PP_i-Dependent Phosphofructokinase from Thermoproteus tenax, an Archaeal Descendant of an Ancient Line in Phosphofructokinase Evolution

Abstract

Flux into the glycolytic pathway of most cells is controlled via allosteric regulation of the irreversible, committing step catalyzed by ATP-dependent phosphofructokinase (PFK) (ATP-PFK; EC 2.7.1.11), the key enzyme of glycolysis. In some organisms, the step is catalyzed by PP_i-dependent PFK (PP_i-PFK; EC 2.7.1.90), which uses PP_i instead of ATP as the phosphoryl donor, conserving ATP and rendering the reaction reversible under physiological conditions. We have determined the enzymic properties of PP_i-PFK from the anaerobic, hyperthermophilic archaeon Thermoproteus tenax, purified the enzyme to homogeneity, and sequenced the gene. The ∼100-kDa PP_i-PFK from T. tenax consists of 37-kDa subunits; is not regulated by classical effectors of ATP-PFKs such as ATP, ADP, fructose 2,6-bisphosphate, or metabolic intermediates; and shares 20 to 50% sequence identity with known PFK enzymes. Phylogenetic analyses of biochemically characterized PFKs grouped the enzymes into three monophyletic clusters: PFK group I represents only classical ATP-PFKs from Bacteria and Eucarya; PFK group II contains only PP_i-PFKs from the genus Propionibacterium, plants, and amitochondriate protists; whereas group III consists of PFKs with either cosubstrate specificity, i.e., the PP_i-dependent enzymes from T. tenax and Amycolatopsis methanolica and the ATP-PFK from Streptomyces coelicolor. Comparative analyses of the pattern of conserved active-site residues strongly suggest that the group III PFKs originally bound PP_i as a cosubstrate.

As first discovered in Entamoeba histolytica (27), in some members of all three domains of life (Bacteria, Eucarya, and Archaea), the first committing step of glycolysis, the phosphorylation of fructose 6-phosphate (Fru 6-P), is catalyzed not by common ATP-dependent phosphofructokinase (PFK) (ATP-PFK; EC 2.7.1.11) but by an enzyme which uses PP_i as a phosphoryl donor (PP_i-PFK; EC 2.7.1.90) (22–34). The only archaeal PP_i-PFK described so far is the enzyme of Thermoproteus tenax, a hyperthermophilic, anaerobic archaeon which is able to grow chemolithotrophically with CO₂, H₂, and S⁰, as well as chemo-organothrophically in the presence of S⁰ and carbohydrates (11, 41). As shown by enzymatic and in vivo studies (pulse-labeling experiments), T. tenax metabolizes glucose mainly via a variation of the Embden-Meyerhof-Parnas pathway distinguished by the reversible PP_i-PFK reaction (34, 35).

In contrast to the virtually irreversible reaction catalyzed by the ATP-PFK, the phosphorylation by PP_i is reversible. Thus, for thermodynamic reasons, the PP_i-PFK should be able to replace the enzymes of both the forward (ATP-PFK) and reverse (fructose-bisphosphatase [FBPase]) reactions. Consistent with the presumed bivalent function of the PP_i-dependent enzyme, in prokaryotes and parasitic protists which possess PP_i-PFK, little, if any, ATP-PFK or FBPase activity is present. Strikingly, the PP_i-PFKs of these organisms proved to be nonallosteric, suggesting that the control of the carbon flux through the pathway is no longer exerted by the PFK in these organisms. A considerably different situation has been described for higher plants and the green alga Euglena gracilis, showing comparable ATP-PFK, FBPase, and PP_i-PFK activities and allosteric regulation of their PP_i-dependent enzyme by fructose 2,6-bisphosphate (12, 22). However, in most cases it is not obvious which physiological role PP_i-PFK performs: reversible catalysis of glycolysis/gluconeogenesis, increase of the energy yield of glycolysis under certain conditions in which the energy charge is low, or ATP-conservation in obligately fermentative organisms (22).

Closely related to questions concerning the biological function of PP_i-PFKs is the matter of their evolutionary origin: are these enzymes the result of a secondary adaptation from ATP-PFKs, or do they represent an original phenotype, as suggested by their specificity for PP_i, which is thought to be an ancient source of metabolic energy (9, 18, 19, 26). Indicated by sequence similarity (3, 4), all known PP_i- and ATP-PFKs are homologous and therefore originated from a common ancestral root. From more recent studies of Streptomyces coelicolor PFK (4), the previous assumption of a single event which separated PP_i- and ATP-PFKs had to be revised in favor of a multiple differentiation, leaving open, however, the question of the primary or secondary origin of PP_i-PFK.

Understanding of PFK evolution has been impaired by a lack of knowledge concerning archaeal PFK, although the existence of ATP-PFK (31), PP_i-PFK (34), and also ADP-dependent PFK (16, 31) in Archaea has been described. To address the evolution of PFK, we describe the PP_i-PFK from T. tenax and compare its sequence and structure to those of known bacterial and eucaryal PFK enzymes.

MATERIALS AND METHODS

Organism and growth conditions.

T. tenax Kra 1 (41), DSM 2078, was cultivated as previously described (34, 35). Escherichia coli strains were grown at 37°C in Luria-Bertani medium in the presence or absence of ampicillin (100 μg/ml) (29). Methods for measuring enzyme activity and protein content and performance of sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), N-terminal sequencing, and molecular mass determination of the native protein were described previously (34, 35).

Purification of PP_i-dependent PFK.

T. tenax cells (15 g [wet weight]) were resuspended in 25 ml of 0.1 M HEPES-KOH buffer (pH 7.5) containing 0.3 M 2-mercaptoethanol. The cells were disrupted by passing the suspension four times through a French pressure cell (35-ml volume, 200 MPa). Cell debris was removed by centrifugation at 40,000 × g for 30 min (4°C). The supernatant was incubated for 30 min at 80°C, centrifuged again, and applied to a Q-Sepharose fast-flow column (volume, 90 ml; diameter, 2.5 cm; Pharmacia) equilibrated with 50 mM HEPES-KOH (pH 7.5) containing 30 mM 2-mercaptoethanol and 100 mM KCl. Protein was eluted with a linear salt gradient (100 to 400 mM KCl; 400 ml). The fractions containing PP_i-PFK activity were pooled, (NH₄)₂SO₄ was added to a final concentration of 0.4 M, and the mixture was applied to a phenyl-Sepharose CL4B column (volume, 20 ml; diameter, 2 cm; Pharmacia) pre-equilibrated in 50 mM HEPES-KOH (pH 7.5) containing 7.5 mM 2-mercaptoethanol, 200 mM KCl, and 0.4 M (NH₄)₂SO₄. The column was washed with 3 volumes of the same buffer, and the protein was eluted with a linear gradient (120 ml) with a decreasing ion concentration [0.4 to 0.0 M (NH₄)₂SO₄]) and an increasing ethylene glycol concentration (0.0 to 50%). Pooled fractions with PP_i-PFK activity were dialyzed against 50 mM HEPES-KOH (pH 7.5) containing 7.5 mM 2-mercaptoethanol and 300 mM KCl, concentrated by membrane filtration (Centricon 30; Amicon), and subjected to gel filtration on HiLoad 26/60 Superdex 200 prep grade (volume, 325 ml; diameter, 2.6 cm; Pharmacia) preequilibrated in the same buffer. Fractions containing the homogeneous enzyme solution were pooled.

Cloning and sequencing of the PP_i-PFK gene pfp.

The genomic DNA of T. tenax was prepared by the method of Weil et al. (38) modified as described by Meakin et al. (21). Southern hybridizations (8) were performed with an oligonucleotide probe (PFK 1: AT[ACT] TA[CT] CA[CT] GG[AGCT] TGG [AC]G) derived from the hexapeptide IYHGWR, corresponding to positions 35 to 40 of the N-terminal amino acid sequence determined by Edman degradation. Hybridization and signal detection were performed with the digoxigenin kit from Boehringer in accordance with the instructions of the manufacturer. A 5.2-kb HindIII fragment was labelled with the oligonucleotide probe. After elution from the agarose gel, fragments in that size range were cloned into pBluescript II KS⁺ (Stratagene) by using E. coli KL1 Blue (Stratagene) as the host (29). Positive clones were analyzed by restriction analysis and sequencing. The sequence of the gene was determined in both directions by the chain termination method (30) as modified by Wiemann et al. (39) with the aid of an Automated Laser Fluorescent DNA Sequencer (Pharmacia).

Tree construction.

Database searching for homologous amino acid sequences was performed by running the BLASTP program (1) with the amino acid sequence database at the National Center for Biotechnology Information. Sets of amino acid sequences were first aligned with the help of the CLUSTAL program (15) by using the default parameters and then edited by eye. Regions of uncertain alignment were deleted from the final alignment. Phylogenetic trees were constructed by using both the maximum-parsimony and distance matrix methods. The maximum-parsimony method was performed with the PAUP (37) and PHYLIP (10) software packages. For the distance matrix method, we used the matrix options Dayhoff and George-Hunt-Barker to create the distance matrices in PROTDIST. The distance trees were inferred by using the NEIGHBOR program (10). The SEQBOOT, PROTPARS, PROTDIST, NEIGHBOR, and CONSENSE programs of the PHYLIP package were used to derive confidence limits, estimated by 100 bootstrapping replicates.

Nucleotide sequence accession number.

The pfp nucleotide and amino acid sequence data reported in this study has been submitted to the EMBL database under accession no. Y14655.

RESULTS AND DISCUSSION

Purification and enzymic properties of PP_i-PFK.

PP_i-PFK was purified to homogeneity, as judged by SDS-PAGE (Fig. 1), from heterotrophically grown cells. The protein proved to be insensitive to oxygen but highly sensitive to low ionic strength. To prevent inactivation, almost all purification steps were performed at KCl concentrations of 100 to 300 mM. From 15 g of wet cells, 2.6 mg of homogeneous protein with a specific activity of 3.5 U/mg was recovered, corresponding to a yield of 23%. The comparatively low specific activity of the PP_i-PFK of T. tenax (for comparison: Amycolatopsis methanolica, 107 U/mg [2]; E. histolytica, 45.7 U/mg [27]) is, however, mainly due to the low assay temperature of 50°C, which was chosen for the use of mesophilic auxiliary enzymes in the coupled optical test.

FIG. 1.

SDS-polyacrylamide gel electropherogram of the fractions of the following purification steps of PP_i-PFK: molecular size standard (lane 1), crude extract (40 μg of protein; lane 2), heat-treated extract (40 μg of protein; lane 3), anion-exchange chromatography product (20 μg of protein; lane 4), hydrophobic interaction chromatography product (20 μg of protein; lane 5), and gel filtration product (3 μg of protein; lane 6).

Molecular mass determinations of the PP_i-PFK of T. tenax yielded 37 kDa for the subunit (SDS-PAGE) and 100 kDa for the native enzyme (gel filtration). Thus, the quaternary structure of the enzyme (homodimer or homotrimer) could not be deduced unequivocally from the mass ratio.

The enzymic properties characterized the PP_i-PFK of T. tenax as a bidirectionally working enzyme with, however, a slight preference for the phosphorylating direction (Table 1). The enzyme displayed Michaelis-Menten kinetics in the catabolic and anabolic directions with similar affinities for PP_i, Fru 6-P, and fructose 1,6-bisphosphate (Fru 1,6-P₂) (K_m values of 23, 53, and 33 μM, respectively). The affinity for P_i was significantly lower (K_m = 1.43 mM), which might be explained by an adaptation to higher intracellular P_i concentrations, as discussed by Reeves et al. (27) for the PP_i-PFK of E. histolytica.

TABLE 1.

Kinetic parameters of PP_i-PFK^a

Variable substrate	Km (mM)	Vmaxf (U/mg of protein)
PPib	0.023	2.7
Fru 6-Pc	0.053	2.9
Pid	1.43	2.1
Fru 1,6-P2e	0.033	2.0

^aAssay temperature, 50°C. 

^bWith 10 mM Fru 6-P and 1 mM MgCl₂. 

^cWith 5 mM PP_i and 1 mM MgCl₂. 

^dWith 20 mM Fru 1,6-P₂ and 5 mM MgCl₂. 

^eWith 5 mM P_i and 5 mM MgCl₂. 

^fV_max, maximum velocity. 

Like other PP_i-PFKs described so far (those of A. methanolica [2], E. histolytica [27], and Propionibacterium freudenreichii [23]), the enzyme of T. tenax showed high specificity for its substrates. In the phosphorylating reaction, PP_i could not be replaced by ATP or ADP. Glucose 6-phosphate (forward direction) and fructose 2,6-bisphosphate (reverse reaction) were not used as substrates. The activity of PP_i-PFK in both directions depended on the presence of Mg²⁺ ions, as shown by inhibition in the presence of EDTA.

The PP_i-PFK of T. tenax is not regulated by the known allosteric modulators of PFKs such as adenine nucleotides (ATP, ADP, and AMP; concentrations tested, 2, 5, and 10 mM), metabolites (glucose, pyruvate, phosphoenolpyruvate, and citrate; concentration tested, 5 mM), and fructose 2,6-bisphosphate (concentrations tested, 0.1 and 1 mM).

Current knowledge indicates that nonallosteric PP_i-PFK is only present in organisms which integrated this enzyme into their basic carbohydrate metabolism, replacing ATP-PFK and FBPase (P. freudenreichii, E. histolytica, etc.). Since one of the main control points of the Embden-Meyerhof-Parnas pathway is lost with this substitution, the flux through the pathway must be regulated in different ways. As reported previously, T. tenax does not show any ATP-PFK or FBPase activity (34). In this respect, the nonallosteric phenotype of PP_i-PFK corresponds to its proposed function as an integral constituent of basic carbohydrate metabolism. Obviously, in this organism the main control point of the pathway, besides pyruvate kinase, is an irreversible, nonphosphorylating glyceraldehyde 3-phosphate dehydrogenase (33).

Nucleotide sequence of the pfp gene.

A clone containing a 5.2-kb HindIII fragment of the genomic DNA of T. tenax was selected by hybridization with an oligonucleotide probe derived from the N-terminal sequence of the purified protein. Sequence analysis revealed an open reading frame whose deduced amino acid sequence corresponds to the determined N-terminal protein sequence. The complete sequence of the T. tenax pfp gene (Fig. 2) comprises 1,011 bp coding for 337 amino acids with a calculated molecular mass of 36.8 kDa, which corresponds to the subunit molecular mass of 37 kDa determined by SDS-PAGE.

FIG. 2.

The nucleotide sequence of the pfp gene of T. tenax and its flanking regions. The deduced amino acid sequence is shown beneath the nucleotide sequence. The ATG start and TGA stop codons are underlined, and the determined N-terminal amino acid sequence of the purified protein is in boldface.

No convincing motifs for transcription or translation signals were recognized in the flanking regions. A partial open reading frame upstream (position 0 to 363) and a short open reading frame downstream (position 1485 to 1709) of the pfp gene were identified with, however, no significant similarities to known proteins. Whether the pfp gene is part of an operon, as found for the pfp and pfk genes of several bacterial species (3, 5, 20, 28), remains to be determined.

Overall sequence similarity of PP_i-PFK and functional residues.

To determine the overall sequence similarities of the T. tenax enzyme with known PFKs and to identify structural features correlated with its PP_i dependence, we aligned the T. tenax PFK with 30 sequences of enzymes whose cosubstrate specificity was clearly defined. The sequence data set comprises 10 sequences of PP_i-PFKs (7 eucaryal sequences, 2 bacterial sequences, and 1 archaeal sequence) and 21 sequences of ATP-PFKs (12 eucaryal and 9 bacterial sequences). To ensure reliable assignment of homologous positions, we selected three well-conserved sequence fragments corresponding to positions 1 to 112, 118 to 192, and 214 to 223 of the T. tenax sequence, comprising 201 residues (Fig. 3).

FIG. 3.

Multiple alignment of ATP- and PP_i-PFKs. The amino acid sequence of the PP_i-PFK of T. tenax was aligned with 12 amino acid sequences of PP_i- and ATP-PFKs in the EMBL and SwissProt databases. Gaps introduced for optimal alignment are marked by hyphens, and functionally important residues are marked by the letters F (Fru 6-P binding site) and A (ATP or PP_i binding site). Abbreviations: T. te, T. tenax; A. me, A. methanolica; S. co, S. coelicolor; Th. th, Thermus aquaticus subsp. thermophilus; B. st, B. stearothermophilus; E. co, E. coli; Ho. sa, Homo sapiens; Sa. ce, Saccharomyces cerevisiae (α and β subunits); G. la, Giardia lamblia; So. tu, Solanum tuberosum (β subunit); E. hi, E. histolytica; P. fr, P. freudenreichii subsp. shermanii.

As shown by the obvious sequence similarity (Fig. 3), all of the PFKs, irrespective of cosubstrate specificity, are homologous and thus have evolved from a common ancestor (3, 4). Although the overall sequence similarity does not reflect basic structural differences between PP_i- and ATP-PFKs, the ATP-dependent enzymes are more similar to each other (42 to 72% identity) than are the PP_i-dependent enzymes, which exhibit considerable heterogeneity (24 to 51% identity). Strikingly, the PP_i-PFK of T. tenax showed, on average, higher similarity to ATP-PFK (39 to 41% identity) than to the PP_i-dependent homologs (20 to 30% identity). The highest similarities found were to the ATP-PFK of S. coelicolor (46.8% identity) and to the PP_i-dependent enzyme of A. methanolica (46.3% identity).

To identify sequence motifs correlated with the cosubstrate specificity and thus suitable for obtaining closer insights into the structural basis of its differentiation, we focussed on active-site residues assigned by homology on the basis of the resolved three-dimensional structure of the bacterial ATP-PFKs of E. coli and Bacillus stearothermophilus (14, 32). As demonstrated by experimental verification, the prediction of functionally important residues by homology proved to be reliable even for distantly related PFK species (e.g., PP_i-PFK from P. freudenreichii [40]), indicating a highly conserved spatial structure in the active-site region of ATP- and PP_i-PFKs.

To assign residues involved in substrate and cosubstrate binding, we considered only residues in unequivocally homologous sequence regions. Residues located in the vicinity of gaps introduced for optimal alignment, indicating major spatial deviations, were disregarded (e.g., residues 70 and 71 in the T. tenax PFK, corresponding to residues 72 and 73 in the B. stearothermophilus PFK). For the alignment, only the catalytically active β subunits of the PP_i-dependent plant enzymes were considered. The residues assigned to the Fru 6-P and ATP or PP_i binding of the various PFKs are listed in Table 2.

TABLE 2.

Active-site residues of PFKs.

Site and positiona	ATP-PFK			PPi-PFK
	Eucarya (12 species)	Bacteria		Eucarya (7 species)	Bacteria		Archaea (T. tenax)
		8 species	S. coelicolor		P. freudenreichii	A. methanolica
Fru 6-P binding
125 (124)	S (D)	T	T	T	T	T	T
127 (126)	D (Q)	D	D	D	D	D	D
129 (128)	D	D	D	D	D	D	D
162 (161)	R	R	R	Y (R, G)	R	R	R
169 (168)	M (Q)	M	M	M	M	M	M
170 (169)	G (V)	G	G	G	G	G	G
171 (170)	R (S)	R (H)	R	R	R	R	R
222 (220)	E	E	E	E	E	E	E
ATP or PPi binding
11 (10)b	G	G	G	G	G	G	G
41 (39)c	Y	Y (F)	W	P (F, Y)	Y	W	W
77 (75)d	R (K)	Kf	K	I (T, A)	L	K	K
103 (101)e	D	D (N)	E	D	D	E	D
104 (102)b	G	G	D	D (G)	D	D	D
105 (103)b	S	S (T)	T	S (T)g	T	T	T
107 (105)d	T (I)	Hf	G	T (F, R)	T	G	G
108 (106)d	G	G	V	N (S, G)	T	V	A

^aThe residues involved in substrate and cosubstrate binding were assigned on the basis of the resolved three-dimensional structures of the ATP-PFKs of E. coli and B. stearothermophilus. The functions of the residues in E. coli and B. stearothermophilus ATP-PFKs are described in references 14 and 32. Numbering is according to the B. stearothermophilus sequence; the numbers in parentheses refer to the T. tenax sequence. Dominant residues are in boldface, and residues in parentheses occur in one species. 

^bBonds to β-P. 

^cContacts ribose. 

^dContacts adenine. 

^eBinds to Mg²⁺. 

^fDeviation in more than two species. 

^gOccurs in two species. 

Virtually all of the residues assigned to Fru 6-P binding are conserved, irrespective of the type and origin of the enzyme, indicating highly similar spatial arrangements of the substrate binding site throughout all PFKs. On the contrary, the conservation of the residues predicted for binding of the phosphoryl donor is significantly lower. Only a Gly at position 11 is common to all of the PFKs. At position 103, an acidic residue (Asp or Glu) is dominant. Strikingly, the PP_i-dependent enzymes and the ATP-PFK of S. coelicolor showed systematic deviations at residues 41, 104, 105, and 108, which were identified as important for ATP binding in the E. coli and B. stearothermophilus enzymes (14, 32), suggesting that these changes correlate with differences in cosubstrate binding.

Although the residues at positions 77 and 107 have been assigned important functions in adenine binding, their variability in ATP-dependent enzymes is too high for them to be indicative of cosubstrate binding specificity. The ATP-PFKs, except for the ATP-PFK of S. coelicolor, are characterized by Tyr at position 41, Gly at positions 104 and 108, and Ser at position 105. The high conservation of these residues is ascribed to their specific contact with ribose (Tyr 41) or adenine (Gly 108) or just the limited space left between the polypeptide chain and the bulky ligand (Gly 104 and Ser 105).

The assigned residues in the PP_i-PFKs deviating from the respective pattern of the ATP-dependent enzymes are a hydrophobic—mostly aromatic—residue at position 41, Asp at position 104, Ser or Thr at position 105, and an unspecified residue at position 108. As shown in Table 2, the T. tenax PFK fits very well into the common pattern of PP_i-PFKs.

Surprisingly, at four of the eight listed positions, the ATP-dependent enzyme of S. coelicolor shows the characteristic residues of PP_i-PFKs. Furthermore, six of the eight residues are identical in the ATP-dependent enzyme of S. coelicolor and the PP_i-dependent enzymes of T. tenax and a complete accordance could be observed with the PP_i-PFK of A. methanolica. This close resemblance—despite the difference in cosubstrate specificity—hinted that structural determinants other than the predicted active-site residues govern the cosubstrate specificity of this enzyme. The high level of correspondence to the structural features characteristic of the PP_i-dependent enzymes, along with the preferred overall sequence similarity to the PP_i-PFKs from A. methanolica (74.1% identity) and T. tenax (46.8% identity), strongly suggests that the S. coelicolor enzyme evolved from a PP_i-dependent precursor and converted more recently to the ATP-dependent phenotype.

Phylogenetic analyses.

The enzyme of T. tenax is the first archaeal PFK available for phylogenetic studies. This offers the opportunity to gain insights into the differentiation of substrate specificity of these enzymes, as well as indications about the phenotype of the common PFK precursor.

As shown by the sequence similarity, the archaeal enzyme is a homolog of the known PFKs. This finding indicates once again that the central carbohydrate metabolism of glycolysis was established before the segregation of the three domains of life and that the present-day variations of the pathway observed in different phylogenetic lineages must be a result of divergence rather than convergence.

Phylogenetic trees were constructed by using a data set of 31 sequences with the aid of the maximum-parsimony and distance matrix methods. Only sequences of PFKs with defined cosubstrate specificity were included in the alignment. The different methods gave us the same tree topology, apart from a minor change. The ATP-PFK from Drosophila melanogaster, which forms a common stem with the ATP-PFK of Haemonchus contortus in the neighbor-joining tree, represents a separate lineage in the maximum-parsimony tree.

Surprisingly, the PFKs cluster into three different monophyletic groups (Fig. 4): (i) PFK group I, consisting of only ATP-PFKs; (ii) PFK group II, consisting of only PP_i-PFKs; and (iii) PFK group III, comprising both ATP- and PP_i-PFKs.

FIG. 4.

Phylogenetic tree of PFKs. The tree is based on distance analysis (neighbor-joining method) of sequences of PP_i- and ATP-PFKs in the EMBL and SwissProt data banks. Only enzymes whose substrate specificity was clearly defined and sequence regions (positions 1 to 112, 118 to 192, and 214 to 223 of the T. tenax sequence) which show unequivocal similarity were included (Fig. 3). Bootstrap values (neighbor-joining and maximum-parsimony methods) are indicated at basal nodes (in percentages) and are based on 100 data sets. Groups: I, Dictyostelium discoideum, H. contortus, D. melanogaster, H. sapiens (muscle type), Rattus norvegicus (muscle type), Oryctolagus cuninculus (muscle type), Schistosoma mansoni, Aspergillus niger, S. cerevisiae, Klyveromyces lactis, T. aquaticus subsp. thermophilus, B. stearothermophilus, L. delbrueckii, L. lactis, B. macquariensis, C. acetobutylicum, Spiroplasma citri, and E. coli; II; P. freudenreichii subsp. shermanii, E. histolytica, Ricinus communis, S. tuberosum, G. lamblia, Naegleria fowleri; III, T. tenax, S. coelicolor, and A. methanolica.

The ATP-PFKs of group I form a coherent cluster divided into two branches of solely bacterial or eucaryal enzymes that are well separated from the other two monophyletic groups (PFK groups II and III) by a common stem. The branching order of the bacterial ATP-PFKs only partially follows the universal tree based on 16S rRNA sequences (24) and strongly suggests the presence of at least two paralogous lines (represented by the gram-positive bacteria [i] B. stearothermophilus, Lactococcus lactis, and Lactobacillus delbrueckii and [ii] Clostridium acetobutylicum and B. macquariensis). The branching point of the eucaryal PFKs is paralleled by a basal gene duplication leading to a more complex regulation capacity of these enzymes (12, 13, 25). Overall, the topology of this eucaryal subtree primarily reflects a vertical heritage of pfk genes.

The branching system of PFK group II includes bacterial and eucaryal PP_i-dependent enzymes and is generally characterized by long branch lengths, probably due to either specific functional adaptation or release from functional constraints. There is no evidence for a gene duplication event between the bacterial nonallosteric PFK of P. freudenreichii and the allosteric eucaryal plant enzyme, as described for ATP-PFK group I. However, the α and β subunits of plants evolved by gene duplication (7, 12).

The PFK group III cluster is characterized by rather short branches comprising not only the PP_i-dependent enzymes of A. methanolica and T. tenax but also the ATP-PFK of S. coelicolor. The topology of the tree is supported by fairly good bootstrap values: the separation of the group III PFKs from the group II enzymes is confirmed by bootstrap values of 86% (neighbor joining) and 89% (maximum parsimony) and its separation from the group I enzymes by bootstrap values of 71% (neighbor joining) and 50% (maximum parsimony), respectively. The affiliation of the T. tenax PFK and the enzymes of A. methanolica and S. coelicolor is ensured by an 82% (neighbor joining) or 70% (maximum parsimony) bootstrap value and further supported by unique sequence signatures (Fig. 3, GWRG [position 38 to 41], SRTNP [position 69 to 73], TLG [position 104 to 106], and AGW [position 172 to 174]).

The unexpectedly close clustering of a bacterial and archaeal PP_i-PFK with a bacterial ATP-dependent enzyme apart from all of the other ATP-dependent representatives and the striking correspondence of the active-site residues lead to the assumption that the change of cosubstrate specificity happened more recently and independently from the basic event which separated the PFKs of groups I and II. Since—as outlined above—the pattern of the active-site residues of group III enzymes coincides with that of group II enzymes, classifying all group III enzymes as primarily PP_i dependent, the change in cosubstrate specificity must have been from PP_i to ATP in the S. coelicolor enzyme. We cannot estimate how complex these structural changes must be to switch the binding specificity from PP_i to ATP. However, we assume that such changes occur more easily in a rather unspecified structure.

Unfortunately, in all three main branches of the PFK tree, only members of two domains are present; thus, we cannot decide whether all three main lineages trace back to a common ancestor. Without further sequence information, we cannot rule out the possibility that the differentiation of the PFK occurred after the three domains had been segregated and the observed relationship is the result of lateral gene transfer events between domains. Under this aspect, the sequence information of other known archaeal ATP-PFKs and ADP-dependent PFKs would be of special relevance. The recently published archaeal genomes of Methanococcus jannaschii (6), Archaeoglobus fulgidus (17), and Methanobacterium thermoautotrophicum (36) comprise no open reading frames homologous to the known pfk and pfp genes.

Given the present stage of knowledge, the nonspecialized features of the PFK group III enzymes (from T. tenax, A. methanolica, and S. coelicolor), reflected by their limited regulatory capacity, their slow evolution rates, and their original specificity for PP_i, which is assumed to be the primary phosphoryl donor in metabolic processes (9, 18, 19, 26), strongly suggest that these enzymes represent the descendants of the most ancient lineage within the known PFKs.