Archaeal fructose-1,6-bisphosphate aldolases constitute a new family of archaeal type class I aldolase.

Fructose-1,6-bisphosphate (FBP) aldolase activity has been detected previously in several Archaea. However, no obvious orthologs of the bacterial and eucaryal Class I and II FBP aldolases have yet been identified in sequenced archaeal genomes. Based on a recently described novel type of bacterial aldolase, we report on the identification and molecular characterization of the first archaeal FBP aldolases. We have analyzed the FBP aldolases of two hyperthermophilic Archaea, the facultatively heterotrophic Crenarchaeon Thermoproteus tenax and the obligately heterotrophic Euryarchaeon Pyrococcus furiosus. For enzymatic studies the fba genes of T. tenax and P. furiosus were expressed in Escherichia coli. The recombinant FBP aldolases show preferred substrate specificity for FBP in the catabolic direction and exhibit metal-independent Class I FBP aldolase activity via a Schiff-base mechanism. Transcript analyses reveal that the expression of both archaeal genes is induced during sugar fermentation. Remarkably, the fbp gene of T. tenax is co-transcribed with the pfp gene that codes for the reversible PP(i)-dependent phosphofructokinase. As revealed by phylogenetic analyses, orthologs of the T. tenax and P. furiosus enzyme appear to be present in almost all sequenced archaeal genomes, as well as in some bacterial genomes, strongly suggesting that this new enzyme family represents the typical archaeal FBP aldolase. Because this new family shows no significant sequence similarity to classical Class I and II enzymes, a new name is proposed, archaeal type Class I FBP aldolases (FBP aldolase Class IA).

Fructose-1,6-bisphosphate (FBP) 1 aldolase (EC 4.1.2.13) catalyzes the reversible aldol condensation of glyceraldehyde 3-phosphate (GAP) and dihydroxyacetone phosphate (DHAP) yielding FBP. The enzyme fulfills an amphibolic function being involved in catabolic (glycolysis) as well as anabolic pathways (gluconeogenesis and Calvin cycle). In spite of this central function in carbohydrate metabolism, up to now no archaeal genes coding for the respective enzyme activities have been analyzed.
Two distinct classes of FBP aldolases occur in nature, which differ in their enzymatic mechanisms (1)(2)(3)(4). Class I FBP aldolases form a Schiff-base intermediate between the carbonyl substrate (FBP and DHAP) and the ⑀-amino group of the active site lysine residue and are inactivated by borohydride (NaBH 4 ), whereas Class II FBP aldolases depend on divalent metal ions to stabilize the carbanion intermediate and are, therefore, inhibited by EDTA. Class II enzymes of bacterial and eucaryal origin generally form dimers with a subunit molecular mass of ϳ40 kDa, whereas the Class I pendants are heterogeneous. Eucaryal aldolases are homomeric tetramers with a subunit molecular mass of ϳ40 kDa, and for bacterial enzymes oligomeric arrangements from monomer to decamer and subunit molecular masses of 27-40 kDa have been described (5,6).
Sequence comparisons of Class I and II FBP aldolases revealed no detectable sequence homology, suggesting convergent evolution (4,5,(7)(8)(9)(10)(11). The latter is supported by comparisons of available crystal structures of rabbit muscle Class I and Escherichia coli Class II FBP aldolases indicating that even though both classes adopt a common folding topology ((␤␣) 8 triose-phosphate isomerase (TIM)-barrel fold) and catalyze identical reactions, they share no conserved catalytic residues, and the location of their active sites is distinct (12). However, more recent analysis combining sequence, structure, and functional information indicates that many of the (␤␣) 8 (TIM) barrel superfamilies, such as aldolases, TIMs, and enolases, share a common evolutionary origin (ancestral ␤/␣ barrel), although they adopt a wide range of enzymatic functions (13,14).
The distribution of FBP aldolases during evolution is complex and still puzzling. Class II aldolases seem to be confined to more simple organisms such as bacteria and a few unicellular eucaryotes (fungi, including yeast), whereas Class I FBP aldolases are present in higher forms of life (animals, higher plants, ferns, and mosses), and only a few bacteria possess a Class I enzyme, sometimes in addition to a Class II enzyme. Earlier branching protists studied so far show a marked diversity of harboring Class I and/or Class II enzymes (for review see Refs. 5 and 10).
Recently, Thomson et al. (6) described a new type of FBP aldolase in E. coli, which belongs to Class I aldolases according to its Schiff-base mechanism but differs significantly from the other members of this class by its low sequence similarity. The E. coli Class I FBP aldolase was originally mis-annotated in the E. coli genome as dehydrin (DhnA, dhnA gene) due to its overall identity (13-20%) to dehydrins in plants, which are stress proteins that are induced in response to dehydration (6).
Although Class I and Class II FBP aldolase activities have been demonstrated in Archaea (15)(16)(17)(18)(19)(20)(21), no genes encoding classical Class I or II enzymes have been identified in any of the sequenced archaeal genomes suggesting that Archaea possess novel types of aldolases that are either absent or not yet recognized as such in Bacteria and Eucarya. The latter is supported by initial data base searches of Galperin et al. (22) who identified gene homologs of the unusual Class I FBP aldolase gene (dhnA) of E. coli in the sequenced archaeal genomes. However, none of this archaeal gene products was examined with respect to its enzymatic function. In order to prove that DhnA homologs in the two major archaeal kingdoms code for FBP aldolases, we expressed the dhnA gene homologs of the crenarchaeote Thermoproteus tenax and the euryarchaeote Pyrococcus furiosus in E. coli, and we analyzed the function of their gene products. The two hyperthermophiles differ from each other not only with respect to phylogeny but also with respect to physiology. T. tenax is a facultative chemoorganotroph (23,24), and P. furiosus is an obligate chemoorganotroph (25). T. tenax uses two different pathways for carbohydrate catabolism, i.e. a modified, non-phosphorylative Entner-Doudoroff pathway and a variant of the reversible Embden-Meyerhof-Parnas pathway (19,26). The latter is characterized by a PP i -dependent phosphofructokinase (PP i -PFK) (27), two different glyceraldehyde-3-phosphate dehydrogenases (28,29), and a pyruvate kinase with reduced allosteric potential (30). P. furiosus possesses one catabolic pathway, a variant of the Embden-Meyerhof-Parnas pathway that differs significantly from the T. tenax variant (21) and involves an ADP-dependent glucokinase (31), an ADP-dependent PFK (32), a canonical glyceraldehyde-3-phosphate dehydrogenase, and a ferredoxin-dependent glyceraldehyde-3-phosphate oxidoreductase (33,34).

EXPERIMENTAL PROCEDURES
Chemicals and Plasmids-DL-GAP was prepared from monobarium salts of the diethyl acetyl, according to the manufacturer's instructions (Sigma). All other chemicals and enzymes were purchased from Sigma, Merck, or Roche Diagnostics GmbH in analytical grade. For heterologous expression the vectors pET-15b and pET-24d (Novagen) and for generating antisense mRNA the vector pSPT 19 (Roche Diagnostics GmbH) were used.
Strains and Growth Conditions-Mass cultures of T. tenax Kra1 (DSM 2078) were grown as described previously (19). P. furiosus (DSM 3638) was grown in CDM medium as described previously (35) with the only exception that yeast extract was omitted and substituted by the individual amino acids (0.25 mM final concentration). Maltose (10 mM) or pyruvate (40 mM) was added as the primary carbon source. E. coli strains DH5␣ (Life Technologies, Inc.), XL1Blue (Stratagene), BL21(DE3), and BL21(DE3)pLysS (Novagen) for cloning and expression studies were grown under standard conditions (36) following the instructions of the manufacturer.
Reactions were started by addition of the substrate FBP, and the enzyme concentrations ranged from 2 to 40 g of protein/ml test volume. To determine the substrate specificity of the FBP aldolases, the standard enzyme assay was used substituting FBP by other substrates, such as fructose 1-phosphate (Fru-1-P). For effector studies citrate was added to an end concentration of 10 mM in the presence of half-saturating concentrations of FBP. To test the metal ion requirement, up to 10 mM EDTA or different metal ions (0.1 and 1 mM) were added to the mixture. Protein concentration was measured according to the method of Bradford (37) using the Bio-Rad Protein-Assay (Bio-Rad) with BSA as standard.
Active Site Labeling-To investigate the involvement of a Schiff-base mechanism, the FBP aldolase of T. tenax (0.09 mg of protein) was incubated at room temperature in 50 mM HEPES/KOH (pH 7.5), 100 mM NaBH 4 (1 M stock solution in 10 mM NaOH) in the presence or absence of saturating concentrations (10 mM) of DL-GAP, DHAP, or FBP (total volume 250 l). After 10 min the samples were dialyzed twice against 2 liters of 20 mM Tris/HCl (pH 8.5, 4°C; overnight) and assayed for FBP aldolase activity. The assay was performed at 70°C using the non-phosphorylating NAD ϩ -dependent glyceraldehyde-3-phosphate dehydrogenase (EC 1.2.1.9) (28) of T. tenax as auxiliary enzyme. The assay (total volume 1 ml) was performed in 100 mM Tris/HCl (pH 7.0, 70°C) in the presence of 5 mM NAD ϩ , 5 mM FBP, and 5 units of NAD ϩ -dependent glyceraldehyde-3-phosphate dehydrogenase. The increase in absorption was measured at 366 nm (⑀ 70°C ϭ 3.15 mM Ϫ1 cm Ϫ1 ).
Cloning and Sequencing of the Coding Genes-The identification of both genes encoding FBP aldolase (fba) was based on significant sequence similarity to the recently described E. coli Class I FBP aldolase (DhnA, GenBank TM accession number P71295). The fba gene of T. tenax (EMBL accession number AJ310483) was identified by sequencing the genomic clone (5.2-kb HindIII fragment) harboring the pfp gene (27). The P. furiosus gene (GenBank TM accession number AF368256, NCBI) was identified in the P. furiosus data base (www.genome.utah.edu).
Expression of the FBP Aldolases in E. coli-For expression of the T. tenax FBP aldolase, the coding region was cloned into pET-15b via two new restriction sites (NcoI and BamHI) introduced by PCR mutagenesis with the primers FBPA-f (GCTCAAGCATCCATGGCAAA, sense) and FBPA-rev (CCCCCGTCAGGGATCCTATC, antisense). The following primer set was designed to amplify the P. furiosus open reading frame in pET-24d (NcoI and BamHI) and to delete an internal NcoI restriction site using the PCR-based overlap extension method (38): BG749 (CGCGCGCGCCATGGAGGCCCCTCAAAATGTTGG, sense), BG750 (CCGTGGTCCATCGCGAAGATTAA, antisense), BG751 (TTAATCT-TCGCGATGGACCACGG, sense), and BG688 (GCGCGGATCCT-CAAATGAGACCTTCTGCCTTAGC, antisense). The introduced mutations are shown in boldface, and introduced NcoI and BamHI restriction sites are underlined. The sequence of both expression clones was confirmed by sequencing both strands. Expression of the T. tenax enzyme in E. coli BL21(DE3)pLysS and of the P. furiosus enzyme in BL21(DE3) was performed following the instructions of the manufacturer (Novagen).
Site-directed Mutagenesis of the P. furiosus FBP Aldolase-The active site mutation was introduced in the P. furiosus fba gene using Pfu polymerase in the PCR-based overlap extension method (38). The following primer set was designed to introduce mutation K191A: BG827 (AGCAGATATGATAGCGACCTATTGGAC, sense) and BG828 (GTC-CAATAGGTCGCTATCATATCTGCT, antisense), the introduced mutations are shown in boldface.
Purification of Recombinant FBP aldolases of T. tenax and P. furiosus-For purification of the recombinant T. tenax enzyme, 10 g of E. coli cells were resuspended in 20 ml of 100 mM HEPES/KOH (pH 7.5) containing 300 mM 2-mercaptoethanol and passed three times through a French press cell at 150 megapascals. After centrifugation (20,000 ϫ g, 45 min, 4°C), the crude extract was heat-precipitated (90°C, 30 min), centrifuged again, and dialyzed overnight against 50 mM HEPES/KOH (pH 7.5) containing 5 mM dithiothreitol (2-liters volume, 4°C). The dialyzed fraction was applied to Q-Sepharose fast-flow (Amersham Pharmacia Biotech) equilibrated in the same buffer and eluted with a linear salt gradient of 0 -500 mM KCl. Fractions containing the homogeneous enzyme solution were pooled.
For the purification of the recombinant FBP aldolase from P. furiosus, 3 g of E. coli cells were resuspended in 10 ml of 50 mM Tris/HCl (pH 7.8). The suspension was passed twice through a French press cell (100 megapascals), and cell debris was removed by centrifugation (10,000 ϫ g, 20 min, 4°C). After heat precipitation (70°C, 30 min) and centrifugation, the supernatant was filtered through a 0.45-m filter and loaded onto a Mono Q HR 5/5 column (Amersham Pharmacia Biotech) equilibrated in 50 mM Tris/HCl (pH 7.8). Proteins were eluted by a linear salt gradient of 0 -1000 mM NaCl. Active fractions were pooled, concentrated by microfiltration (Centricon 30, Amicon), and applied to a Superdex 200 prep grade column (Amersham Pharmacia Biotech), equilibrated in 50 mM Tris/HCl (pH 7.8), 100 mM NaCl. Fractions containing the homogeneous enzyme were pooled.
Analytical Ultracentrifugation of the T. tenax FBP Aldolase-Sedimentation velocity and equilibrium analyses were conducted using an analytical ultracentrifuge Optima X-LA (Beckman Instruments, Palo Alto, CA) equipped with double sector cells and an AnTi 50 rotor. The protein was dissolved in 50 mM HEPES/KOH (pH 7.5) containing 100 mM KCl and 2 mM dithiothreitol at a concentration of 0.48 mg of protein/ml. Sedimentation velocity experiments were performed at 30,000 rpm (20°C), and the data were analyzed according to the sedimentation time derivative method (39). Sedimentation equilibrium was analyzed at 6,000 rpm (20°C) using the software provided by Beckman Instruments. Gel filtration experiments were performed as described previously (27).
Northern Blot Analyses of the T. tenax fba Transcript-Preparation of total RNA from auto-and heterotrophically grown T. tenax cells and Northern blot analyses were performed as described before (30). Digoxigenin-labeled antisense mRNA of FBP aldolase and PP i -PFK were obtained by in vitro transcription from the T7 promoter of vector pSPT 19 (Roche Diagnostics GmbH). A part of the coding region of FBP aldolase (502 bp) and the coding region of PP i -PFK (1011 bp) was cloned into the EcoRI and BamHI restriction sites of the vector by PCR mutagenesis using the primer sets CGAGGAGGGGGAATTCCATA (sense) and GAAGGTCTTGGGATCCCCCG (antisense) for FBP aldolase and GCTGGCCGAGCCTCTGAATTCATGAAGATAG (sense) and CTAG-GCAAAGAGGGATCCGGGGCCTAGC (antisense) for PP i -PFK. The introduced mutations are shown in boldface, and the EcoRI and BamHI restriction sites are underlined.
Primer Extension Analyses-Primer extension analyses for T. tenax were performed as described previously (30). To map the transcription start site of the fba-pfp transcript the 5Ј-32 P-labeled antisense oligonucleotide (5Ј-CCGTGCTCAATGCCGTGG-3Ј, position 72-89 of the fba gene) was used as primer for cDNA synthesis. For P. furiosus total RNA was isolated from maltose and pyruvate grown cells as described previously (40), and the transcription start was determined with a fluorescence (IRD800)-labeled antisense oligonucleotide (5Ј-CAAAGTCCG-TAGGGCCGTGC-3Ј (MWG), position 99 -118 of the fba gene). The primer extension reaction was performed using the Reverse Transcription System (Promega) according to the instructions of the manufacturer with the following modifications. Hybridization of total RNA (15 g) and oligonucleotide (5 pmol) was performed for 10 min at 68°C before allowing to cool to room temperature. The reaction (20-l final volume) was started by addition of dNTPs (1 mM), MgCl 2 (5 mM), RNAsin (20 units), and Avian Myeloblastosis Virus reverse transcriptase (22.5 units). After incubation for 30 min at 45°C, the reaction volume was diluted to 50 l with 10 mM Tris/HCl (pH 8.5), 1 l of RNase A (5 mg/ml) was added, and the sample was incubated for 10 min at 37°C. cDNA was precipitated with ethanol and dissolved in 3 l of loading buffer, and 1 l was applied to a sequencing gel in parallel with the sequencing reactions obtained with the same oligonucleotide.
Sequence Retrieval and Phylogenetic Analyses-Protein sequences were extracted from GenBank TM and the TIGR microbial data base using BLAST and first aligned with ClustalW (41); this alignment was manually refined using the MUST program package (42). Regions of uncertain alignment and partial sequences were omitted from the analyses leaving a total of 27 sequences and 172 amino acid positions. The topology of the phylogenetic tree was inferred using the PROTML program of the Molphy version 2.3 package (43), starting with the NJDIST tree using the local rearrangement and the JTT-F options. A gamma parameter-based maximum likelihood estimate of the branch length of the tree as well as of the statistical support for internal nodes (quartet puzzling support values) was performed using the program puzzle version 5 (44). Distance analyses including 1000 bootstrap replicates were performed with the MUST package using the Kimura correction and the neighbor joining method (45). Parsimony bootstrap analysis was performed using PAUP* with 2000 bootstrap replicates and 10 times random addition (46). Secondary structure prediction was performed using the predictprotein program (www.embl-heidelberg/ predictprotein/) (47, 48).

RESULTS
Nucleotide Sequence of the fba Genes of T. tenax and P. furiosus-Both fba genes were identified due to their sequence similarity with the recently characterized Class I FBP aldolase from E. coli (DhnA, dhnA gene) (6). The T. tenax enzyme was identified by sequence analysis of the genomic clone comprising the pfp gene (5.2-kb HindIII fragment), which revealed an additional open reading frame of 792 bp (Fig. 1) preceding the pfp gene (1014 bp) (27). This open reading frame codes for a polypeptide of 263 amino acid residues with a calculated molecular mass of 28.7 kDa and showed high overall similarity (26% identity, blast data base search) to the Class I FBP aldolase (DhnA) of E. coli (6). Strikingly, the coding regions of both T. tenax genes fba and pfp overlap by 1 bp with the A of the start codon (ATG) of the pfp gene being the last nucleotide of the triplet encoding the C-terminal valine (GTA) of the fba gene (Fig. 1). The fba gene of P. furiosus (846 bp) was identified in the P. furiosus data base by similarity of the translated 31.1-kDa polypeptide (282 amino acid residues) to E. coli DhnA (26% identity, blast data base search). Contrary to T. tenax, the P. furiosus fba gene is separated from the next neighbored downstream open reading frame with similarity to agmatinase (speB gene) by 61 nucleotides and therefore is presumably not organized in an operon structure (Fig. 1).
Expression of the fba Genes from T. tenax and P. furiosus in E. coli and Purification of Recombinant FBP Aldolases-The fba gene products of T. tenax and P. furiosus were expressed in E. coli, and their FBP aldolase activity was confirmed for both enzymes using a coupled enzyme assay. For further biochemical studies both recombinant enzymes were purified.  Table I) (6). Like the E. coli enzyme both archaeal FBP aldolases showed additional activity with Fru-1-P, although the much higher K m value for Fru-1-P (T. tenax 498-fold, P. furiosus 197-fold, E. coli 1650-fold) of all three enzymes strongly suggests that FBP is the physiological substrate (Table I). As shown for the FBP aldolase of T. tenax other sugar phosphates such as fructose 6-phosphate, glucose 6-phosphate, fructose 2,6-bisphosphate, and 6-phosphogluconate (concentration range of 5-10 mM) do not serve as substrates in the catabolic direction. Both archaeal FBP aldolases, however, like the E. coli enzyme, were activated in presence of saturating concentrations of citrate (10 mM) by a factor of 2.2 and 2.4, respectively (Table I).
The involvement of a Schiff-base mechanism in the FBP aldolase reaction was examined for the T. tenax enzyme by treating the enzyme with sodium borohydride in the presence and absence of the substrates GAP, DHAP, and FBP. The significant reduction of the specific activity in the presence of the carbonyl substrates DHAP (38% residual activity) and FBP (29% residual activity) as compared with the presence of GAP (80% residual activity) and the control, after NaBH 4 treatment (100% activity, 0.8 units/mg protein, 70°C), accounts for the formation of a Schiff-base in the enzyme reaction. In accordance with these results, a lysine residue is conserved at position 177 in the T. tenax sequence (Fig. 4), which corresponds to the active site Lys-237 (falsely marked as Lys-236) in the E. coli Class I FBP aldolase (DhnA) (6). Finally, the observation that neither metal ions such as Mn 2ϩ , Mg 2ϩ , Zn 2ϩ , Ca 2ϩ , and Fe 2ϩ (concentrations tested, 0.1 and 1 mM) nor EDTA (concen-trations tested, 0.1, 1, and 10 mM) affects the enzyme activity supports the biochemical classification of the T. tenax enzyme as a Class I aldolase. As shown in Fig. 4 also the P. furiosus FBP aldolase exhibits the active site lysine residue (position 191), and the assumed involvement of a Schiff-base mechanism was supported by site-directed mutagenesis of the active site lysine to alanine (K191A) resulting in a virtually inactive mutant enzyme (not shown).
Molecular Mass-The homogeneous FBP aldolases from T. tenax and P. furiosus revealed similar subunit sizes in SDSpolyacrylamide gel electrophoresis of ϳ30 and 33 kDa, respectively, thus being in good agreement with the calculated molecular mass of 28.7 and 31.1 kDa. However, differences between the two enzymes are obvious concerning their oligomeric state under native conditions (Table I). Gel filtration experiments revealed for the recombinant P. furiosus enzyme an apparently uniform oligomer with a molecular mass of 272 kDa (representing presumably octamers), whereas for the T. tenax FBP aldolase two different oligomeric forms were identified. As shown by repeated chromatography of the separated oligomers, both forms are convertible to one another. Sedimentation velocity experiments revealed two distinct oligomers with apparent sedimentation coefficients of 9.34 S and 14.5 S indicating a slow equilibration reaction between the two forms of the T. tenax FBP aldolase. For the smaller association form an apparent molecular mass of 237-245 kDa was determined by sedimentation equilibrium centrifugation suggesting a stoichiometry of eight monomers per oligomer.
Transcript Analyses-To determine if the expression of FBP aldolase of T. tenax and P. furiosus is controlled at transcriptional levels, we examined the effect of the carbon source on fbp transcription. Since the juxtaposition of the fba and pfp gene in T. tenax suggests an operon organization-specific antisense mRNA probes for the pfp and fba gene were used to test for the formation of co-transcripts (Fig. 2). Northern blot experiments were performed with total RNA from autotrophically (in the presence of CO 2 and H 2 ) and heterotrophically (in the presence of glucose) grown T. tenax cells. They revealed a strong hybridization signal for both probes at 1.9 kb and two additional, weaker, probe-specific signals at 1.2 kb for the pfp probe and 0.8 kb for the fba probe, thus indicating the presence of bicistronic as well as monocistronic messages. The signals of both probes were much stronger with mRNA from heterotrophically compared with autotrophically grown cells (Fig. 2). Densitometric quantification of slot blot analysis using the pfp probe and different concentrations of total RNA (10 -0.625 g) from auto-or heterotrophically grown cells revealed a 6-fold higher transcript abundance in the latter (data not shown). Also in P. furiosus cells grown on maltose or pyruvate the transcript level of the fba gene varied similarly (dot blot analysis, data not shown). Like in T. tenax conditions favoring the catabolic direction (growth on maltose) induce a higher transcript amount (2-3-fold increase) as compared with anabolic conditions (growth on pyruvate).
For a more accurate assignment of the promoter region in T. tenax and P. furiosus, the transcription starts of the fba-pfp mRNA and the fba mRNA, respectively, were determined by primer extension analyses. For the T. tenax fbp-pfp operon an antisense oligonucleotide binding at positions 72-89 of the fba gene was used. As shown in Fig. 3A transcription is initiated at the adenosine (A) immediately in front of the start codon (ATG) of the fba gene (position ϩ1). A similar proximity of transcription and translation start site was already observed for the pyk gene, coding for the pyruvate kinase of T. tenax (30) and corresponds with the observation that some Archaea contain a high portion of mRNAs lacking Shine-Dalgano sequences in front of their coding genes (49,50). In accordance with the Northern analyses the amount of copy DNA in the primer extension studies was by a factor 4.5-7.1 higher in hetero-than in autotrophically grown T. tenax cells. The transcription start of the P. furiosus fba mRNA (Fig. 3B) was initiated at the guanosine 10 bp upstream of the ATG start codon (position ϩ1), and in contrast to T. tenax a putative RBS was identified.  Inspection of the 5Ј-flanking regions (Fig. 3C) revealed for the fba genes of T. tenax and P. furiosus AT-rich regions 20 -30 nucleotides upstream of their transcription start sites, which correspond well with the archaeal promoter consensus sequences (51)(52)(53). In T. tenax the TATA box (crenarchaeal consensus sequence (C/T)TTTTAAA) is centered around position Ϫ25/Ϫ26, and 2 bp (Ϫ30 GA Ϫ31) upstream of the TATA box is the putative transcription factor B recognition element (BRE site, consensus sequence (A/G)N(A/T)AA(A/T)). A putative ribosome-binding site (RBS, GGAGG) seems to be absent. In P. furiosus a putative RBS (GGTGA) is identified at position ϩ1 to ϩ5, and the TATA box is positioned around Ϫ24/Ϫ25, and 2 bp upstream is the putative purine-rich BRE site (54).
Phylogenetic Analyses-Data bank searches with the fba genes of T. tenax and P. furiosus revealed sequences with apparent similarity to the Class I FBP aldolases of E. coli (DhnA) in some bacterial and all archaeal genomes, with the only exception being Thermoplasma acidophilum. Whereas most of the genomes analyzed contain only a single dhnA-like gene, Archaeoglobus fulgidus, Methanococcus jannaschii, Halobacterium sp. NRC-1, and E. coli possess two paralogous genes (22). This new FBP aldolase family represents a divergent group with sequence similarities as low as about 20% identity (based on the 172-amino acid core region used for the phylogenetic analyses) between the different groups. Nevertheless, despite this substantial divergence, the universal conservation of the active site lysine (Lys-177, T. tenax; Lys-191, P. furiosus; and Lys-237, E. coli DhnA) and an additional conserved sequence motif preceding the active site lysine (positions 171-176 T. tenax) as well as three further conserved regions, ranging from positions 20 -27, 98 -109, and 199 -204 (numbering of T. tenax fba gene), characterize them unequivocally as homologs of E. coli class I FBP aldolase (DhnA) (Fig. 4).
Strikingly, DhnA homologs do not display significant overall similarity with the members of the classical Class I and Class II FBP aldolases as deduced from automated sequence comparison programs (e.g. Blast search). However, by closer inspection, sequence signatures could be identified resembling the active site region (position 177, T. tenax) and the phosphatebinding motif (position 203-204, T. tenax) of some members of the (␣␤) 8 TIM barrel superfamilies (13), strongly suggesting that this new family of Class I FBP aldolases is at least distantly related to classical Class I FBP aldolases. Moreover, secondary structure predictions (47, 48) performed with the aldolase sequences of T. tenax, P. furiosus, and Sulfolobus solfataricus not only identified these enzymes as (␣␤) 8 barrel proteins but also locate the functionally important residues at equivalent positions to the ones found in classical Class I FBP aldolases as well as in other enzymes of the (␣␤) 8 TIM barrel superfamilies (active site lysine in ␤ 6 , phosphate binding region at the end of ␤ 7 ; Fig. 4) (13). From the high conservation of these key residues we further conclude that the new type of Class I FBP aldolase generally functions as a Schiff-base aldolase acting on phosphorylated substrates.
To analyze the phylogenetic relationships between the various DhnA homologs of Bacteria and Archaea, we aligned 27 sequences of 23 different species and selected a sequence fragment of 172 amino acid residues (Fig. 4) for construction of phylogenetic trees (Fig. 5). The phylogenetic analyses include the three mostly used methods (maximum likelihood, maximum parsimony, and distance-based neighbor joining) and resulted in a complex tree topology with at least 7 deeply rooting branches. Two of them bear exclusively bacterial (branch 1B and 4B) or archaeal sequences (branch 2 and 3) and three compose both archaeal and bacterial sequences (branch 1A, 1C, and 4A).

Aldolases of T. tenax and P. furiosus, Members of a New Type of Class I FBP Aldolase-
The FBP aldolases of T. tenax and P. furiosus resemble specifically the Class I FBP aldolase of E. coli (DhnA) not only on sequence level but also in regard to biochemical properties. In common with E. coli Class I FBP aldolase (DhnA), catalysis of both archaeal enzymes proceeds via a Schiff-base mechanism. The archaeal enzymes, like the E. coli enzyme, exhibit (i) additional enzyme activity with Fru-1-P, albeit at a much higher K m value than for FBP, and (ii) maximal turnover rates that are stimulated by citrate (Table I). Finally, also with respect to quarternary structure both archaeal aldolases show specific resemblance to the Class I enzyme of E. coli (DhnA). All three enzymes tend to form higher oligomerization states representing octa-/decamers or even higher oligomers, whereas the members of the classical Class I and II FBP aldolases form mostly tetramers or dimers, respectively. Thus, structural features and mode of enzyme mechanism classify the FBP aldolases of T. tenax and P. furiosus as members of a new type of Class I FBP aldolase, distinct from classical Class I enzymes, which consists of homologs in almost all Archaea and some Bacteria.
Transcription of the fba Genes of T. tenax and P. furiosus, Integration of the FBP Aldolases in the Physiological Framework-The PP i -PFK (27) and the FBP aldolase catalyze reversible reactions of successive steps in the variant of the Embden-Meyerhof-Parnas pathway of T. tenax, and as such both enzymes fulfill equivalent functions in the anabolic as well as catabolic direction of the pathway. Therefore, the co-transcription of the fba and pfp gene gives rise to the coordinated expression of both enzymes in T. tenax. On the contrary, in most organisms using pathways characterized by a unidirectional working PFK, either dependent of ATP or like in P. furiosus of ADP (21,32), a linkage of FBP aldolase and PFK coding genes does not seem to be meaningful. Sometimes FBP aldolase genes are co-transcribed with genes coding for other reversible enzymes of glycolysis (e.g. glyceraldehyde-3-phosphate dehydrogenase and phosphoglycerate kinase) or of the Calvin cycle (e.g. ribulose bisphosphate carboxylase/oxygenase and phosphoribulokinase) as shown for classical Class II FBP aldolases (5,55,56). Because FBP aldolase is an essential constituent of glycolysis as well as gluconeogenesis, it is remarkable that the fba expression in both organisms T. tenax and P. furiosus is significantly higher under catabolic than under anabolic growth conditions (T. tenax, glucose/CO 2 ; P. furiosus, maltose/pyruvate). An explanation might be that the higher transcript level under catabolic conditions is caused by the necessity of higher carbon flux rates through the pathway for energy conservation than is required for biosynthesis.
A New Family of Aldolases, the Archaeal Type Class I FBP Aldolases-Despite functional similarity with the classical Class I FBP aldolases, the new family of Class I aldolases differs significantly at sequence level. These non-significant average sequence similarities as well as the absence of certain DhnA-typical motifs in classical Class I enzymes characterize this new family of Class I FBP aldolases as a very divergent, new type in addition to classical Class I aldolases. However, both types of Class I FBP aldolases like other (␤␣) 8 (TIM) barrel proteins share, beside the predicted similar secondary structure arrangement, basic common sequence features in regions flanking the active site lysine or engaged in phosphate binding (13,57).
Strikingly, all completed archaeal genomes contain at least one homolog of this new type of Class I FBP aldolases, with the only exception of T. acidophilum, which is supposed to use only the non-phosphorylative Entner-Doudoroff pathway for carbohydrate metabolism (58,59). In contrast to Archaea, only in about 50% of completely sequenced bacterial genomes DhnArelated open reading frames have been identified, and no eucaryal homolog has been assigned yet. At the moment we do not know whether this new type of Class I FBP aldolases is the only enzyme responsible for aldolase activity in Archaea. Reports of metal-dependent Class II aldolase enzyme activity in Haloarchaea (eg. Halobacterium halobium) (16) suggest that additional enzymes might be present, which have not been identified yet in the sequenced genomes, due to their low sequence similarity to known Class I and II aldolases. Because of this so far obviously exclusive occurrence of this new type of aldolase, together with the absence of classical Class I and II aldolases, in Archaea and the non-significant amino acid sequence homology to classical Class I enzymes, we propose to classify this new family as archaeal type Class I FBP aldolases (Class IA) to oppose them to classical Class I aldolases only found in Eucarya and Bacteria.
Phylogenetic Implications-The phylogenetic tree (Fig. 5) is composed of seven deeply branching lineages each bearing members of one or both prokaryotic domains, whose relationships among each other are rather poorly resolved. The presence of Class IA FBP aldolases from Bacteria and Archaea, from Euryarchaeota and Crenarchaeota (e.g. aldolases of Euryarchaeota in branch 1A, 2, 3, 4A; enzymes of Crenarchaeota in branch 2 and 3), or even from one organism (e.g. enzymes of E. coli in branch 1B and 4B) in at least two different deeply rooting main branches suggests that early gene duplication events confer mainly to the characteristic topology of the tree. Probably an early, first gene duplication event (before the separation of Archaea and Bacteria) has led to a segregation into two main paralogous lineages (1A, B, C, 2, and 3, 4A, B). Subsequent gene duplication events in each of the two main lineages combined with differential loss might have created four lineages exhibiting only Archaea (branch 2 and 3) or Bac- teria and Archaea (branch 1A-C and branch 4A, B). This scenario seems to be supported by the fact that the separation of the main branches is in all cases clearly older than the corresponding speciation events in the bacterial or archaeal lineages (e.g. divergence of Euryarchaeota and Crenarchaeota, branch 2 and 3) and in two cases even older than the divergence of Bacteria and Archaea (branch 1A-C and branch 4A, B). An example of a more recent gene duplication is found in branch 1A leading to the isoenzymes of A. fulgidus.
Phylogenetic analyses identified at least 4 lineages in Archaea and 2 lineages in Bacteria. The only limited presence of the archaeal type Class I FBP aldolases in Bacteria might be explained by the assumption of differential loss of the coding genes in some bacterial lineages. Several independent lateral gene transfers from Archaea to Bacteria seem to be unlikely since the branching order of archaeal enzymes (branch 1A and 2) as well as of bacterial enzymes (branch 1B and 4B) fits quite well with the archaeal and bacterial radiation. However, lateral gene transfer events, most likely from Archaea to Bacteria, may be responsible for the surprising occurrence of the aldolases of Aquifex aeolicus and Desulfovibrio vulgaris in branch 1A as well as of Thiobacillus ferrooxidans in branch 4A.
Surprisingly, the FBP aldolase of Treponema denticola shows no affinity to the bacterial branch 1B but is rather closely related to the enzyme of the Crenarchaeote S. solfataricus, with which it forms a highly supported monophyletic group. This close association may perhaps also be due to a lateral gene transfer event. If the T. denticola sequence is omitted from the tree calculation, the S. solfataricus sequence is placed at the basis of branch 1A (albeit with a low bootstrap support of 50), thus complementing the archaeal lineage by a crenarchaeal member.
In summary, the archaeal type Class I aldolase genes seem to be descendants of an ancient lineage separated very early from the gene lineages of the classical Class I and Class II aldolases, probably already in the common ancestor of extant life. The evolution of the new family seems to be mainly the result of non-lateral orthologous evolution including a complex mixture of gene duplications with subsequent differential loss and probably some late lateral gene transfer elements. Surprisingly, up to now no homologs of that gene family have been identified in Eucarya, questioning whether they never possessed these genes or secondarily lost them.