From ATP as substrate to ADP as coenzyme: functional evolution of the nucleotide binding subunit of dihydroxyacetone kinases.

Dihydroxyacetone kinases are a family of sequence-related enzymes that utilize either ATP or a protein of the phosphoenolpyruvate:sugar phosphotransferase system (PTS) as a source of high energy phosphate. The PTS is a multicomponent system involved in carbohydrate uptake and control of carbon metabolism in bacteria. Phylogenetic analysis suggests that the PTS-dependent dihydroxyacetone kinases evolved from an ATP-dependent ancestor. Their nucleotide binding subunit, an eight-helix barrel of regular up-down topology, retains ADP as phosphorylation site for the double displacement of phosphate from a phospho-histidine of the PTS protein to dihydroxyacetone. ADP is bound essentially irreversibly with a t((1/2)) of 100 min. Complexation with ADP increases the thermal unfolding temperature of dihydroxyacetone L from 40 (apo-form) to 65 degrees C (holoenzyme). ADP assumes the same role as histidines, cysteines, and aspartic acids in histidine kinases and PTS proteins. This conversion of a substrate binding site into a cofactor binding site reflects a remarkable instance of parsimonious evolution.

Few compounds are as ubiquitous and highly connected as adenine nucleotides (1). ATP functions as a carrier of chemical energy, ADP, AMP, ADP-ribosyl, adenylyl moieties as enzyme regulators, and cAMP as a second messenger. The nucleotide coenzymes NADH, FAD, and coenzyme A contain an adenosyl group that without direct participation in catalysis assists in binding to the apoenzyme. Here we report on ADP acting as phosphorylation site in the double displacement phosphoryl transfer reaction catalyzed by the Escherichia coli dihydroxyacetone (Dha) 1 kinase. Dha kinases are a family of sequencerelated enzymes that can be divided into two groups according to their phosphate donor, namely ATP or a phosphoprotein of the bacterial phosphoenolpyruvate:sugar phosphotransferase system (PTS) (2) (Fig. 1A). ATP-dependent kinases occur in bacteria, yeast, animals, and plants; Dha kinases dependent on the PTS, a dedicated energy transducing system involved in carbohydrate uptake and control of carbon metabolism (3,4), occur only in bacteria. In methylotrophic yeast, free Dha is the product of a transketolase reaction between ribulose-5-phosphate and formaldehyde derived from methanol. In bacteria, Dha is formed by oxidation of glycerol (5)(6)(7). For yeast it has been shown that Dha kinases fulfill a "housecleaning" function by removing chemically reactive (8) and potentially hazardous short chain carbohydrates (9).
The Dha kinases of C. freundii (DAK) and of E. coli (DhaK, DhaL) are prototypes of ATP-and PTS-dependent kinase, respectively (2,10,11). The former consists of two domains that are connected by a long flexible linker, the latter of two subunits (DhaK and DhaL) that show homology with DAK throughout their combined lengths. DhaK contains the Dha binding site, DhaL the nucleotide binding site (12,13). The DhaL fold, an eight-helix barrel of regular up-down topology, constitutes a new scaffold of nucleotide-binding proteins (Fig.  1, B and C). The PTS-dependent kinases utilize an additional subunit (DhaM) that serves as a phosphate shuttle between the general phosphoryl carrier protein HPr of the PTS and the kinase (Fig. 1A). DhaM subunits are made up of one domain that delivers the phosphate to the kinase and (species-dependent) one or two additional domains that form a phospho-histidine relay to the PTS. All three domains are homologous to known proteins of the PTS (2).
The sequence and functional similarity between PTS-and ATP-dependent kinases points to their common origin. A phylogenetic tree generated from 41 DhaL subunits and domains shows three distinct protein clusters (Fig. 1D). The first contains the DhaL-like domains from ATP-dependent kinases of animals, plants, yeast, and bacteria. The second contains three types of DhaL subunits: (i) DhaL encoded by operons containing genes for DhaK subunits and multidomain (M) or singledomain (m) DhaM, (ii) DhaL encoded in dhaKL operons without a dhaM gene (x), and (iii) DhaL paralogs encoded in an operon together with a protein that itself consists of a SorCtype transcription factor and a DhaK-like domain (12). The two subunits are presumed to control Dha kinase expression (14). The third cluster contains domains from proteins of unknown function that are related to DhaL only by similarity of fold (12). The evolutionary relationship between ATP-and PTS-dependent kinases is elucidated by the observation of an isotope exchange between Dha and Dha phosphate (DhaP). It pointed to a double displacement mechanism, and together with our previous failure to detect a covalent phospho-histidine or phosphoaspartate intermediate (2) it eventually led to the identification of a tightly bound ADP acting as phosphate acceptor in the kinetically active intermediate.

MATERIALS AND METHODS
Protein Purification and Activity Assays-Proteins were purified as previously described (2, 15). All activity assays and analytical gel filtrations of DhaL were done in salt buffer (1 mM MgCl 2 , 50 mM NaCl, 10 mM HEPES, pH 7.5, and 10 M ADP unless indicated otherwise). Equilibrium phosphate exchange assays contained per 0.1 ml: 1 M DhaK, 0.1 mM [ 14 C]Dha, 0.8 mM DhaP, and DhaL or DhaM as indicated. After 30 min of incubation at 37°C, the reaction was stopped by dilution with 2 ml of ice-cold water, and [ 14 C]DhaP was separated by ion exchange chromatography and determined by liquid scintillation counting as described (16). Dha kinase assays contained per 0. [ 32 P]P-enolpyruvate was prepared as described (18). [␣-32 P]ADP was prepared from [␣-32 P]ATP by incubation with hexokinase and an excess of glucose. The incubation mixture contained 0.1-0.1 mM ATP, 0.45 Ci of [␣-32 P]ATP (Amersham Biosciences), 1 mM glucose, 20 g of hexokinase/1 ml of buffer A (20 mM Tris-HCl, pH 8.3, 150 mM NaCl, 5 mM MgCl 2 , 2 mM dithiothreitol). Hexokinase was inactivated by heating at 80°C for 10 min, and [␣-32 P]ADP was used without further purification.
Protein Unfolding-Temperature-induced unfolding was monitored by circular dichroism spectroscopy at 222 nm in a Jasco spectropolarimeter (J715-A) using a 0.5-mm light-path cell. 20 M DhaL and 100 M indicated nucleotide were heated at a rate of 60°C h Ϫ1 in buffer A (5 mM MgCl 2 , 150 mM NaCl, 10 mM Tris-HCl, pH 8.5, 1 mM dithiothreitol, 10% glycerol). Data were normalized to the difference of the linear changes above and below the transition region.
Isotope Exchange Experiments-The DhaL⅐[␣-32 P]ADP complex was prepared by incubation of 0.67 ml of 80 M DhaL⅐ADP in buffer A containing 100 M free ADP and 2 Ci of [␣-32 P]ATP overnight at 20°C. During this time [␣-32 P]ADP was formed by the intrinsic ATPase activity of DhaL, and isotopic equilibrium with ADP was established. The DhaL⅐[␣-32 P]ADP complex was then separated from free [␣-32 P]ADP by gel filtration on a Superdex-75 column (Amersham Biosciences) in buffer A at room temperature. Peak fractions containing DhaL⅐[␣-32 P]ADP were pooled. The ADP/DhaL molar ratio in the peak fractions was between 0.9 and 0.95. ADP exchange was initiated by the addition of a 166-fold molar excess of cold ADP (10 l of 20 mM ADP) to 190 l of 6 M DhaL⅐[␣-32 P]ADP. After incubation at 37°C for the time indicated, the DhaL⅐[␣-32 P]ADP complex was separated from free [␣-32 P]ADP by gel filtration on a HI Trap desalting column (Amersham Biosciences) in buffer A. The k off rate was calculated by non-linear least square fitting of the radioactivity counts in the two fractions to the equations cpm (t) ϭ cpm 0 exp Ϫ {k off ⅐t} and cpm (t) ϭ cpm 0 (1 Ϫ exp Ϫ {k off ⅐t}), respectively.
Sequence Analysis-Amino acid sequences were aligned with CLUSTALW. Analyses were performed with 41 of 61 representative sequences from bacteria and eukaryotes. Sequences of paralogs were included; orthologs of closely related organisms were omitted. The phylogenetic tree was constructed with Gene-Bee (13) using the default parameters. Swiss-Prot primary accession numbers are given. The sequences of Klebsiella pneumoniae and Mycobacterium smegmatis were obtained at genome.wustl.edu/projects/bacterial/kpneumoniae/ and tigrblast.tigr.org/ufmg/index.cgi?data base ϭ m_smegmatis seq.

RESULTS AND DISCUSSION
The DhaL Subunit of the PTS-dependent Dha Kinase Is Transiently Phosphorylated-The Dha binding subunit DhaK,  (12). C, hydrogen-and metal-binding interactions between AMPPNP and conserved residues of the nucleotide binding site (12). Numbers refer to residues of the C. freundii kinase, numbers in parentheses to the E. coli DhaL subunit. His-38 is invariant in the PTS-dependent kinases, Thr-388 is conserved in most ATP-dependent kinases. D, phylogenetic tree of the DhaL domains of Dha kinases. See "Results" for details. DAK refers to two-domain Dha kinases (solid bar ATP-and PTS-dependent; broken bar putative PTS-dependent). PgdK is encoded in a genetic island of meningitic E. coli. DHBK, 3,4-dihydroxy-2-butanone kinase (21). Bootstrap percentages are given for 100 resampled alignments. Bar, 0.1 replacement per position.  (Fig. 2), indicating that DhaL was necessary and sufficient as the phosphate transferring subunit. His-38 of DhaL appeared as a prime candidate for a protein phosphorylation site (12) because (i) it is invariant in all PTS-dependent kinases, (ii) the H38T and the H38A mutants were inactive, (iii) His-38 corresponds to the nucleotide binding Thr-388 of the Citrobacter freundii kinase (Fig. 1, B and C), and (iv) PTS proteins generally are phosphorylated at histidines (3,4).
After incubation of DhaL with [ 32 P]P-enolpyruvate (phosphoenolpyruvate) and catalytic amounts of EI, HPr, and DhaM, the 32 P label comigrated with DhaL on gel filtration in EDTAfree buffer while it was released in the presence of EDTA (Fig.  3, A and B). 32 P-labeled, histidine-tagged DhaL could be affinity captured with Ni 2ϩ -nitrilotriacetic acid beads but could not be detected in non-denaturing polyacrylamide gels (not shown). The EDTA sensitivity, and the fact that the nucleotide is coordinated by two well ordered magnesium ions in the x-ray structure of the C. freundii kinase (Fig. 1C) (12), suggested that a nucleotide may also play a role in the PTS-dependent Dha kinases. To demonstrate that the [ 32 P]phosphate comigrating with DhaL is the ␥ phosphate of ATP and that this ␥ phosphate is actually transferred to Dha, the peak fraction from the gel filtration ( Fig. 3A) was split in two. DhaK and Dha were added to one half and EDTA (5 mM) to the other. Phosphate was transferred quantitatively to Dha in the presence of DhaK while [␥-32 P]ATP was released in the presence of EDTA (Fig.  3C, lanes a and b). This demonstrates that DhaL⅐ATP is the catalytically competent intermediate and that hydrolysis of ATP to ADP does not represent a control mechanism, for instance to turn off the enzyme.
DhaL Contains ADP as a Coenzyme-To further support this proposition, histidine-tagged DhaL was purified by metal chelate affinity chromatography followed by gel filtration in the presence of EDTA. DhaL thus purified is inactive, severely unstable, and prone to precipitate at room temperature. Addition of ADP or ATP restored activity, whereas the non-hydrolyzable analogue AMPPNP, the structural analogue ADP␤S, AMP-vanadate, and AMP did not (Fig. 4A). The ADP concentration necessary for half maximal activity (AC50) is 18 nM. The H38T mutant also could be activated by ADP, but it had a 1000-fold higher AC50 of 38 M (Fig. 4B), indicating that His-38 is critical for strong binding of the ADP cofactor and not, as presumed, as a phosphorylation site. The dissociation rate of the DhaL⅐ADP complex (k off rate) was measured by isotope exchange between the purified DhaL⅐[␣-32 P]ADP complex and a 166-fold molar excess of unlabeled ADP (Fig. 5A). The k off rate calculated from the decay curve is 11.2 Ϯ 1.3⅐ 10 Ϫ5 s Ϫ1 . This corresponds to a half-life (t1 ⁄2 ) of 100 min for the Dha⅐ADP complex, which is three times the generation time of a rapidly dividing E. coli cell. Taking the AC50 value of 18 nM (Fig. 4B) as an estimate for K d , the calculated k on rate for the DhaL ADP association thus is 6.2⅐10 3 s Ϫ1 M Ϫ1 . This value is three orders of magnitude slower than the average association rate constants of enzymes that bind nucleotides as substrates and after each turnover release them as products (19). The slow association rate may be due to desolvation requirement (19) or, more likely, to a major conformational adjustment required for the opening of the ADP binding site.
Complexation with ADP increases the thermal unfolding temperature of the DhaL apoenzyme from 41.5 to 65.7°C (Fig.  5B), providing additional evidence for an essentially irreversible apoenzyme-coenzyme complex and the important role of ADP in stabilizing the protein. AMPPNP and ADP␤S increase the unfolding temperature of DhaL by 9 and 12°C, respectively, relative to the nucleotide-free form. The H38T and H38A mutants, which have a 1000-fold lower affinity for ADP, are moderately stabilized by the nucleotide (Fig. 5C).
The fact that ADP and ATP have the same cofactor activity whereas AMPPNP has none (Fig. 4A) supports the proposition that ADP is the active moiety. DhaL has a weak ATPase activity that produces a burst of one [␣-32 P]ADP/DhaL upon incubation of apo-DhaL with [␣-32 P]ATP (Fig. 6A). Conversely, a stoichiometric amount of [␥-32 P]ATP is rapidly formed but not released when the DhaL⅐ADP complex is incubated with [ 32 P]P-enolpyruvate in the presence of DhaM and the general phosphoryl carrier proteins of the PTS (Fig. 6B). The two burst reactions indicate that ADP and ATP are not exchanged on a time scale of several minutes. In contrast, the H38T mutant with low affinity to ADP catalyzed the slow and continuous hydrolysis of ATP as well as a PTS-dependent production of ATP from ADP (Fig. 6, A and B). Congruent with the observed nucleotide exchange is the weak ATP-dependent Dha kinase activity of the H38T mutant (Fig. 6D). Not unexpected for a single site mutation, this increase is modest and the activity remains 400 times lower than the PTS-dependent activity (Fig. 6C).

ATP-and ADP-dependent DhaL Have a Common Origin-
The tightly bound ADP participates in a double displacement phosphoryl transfer reaction and thus plays the same role as histidines, cysteines, and aspartic acids in other phosphoprotein intermediates. Such a cofactor function of ADP has to the best of our knowledge not been described before. Retaining ADP as cofactor instead of evolving an amino acid such as a histidine into a phosphorylation site is an unprecedented example for parsimony in functional evolution during the switch of a kinase from ATP to the PTS as source of high energy phosphate. Which modifications of the nucleotide binding site accompanied this switch? In loop 1 (LDXXXGDGD(T/ H)GXNM) of the helix barrel, the threonine that is in hydrogen bonding distance to the ␣ phosphate of ATP (12) in the ATPdependent kinase of C. freundii is replaced by a histidine that is invariant in the PTS-dependent forms. Although interactions of nucleotide oxygens with aspartate, serine, threonine, lysine, and arginine are frequent, interactions with histidine were not detected in over 3000 contacts analyzed (1), adding evidence to our finding (Fig. 4B) that this histidine may be important for tight binding of ADP. In loop 3 (GG(S/A)SGXLYG) a serine/ threonine that coordinates with the ␥ phosphate in the ATPdependent kinase (12) is replaced by an alanine in most PTSdependent kinases. In the conserved sequence of loop 5 (GXXAXXGDKTM) no conspicuous differences between the two forms can be detected. The most remarkable of the four loops surrounding the nucleotide binding site is loop 7 (GRASYL(X/ G)(X/E)(X/R)(X/S)(X/L)(X/G)XXDPGA). The middle section of this long loop is highly variable in the ATP-dependent kinases, and the loop is disordered in the x-ray structure of the C. freundii kinase in complex with AMPPNP (12). In the PTSdependent DhaL, however, not only the flanking region (GRASYL and DPGA) but also the central portion (GERSLG) is almost invariant. Loop 7 is long enough to latch over the nucleotide binding pocket, and the rigidity of this loop could determine the rate of nucleotide exchange. Instead of being exchanged after each round of phosphate transfer, ATP is regenerated from ADP in situ by DhaM, a dedicated phosphoprotein of the PTS. Evolution of DhaM must have occurred by the more "conventional" path involving gene duplication, fusion, and specialization of protein modules of the PTS that already have the intrinsic property of multiply interacting with proteins as diverse as carbohydrate transporters, enzymes, and transcriptional regulators. It is noteworthy that some bacteria (Sinorhizobium, Mycobacterium, Listeria, and Klebsiella) have genes for more than one DhaL-like subunit or domain that according to the phylogenetic tree (Fig. 1D) arose by gene duplication from a common ancestor. The association of these paralogs with transcription regulators indicates that they act as coactivators or corepressors of transcription. It thus appears that the use of ADP as cofactor has been invented only once in evolution and is now being used for both catalysis and signaling. A double function as enzyme subunit and autoactivator of transcription has been described for DhaL of E. coli (20).