Precision substrate targeting of protein kinases v-Abl and c-Src.

The active site substrate specificities of v-Abl and c-Src are compared and contrasted. Both enzymes catalyze the phosphorylation of a broad assortment of peptide-bound aliphatic and aromatic alcohols, such as achiral and simple straight chain residues. In addition, both protein kinases exhibit a “dual specificity” with respect to the ability to utilize D- and L-configurational isomers as substrates. However, c-Src and v-Abl are extremely inefficient as catalysts for certain structural arrangements, including secondary alcohols and primary alcohols containing large substituents in close proximity to the hydroxyl moiety. In addition to these similarities, these enzymes also display noteworthy differences in catalytic behavior. Whereas c-Src exhibits a modest preference for aromatic versus aliphatic alcohols, v-Abl does not. Most dramatic is the ability of c-Src to utilize short chain alcohols as substrates, an activity virtually absent from the catalytic repertoire of v-Abl. The implications of these observations are 2-fold. First, because both enzymes are able to accommodate a wide variety of structural variants within their respective active site regions, there exists a substantial degree of flexibility with respect to inhibitor design. Second, because these enzymes exhibit disparate active site specificities, it is possible that other tyrosine-specific protein kinases will display unique substrate specificities as well. Consequently, it may ultimately be possible to exploit these differences to generate inhibitors that precisely target specific protein kinases.

Protein phosphorylation plays a pivotal role in signaling pathways, which regulate such fundamental processes as cell growth, differentiation, and division (1,2). Consequently, protein kinase inhibitors that can be precisely directed to specific members of the kinase family could prove to be of decided utility in understanding and controlling these signaling pathways. Much of the effort devoted to the design and construction of agents that are targeted to individual protein kinases has focused on the identification of characteristic substrate sequence specificities. Because most protein kinases utilize short peptides as substrates (3,4), it may be possible to create inhibitors for specific protein kinases by employing peptides containing appropriate primary sequences. Unfortunately, this approach suffers from the limitation that many kinases exhibit broad overlapping sequence specificities with other family members. One way to overcome this obstacle is to employ additional recognition motifs in conjunction with the specific primary sequence to narrow the range of kinases that are impaired by inhibitory species.
We have developed a facile method to rapidly assess the active site substrate specificity of protein kinases (5)(6)(7)(8)(9)(10)(11). Although protein kinases will generally phosphorylate only serine, threonine, and/or tyrosine residues in intact protein substrates, we have found that these enzymes will phosphorylate a diverse array of alcohol-containing non-amino acid residues in synthetic peptides. This approach has allowed us to investigate the range of functionality that can be accommodated within and phosphorylated by the protein kinase active site, an issue particularly germane to the development of such inhibitory species as transition state analogs, affinity labels, and mechanism-based inhibitors. In addition, although many serine/threonine-specific protein kinases exhibit overlapping specificities for conventional peptides, we have recently shown that several of these enzymes will display remarkably different behavior toward peptides bearing exotic alcohol-containing residues (9). 1 In short, it may ultimately be possible to target specific members of the protein kinase family by coupling the active site substrate specificity of an individual family member with its characteristic sequence specificity.
We recently described the first active site substrate specificity analysis of a tyrosine-specific protein kinase (11). The specificity of c-Src, like that of several serine/threonine-specific protein kinases, is not limited to amino acid residues found in proteins. Indeed, c-Src phosphorylates a broad assortment of aliphatic and aromatic alcohol-containing residues that are attached to active site-directed peptides. Do members of the tyrosine-specific protein kinase family, like their serine/threonine-specific counterparts, exhibit dissimilar active site specificities? If the answer to this question is in the affirmative, it may be possible to exploit these differences in specificity to create inhibitory agents that distinguish between otherwise closely related enzymes. As a consequence, we have investigated the active site specificity of a second "tyrosine-specific" protein kinase.
The protein product of v-abl, like that of c-src, specifically phosphorylates tyrosine residues in protein substrates. The oncogenic form of abl was first discovered as the transforming gene of Abelson murine leukemia virus (12). In humans, the normal cellular homolog, c-abl, is involved in chronic myelogenous leukemias and some acute lymphocytic leukemias (reviewed in Ref. 13). In these leukemias, a chromosomal translocation takes place in which 5Ј sequences of the bcr gene become fused to abl, generating an oncogenic Bcr-Abl fusion protein (14). In all cases, there is a strong correlation between oncogenic transformation and tyrosine kinase activity. Like abl, src was originally identified as a retroviral oncogene; both transforming genes are altered versions of normal cellular proto-oncogenes (15). v-Abl and c-Src are members of the nonreceptor tyrosine kinase family and, as such, share many common features. Both contain SH2 and SH3 domains (16,17), and both Src and several forms of Abl undergo myristoylation (18,19), which promotes membrane association. These enzymes also share a strong sequence homology within their respective active site regions (20,21). Finally, both kinases exhibit overlapping substrate specificities in vitro, 2 a feature that can render the design of kinase-specific inhibitors problematic. In spite of these similarities, we now report that these tyrosinespecific kinases can be distinguished from one another by their active site substrate specificities.
Peptide Synthesis-Fmoc-Arg(Mtr)-Arg(Mtr)-Arg(Mtr)-Arg(Mtr)-Arg(Mtr)-Leu-Glu(t-butyl)-Glu(t-butyl)-Leu-Leu-resin was prepared on the 2-methoxy-4-alkoxybenzyl alcohol resin (substitution level, 0.45 mmol/g resin) with Fmoc amino acids (22,23). A standard Fmoc peptide synthesis protocol was employed using an automated peptide synthesizer (Advanced ChemTech Act 90). Upon completion of peptide synthesis, the side chain-protected peptide was cleaved from the resin using 1% trifluoroacetic acid in CH 2 Cl 2 (20 ml/g resin, 15 min). The resin was filtered, and the trifluoroacetic acid/CH 2 Cl 2 solvent (containing the peptide) was collected. This process was repeated three additional times. 100 ml of water was subsequently added to the combined trifluoroacetic acid/CH 2 Cl 2 extracts. Triethylamine was added to furnish a neutral solution and then all but 15 ml of the solvent was removed under reduced pressure. The precipitated peptide was collected via filtration and washed with ethyl ether (50 ml). Typically, 1 g of resin provided 400 mg of side chain-protected peptide. The free C terminus of Fmoc-Arg(Mtr)-Arg(Mtr)-Arg(Mtr)-Arg(Mtr)-Arg(Mtr)-Leu-Glu(t-butyl)Glu(t-butyl)-Leu-Leu (40 mg, 0.014 mmol) was activated via treatment with Bop (0.028 mmol) in 3 ml of CH 2 Cl 2 /N,N-dimethylformamide (1:1) for 3 min. Individual amino alcohols (0.14 mmol) were introduced and allowed to react with the activated peptide for 1 h. Subsequently, the N-terminal Fmoc protecting group was removed via treatment with piperidine for 30 min, and the N terminus deprotected peptide alcohol was then precipitated from solution by the addition of ethyl ether (50 ml). The heterogeneous solution was filtered in order to isolate the peptide. The remaining protecting groups (Mtr and t-butyl ester) were removed by dissolving the peptide alcohol in 90% trifluoroacetic acid/ 10% thioanisol (6 ml). After stirring for 6 h, the peptide was precipitated from solution (50 ml ethyl ether) and subsequently isolated via filtration. Each peptide alcohol was then purified via a two-step sequence. First, the individual peptides were ion exchanged on CM-Sephadex C-25 (0.4 M KCl to a 1.2 M KCl gradient using a 50 mM potassium acetate pH 3.5 buffer). The desired peptide alcohol typically eluted at 0.8 -1.0 M KCl and was subsequently collected and lyophilized. The peptides were then desalted via preparative high pressure liquid chromatography using three Waters radial compression modules (2.5 ϫ 10 cm) connected in series (gradient containing solvent A (0.1% trifluoroacetic acid in water) and solvent B (0.1% trifluoroacetic acid in acetonitrile): 0 -3 min (100% A), a linear gradient from 3 (100% A) to 5 min (80% A and 20% B), 5-20 min (20 -50% B), 20 -25 min (50 -80% B)). All of the purified peptides gave satisfactory fast atom bombardment mass spectral analyses.
c-Src-Human c-Src was purchased from Upstate Biotechnology, Inc. The enzyme was expressed by recombinant baculovirus containing the human c-src gene in SF9 insect cells.
v-Abl-The catalytic domain of v-Abl was expressed in Escherichia coli as a fusion protein with glutathione S-transferase as described previously (25). The enzyme was purified to homogeneity and characterized according to previously described methods (25) with the following modification: E. coli cells expressing the glutathione S-transferaseabl fusion protein were lysed in a French pressure cell. After centrifugation of the cell lysates (13,000 ϫ g for 30 min), cell pellets were resuspended in 1.5% N-lauroylsarcosine, 25 mM triethanolamine, and 1 mM EDTA, pH 8.0. This mixture was rocked at 4°C for 30 min and then recentrifuged at 4°C (10,000 ϫ g for 10 min). The supernatant was combined with that from the original centrifugation and applied to glutathione-agarose.
c-Src Assay-Assays were performed in triplicate at pH 7.5 and thermostatted in a water bath maintained at 22°C. . Reactions were terminated after 20.0 min by spotting 25-l aliquots onto 2.1-cm diameter phosphocellulose paper discs. After 10 s the discs were immersed in 10% glacial acetic acid and allowed to soak with occasional stirring for at least 1 h. The acetic acid was decanted, and the discs were collectively washed with four volumes of 0.5% H 3 PO 4 and one volume of water followed by a final acetone rinse. The discs were air dried and placed in plastic scintillation vials containing 6 ml of Liquiscint prior to scintillation counting for radioactivity. Reactions were terminated after 20.0 min by spotting 25-l aliquots onto 2.1-cm diameter phosphocellulose paper discs. After 10 s the discs were immersed in 10% glacial acetic acid and allowed to soak with occasional stirring for at least 1 h. The acetic acid was decanted, and the discs were collectively washed with four volumes of 0.5% H 3 PO 4 and one volume of water followed by a final acetone rinse. The discs were air dried and placed in plastic scintillation vials containing 6 ml of Liquiscint prior to scintillation counting for radioactivity.
Determination of Kinetic Constants-The apparent K m (Ϯ S.D.) and V max (Ϯ S.D.) values were determined from initial rate experiments. The data from these experiments were plotted using the Lineweaver-Burke procedure, and the corresponding plots proved to be linear.

RESULTS AND DISCUSSION
Peptide substrates containing structurally disparate alcoholbearing residues, positioned at the site of phosphorylation, can provide critical information concerning the range of compounds that can be accommodated within and phosphorylated by the active site of protein kinases (5). Although it can be laborious to synthesize peptides that contain exotic amino acid residues at internal sites, it requires significantly less effort to prepare analogous peptides containing nonstandard residues at either the N or the C terminus (5). Both c-Src and v-Abl phosphorylate the C-terminal tyrosine residue in peptide 1 ( Table I). The five arginine residues on the N terminus of this peptide are present in order to promote binding of the phosphorylated peptide to phosphocellulose filter paper. This provides a convenient method for following the kinetics of phosphorylation. The Cterminal portion of the peptide, Leu-Glu-Glu-Leu-Leu, is negatively charged, a feature that appears to be important for substrate recognition by some tyrosine kinases (15,26,27). 2 However, whereas 1 is a substrate for both c-Src and v-Abl, it is clear that this peptide is more efficiently phosphorylated by the former than by the latter. This may be due, in part, to real differences in sequence specificities between the two enzymes. For example, Src has a special preference for peptide substrates that contain acidic residues at the P-3 and P-4 positions (27). 2 In contrast, v-Abl is much less dependent upon the presence of negatively charged residues for substrate recognition (26,27). However, it is also important to note that kinetic constants obtained for the v-Abl-catalyzed phosphorylation of 1 compare favorably with those associated with the best peptide substrates reported for this enzyme (15). 2 In short, the fact that peptide 1 is a 42-fold more efficient substrate for c-Src than for v-Abl may simply be due to the innate ability of the former to serve as more potent phosphoryl transfer catalyst than the latter. The salient point is that both enzymes phosphorylate common substrates, a feature that has been previously noted (15, 28). 2 Both v-Abl and c-Src phosphorylate aromatic and benzylic alcohols (Table I). For example, peptide 2, which contains an tyramine residue, is a substrate for both enzymes. Although it is significant that these tyrosine-specific kinases will phosphorylate an achiral residue, it is equally noteworthy that 2 displays an enhanced specificity for c-Src versus v-Abl relative to that of the tyrosine-containing peptide 1 (Table II). In contrast, the aminomethylbenzyl alcohol 3 is almost identical to L-tyrosine in its ability to discriminate between these enzymes. Finally, the benzylic alcohol in 4 is less efficient than L-tyrosine in distinguishing between c-Src and v-Abl. The decrease in c-Src/ v-Abl specificity as one moves from compound 2 to compound 3 to compound 4 is largely due to changes in activity of c-Src toward these substrates. All three peptides are virtually identical as substrates for v-Abl. In short, although c-Src does discriminate between aromatic and aliphatic alcohols (i.e. 2 versus 3 or 4), v-Abl does not.
We previously demonstrated that the serine/threonine-specific cAMP-dependent protein kinase will not catalyze the phosphorylation of tyrosine in an active site-directed peptide (29). However, this protein kinase will catalyze the phosphorylation of aromatic alcohols if the hydroxyl moiety can be positioned in the active site in a manner comparable with that of serine (8,10). In addition, although cAMP-dependent protein kinase (as well as protein kinase C) will phosphorylate achiral aliphatic alcohols, this activity is highly dependent upon chain length (5,9). There is a dramatic loss in substrate efficacy if the number of methylene groups on the residue to be phosphorylated is greater than two. This is not particularly surprising, because the hydroxyl and amino groups in serine are separated by two methylene groups as well. The active site substrate specificity of serine/threonine-specific protein kinases is therefore strictly determined by the relative orientation of and the distance between the phosphorylatable alcohol moiety and the peptide backbone to which it is attached. Is the active site specificity of tyrosine-specific protein kinases as tightly regulated? The results in Table I clearly demonstrate that both c-Src and v-Abl will phosphorylate aromatic and benzylic (i.e. aliphatic) alcohols. However, each of these substrates possesses an aromatic moiety. Furthermore, the hydroxyl functionality in species 2-4 is oriented, relative to the peptide backbone, in a manner reminiscent of that found in the tyrosine-containing peptide 1.

TABLE I K m (M) and V max (nmol/min ⅐ mg) values for the aromatic and
benzylic alcohol-containing peptides 1-4 Kinetic constants were determined as described under "Materials and Methods." The results with c-Src have been previously reported (11). The values are given as the averages Ϯ S.D.

TABLE II A comparison of the substrate specificity of peptides 1-19 for c-Src
and v-Abl The c-Src:v-Abl specificity is calculated by dividing the k cat /K m value obtained with c-Src by the corresponding value obtained with v-Abl. Is this structural alignment of functionality a requirement for recognition by tyrosine-specific protein kinases? In order to address this question, we prepared a series of peptides containing straight chain aliphatic alcohols. In a broad sense, both v-Abl and c-Src utilize these species as substrates (Table III). Therefore, it is evident that an aromatic moiety is not required for substrate recognition by either of these enzymes. Furthermore, the distance between the aromatic alcohol in tyrosine and the ␣-amino group is roughly equivalent to a chain containing 5-7 methylene groups. Consequently, one might predict that compounds 8-10 should serve as the most efficient substrates for v-Abl and c-Src and that the remainder of the compounds listed in Table III would be extraordinarily inefficient substrates (if they are phosphorylated at all). Somewhat surprisingly, this is not the case (Table II). For example, the octanolamine derivative 11 is the most effective c-Src substrate. However, the k cat /K m values for peptides 5-12 do not differ to any significant extent with this enzyme. Only at a chain length of 12 carbon atoms is a drop-off in substrate efficacy apparent. In this particular case, we were only able to obtain a lower limit of the maximal velocity. This is due to the fact that substrate saturation appears to occur at a relatively low concentration (Ͻ150 M) and that at these concentrations the rate of phosphorylation of 13 is exceedingly weak. This implies that the concentrations of 13 employed may exceed the K m . As a consequence, we are able to provide only an estimate of V max for this peptide, as well for peptides 15-18 in Table IV.
In contrast to the behavior displayed by c-Src toward peptides 5-13, v-Abl exhibits a definite dependence upon chain length. Short chains, such as the ethanolamine (5) and propanolamine (6) derivatives, are extremely poor v-Abl substrates. Interestingly, our inability to obtain K m and V max values in these two instances is not due to the difficulties that we encountered with the c-Src-catalyzed phosphorylation of peptide 13 described above. In the case of v-Abl, although high concentrations (1-2 mM) of peptides 5 and 6 provide weak rates of phosphorylation, we did not detect any evidence of substrate saturation. This implies that the K m values for peptides 5 and 6 are quite large, a notion consistent with the results obtained for peptide 7, the next higher homolog. At substrate concentrations significantly below the K m , the Michaelis-Menten equation reduces to v ϭ (k cat /K m )[E][S], which allows the ready acquisition of k cat /K m . However, because we are uncertain of the actual K m , the k cat /K m values indicated for peptides 5, 6, and 14 in Table II should be viewed as approximations. Finally, the v-Abl-catalyzed rates of phosphorylation for peptides 15-18 are so weak that we could only provide an estimate of k cat /K m in these cases.
Based upon the k cat /K m value extracted for peptide 5, it is evident that this species is a significantly more potent substrate (Ͼ2,000-fold) for c-Src than for v-Abl. As such, peptide 5 is the most accurate c-Src-targeted substrate that we have evaluated to date. However, this bias in favor of c-Src decreases in a dramatic fashion with increasing chain length. This decrease is a consequence of an improvement in k cat /K m for v-Abl with the longer aliphatics, which in turn is due to a decrease in K m . Indeed, the K m for peptide 13 is even smaller than that obtained for the L-tyrosine peptide 1. Interestingly, these results are consistent with previous sequence specificity studies reported for these two kinases. v-Abl has been shown to prefer hydrophobic residues on the N-terminal side of tyrosine to a greater extent than c-Src (15). This appears to correlate with the preference of v-Abl for the longer and more hydrophobic straight chain alcohols of Table III.
Because the ethanolamine-containing peptide 5 serves as a substrate for c-Src it is not too surprising that Arg-Arg-Arg-Arg-Arg-Leu-Glu-Glu-Leu-Leu-Ser-amide (14) is phosphorylated by this enzyme as well (Table IV). However, the ethanolamine derivative (5) is 4.5-fold more efficient than its serinecontaining counterpart (14). One possible explanation for this observation is that the amide moiety, which is present in 14 but absent in 5, encounters unfavorable steric interactions with active site residues. If this supposition is correct, then substituents more sterically demanding than the amide functionality should produce even less effective substrates (Table IV). Indeed, this appears to be the case for the benzyl-substituted derivatives 15 and 16. These results are not particularly unexpected given the structural characteristics of the customary

M) and V max (nmol/min ⅐ mg) values for the aliphatic alcohol-containing peptides 5-13
Kinetic constants were determined as described under "Materials and Methods." The results with c-Src have been previously reported (11). The values are given as the averages Ϯ S.D. Kinetic constants were determined as described under "Materials and Methods." The results with c-Src have been previously reported (11). The values are given as the averages Ϯ S.D.

RRRRRLEELL-amino alcohol
substrate of c-Src. In the case of tyrosine, the hydroxyl moiety protrudes from an aromatic framework, which is tied back in a sterically uncompromising cyclic fashion. In contrast, the amide and benzyl substituents of peptides 14-16 are more sterically prominent and obtrusive than the aromatic moiety of tyrosine, particularly near the hydroxyl functionality. Furthermore, secondary alcohols (17 and 18) also fail to serve as efficient c-Src substrates. All of these observations are consistent with the notion that sterically demanding functionality may encounter unfavorable interactions when forced to reside within the active site region of c-Src.
Although the ethanolamine-derivatized peptide 5 is an effective c-Src substrate, it is almost imperceptibly phosphorylated by v-Abl. If the steric arguments outlined above apply to v-Abl as well, then at the very least this enzyme should not be able to phosphorylate peptides 14-18 any more efficiently than 5. Indeed this is the case. Nevertheless, v-Abl does phosphorylate all of these peptides, albeit at an nearly indiscernible level that is barely above background.
Protein kinases are commonly classified by their ability to phosphorylate aliphatic (i.e. serine/threonine) or aromatic (i.e. tyrosine) hydroxyl groups in naturally occurring proteins. However, as is apparent from this as well as earlier (5-11) active site substrate specificity studies, these categories break down with synthetic substrates. Nonetheless, a new pattern is beginning to emerge with respect to the classification of protein kinases. We, as well as others, have shown that although cAMP-dependent protein kinase will only phosphorylate Lamino acid residues, protein kinase C will catalyze the phosphorylation of both L-and D-stereoisomers (5,9,30,31). More recently, cGMP-dependent protein kinase and c-Src (11) have been added to the latter category. We have now found that peptide 19, which contains a D-tyrosine residue, will also serve as a substrate for v-Abl. The following features concerning this activity are notable. First, although v-Abl will phosphorylate peptides containing either D-or L-tyrosine, the latter is a significantly more efficient (65-fold) substrate than the former. Second, the D-tyrosine-containing peptide (19) is a 5-fold poorer substrate than the corresponding achiral species (2). Consequently, it is clear that the inverted configuration associated with D-tyrosine in conjunction with the amide substituent actively interferes with the ability of this species to be recognized as a substrate for v-Abl. Third, peptide 19 serves as a 230-fold more efficient substrate for c-Src than for v-Abl, which may indicate that the former is able to accommodate D-amino acid residues more readily than the latter.
We have not yet completely unraveled the molecular mechanism that enables certain protein kinases to phosphorylate both L-and D-configurational isomers. However, we have previously proposed that the alcohol functionality of a D-residue can be presented to the target kinase in a manner analogous to that of an L-residue if the peptide were to bind to the enzyme surface in a aberrant C to N terminus fashion (9). Although an analogous binding mode may be in operation for v-Abl and c-Src, other mechanisms may account for this form of active site specificity as well.
In summary, we have compared and contrasted the active site substrate specificities of v-Abl and c-Src. Although these enzymes are commonly referred to as tyrosine-specific, both species are able to phosphorylate a wide variety of aromatic and aliphatic alcohols, including a D-tyrosine residue. Both are also unable to phosphorylate certain structural arrangements, such as secondary alcohols and primary alcohols containing large substituents within close proximity to the hydroxyl moiety. Nevertheless, v-Abl and c-Src do exhibit certain subtle and gross differences in catalytic behavior as well. Although c-Src displays a moderate preference for aromatic versus benzylic alcohols, v-Abl does not. Although c-Src utilizes short straight chain alcohols as substrates, the v-abl-encoded protein kinase does not. Indeed, the former does not discriminate against aliphatic alcohols on the basis of chain length, whereas the latter exhibits a clear bias on these grounds. In short, these results are encouraging with respect to inhibitor design, because both enzymes are able to accommodate a range of variants within their active site regions. Also encouraging are the clear differences in the active site specificities of these enzymes. This particular feature, when employed in combination with preferred sequence specificities, may ultimately provide a means to precisely target and inactivate specific protein kinases.