The role of the catalytic base in the protein tyrosine kinase Csk.

A potential distinguishing feature between protein tyrosine kinases and homologous serine/threonine kinases is the function of the catalytic base in these enzymes. In this study, we show that a peptide containing the unnatural amino acid trifluorotyrosine shows remarkably similar efficiency as a substrate of the tyrosine kinase Csk (C-terminal Src kinase) compared with the corresponding tyrosine-containing peptide despite a 4-unit change in the phenolic pKa. These results argue against the importance of early tyrosine deprotonation by a catalytic base in Csk. To further explore the role of the proposed catalytic base, the Csk mutant protein D314E was produced. This mutant displayed a significant reduction in km (approximately 104) but relatively little effect on substrate Km values compared with wild-type Csk. Examination of the thio effect (km-ATP/km-adenosine 5′-O-(thiotriphosphate)) for D314E Csk led to the suggestion that a role of aspartate 314 may be to enhance the reactivity of the γ-phosphate of ATP toward electrophilic attack. These results may have significant impact on protein tyrosine kinase inhibitor design.

Protein tyrosine kinases play central roles in cell proliferation, cell differentiation, and signaling processes (1). While their enzymatic activity, transfer of a phosphoryl group from ATP to tyrosines in proteins, was first described over 15 years ago, details of tyrosine kinase catalytic mechanism have largely remained elusive. In contrast, their homologous family member protein serine/threonine kinases have been more intensively studied and are better characterized enzymologically (2)(3)(4)(5)(6)(7)(8). In order to rationally develop specific inhibitors for these enzymes (9) as well as to delineate the biochemical consequences of their in vivo regulation, a more complete understanding of tyrosine kinase catalysis is needed.
Recently, we carried out a preliminary mechanistic evaluation of the protein tyrosine kinase Csk (C-terminal Src kinase) from humans (10), overproduced in and purified from Esche-richia coli (11). It was shown that, like the better studied serine/threonine kinase protein kinase A, the Csk-catalyzed kinase reaction obeys a ternary complex mechanism, most likely with rapid and random binding of the substrates ATP and 4:1 poly(Glu,Tyr). This argued against a covalent phosphoenzyme intermediate. Moreover, using ATP␥S 1 and microviscosity effects, it was shown that in the Csk ATP reaction, product ADP diffusional release is partially rate-determining for the overall reaction (10). The lack of a large deuterium solvent kinetic isotope effect (actually a slightly inverse, k cat H/ k cat D ϭ 0.74, was seen) in the ATP␥S reaction (where the chemical step appeared fully rate-determining) suggested that tyrosine hydroxyl deprotonation is asymmetric in the reaction transition state. Because a tetrafluorotyrosine-containing peptide was previously reported to be a tight binding inhibitor (and not detectably a substrate) for insulin receptor tyrosine kinase (12), a phenoxide anion was reasoned to be transition state like (10). The phenoxide anion was proposed to be the species attacking the ␥-phosphate of ATP (see Fig. 1). In principle, the role of a catalytic base may be a mechanistically distinguishing feature between serine/threonine and tyrosine kinases because the pK a values of tyrosine hydroxyls (approximately 10) are significantly lower than those of serine or threonine hydroxyls (approximately 13-15) (2-5, 7, 8).

MATERIALS AND METHODS
Synthesis of Trifluorotyrosine (F 3 Y)-Reaction between trifluorophenol, pyruvate, and NH 4 OH catalyzed by tyrosine phenol-lyase was carried out analogous to previously described methods (13,14). Pure trifluorotyrosine (1.4 g/liter) was obtained and gave satisfactory spectroscopic and analytical data.
Peptide Synthesis-Peptides were made on a Biosearch 9600 automated peptide synthesizer. They were purified by high pressure liquid chromatography using C-18 reverse-phase chromatography (with a H 2 O/acetonitrile/trifluoroacteic acid solvent system) and were characterized by mass spectrometry.
Inhibition of Csk Phosphorylation of Poly(Glu,Tyr) by Peptides ED-NEYTA and EDNE(F 3 Y)TA-Enzyme reactions were carried out as previously described (10). Kinase Assays of Csk with EDNE(F 3 Y)TA and EDNE(Y)TA as Substrates-Assays were carried out with 0.5 mM ATP, 3 mM MnCl 2 , PEP (1 mM), DTT (10 mM), NADH (0.19 mM), pyruvate kinase-lactate dehydrogenase (20 g, 3:1), Csk (800 nM), Tris -HCl (60 mM, pH 7.4), and variable peptide concentrations (for EDNEYTA, 1-30 mM; for EDNE(F 3 Y)TA, 1-16 mM) in 0.4 ml at 30°C and quantitatively monitored by UV absorbance change at 340 nm (15). Calibration experiments with poly(Glu,Tyr) showed that the radioactive assay gave results that were quantitatively similar (within 30%) compared with this coupled assay, with no evidence of "uncou-* This work was supported in part by a grant from Hoffmann-La Roche (to C. T. W.) and by Grant GM42588 from the National Institutes of Health (to R. S. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Csk-mediated Phosphorylation of EDNE(F 3 Y)TA Monitored by 19 F NMR-Reactions
§ Supported by a postdoctoral fellowship from the Howard Hughes Medical Institute.
pling" (ADP formation in the absence of phosphorylated poly(Glu,Tyr)). Kinetic constants were calculated using the rectangular hyperbola curve fit in Kinetasyst and displayed in the text Ϯ S.E.
D314E Csk Construction-The D314E Csk mutant was generated by PCR mutagenesis using standard molecular biological procedures and confirmed by DNA sequencing. Expression and purification to near homogeneity of the D314E Csk were carried out analogous to previously described methods (11). Enzyme concentrations were measured by Bradford assay, which for wild-type Csk has been standardized to the method of von Hippel (16).
D314E Csk Kinase Assays-These assays were performed as previously described (10) for wild-type Csk with minor modifications. Enzyme activity proved to be linear versus time up to 45 min and linear versus enzyme concentration up to 10.5 M, and autophosphorylation was insignificant.

RESULTS AND DISCUSSION
To further explore the role of a proposed active site catalytic base, we decided to pursue studies on Csk using an unnatural fluorotyrosine-substituted peptide. The goal was to determine the inhibitory properties toward Csk of an isosteric tyrosine analog with a reduced pK a . Using tyrosine phenol-lyase (13,14), trifluorotyrosine was generated (see Fig. 2A) and after conversion to the Fmoc derivative (17) incorporated into the Src autophosphorylation peptide EDNE(F 3 Y)TA by solid phase peptide synthesis. EDNEYTA has previously been reported to be a Csk substrate (18), and we confirmed that here (k cat ϭ 14 Ϯ 0.4 min Ϫ1 , K m ϭ 6.1 Ϯ 0.4 mM, see Fig. 2D). The enzymatically synthesized trifluorotyrosine was presumed enantiomerically pure (with L-configuration) and was measured by spectrophotometric titration to have a phenolic pK a of 6.2 Ϯ 0.3. The pK a of the trifluorotyrosine in EDNE(F 3 Y)TA was found to be 6.5 Ϯ 0.2. These pK a values are approximately 4 units below that of standard tyrosine. Although they are marginally higher than the pK a measured for tetrafluorotyrosine (pK a ϭ 5.4) (19) used for study with the insulin receptor tyrosine kinase, the pK a of 6.5 implies that approximately 90% of EDNE(F 3 Y)TA exists as the phenoxide anion at pH 7.4 (the enzyme assay pH).
Initially, EDNE(F 3 Y)TA was tested as a Csk competitive inhibitor with poly(Glu,Tyr) as the tyrosine-containing substrate. As shown, EDNE(F 3 Y)TA exhibited similar inhibitory properties to EDNEYTA (see Fig. 2B). This similarity was surprising given the potency reported for the tetrafluorotyrosine-containing peptide toward insulin receptor kinase (K i ϭ 4 M; K m for the tyrosine peptide ϭ 400 M) (12). A careful examination of the reported data in this paper reveals that the D-and L-tetrafluorotyrosine analog-containing peptides show very similar inhibitory properties (12) when plotted against varying lysozyme substrate concentrations. This lack of stereospecificity, as well as the lack of reported data comparing the corresponding tyrosine and tetrafluorotyrosine-containing peptides as inhibitors of lysozyme phosphorylation (12), led us to question the K i determinations and mechanistic interpretations.
The peptide EDNE(F 3 Y)TA was next investigated as a Csk substrate. As mentioned, the tetrafluorotyrosine-containing peptide was reported not to be a substrate for insulin receptor tyrosine kinase (12). Much to our surprise, EDNE(F 3 Y)TA was found to be a substrate for Csk by monitoring ADP formation using a spectrophotometric coupled assay (15). A large scale reaction with EDNE(F 3 Y)TA was followed by 19 F NMR. As shown in Fig. 2C, there is nearly complete conversion of unphosphorylated EDNE(F 3 Y)TA to phosphorylated EDNE(F 3 Y)TA catalyzed by Csk as evidenced by the downfield shifting of the three fluorine signals. Note that the signals at Ϫ141 and Ϫ160 ppm of EDNE(F 3 Y)TA represent the fluorines closest to the hydroxyl because their chemical shifts are moved most downfield by phosphorylation.

Catalytic Base and Csk 22106
similar to the tyrosine-containing peptide (Fig. 2D). This result is particularly striking in light of the reported results with the tetrafluorotyrosine-containing gastrin peptide and the insulin receptor tyrosine kinase (12). Although the lack of the tetrafluorotyrosine-containing gastrin peptide to act as a tyrosine kinase substrate may have been due to the fact that a different (insulin receptor) tyrosine kinase was being studied, several other possible explanations exist. One is the instability of the phosphorylated product in their assay (which involved 5 washes with the strong acid phosphoric acid). A second is that the equilibrium constant for the reaction may lie far to the left (favoring ATP). Note that the free energy of hydrolysis of the adenylyl-O-tyrosine-phosphodiester bond was measured to be similar (within 1 kcal/mol) to the free energy of hydrolysis of the pyrophosphate bond of ATP (20). In our case with EDNE(F 3 Y)TA, the reaction is driven toward phosphorylation by coupling to pyruvate kinase and lactate dehydrogenase.
The implications of EDNE(F 3 Y)TA serving as an efficient Csk substrate (and not a potent inhibitor) significantly altered our perception of the catalytic mechanism of Csk. It is assumed that the chemical step in EDNE(Y/F 3 Y)TA phosphorylation reactions is at least partially rate-determining since their k cat values are 2-3-fold lower than for Csk-mediated poly(Glu,Tyr) phosphorylation (10). If stoichiometric deprotonation of the tyrosine phenol prior to ATP attack were occurring, one would expect that the tyrosine-containing substrate would be significantly more reactive as a nucleophile compared with the trifluorotyrosine-containing substrate. In this case, the nucleophilicity of the trifluorophenoxide anion versus the unsubstituted phenoxide anion could differ by as much as 10 4 . Alternatively, enzymatic phosphoryl transfer could follow a dissociative mechanism, whereby ADP-phosphoryl covalent bond breaking is slow (forming a metaphosphate) and precedes nucleophilic attack. A dissociative mechanism, while not inconceivable, is inconsistent with the thio effect observed with ATP␥S and the wild-type Csk reaction (see below). It would also be contrary to the generally accepted view about how kinase enzymes work (21). Furthermore, if Csk obeyed a dissociative mechanism, early tyrosine deprotonation by a catalytic base would serve no useful role in facilitating phosphoryl transfer.
The most plausible interpretation of the trifluorotyrosine/Csk experiments is that deprotonation of the natural tyrosine substrate occurs principally after Tyr O-P bond formation in the enzymatic reaction. In this way, the nucleophilicity of the trifluorophenoxide anion could be expected to be comparable with the largely protonated phenol. 2 As the above studies led us to question the importance of catalytic base function in the Csk kinase reaction, the site-directed Csk mutant D314E was generated. Aspartate 314 is one of 4 invariant protein kinase amino acid residues (22). In protein kinase A, the corresponding aspartate (Asp-166) has been identified near the substrate serine hydroxyl and predicted to be the catalytic base for this enzyme from x-ray crystallographic analysis (23), although additional roles for this residue have been considered (3). Note that although Csk DNA constructs for D314N and D314A were also generated, protein expression was not detectable with these mutants.
The Csk mutant D314E had greatly reduced enzyme activity vs. the wild-type enzyme. K m and k cat values were measured for D314E Csk. The K m values for ATP (35 Ϯ 8 M) and poly(Glu,Tyr) (88 Ϯ 12 g/ml) are rather similar to those of wild-type Csk (ATP K m ϭ 12 Ϯ 1 M; poly(Glu,Tyr) K m ϭ 48 Ϯ 2 g/ml), but the k cat is reduced by a factor of approximately 10 4 (k cat for wild-type Csk Ϸ40 min Ϫ1 ; k cat for D314E Csk Ϸ0.005 min Ϫ1 ) (10). That this enzyme activity in D314E Csk was not due to trace wild-type contamination was most convincingly demonstrated by examining the sucrose microviscosity effect. Behaving similar to the microviscosity effect on the ATP␥S, wild-type Csk reaction (slope ϭ Ϫ0.22 Ϯ 0.03) (10), there is a modest fall in k cat -control/k cat -viscogen with increasing levels of sucrose for the ATP, D314E Csk reaction (slope ϭ Ϫ0.28 Ϯ 0.06) (see Fig. 3). This is clearly different from the sucrose microviscosity effect on the ATP Csk reaction (slope ϭ ϩ0.42 Ϯ 0.04) (see Fig. 3). It further suggests that as expected the chemical step is fully ratedetermining in both the Csk mutant D314E, ATP reaction, and the ATP␥S, wild-type Csk reaction.
The substantial rate loss associated with D314E Csk compared with wild-type Csk was very surprising given the trifluorotyrosine results described above. That large structural perturbations in D314E Csk could account for the larger than anticipated k cat effect cannot be ruled out. The rather minor K m changes for both ATP and poly(Glu,Tyr) with D314E Csk as well as a similar sucrose microviscosity effect to the wild-type Csk ATP␥S reaction weaken such an argument, however. Further support for the structural integrity of D314E Csk comes from data using phospho-Ultrogel affinity chromatography (11) and native polyacrylamide gel electrophoresis, where behavior proved to be identical to wild-type Csk (data not shown). 2 Another way to envisage this interpretation is that the high relative concentration of trifluorophenoxide anion (90%) offsets the decreased nucleophilicity of this anion compared with the phenoxide anion, which would only be present in 0.2% concentration at pH 7.4 (assuming a pK a of 10 for tyrosine). In this variant, which is experimentally indistinguishable from a late deprotonation model, both of the attacking species are anions. Both interpretation variants are still fundamentally different from the proposal of Fig. 1, where there is enzyme-assisted, early deprotonation.

Catalytic Base and Csk
As the trifluorotyrosine studies argued against early tyrosine deprotonation in the Csk catalytic mechanism, it was not surprising that, like the apparent k cat of the wild type Csk reaction, k cat is unaffected by increasing the pH of the D314E Csk kinase reaction from 7.4 to 8.8. Furthermore, the deuterium solvent kinetic isotope effect on k cat measured with D314E Csk (at pH 7.4) is actually modestly inverse (k cat H/k cat D ϭ 0.55 Ϯ 0.08), arguing against a proton transfer step becoming dominant in the transition state of the mutant. A significant, standard kinetic isotope effect (k cat H/k cat D Ͼ 2) might have been expected with disruption of catalytic base function.
At this point, we were unable to fully explain the rather dramatic 10 4 -fold rate reduction in the D314E Csk mutant. Clearly, aspartate 314 could be important in hydrogen bonding and orienting the tyrosine hydroxyl toward the ␥-phosphate. Results of the D314E Csk mutant with ATP␥S suggested another important function, however. The thio effect (k cat -ATP/k cat -ATP␥S) (24) in the wild-type reaction is 16.9 Ϯ 0.6 (corrected ϭ 34 assuming that the rate of ADP release is equal to the chemical step) (10). In contrast, the thio effect in the mutant reaction is 4.8 Ϯ 0.6. Recently, it has been pointed out that the thio effect for non-enzymatic (non-acid catalyzed) phosphoryl transfers is different for phosphotriesters versus phosphodiesters (24). For phosphotriester reactions, the usual thio effect is in the range of 10-200 (25), whereas for phosphodiester reactions, the range is 4-11 (26). While enzymatic thio effects can be more complex to interpret because of subtle changes in substrate-active site interactions, it is tempting to rationalize the difference in thio effects for wild-type and mutant Csk as being related to a change in transition state from triester-like to diester-like.
Triesters typically react much faster (10 3 -10 5 -fold) under nonacidic conditions with nucleophiles than diesters, probably because of the enhanced electrophilicity of the phosphorus in the former (27,28). Thus one might anticipate a large rate reduction to be associated with rendering the ␥-phosphate of ATP more diester-like in the enzyme active site. For example, one hydrogen bond donor to one of the oxyanions on the ␥-phosphorus could be disrupted in Csk D314E. Extending this proposal, Asp-314 could actually serve as this hydrogen bond donor in the wild-type enzyme, but because of its altered side, chain, Glu-314 is unable to fulfill this role. An alternative possibility is that the wild-type aspartate 314 enforces a structure that allows another active site functional group to hydrogen bond to the ␥-phosphate. In this scenario, D314E mutation indirectly destabilizes a hydrogen bond to the ␥-phosphate by altering the active site organization.
Although a three-dimensional structure of a catalytically active tyrosine kinase with bound substrates or inhibitors remains to be reported (29), it seems reasonable to suggest that the "catalytic base" in Csk and other tyrosine kinases may have a dual role. The "catalytic base" in Csk may still be important in hydrogen bonding the tyrosine phenol and orienting it in the active site (3). Concomitantly it could serve to render the ␥-phosphate more reactive by hydrogen bonding either directly (X ϭ Hϩ) or indirectly (X ϭ metal, amino acid side chain) (see Fig. 4). ATP activation may be particularly important in tyrosine kinase reactions where the nucleophilicity of the attacking oxygen may be reduced by aryl conjugation compared with alkyl alcohols of serine/threonine kinases.