidophilum (Ta1419) was overexpressed in E. coli, and the recombinant enzyme was characterized as bifunctional PGI/PMI. PGI/PMIs showed low sequence identity to the PGI superfamily and formed a distinct phylogenetic cluster. However, secondary structure predictions and the presence of several conserved amino acids potentially involved in catalysis indicate some structural and functional similarity to the PGI superfamily. Thus, we propose that bifunctional PGI/PMI constitutes a novel protein family within the PGI superfamily.

Phosphoglucose isomerase (PGI 1 ; EC 5.3.1.9) catalyzes the reversible isomerization of glucose 6-phosphate to fructose 6-phosphate. PGI plays a central role in sugar metabolism of eukarya, bacteria, and Archaea both in glycolysis via the Embden-Meyerhof pathway in eukarya and bacteria and in its modified versions found in Archaea. PGI is also involved in gluconeogenesis where the enzyme operates in the reverse direction (for the literature see Refs. [1][2][3]. PGIs from the domains of eukarya and bacteria are well studied enzymes. A variety of PGIs have been purified and biochemically characterized, and the encoding genes have been cloned and sequenced (e.g. Refs. 4 -11). Crystal structures have been determined for the eukaryotic PGIs from pig, rabbit, human, and from the bacterium Bacillus stearothermophilus, and conserved amino acids proposed to be involved in substrate binding and/or catalysis have been identified (12-15, 17, 18, 20 -25).

The eukaryal and bacterial PGIs belong to the PGI superfamily defined by its two conserved signature patterns [DENS]-X-[LIVM]-G-G-R-[FY]-S-[LIVMT]-X-[STA]-[PSAC]-[LIVMA]-Gand [GS]-X-[LIVM]-[LIVMFYW]-X 4 -[FY]-[DN]-Q-X-G-V-E-
To date this superfamily includes more than 300 PGI sequences (see www.sanger.ac.uk/cgi-bin/Pfam/getacc?PF00342) (26) from bacteria and eukarya but only three from the archaeal domain. These include the PGI from the hyperthermophilic euryarchaeon Methanococcus jannaschii (MjPGI). This PGI has recently been characterized as the first archaeal member of the PGI superfamily. 2 So far little is known about other archaeal PGIs. Recently, two other euryarchaeal PGIs have been characterized, one from the hyperthermophilic euryarchaeon Pyrococcus furiosus (2,27) and from closely related Thermococcus litoralis (28). Both PGIs belong to the cupin superfamily (29) and, thus, represent a convergent branch of PGI evolution. The cupin superfamily is present in all three domains of life, eukarya, bacteria, and Archaea, and comprises a group of structurally conserved but functionally diverse proteins ranging e.g. from mannose 6-phosphate isomerases and epimerases involved in formation of bacterial cell wall carbohydrates to nonenzymatic storage proteins in plant seeds and transcription factors (29,30).
So far PGIs of any crenarchaeota have not been characterized, and the encoding genes have not been identified. Recently, cell extracts of the aerobic crenarchaeon Aeropyrum pernix have been shown to contain high PGI activities together with other enzymes of a modified Embden-Meyerhof pathway. 3 Several unusual enzymes of this pathway, i.e. an archaeal ATP-dependent repressor, open reading frame, and kinase glucokinase (31), a non-regulatory ATP-dependent 6-phosphofructokinase of the phosphofructokinase-B family (32), and a non-regulatory pyruvate kinase (33), have been characterized. Despite the fact that A. pernix contained high PGI activity, a gene with significant similarity to either the PGI superfamily or the cupin-type PGIs could not be identified in the completely sequenced genome of A. pernix (34). This suggests that the A. pernix PGI might be significantly different from all known PGIs analyzed so far.
In this communication we report the purification and characterization of PGI activity of the crenarchaeon A. pernix. Surprisingly, unlike all known PGIs, the purified enzyme catalyzed the isomerization of both glucose 6-phosphate and epimeric mannose 6-phosphate at a similar catalytic efficiency, thus defining the enzyme as a novel bifunctional phosphoglucose isomerase/phosphomannose isomerase (PGI/PMI). The encoding gene pgi/pmi was identified in the genome of A. pernix and functionally expressed in Escherichia coli. Putative homologs were found in several archaeal and two bacterial genomes. The homolog of the euryarchaeon Thermoplasma acidophilum was functionally expressed in E. coli and characterized as bifunctional PGI/PMI. Sequence comparison and phylogenetic analyses indicate that these bifunctional PGI/ PMIs constitute a novel protein family within the PGI superfamily.

MATERIALS AND METHODS
Growth of A. pernix K1-A. pernix K1 (DSM 11879) was grown aerobically at 90°C in a 100-l Biostat fermenter on a complex medium as described (35) except that artificial sea water was used instead of Biomaris water, and 1 g of starch was added per liter. Cells were grown and harvested (after 17 h) at the late exponential growth phase.
Purification of PGI Activity (PGI/PMI) from A. pernix-Because the enzyme was not sensitive to oxygen, all steps of the purification procedure were carried out in the presence of oxygen at 4°C. During the purification procedure, the PGI activity of bifunctional PGI/PMI was followed as described below. Cell extracts were prepared from 214 g (wet weight) of frozen cells that were suspended in 100 mM Tris-HCl, pH 7.5, containing 2 mM dithioerythritol, 5 mM EDTA (buffer A). Cells were disrupted by passing through a French pressure cell at 1.3 ϫ 10 8 pascals. Cell debris and unbroken cells were removed by centrifugation for 90 min at 100,000 ϫ g at 4°C. The 100,000 ϫ g supernatant was applied to DEAE-Sepharose (100 ml) that had been equilibrated with buffer A. Protein was eluted by increasing NaCl gradients in buffer A: 100 ml of 0 -0.06 M, 30 ml of 0.06 M, 450 ml of 0.06 -0.7 M, 10 ml of 0.7 M, 200 ml of 0.7-2 M. Fractions containing the highest PGI activity were pooled (0.14 -0.36 M), adjusted to 2 M ammonium sulfate and to pH 8.0, and applied to a phenyl-Sepharose column (16 ml) previously equilibrated with buffer B (50 mM Tris/HCl, 2 M ammonium sulfate). Protein was desorbed by a decreasing ammonium sulfate gradient in buffer B: 130 ml of 2-0.3 M, 15 ml of 0.3 M, 60 ml of 0.3-0 M. PGI-containing fractions (0.1-0 M) were concentrated to a volume of 0.9 ml by ultrafiltration (exclusion size 10 kDa) and applied to a Superdex 200 HiLoad 16/60 gel filtration column equilibrated with buffer C (150 mM NaCl, 50 mM Tris/HCl, pH 7.5). The PGI activity-containing fractions (85-92 ml) were pooled, diluted 3-fold with buffer D (50 mM Tris/HCl, pH 7.5), and applied to an Uno-Q1 column (1 ml) equilibrated with buffer D. Protein was eluted by increasing NaCl gradients in buffer D: 4 ml of 0 -0.06 M, 1.5 ml of 0.06 M, 8 ml of 0.06 -0.55 M, 1.5 ml of 0.55 M, 3 ml of 0.55-1 M, 1 ml of 1 M, 2 ml of 1-2 M, 5 ml of 2 M. Fractions containing the highest PGI activity (0.08 -0.56 M) were pooled, concentrated to a final volume by vacuum centrifugation (55°C, Christ , and applied to a Bio-Silect SEC 250 -5 gel filtration column equilibrated with buffer C. Fractions with the highest PGI activity eluted at 12-13 ml and were essentially pure and stored at Ϫ20°C. Under these conditions the activity remained about constant. Analytical Assays-The purity of the preparation was checked by SDS-PAGE in 12% gels followed by staining with Coomassie Brilliant Blue R 250 according to standard procedures (36). Protein concentrations were determined by the method of Bradford (37) with bovine serum albumin as standard. The protein concentration of purified recombinant protein was determined using the molar extinction coefficients at 280 nm of the enzymes as calculated by the ProtParam (us. expasy.org/tools/protparam.html): ApPGI/PMI ϭ 47650 M Ϫ1 cm Ϫ1 , TaPGI/PMI ϭ 25010 M Ϫ1 cm Ϫ1 . Both methods deviated by less than 5%.
Identification of ORFs encoding PGI/PMIs-Pure enzyme was excised from Coomassie-stained SDS-PAGE and analyzed by MALDI-TOF spectroscopy after tryptic cleavage as previously described (38).
Other putative ORFs encoding for PGI/PMIs were identified by BLASTP searches using the A. pernix PGI/PMI sequence.
Molecular Mass Determination-Molecular mass determinations were performed by both gel filtration on Superdex 200 (150 mM NaCl, 50 mM Tris/HCl, pH 7.5) and by analytical ultracentrifugation. For analytical ultracentrifugation the TaPGI/PMI was dialyzed against 20 mM Tris/HCl pH 7.0. Centrifugation was done at 20°C in a Beckman Optima XL-A analytical ultracentrifuge equipped with a Titan AN 50 rotor and absorption optics. Sedimentation velocity experiments were performed for an approximation of the molecular mass at 50,000 rpm, and sedimentation-diffusion equilibrium runs were performed at 10,000 rpm for determining the molecular mass in a 150-l volume. In sedimentation velocity experiments the apparent sedimentation coefficient s 20,w app and an approximated molecular mass were evaluated from the velocity and shape of the sedimenting boundary by fitting the time-dependent concentration profiles calculated with Lamm's differential equation for a single sedimenting species to the measured data using the program package AKKUPROG (40). When the measured concentration profile of sedimentation-diffusion equilibrium runs remained unchanged for 12 h, we assumed equilibrium to be attained. AKKUPROG was used to calculate the apparent molecular masses by fitting the ideal distribution for a single species to the measured concentration profiles. The partial specific volume of the protein was calculated from the amino acid sequence.
Enzyme Assays and Determination of Kinetic Parameters-Activity of PGI/PMI s of A. pernix and T. acidophilum was measured both as PGI or PMI activity. PGI activity was used to determine the pH and the temperature optima of the enzymes. pH optima were determined using acetate, Tris/HCl, ethanolamine, CHES, and CAPS at 100 mM each. For the determination of enzyme activity the following standard conditions were used: ApPGI/PMI, 80°C, pH 7.4, 2 g of enzyme; TaPGI/PMI, 50°C, pH 7.6, 5 g of enzyme. PGI activity (G6P 7 F6P) was determined in both directions. The formation of glucose 6-phosphate (G6P) was investigated by coupling it to the reduction of NADP ϩ either via yeast (20 -50°C) or via hyperthermophilic Thermotoga maritima glucose 6-phosphate dehydrogenase (50 -95°C) (41). The standard assay I contained 100 mM Tris/HCl, 0.5 mM NADP ϩ , 5 mM fructose 6-phosphate (F6P), and 1.1 units of glucose 6-phosphate dehydrogenase. This assay was routinely used for most PGI activity assays. The formation of F6P was measured by coupling it to the oxidation of NADH via mannitol-1phosphate dehydrogenase from Klebsiella pneumoniae, which was ex-pressed and purified as previously described (42). The standard assay II contained 100 mM Tris/HCl, 0.3 mM NADH, 0.1-15 mM G6P, and 0.6 units of mannitol-1-phosphate dehydrogenase. At temperatures higher than 50°C a discontinuous assay was used as previously described (2). PMI activity of PGI/PMIs was analyzed in the direction of F6P formation as described above with the modified standard assays I or II using 0.1-15 mM mannose 6-phosphate (M6P) as substrate. At 50°C and below F6P formation was detected using modified standard assay II. Above 50°C, F6P formation was detected by using modified standard assay I with the addition of 1 units of hyperthermophilic P. furiosus PGI (2). The following substances were tested for potential inhibitory effect on PGI activity: erythrose 4-phosphate, 6-phosphogluconate, and phosphate. The date points given in the figures are mean values of at least three experiments; S.E. are given. Kinetic data were fitted to the Michaelis-Menten equation; the curves drawn represent best fits to the Michaelis-Menten equation (see Fig. 1) or to first order exponential decay equations (see Fig. 2, C and D) with the MicroCal™ software 5.0 using the Levenberg-Marquard algorithm; respective S.E. are given.
Temperature Dependence and Thermal Stability-The temperature dependence of enzyme activity was measured between 20 and 96°C in 50 mM sodium phosphate with F6P as substrate (5 mM). The thermostability of the purified enzyme (1 g of enzyme, pH 7.0) was tested in sealed vials that were incubated at temperatures between 70 and 110°C for 2-120 min. The vials were then cooled on ice for 10 min, and residual enzyme activity was tested and compared with the controls (unheated samples).
Circular Dichroism (CD) Spectroscopy-CD spectroscopy analyses were performed on Jasco J-715 CD spectrometer. Spectra were recorded in 0.1-mm cuvettes and corrected for the signal of the solvent (10 mM sodium phosphate, pH 7.0). At least five spectra were averaged. Secondary structure analysis and assignments to different secondary structure types were performed by the experimentally established spectra-structure correlation using the Varselec option of Dicroprot (43). The program Predict Protein was used for sequence-based secondary structure predictions (cubic.bioc.columbia.edu) (44,45). Heat-induced unfolding of PGIs were analyzed in temperature gradient experiments using closed cuvettes (0.1 cm). The protein samples were dialyzed against 10 mM sodium phosphate buffer, pH 7.0, and the protein concentrations were set to 100 g/ml. The temperature of the samples was raised at a rate of 1°C/min (40 -100°C). Protein unfolding was followed by changes of the ␣-helical CD ellipticity at 221 nm. The observed ellipticity at a given temperature (⌽) was corrected for the temperaturedependent base line, giving ⌽Ј. The corrected values of the folded state ⌽Ј f and the unfolded state ⌽Ј uf were used for calculating the fraction of folded protein (x uf ): x uf ϭ (⌽Ј Ϫ ⌽Ј f )/(⌽Ј uf Ϫ ⌽Ј f ). Spectra were recorded before and after each temperature gradient. The spectra as well as the ellipticity of the unfolded state was obtained at pH 2 and 12 when complete thermal unfolding was not obtained up to 100°C. Temperature gradient experiments were performed in triplicate; the respective S.E. are given.
Sequence Handling-Protein sequences were extracted from the NCBI, GenBank TM , and Pfam data bases. Sequence alignments were constructed with the neighbor-joining (NJ) method of ClustalX using the GONNET matrix (46). Phylogenetic trees were constructed using the neighbor-joining option of ClustalX, the maximum likelihood (ML) method of PROML (Phylip, version 3.6), and the maximum parsimony (MP) method of PROTPARS (Phylip, version 3.6). Confidence limits were estimated by 100 bootstrapping replicates. The PGI signature patterns were updated using an alignment provided by Pfam (see www.sanger.ac.uk/cgi-bin/Pfam/getacc?PF00342), which included all PGI sequences presently available (20.8.03).

Purification of PGI Activity of A. pernix and Its Identification as Bifunctional PGI/PMI-Extracts from
A. pernix grown on media with peptone and starch exhibited PGI activity (0.27 units/mg at 75°C). This PGI activity was purified about 800fold using five chromatographic steps to a specific activity of 200 units/mg with a yield of 3%. The purified protein was electrophoretically homogenous, as indicated by one band on SDS-PAGE. The purified enzyme catalyzed the reversible isomerization of G6P to F6P, i.e. showed PGI activity. The V max values (at 80°C) with G6P and F6P were 225 and 195 units/mg; the corresponding K m values were 3.5 and 0.44 mM, respectively (see also Table I). Surprisingly, the enzyme also catalyzed the isomerization of M6P to F6P at almost similar catalytic efficiency; with M6P the V max and K m values were 209 units/mg and 1.1 mM, respectively. Thus, the enzyme of A. pernix exhibited bifunctional activities as PGI and PMI, and it was, therefore, assigned as bifunctional PGI/PMI. The PGI/ PMI differs from all known PGIs that show strict substrate specificity for G6P and F6P and did not use M6P. Rather, M6P has been shown to be an inhibitor of conventional PGIs (e.g. Refs. 1, 47, and 48).
Identification of the Gene pgi/pmi Encoding Bifunctional PGI/PM of A. pernix and Functional Overexpression in E. coli-By MALDI-TOF spectroscopy analysis of purified PGI/ PMI from A. pernix ORF APE0768 was identified as putative pgi/pmi gene. ORF APE0768 contains 1008 bp coding for a polypeptide of 335 amino acids with a calculated molecular mass of 36.112 kDa. The ORF was cloned and expressed in E. coli. The recombinant protein was purified by one heat treatment and two chromatographic steps to apparent homogeneity. The purified recombinant protein showed bifunctional PGI/PMI activity, thus proving ORF APE0768 to represent the functional pg/ipmi gene. The recombinant PGI/PMI showed almost identical properties as the enzyme purified from A. pernix.
Putative PGI/PMI Homologs and Functional Overexpression of PGI/PMI from T. acidophilum-Putative homologs of the A. pernix PGI/PMI were identified in the genomes of eight (hyper)thermophilic Archaea and bacteria: Pyrobaculum aerophilum, T. acidophilum, Thermoplasma volcanium, Ferroplasma acidarmanus, Sulfolobus tokodaii, Sulfolobus solfataricus, Aquifex aeolicus, Anaerocellum thermophilum. An alignment of ApPGI/PMI and its putative homologs exhibited significant amino acid sequence identity (17-70%) (see Fig. 5 and "Discussion"). One putative PGI/PMI, the gene product of ORF Ta1419 of the thermophilic euryarchaeon T. acidophilum, was characterized. ORF Ta1419 contains 933 bp coding for a polypeptide of 310 amino acids with a calculated molecular mass of 35.157 kDa. The ORF was cloned and expressed in E. coli. The recombinant protein was purified by a heat treatment and two chromatographic steps. The purified protein also showed bifunctional PGI/PMI activity and was characterized in detail (see "Catalytical Properties and Substrate Specificity").
Characterization of PGI/PMIs from A. pernix and T. acidophilum-The biochemical and biophysical properties of PGI/ PMIs from A. pernix (ApPGI/PMI) and from T. acidophilum   P. aerophilum, S. solfataricus, S. tokodaii, T. volcanium, T. acidophilum, F. acidarmanus, A. thermophilum, and  A. aeolicus (for accession numbers see Fig. 6. The alignment was generated with ClustalX (46). The signature pattern of the PGI/PMI family

S-Y-S-G-[NT]-T-[ESTIL]-E-T-[LIV]
is highlighted by a green box. The predicted secondary structure for the ApPGI/PMI according to the program Predict Protein (cubic.bioc.columbia.edu) (44, 45) is given above the sequences, helices are indicated as bars, and ␤-sheets are indicated as arrows. Residues of PGI/PMI sequences that correspond to equivalent residues in PGIs are marked as follows. Residues proposed to be important for substrate binding and/or catalysis of PGIs as deduced from x-ray structures (14,21,22)  (TaPGI/PMI) were characterized (Table I).
Molecular Properties-Apparent molecular masses of ApPGI/PMI and TaPGI/PMI, as analyzed by gel filtration, were 45 and 48 kDa, respectively. In addition, the molecular mass of TaPGI/PMI was determined by analytical ultracentrifugation. The measured concentration profile of sedimentation-diffusion equilibrium runs were best fit using a molecular weight of 67 kDa, which is significant higher than the value (48 kDa) measured by gel filtration. A significant underestimation of molecular mass by gel filtration has also been reported for several hyperthermophilic proteins, e.g. pyrophosphatase from Sulfolobus acidocaldarius (49). SDS-PAGE revealed one subunit of ApPGI/PMI and TaPGI/PMI, each with apparent molecular masses of 36 and 35 kDa, respectively, indicating a dimeric structure for both PGI/PMIs.
Catalytical Properties and Substrate Specificity-ApPGI/ PMI and TaPGI/PMI exhibited bifunctional activities as PGIs and PMIs. In addition to G6P, M6P was isomerized to F6P at about the same catalytic rate. Kinetic properties of both enzymes were determined for both reaction directions for the PGI activity (G6P and F6P) and in the direction of F6P formation for PMI activity. The rate dependence on all substrates of the enzymes followed Michaelis-Menten kinetics; the respective K m and V max values are given in Table I. The rate dependence of both enzymes on M6P concentration is shown in Fig. 1, A and B. The respective K m values for all three substrates are 2-5fold higher for the ApPGI/PMI (assay temperature 80°C) as compared with those obtained for the TaPGI/PMI (assay temperature 50°C) but were about identical for F6P when both enzymes were assayed at 50°C, indicating a lower substrate for the ApPGI/PMI at higher temperatures.
The bifunctional PGI/PMIs from A. pernix and T. acidophilum were inhibited by low concentrations of erythrose 4-phosphate and 6-phosphogluconate (Table I), i.e. effective inhibitors of PGIs of the PGI superfamily (10, 50 -53). For the ApPGI/PMI a competitive inhibition with 6-phosphogluconate could be demonstrated, i.e. the addition of the inhibitor (50 M) caused an apparent 4.3-fold increase of the K m value, whereas the respective V max values were almost identical in the presence and the absence of the inhibitor (61 and 59 units/mg, respectively, 50°C). The pH optima of both ApPGI/PMI and TaPGI/ PMI were at 7.6.
Temperature Optimum and Thermostability-The thermophilic properties of the PGI/PMIs were analyzed by following heat-induced unfolding as detected by CD spectroscopy at 221 nm, the temperature dependence of the specific activity, and the heat resistance of the proteins (Figs. 2 and 3). ApPGI/PMI activity showed a temperature optimum of above 98°C (the highest temperature tested) and was extremely thermostable. Neither significant loss of enzymatic activity upon 120-min incubation at 100°C nor unfolding could be detected up to 100°C, indicating a melting temperature higher than 100°C. At 110°C the ApPGI/PMI still showed a half-life of about 35 min. The TaPGI/PMI also exhibited remarkable thermophilic properties. The activity of the enzyme showed a temperature optimum of 75°C, which correlates with the onset of unfolding (about 75-78°C). Above that temperature the heat resistance of the protein was lost. TaPGI/PMI was not inactivated upon incubation at 70°C for 120 min; the half-life of the enzyme at 80°C was about 30 min. The enzyme showed a melting temperature of 86°C; at 90°C almost complete unfolding as well as the complete loss of heat resistance could be detected.
CD Spectra-To get information about the overall fold of the PGI/PMIs, CD spectra were recorded for the PGI/PMIs from A. pernix and T. acidophilum. The spectra of both enzymes were nearly superimposable (Fig. 4). For the ApPGI/PMI and the TaPGI/PMI, ␣-helical contents of 37 and 41% and ␤-sheet content of 16 and 17%, respectively, were estimated, which closely match the secondary structure predictions (40 and 44% ␣-helical, 19 and 22% ␤-sheet, respectively). The secondary structure predictions were comparable with those derived from the x-ray structures of the human PGI when aligned regions were compared (not shown).

DISCUSSION
In this report we describe bifunctional PGI/PMIs from A. pernix and T. acidophilum. The enzymes represent the first characterized members of a novel bifunctional PGI/PMI family.
Bifunctionality and Physiological Function of PGI/PMI-The ApPGI/PMI and the TaPGI/PMI exhibited very similar kinetic properties. The most striking feature of these enzymes is their bifunctionality, which is unique among the PGIs and PMIs described so far. All characterized PGIs and PMIs are highly specific for either G6P/F6P and M6P/F6P, respectively. In contrast, the PGI/PMIs described in this study used G6P, M6P, and F6P as substrates at almost equal rates and catalytic efficiency. The kinetic constants for G6P and M6P of these bifunctional enzymes were in the same range as described for PGIs and PMIs (e.g. Refs. 7, 9, 11, and 53). In contrast, M6P has been demonstrated to be a competitive inhibitor for several conventional PGIs (47,54). Low rates of M6P formation from G6P via F6P could be demonstrated for the rabbit PGI; however, this (epimerase) activity proceeds at a 2 ϫ 10 5 -fold lower rate as compared with the isomerization rate of F6P (55). Thus, this reaction should be considered as a side reaction rather than being of physiological relevance. In contrast, the PGI/ PMIs of this study catalyzed M6P conversion to G6P via fructose at similar rates as F6P isomerization. Neither A. pernix nor T. acidophilum contain an additional pgi nor a pmi gene in their genomes, which indicates that both enzymes have physiological roles as PGIs and as PMIs in these organisms. As reported for various conventional PGI (10, 50 -53), both the ApPGI/PMI and the TaPGI/PMI were inhibited by micromolar concentrations of erythrose 4-phosphate and 6-phosphogluconate, indicating a similar active site (see "Sequence Comparison").
Molecular and Thermophilic Properties-The PGI/PMIs from the Archaea A. pernix and T. acidophilum were characterized as homodimers with subunits of about 36 kDa. When compared with conventional PGIs, PGI/PMIs (including the putative homologs) are significantly smaller (296 -354 amino acids versus more than 401 amino acids (M. jannaschii for conventional PGIs (Table II; www.sanger.ac.uk/cgi-bin/Pfam)) and might represent the minimal core structure necessary for a phosphoglucose isomerase of the PGI superfamily. In contrast, the cupin type PGIs from Thermococcales have subunits of about 23 kDa and are significantly smaller (Table II) (2,28). In accordance with the optimal growth temperature of A. pernix, the hyperthermophilic ApPGI/PMI is the most thermophilic PGI and PMI described so far. In comparison, the PGIs from the hyperthermophiles M. jannaschii and P. furiosus had a temperature optimum of 89°C and above 96°C, respectively (2). 2 P. furiosus PGI had a half-life of about 90 min at 100°C, whereas the A. pernix enzyme was heat-resistant at that temperature for up to 2 h. The TaPGI/PMI showed a temperature optimum of 75°C, which is about 16°C above the growth optimum of the T. acidophilum (59°C).
Sequence Comparison-All PGIs characterized so far belong either to the PGI superfamily or to the cupin superfamily. All characterized PMIs belong to the cupin superfamily, which includes class I and class II PMIs; type I are bicupin PMIs present in eukarya and bacteria (Ͼ80 sequences, see www. sanger.ac.uk/cgi-bin/Pfam), whereas type II phosphomannose isomerases are bifunctional enzymes comprising a PMI and guanosine diphospho-D-mannose pyrophosphorylase domain and are found in bacteria and Archaea. In addition, the Sinorhizobium meliloti pmi gene might represent a third type of PMI (56) which, however, has not yet been biochemically characterized. Although more than 200 sequences (see www. sanger.ac.uk/cgi-bin/Pfam) are available on cupin PMIs, no archaeal PMI has been characterized so far.
Bifunctional PGI/PMI from A. pernix and T. acidophilum did not show any similarity to members of the cupin superfamily nor did they contain its two consensus pattern (motif 1, GX 5 HXHX (3/4) EX 6 G; motif 2, GX 5 PXGX 2 HX 3 N). Because all characterized PMIs fall into the cupin superfamily, the bifunctional archaeal PGI/PMIs, which belong to the PGI superfamily (see below), represent a novel convergent line of PMI evolution. Below PGI/PMIs were compared with the PGI superfamily, the conclusion being that PGI/PMIs represent a novel enzyme family within the PGI superfamily; at a first glance, the homology between ApPGI/PMI and TaPGI/PMI and conventional PGIs from the PGI superfamily is quite low (5-14% identity). However, at a closer look, including putative PGI/PMI homologs, some features of this PGI/PMIs become evident that strongly suggest a common origin of both conventional PGIs and PGI/PMIs.
Putative homologs of the ApPGI/PMI and the TaPGI/PMI were identified in the genomes of seven (hyper)thermophilic Archaea and bacteria: P. aerophilum, T. volcanium, F. acidarmanus, S. tokodaii, S. solfataricus, A. aeolicus, A. thermophilum. An alignment of ApPGI/PMI, TaPGI/PMI, and the putative homologs exhibited significant amino acid sequence identity (17-70%). 27 residues are completely conserved among these nine sequences ( Fig. 5; for the accession numbers see Fig.  6). Although the coding function of the hypothetical ORFs has to be demonstrated, we expect these putative homologs of the ApPGI/PMI and TaPGI/PMI to encode for bifunctional PGI/ PMIs as well, representing members of a the novel PGI/PMI family. All PGI/PMI sequences exhibit significant changes from an updated version of the two PGI signature patterns, However, signature patterns only reflect similarities within certain conserved regions of PGIs and PGI/PMIs but do not provide information about a common fold or catalytic mechanism, which is more relevant. PGI structures have been solved from a variety of mammalian sources and from B. stearothermophilus in native, inhibitor, and substrate (F6P)-bound forms (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25). CD spectra analysis and secondary structure predictions suggest a similar fold of the PGI/PMI from A. pernix and T. acidophilum and e.g. human PGI when aligned regions were compared. The protein architecture is very similar between mammalian and the bacterial PGI and comprises a large and a small domain, each with a central ␤-sheet surrounded by ␣-helices. The active site of PGI, which is located in a crevice between the domains close by the subunit boundary, comprises residues that presumably play a role in the catalytic mechanism of PGI during substrate binding and/or catalysis. Most of these residues, including the four most crucial residues, Glu-357, Arg-272, His-388, and Lys-518 (mammalian numbering), are completely or functionally conserved in the PGI/PMIs as well (Table III). This implies that PGIs and PGI/PMIs might share a very similar catalytic mechanism. However, one might expect differences in substrate binding and/or the catalytic mechanism, which accounts for the extra PMI activity in PGI/PMIs. The presence of functional conserved residues between PGIs and PGI/PMIs might indicate functional divergence from an ancestral PGI. To analyze the phylogenetic relationship of PGI/ PMIs, of which ApPGI/PMI and TaPGI/PMI have been characterized, phylogenetic analyses were performed using three tree construction methods, NJ, ML, and MP. A phylogram including both characterized and putative PGI/PMIs as well as PGIs from eukarya, bacteria, and Archaea based on a NJ tree is given in Fig. 5. The overall topology was achieved by the three methods used and is supported by good bootstrapping values. The most striking feature of this tree is its dichotomic structure. This topology goes along with two different enzyme functions as described above, separating the PGI superfamily into the PGI/ PMI family and the PMI family. Within these families the topology is quite puzzling, which might be explained by assuming several independent lateral gene transfers between organisms of the three kingdoms. PGI has sometimes been described as a "workhouse" enzyme of sugar metabolism (14), and the uptake of pgi genes from environmental sources might be of advantage in improving the efficiency of sugar metabolism. Recently, the PGI family has been divided into three major subfamilies: I, eukaryotic; II, cyanobacterial/chloroplasts; III, bacterial (57,58). Most PGI sequences fall into the subfamily I and the monophyly of the major eukaryotic groups is reflected in the topology of the reconstruction. However, several eubacterial clades (Zymomonas mobilis, Xanthomonas campestris, E. coli, Haemophilus influenzae) are interspersed between the eukaryotic groups. Their presence might be explained as a result of several independent lateral gene transfers from eukarya to its symbionts and parasites. PGI subfamily II comprises PGIs from chloroplasts, cyanobacteria, and amitochondriate protists. A common origin of these PGIs has been suggested (57). PGIs from various bacterial lineages are found in PGI subfamily III. In addition, PGIs from three euryarchaeota, M. jannaschii, Halobacterium NRC1, Haloarcula marismortui, were included in the bacterial cluster rather than in a separate cluster because they show a high degree of similarity to bacterial PGIs (35-48%), and these organisms might have obtained their pgi gene by a lateral gene transfer. 2 Whereas the PGI family includes almost exclusively bacterial and eukaryal sequences (Ͼ300), the PGI/PMI family comprises nine sequences from thermophilic and hyperthermophilic species: four crenarchaeal sequences, three euryarchaeal sequences from Thermoplasmatales, and two bacterial sequences. Thus, this family is presumably of archaeal or even crenarchaeal origin. The clustering within this group is supported mostly by good bootstrap values, subdividing it into two or three groups, A. pernix, P. aerophilum, Sulfolobales, Thermoplasmatales, including two bacterial sequences. The two bacterial PGI/PMI sequences branch from the line leading to the Thermoplasmatales according to all three tree construction methods used (NJ, ML, and MP); however, the position of the respective basal node differs. This suggests a common origin of these bacterial and the Thermoplasmatales PGI/PMIs sequences. At least for the respective gene from Anaerocellum thermocellum, a lateral gene transfer has to be assumed since the PGIs from closely related Clostridium species (Clostridium thermophilum, Clostridium acetobutylicum, Clostridium perfringens, and Thermoanaerobacter tengcongensis) fall into the PGI subfamily III close to the PGIs from Bacillus species. Whether the PGI/PMI family originated in the early archaeal or even crenarchaeal evolution cannot be clarified due to the limited number of archaeal genomes sequences available so far. Thus, analysis of its distribution in other archaeal organisms will help to address this question. To date PGI/PMIs are present in all four crenarchaeal genomes available. In comparison, in euryarchaeota the presence of PGI/PMIs is limited so far to the Thermoplasmatales; Thermococcales species have been shown to contain PGIs of the cupin type, representing a convergent line of PGI evolution.
According to NJ and MP tree reconstruction analysis, the Thermoplasmatales group branches from the line leading to the Sulfolobales sequences, suggesting an exchange of pgi/pmi gene between ancestors of these organisms, a phenomenon that has been proposed to occur at high frequency between these organisms due to the common natural (acidic) biotopes (59). This would indicate an crenarchaeal origin of the PGI/PMI family, which is presumably a result of diversification of an ancestral PGI to PGI/PMI in the early (cren)archaeal evolution with the gain of the additional PMI function at the cost of a less regulatory potential as compared with two separately transcribed enzymes. On the other hand, ML analysis revealed two PGI/PMI groups, a crenarchaeal group and a Thermoplasmatales group, suggesting an archaeal origin of this family, which would imply that all other euryarchaeota have lost their PGI/ PMI sequences during evolution, which is unlikely. Also, because none of the available genomes contains both a pgi and a pgi/pmi gene, a scenario including an early gene duplication leading to PGIs and PGI/PMIs with a subsequent loss of one of the genes in all organisms appears unlikely as well.
In summary, the first characterized bifunctional PGI/PMIs from A. pernix and T. acidophilum belong to the PGI superfamily. Because of the bifunctional character of these enzymes and their low sequence identity to conventional PGIs and due to their separate phylogenetic clustering, bifunctional PGI/ PMIs were established as a novel protein family within the PGI superfamily. Thus, this superfamily now comprises both a PGI family and a PGI/PMI family.