The ATP-binding site in the 2-kinase domain of liver 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase. Study of the role of Lys-54 and Thr-55 by site-directed mutagenesis.

All known 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase isozymes contain a sequence (GX4GK(S/T)) in the 6-phosphofructo-2-kinase domain corresponding to the so-called nucleotide binding fold signature or Walker A motif. Mutagenesis and crystal structure data from several nucleotide binding proteins, which also contain this sequence, showed the importance of the lysine and serine/threonine residues in nucleotide binding. We have studied the role of Lys-54 and Thr-55 in MgATP binding in the 6-phosphofructo-2-kinase domain of rat liver 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase by site-directed mutagenesis. Lys-54 was mutated to methionine, whereas Thr-55 was mutated to valine, serine, and cysteine. Three mutants, Lys-54 to Met and Thr-55 to Cys or Val, displayed more than a 5000-fold decrease in 6-phosphofructo-2-kinase activity compared with the wild type. The mutations had no effect on fructose-2,6-bisphosphatase activity and did not affect the activation of fructose-2,6-bisphosphatase after phosphorylation by cyclic 3′,5′-AMP-dependent protein kinase. Binding experiments with ATP, ADP, and their analogs (3′-N-methylanthraniloyl derivatives) showed that these two residues do not play the same role. Lys-54 is involved in ATP binding, whereas Thr-55 is important for catalysis.

All PFK-2/FBPase-2 isozymes are homodimers, and each subunit contains two separate catalytic sites. The FBPase-2 reaction is catalyzed in the COOH-terminal half of the subunit in a classical ping-pong reaction mechanism (6). The FBPase-2 domain is homologous with the so-called histidine phosphatases (7), and residues involved in catalysis and substrate binding in the FBPase-2 domain have been studied by site-directed mutagenesis (8).
The PFK-2 reaction is catalyzed in the NH 2 -terminal half of the subunit and involves ternary complex formation (9,10). Site-directed mutagenesis studies have identified some amino acid residues that are important in substrate binding. For example, Arg-104 (11), , and Arg-225 (11) in the L isozyme have been proposed to bind fructose 6-phosphate (Fru-6-P), whereas Arg-230 and Arg-238 might be involved in ATP binding (12). The PFK-2 domain contains a sequence (residues Arg-121-Asp-130 in the L isozyme), which shares sequence similarity with the so-called B motif defined by Walker et al. (13) as being important for ATP binding. This motif contains a conserved aspartate residue, which has been proposed to bind ATP via Mg 2ϩ . Indeed mutation of Asp-130 to Ala in PFK-2 greatly decreased the K cat , decreased the sensitivity toward Mg 2ϩ , and increased the K m for MgATP (14).
The PFK-2 domain also contains a sequence (residues Gly-48 -Thr-55 in the L isozyme) similar to the A motif defined by Walker et al. (13) for nucleotide-binding proteins. This motif is conserved in the PFK-2 domain of all known PFK-2/FBPase-2 sequences. Interestingly, a second A motif is present in the FBPase-2 domain (residues Gly-274 -Lys-280 in the L isozyme) and could play a role in the stimulation of FBPase-2 activity by GTP (15). The A motif consensus sequence (GX 4 GK(S/T)) was first described in adenylate kinase (16). It has the following structure: (␤-strand)-GX 4 GK(S/T)-(␣-helix) with the glycinecontaining loop forming a tight turn between the ␤-strand and the ␣-helix. Mutagenesis of Gly-48 in the L isozyme decreased PFK-2 activity at least 100-fold without affecting FBPase-2 (12). Whether this loss of PFK-2 activity resulted from a change in structure or the affinity for MgATP is not known. The three-dimensional structures of adenylate kinases from various organisms revealed the importance of the "invariant" Lys in ATP binding (17). Site-directed mutagenesis of Lys-21 to Met in adenylate kinase decreased the K cat 1000-fold (18). The Ser/Thr residue adjacent to the conserved Lys has also been implicated in nucleotide binding. From the crystal structures of p21 ras (19), EF-Tu (20), and RecA (21), Mg 2ϩ has been proposed to bind the ␤ and ␥ phosphates of ATP and is itself bound by the OH group of the Ser/Thr residue in the A motif.
We have studied the effect of site-directed mutagenesis of Lys-54 and Thr-55 in the rat recombinant L (rL) isozyme on the kinetic properties of PFK-2 and FBPase-2 and particularly on nucleotide binding in the 2-kinase domain. Lys-54 was mutated to Met (K54M mutant), because the sizes of the side chains are similar and because the same mutation in adenylate kinase drastically decreased K cat (18). Thr-55 was first mutated to Val (T55V mutant), because a similar mutation (Thr to Ile) in EF-Tu abolished nucleotide binding (22). Two other mutations were also made. Thr-55 was mutated to Ser (T55S mutant), because of the physicochemical similarities of Ser with Thr and because mutation of this Thr into Ser in F 1 -ATPase (23) changed its kinetic properties. Thr-55 was also mutated to Cys (T55C mutant), because its pK a and hydrogen bonding capacities should be different from those of Thr.
Mass Spectrometry-Purified proteins (20 -50 g) were separated by reverse-phase narrowbore high pressure liquid chromatography on a Vydac C18 column (2.1 ϫ 250 mm) with an acetonitrile gradient in 0.1% (v/v) trifluoroacetic acid (solvent A). Elution was performed with the following gradient program: 5-100% B in 100 min (solvent B is 70% acetonitrile, 0.1% (v/v) trifluoroacetic acid) at a flow rate of 80 l/min generated by a model 140B Applied Biosystems solvent delivery system. The samples were concentrated in a Speed Vac to 5-10 l. Molecular masses were then measured by matrix-assisted laser desorption ionization time-of-flight mass spectrometry using a Lasermat 2000 (Finnigan Mat). Protein samples (2 l) were mixed with 2 l of 3,5-dimethoxy-4-hydroxycinnamic acid matrix. The mixture (1 l) was spotted onto the target. Molecular masses were determined in triplicate for each preparation with bovine serum albumin for calibration.
Circular Dichroism-Circular dichroism spectra were recorded in a Jobin-Yvon CD6 dichrograph and analyzed using the Jobin-Yvon software (version 1. 1,1989). The proteins (50 l of a 0.5-1 mg/ml solution) were introduced with a long-tipped micropipette into the quartz cell (path length, 0.2 mm), which had been cleaned with acetonitrile and dried under a stream of N 2 . Spectra were recorded (0.2 nm/s) between 195 and 250 nm with a band width of 2 nm. For each preparation, five to six scans were recorded and averaged using the computer software. Mean residue ellipticities were calculated using the computer software taking the protein concentration, measured by the modified Lowry method described below and the molecular mass of the holoenzyme (rat recombinant L isozyme ϩ 6 His residues) of 111,171 Da.
Measurement of Ligand Binding by Fluorescence-The binding of ATP, ADP, and their Mant derivatives were measured by fluorescence. Spectra were recorded at 25°C with continuous magnetic stirring in a SLM 8000C spectrofluorimeter (SLM Aminco) with spectral band widths of 2 and 4 nm for excitation and emission, respectively. The nucleotides were added stepwise in small volumes (1-2 l) to 7-15 g of recombinant (H) 6 -tagged wild type or mutant proteins in 1 ml of a buffer containing 50 mM Hepes, pH 7.5, 100 mM KCl, 0.5 mM EDTA, 1 mM potassium phosphate, 1 mM dithiothreitol in the presence or the absence of 5 mM MgCl 2 . Upon excitation at 295 nm, the intrinsic fluorescence spectra were scanned from 310 to 380 nm and integrated. Corrections were made for buffer fluorescence and for the inner filter effect of the nucleotides from the quenching of N-acetyltryptophanamide (4 M) under the same conditions (30).
The high affinity binding of Mant-ATP and Mant-ADP was also measured by the increase in extrinsic fluorescence of the analog. Excitation was performed at 350 nm, and emission was scanned from 400 to 500 nm. Curve fitting of concentration-dependent changes in fluorescence was accomplished using Grafit (Erithacus software) as described (30,31).
Other Methods-Protein was measured after trichloroacetic acid precipitation in the presence of deoxycholate (32) by the Lowry procedure (33) using bovine serum albumin as a standard. PFK-2 and FBPase-2 activities were measured (29) under the conditions described in the legends to the figures and tables. SDS-polyacrylamide gel electrophoresis analysis in 10% (w/v) acrylamide was as described (34). Kinetic constants were calculated by fitting the data to a hyperbola by nonlinear least squares regression using a computer program (Ultrafit, Biosoft, Cambridge, UK).

RESULTS AND DISCUSSION
Purification-The purification of the recombinant K54M and T55V mutants by the classical procedure was not possible because neither was bound to a blue Sepharose affinity column (see "Materials and Methods"). Therefore, we had to resort to an alternative strategy. A carboxyl-terminal polyhistidinehexapeptide tail was engineered in the wild type and mutant enzymes (see "Materials and Methods") for purification by affinity chromatography on Ni 2ϩ -nitrilotriacetic acid-agarose after polyethylene glycol fractionation and ion-exchange chromatography. The chromatographic behavior of the wild type-(H) 6 and the four mutants was the same. Each eluted from Q-Sepharose as a single peak at about 250 mM KCl and from the Ni 2ϩ -nitrilotriacetic acid-agarose at about 160 mM imidazole. For the wild type-(H) 6 and T55S mutant, PFK-2 activity was measured to monitor the enzyme through the purification, whereas for the K54M, T55V, and T55C mutants, FBPase-2 activity was measured, because their PFK-2 activity was undetectable (see below).
SDS-polyacrylamide gel electrophoresis analysis of the purified preparations of the unmodified wild type and the polyhistidine-tagged proteins showed single bands with M r of 55,500 and 56,600, respectively (not shown). The molecular masses of the monomers, measured by mass spectrometry, were 54,658 Ϯ 95 Da for the wild type and 55,486 Ϯ 20 Da for the recombinant-(H) 6 enzymes. The difference is consistent with the introduction of six His residues.
Kinetic Properties of the Unmodified and Polyhistidinetagged Wild Type rL PFK-2/FBPase-2-Insertion of a polyhistidine-tail to the rL PFK-2/FBPase-2 had no effect on the kinetic properties of PFK-2 and FBPase-2 (Tables I and II). Furthermore, phosphorylation by the cyclic 3Ј,5Ј-AMP-dependent protein kinase of the rL PFK-2/FBPase-2-(H) 6 inactivated PFK-2 and activated FBPase-2 to the same extent as the unmodified enzyme (Table III). The pH curves of PFK-2 activity and FBPase-2 activity were similar for the two enzymes (not shown). This similarity in kinetic properties justifies the use of this new procedure for the study of PFK-2/FBPase-2 isozymes by site-directed mutagenesis.
Kinetic Properties of the Mutants-The (H) 6 recombinant mutated enzymes K54M, T55V, T55C, or T55S had normal kinetic parameters of FBPase-2 (Table II). Furthermore, phosphorylation with the cyclic 3Ј,5Ј-AMP-dependent protein ki-nase doubled FBPase-2 activity as expected (Table III). Therefore, the phosphorylation site was still able to interact with the FBPase-2 domain in all the four mutants. This suggests that the conformation of the phosphorylation domain, which contains Ser-32 and which is N-adjacent to the GX 4 GKT motif, is unaffected in all the mutants.
PFK-2 activity was undetectable in the K54M, T55V, and T55C preparations (Table I). We calculated that PFK-2 activity was less than 0.01 milliunits/mg of protein when measured at pH 7.1 with 5 mM MgATP and 2 mM Fru-6-P, which are optimal conditions for the wild type-(H) 6 . This represents a more than 5000-fold decrease in PFK-2 activity compared with the wild type-(H) 6 measured under identical conditions. By contrast with the three inactive mutants, the T55S mutant showed a slight increase (1.7-fold) in PFK-2 V max and a higher K m for MgATP (4-fold) without change in the K m for Fru-6-P. Moreover, the T55S mutant was 10-fold less sensitive to the inhibition by MgADP when IC 50 was measured with MgATP concentrations equivalent to the K m (500 M MgADP for the mutant versus 50 M for the wild type-(H) 6 ). This would indicate that in this mutant, the increase in V max results from an increased rate of ADP release. This is consistent with a compulsory order ternary complex mechanism (9).
We also tested whether Mant-ATP was a substrate for PFK-2. Although the V max was 10 times lower with Mant-ATP than with authentic ATP, the low K m of 260 nM was comparable with the K d measured by fluorescence quenching (Fig. 1 and  see below). This indicates that for wild type-(H) 6 PFK-2, Mant-ATP binds to the active site of the 2-kinase domain and is a substrate in the PFK-2 reaction.
Substrate Binding-The fact that PFK-2 activity was undetectable in the K54M, T55V, and T55C preparations meant that kinetic measurements could not be used to investigate the effects of the mutations on the K m for the MgATP. Therefore, substrate binding was measured by quenching of the enzyme intrinsic fluorescence, taking advantage of the fact that the rL isozyme contains four tryptophane residues, one being located in the NH 2 -terminal domain (Trp-16), one in the PFK-2 domain (Trp-67), and two in the FBPase-2 domain (Trp-301 and Trp-322). Upon excitation at 295 nm, the fluorescence emission spectrum of the wild type-(H) 6 enzyme was maximal at 325 nm as illustrated in Fig. 2A. All the mutants displayed a similar spectrum (not shown).
The addition of micromolar concentrations of ATP or ADP, with or without magnesium, to the wild type-(H) 6 , T55S, and T55C mutants caused a maximum quenching of fluorescence of unmodified and polyhistidine-tagged wild type and mutant preparations PFK-2 activity was measured in buffer containing 50 mM Hepes, 100 mM KCl, 20 mM KF, 1 mM dithiothreitol, 5 mM potassium phosphate, 1 mg/ml bovine serum albumin, 5 mM MgATP, and 2 mM Fru-6-P at pH 7.1 (29). For the Fru-6-P and MgATP saturation curves, the concentrations of MgATP and Fru-6-P were 5 and 2 mM, respectively, and the concentration of the other substrate was varied up to 5-10 ϫ K m . The results are the means Ϯ S.E. for the number of determinations shown in parentheses; otherwise individual values are given.   6 14.4 Ϯ 1.6 (3) 0.41 Ϯ 0.12 (3) 4.9 T55V-(H) 6 13.0 Ϯ 1.3 (3) 0.35 Ϯ 0.09 (3) 5.1 T55C-(H) 6 9.2 Ϯ 1.6 (4) 0.21 Ϯ 0.08 (4) 3.3 T55S-(H) 6 10.2 Ϯ 0.8 (4) 0.24 Ϯ 0.07 (4) 1.3 about 25%. Binding constants were calculated and are given in Table IV. MgATP and MgADP binding were similar for the wild type-(H) 6 , T55S, and T55C mutants, with K d values ranging from 38 to 76 M. No quenching was observed with the T55V and K54M mutants. We also used Mant derivatives of ATP and ADP to monitor nucleotide binding. In these analogs, the fluorescent moiety is attached to the ribose, leaving the purine ring and phosphate groups free for binding. They have been successfully used to measure nucleotide binding to a number of proteins, which like the 2-kinase domain contain the GX 4 GK(S/T) Walker A motif (13). These include myosin (35), replicative helicase (36), and mitochondrial F 1 -ATPase (30,31).
MgMant-ATP and MgMant-ADP markedly quenched the fluorescence of the wild type-(H) 6 protein up to nearly 50% showing two components (Fig. 3). The same results were obtained with the T55S and T55C mutants (not shown). Binding constants for MgMant-ATP were calculated to be in the 10 Ϫ8 M range for a high affinity site and in the 10 Ϫ5 M range for a much lower affinity site. The K d values (high affinity site) are listed in Table IV. The K d values for the low affinity site were not affected by the mutations. Mutation of Lys-54 to Met and Thr-55 to Val abolished MgMant-ATP and MgMant-ADP binding, whereas mutation of Thr-55 to Ser or Cys did not affect binding (Table IV and Fig. 3).
Similar results were obtained when the quenching of fluorescence by the nucleotides was measured without Mg 2ϩ (Table  IV). However, Mg 2ϩ decreased the K d for Mant-ATP and ATP 5-fold, at least in the wild type-(H) 6 , which could correspond to a more favorable interaction due to partial charge compensation by Mg 2ϩ . In the T55C and T55S mutants, high affinity binding was observed whether the Mg 2ϩ complex was present or not. This could indicate that the partial charge compensation is not present because of the different nature of the side chain in serine and cysteine. The addition of Mant-ATP lowered the intrinsic fluorescence on excitation at 295 nm, whereas a new peak appeared at 432 nm ( Fig. 2A). The fact that the K54M and T55V mutations abolished the high affinity binding component for Mant-ATP and Mant-ADP suggests that it is associated with PFK-2. The low affinity component probably reflects nucleotide binding to the FBPase-2 domain.
The high affinity binding of Mant-ATP and Mant-ADP to the wild type-(H) 6 enzyme was also monitored by the enhancement of extrinsic fluorescence. Upon excitation at 350 nm, Mant-ATP in solution exhibited a characteristic fluorescence emission with a maximum at 445 nm (Fig. 2B). The addition of the wild type-(H) 6 enzyme markedly increased the probe fluorescence and blue-shifted the maximum to 438 nm. The differential spectrum corresponding to bound Mant-ATP was maximal at 432 nm, indicating that a 13 nm blue shift was produced on binding (Fig. 2B). The saturation curve for Mant-ATP was determined by enhancement of fluorescence at 432 nm (not

and FBPase-2 in the phosphorylated recombinant rL PFK-2/FBPase-2 with and without the polyhistidine tag
To obtain phosphorylated forms (phospho-), aliquots (50 g) of recombinant PFK-2/FBPase-2 preparations were incubated with 20 microunits of the cyclic 3Ј,5Ј-AMP-dependent protein kinase for 30 min at 30°C as described (39). Measurements of kinetic parameters were then performed as described in Tables I and II  shown); the calculated apparent K d was 60 nM. The same spectral change was produced by Mant-ADP with a comparable maximal increase at 432 nm, whereas the affinity was 2-fold greater (K d ϭ 30 nM).
The K d for Mant-ATP binding to the high affinity site (50 nM) was about 3 orders of magnitude lower than the K m for ATP (180 M) of PFK-2 activity in the wild type-(H) 6 enzyme. A large increase in affinity for Mant analogs compared with unmodified nucleotides has also been observed for a number of proteins (30,31,(35)(36)(37).
In conclusion, the results of fluorescence quenching suggest that Lys-54 is involved in ATP binding in the 2-kinase domain. The fact that the high affinity site for Mant-ADP was also abolished in the K54M mutant suggests that Lys-54 binds at least the ␤-phosphate of ATP. Binding results obtained with the T55V mutant should be regarded with caution, because the structure of the enzyme was locally disrupted (see below). From binding experiments made with the T55S and T55C mutants, one can conclude that these mutations did not affect nucleotide binding, because both mutants have the same K d values for the nucleotides.
Structure-Because the Lys and Thr residues of the Walker A motif form the start of an ␣-helix after the Gly-containing loop (17,19,20), mutation of Lys-54 and Thr-55 could disrupt the structure. We therefore studied the secondary structures of the K54M, T55V, T55S, and T55C mutants by circular dichroism. The spectrum of the wild type-(H) 6 protein was similar to that of the K54M, T55S, and T55C mutants, whereas that of the T55V mutant was notably different (not shown). The spectra were analyzed and the ratio (R) of the mean residue ellipticities at 208 and 222 nm were calculated (Table V). This ratio has often been taken as an index of helical structures and as a comparative tool for estimating the ␣-helical content (38). Mutations of Lys-54 to Met and of Thr-55 to Ser and Cys had no effect on the R values compared with the wild type-(H) 6 . By contrast, the T55V mutant had a different R value, representing a loss of ␣-helical content. Therefore, the abolition of the high affinity binding site for ATP in the T55V mutant could be due to a structural change in the ATP binding site. Nevertheless, mutation of Thr-55 into Ser or Cys did not affect the structure, and the binding affinity for nucleotides was likewise unaffected.
Conclusions-The site-directed mutagenesis studies reported in this paper show that Lys-54 and Thr-55 do not play the same role in the PFK-2 domain. The fact that nucleotide binding was abolished in the K54M mutant suggests that Lys-54 binds the ␤or ␥-phosphate of the nucleotide. Lys-54 in PFK-2 might play a similar role to Lys-21 in adenylate kinase, which wraps around the glycine-containing loop and hydrogen bonds to the ␤and ␥-phosphates of ATP. In adenylate kinase, it also participates in catalysis by stabilizing the penta-coordinate transition state during in-line transfer of the ␥-phosphate   6 enzyme and mutant preparations Circular dichroism spectra were recorded as described under "Materials and Methods." Mean residue ellipticities were measured at 208 and 222 nm, and the ratio (R) was calculated. The results are the means Ϯ S.E. for the number of determinations shown in parentheses; otherwise individual values are given. The number of different enzyme preparations is also given.

Enzymes
Number of preparations R (222/208) rL-(H) 6 4 0.92 Ϯ 0.02 (4) K54M-(H) 6 2 0.91 Ϯ 0.05 (4) T55V-(H) 6 3 1.70 Ϯ 0.25 (3) T55C-(H) 6 1 0.91-0.93 T55S-(H) 6 1 0.88-0.90 of ATP (17). By analogy, Lys-54 in PFK-2 would thus be involved in ATP binding and in catalysis, as indicated by the drastic change in substrate binding and V max . Thr-55 is more likely to be involved in catalysis because mutation of Thr-55 to Cys abolished PFK-2 activity without affecting the binding affinity of the enzyme for nucleotides. The role of Thr-55 in catalysis is also illustrated by the consequences of the mutation of Thr into Ser on the kinetic properties. This residue might participate in ADP release as suggested for the F 1 -ATPase, where the same mutation also increased the V max by modifying the rate of ADP dissociation (23), or it could play a role in catalysis by stabilizing the transition state via hydrogen bonding to the ␥-phosphate of ATP or by co-ordinating Mg 2ϩ .