Pyridinyl imidazole inhibitors of p38 mitogen-activated protein kinase bind in the ATP site.

The site of action of a series of pyridinyl imidazole compounds that are selective inhibitors of p38 mitogen-activated protein kinase in vitro and block proinflammatory cytokine production in vivo has been determined. Using Edman sequencing, 125I-SB206718 was shown to cross-link to the nonphosphorylated Escherichia coli-expressed p38 kinase at Thr175, which is proximal to the ATP binding site. Titration calorimetric studies with E. coli-expressed p38 kinase showed that SB203580 bound with a stoichiometry of 1:1 and that binding was blocked by preincubation of p38 kinase with the ATP analogue, FSBA (5'-[p-(fluorosulfonyl)benzoyl]adenosine), which covalently modifies the ATP binding site. The intrinsic ATPase activity of the nonphosphorylated enzyme was inhibited by SB203580 with a Km of 9.6 mM. Kinetic studies of active, phosphorylated yeast-expressed p38 kinase using a peptide substrate showed that SB203580 was competitive with ATP with a Ki of 21 nM and that kinase inhibition correlated with binding and biological activity. Mutagenesis indicated that binding of 125I-SB206718 was dependent on the catalytic residues K53 and D168 in the ATP pocket. These findings indicate that the pyridinyl imidazoles act in vivo by inhibiting p38 kinase activity through competition with ATP and that their selectivity is probably determined by differences in nonconserved regions within or near the ATP binding pocket.

Several novel intracellular signaling pathways associated with heat; chemical, osmotic, and oxidative stress; UV; and the proinflammatory cytokines interleukin 1 and tumor necrosis factor have recently been discovered in mammalian cells (1)(2)(3)(4)(5)(6)(7). Because these signals are commonly associated with the early stages of host responses to injury and infection, they have generated significant interest for their possible role in various pathological conditions and consequent potential as targets for novel therapeutics.
The key components of these pathways are members of the MAP 1 kinase family of serine-threonine protein kinases (8).
The MAP kinases are proline-directed serine-threonine protein kinases that are activated by phosphorylation on both a threonine and tyrosine in a Thr-X-Tyr motif found in an activation loop proximal to the ATP and substrate binding sites (9). This is accomplished in vivo by a dual specificity MAP kinase kinase, which in turn is activated by phosphorylation in response to an appropriate extracellular or intracellular signal. There are three main classes of MAP kinase, the ERKs, the c-jun N-terminal kinase/stress-activated protein kinases, and the p38 kinases (also known as CSBP, RK, MPK2, and HOG1 (1, 5-7)), which differ in the size of the activation loop and the nature of the X amino acid in the TXY motif. This is Glu in the ERKs, Pro in the c-jun N-terminal kinase/stress-activated protein kinases, and Gly in the p38 kinases. As a result of this and additional proximal differences, each MAP kinase is recognized and activated by a different set of MAP kinase kinases and in turn has different in vivo substrates (7,10,12,13). These differences are also reflected in the different activation stimuli for each MAP kinase, with the c-jun N-terminal kinase/stressactivated protein kinases and p38 kinases and their homologues (6, 14 -18) being particularly associated with stress and inflammatory stimuli (1942).
We recently discovered that a series of pyridinyl imidazoles that inhibited the production of interleukin 1 and tumor necrosis factor from lipopolysaccharide-activated human monocytes bound to and inhibited p38 kinase, which we originally called CSBP (6). These compounds inhibit two splice forms of p38, and a newly discovered homologue of p38 kinase, p38␤ (16), but they do not inhibit the closely related ERK or c-jun N-terminal kinase/stress-activated protein kinases or other serine-threonine protein kinases (20).
The pyridinyl imidazoles, typified by SB203580, have proven useful in investigating the role of p38 kinase in regulating transcription, translation, and cytoskeletal elements in response to various stress and cytokine stimuli, as well as its potential role in animal models of inflammatory disease (21)(22)(23)(24)(25)(26). However, there has not been to date a detailed understanding of how SB203580 binds and subsequently inhibits p38 kinase catalytic activity. In the present study we show that the pyridinyl imidazoles inhibit p38 kinase activity by binding to the ATP binding site, a region that until recently was not thought to provide enough specificity for the design of protein kinase inhibitors.
Expression of Met-Ala-His 6 p38 in Saccharomyces cerevisiae-An expression vector encoding Met-Ala-His 6 p38 was constructed as follows. A 1.4-kilobase BamHI (Klenow filled)-Asp 718 Ile fragment encoding Met-Ala-His 6 p38 (6) was subcloned into p138NBU (30) that had been digested with XhoI (Klenow filled) and Asp 718 Ile, creating p138NBU-Ala-His 6 p38. Briefly, the plasmid contains the URA3-selectable marker, partial 2 sequences for maintenance at high copy number, with Met-Ala-His 6 p38 expression driven by the copper-inducible CUP1 promoter and terminated by the CYC1 transcriptional terminator. p138NBU-Ala-His 6 p38 was introduced into JBY10 (1), a hog1⌬ strain of S. cerevisiae, by the lithium acetate method (31). Cells were grown in synthetic complete medium minus uracil (32) at 30°C and 225 rpm to A 540 ϭ 1 and induced for expression with 150 M CuSO 4 for 4 h.
Mutagenesis of p38 Kinase-Site-directed mutagenesis was performed on a FLAG-p38 construct (12). The mutants were subcloned into p138NBU and propagated in JBY10, and extracts were prepared as described previously (30).
Purification of Met-Ala-His 6 p38 Kinase from Escherichia coli-Met-Ala-His 6 p38 kinase was purified from clarified E. coli lysates by chromatography on a Qiagen Ni 2ϩ -NTA column equilibrated in 100 mM Tris, pH 8.0, 1 mM phenylmethylsulfonyl fluoride at 4°C and eluted with a 10 -250 mM imidazole gradient in the same buffer. The p38 kinase was further purified by sequential passage over a Superdex 75 column equilibrated in 20 mM Tris, 150 mM NaCl, 5 mM dithiothreitol, and a Mono Q column equilibrated in 100 mM Tris, pH 8.0, 1 mM phenylmethylsulfonyl fluoride, 5 mM dithiothreitol and eluted by a 0 -500 mM NaCl gradient. The purity was greater than 95% and yield was 5 mg/g of E. coli.
Purification of Yeast-expressed Met-Ala-His 6 p38 Kinase-Yeast cells were thawed and lysed with glass beads in 100 mM Tris buffer containing 250 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml pepsatin A, 10 g/ml leupeptin, 20 mM NaF, 2 mM Na 3 VO 4 , 2 mM sodium pyrophosphate, 25 mM ␤-glycerophosphate, 2 mM phenyl phosphate, pH 7.4, and the lysate was clarified by centrifugation. After addition of an equal volume of buffer A (20 mM Tris, pH 7.4), the supernatant was purified by sequential chromatography on Q-Sepharose Fast Flow and Ni 2ϩ -NTA columns (Qiagen) and elution in both cases with 250 mM imidazole in buffer A. Purified p38 kinase was exhaustively dialyzed against 20 mM HEPES, 2 mM NaF, 0.2 mM Na 3 VO 4 , pH 7.4, and stored frozen at Ϫ70°C. The purity was greater than 95% and yield was 20 g/g of yeast.
Analysis of Protein Phosphorylation-50 pM recombinant E. coliexpressed Met-Ala-His 8 p38 was reduced, pyridylethylated, digested with trypsin (6 h at 37°C), and analyzed using negative ion liquid chromatography electrospray mass spectrometry and stepped collision energy scanning for selective identification of phosphorylated peptides (33). No phosphorylated peptides were detected using the phosphateselective scan, and the putative phosphorylation-site peptide (His 182 -Thr-Asp-Asp-Glu-Met-Thr-Gly-Tyr-Val-Ala-Thr-Arg) was observed in its nonphosphorylated state. Approximately 300 pM Met-Ala-His 8 p38 expressed in yeast was reduced, pyridylethylated, and desalted on a C18 guard column. The sample was then dried, reconstituted in 25 mM NH 4 HCO 3 , and cleaved with trypsin for 6 h at 37°C. A 1-l aliquot of the digest containing approximately 30 pM Met-Ala-His 6 -p38 was mixed with either 2 l of 4% NH 4 OH or 10% formic acid. 1 l of each of these solutions was loaded in separate experiments into a nanospray tip, and the sample was analyzed by electrospray mass spectrometry at a flow rate of approximately 30 nanoliters/min (34). A number of experiments were carried out on this single loading of approximately 10 pM sample, including a full scan negative ion electrospray mass spectrometry, precursor scan of m/z 79 (phosphate-selective scan), full scan positive ion electrospray mass spectrometry to locate parents of phosphopeptides for sequencing by tandem MS (MS/MS), and MS/MS of the phosphorylated forms of the peptide HTDDEMTGYVATR from the digest. Signals were observed for this peptide with zero, one, and two phosphates bound. MS/MS of the nonphosphorylated form corroborated the presumed sequence. MS/MS of the monophosphorylated parent indicated that this species was primarily the peptide phosphorylated on Thr 188 . We were not able to obtain useful MS/MS data on the putative di-phosphorylated species.
Sequencing of Radiophotoaffinity-labeled p38 Kinase-Purified E. coli-expressed p38 kinase was UV cross-linked with 125 I-SB206718 as described previously (6) and was fragmented using cyanogen bromide (Pierce) and lysyl endopeptidase (Wako Biochemicals). Cyanogen bromide digests were performed in 70% formic acid using 10 l of a fresh 70 mg/ml cyanogen bromide solution added per 10 g of protein digested. Digests with lysyl endopeptidase were performed in 20 mM Tris-HCl, pH 9.0, at 1:20 (w/w) enzyme:substrate. Digest fragments were separated by 15% reducing Laemmli SDS-polyacrylamide gel electrophoresis gels and electroblotted onto Problott membrane (Applied Biosystems) for sequencing.
Protein sequence analysis was performed on a Beckman LF 3400 TC gas-phase protein sequencer equipped with a Beckman 126/166 system for on-line phenylthiohydantoin analysis. For determining the site of radiophotoaffinity cross-linking, a modified sequencing cycle was used in which the anilinothiazolinone derivatives were collected in S1 (ethyl acetate) directly for 125 I determination on a Beckman Gamma 8500 scintillation counter. The cross-linking site of the related compound, 125 I-SB227931, could not be determined but was clearly distinct from that of 125 I-SB206718.
Measurements of p38 Kinase Activity-Kinase activity was determined by measuring the incorporation of 32  For the immune-complex kinase assay, cells were treated with SB203580 or SB202474 for 30 min prior to activation with 0.4 M sorbitol for 30 min. Cells were then washed twice in PBS and solubilized on ice in lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 2 mM EDTA, 25 mM ␤-glycerophosphate, 20 mM NaF, 1 mM sodium orthovanadate, 2 mM sodium pyrophosphate, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 5 units/ml aprotinin) and centrifuged at 15,000 ϫ g for 20 min at 4°C. Endogenous kinases were precipitated from cell lysates using anti-p38 kinase (6) or anti-MAP-KAP kinase-2 antibodies (kindly supplied by Dr. Nick Morrice, University of Dundee, Dundee, United Kingdom). The beads were washed twice with lysis buffer and twice with kinase buffer (25 mM HEPES, pH 7.4, 25 mM MgCl 2 , 25 mM ␤-glycerophosphate, 100 M sodium orthovanadate, 2 mM dithiothreitol). The immune-complex kinase assays were performed as described (12,35) The amount of radioactivity was quantitated in a B603 Betascope blot analyzer (Betagen). The amount of p38 kinase present in the immunoprecipitates and its extent of activation was determined by immunoblot using anti-phosphotyrosine (PY20, Santa Cruz Biotechnology) and rabbit polyclonal antibodies generated against recombinant p38 kinase (6).
Isothermal Calorimetry-Isothermal titration calorimetry experiments were performed with a Microcal, Inc. MCS microcalorimeter (36). Data were analyzed with an equilibrium binding model having a single equilibrium constant and enthalpy change. p38 kinase was dialyzed against the experimental buffer just prior to the titration with SB203580, and its concentration was determined from absorbance at 280 nm using an extinction coefficient of 1.12 (mg/mL) Ϫ1 cm Ϫ1 and a molecular mass of 42 kDa. The concentration of SB203580 was established gravimetrically. SB203580 was dissolved to 5 mM in 100% DMSO and then diluted into experimental buffer to 100 M (2% DMSO final).
Reaction of p38 kinase with FSBA was carried out under conditions similar to those reported for cAMP-dependent protein kinase (37,38). p38 kinase was incubated overnight at room temperature with 1 mM FSBA, 0.1 M Tris, 150 mM NaCl, 0.1 M MgCl 2 , at pH 7.4, and was dialyzed against experimental buffer prior to titration calorimetry.
Confirmation that p38 kinase reacted with FSBA was done by evaluating its conformational stability. The thermal stability of p38 kinase was evaluated by monitoring the circular dichroism spectral signal at 220 nm as a function of temperature. p38 kinase was unfolded as a sharp, homogeneous transition before and after reaction with FSBA, but the FSBA-reacted p38 kinase showed enhanced thermal stability (T m ϭ 51°C) over the unreacted (T m ϭ 46°C).

RESULTS
Binding to Unactivated p38 Kinase-To determine where the pyridinyl imidazole inhibitors bind to p38 kinase, we determined the cross-linking site of the radiophotoaffinity compound, 125 I-SB206718 (Fig. 1), previously used to identify and purify p38 kinase. Sequencing of peptide fragments from crosslinked, E. coli-expressed p38 kinase showed that radioactivity was associated predominantly with Thr 175 and at much lower levels with Leu 171 and Asp 177 (Fig. 2), all residues in the activation loop of p38 kinase. This suggested that the compounds may be inhibitors of ATP or substrate binding.
In support of this, the previously demonstrated binding of 3 H SB202190 to p38 kinase in the lysates from unstimulated THP.1 cells (6) was competed by ATP with an IC 50 of approximately 4 mM (data not shown). To obtain more quantitative data, we examined compound binding and ATP competition to purified p38 kinase via microcalorimetry. SB203580 bound to E. coli-expressed p38 kinase at 30°C with a K d ϭ 15 nM and a ⌬H ϭ Ϫ12 kcal (Fig. 3, left panel). The molar binding ratio was equal to 0.75 Ϯ 0.1 SB203580/p38 kinase, which is within experimental error of a 1:1 molar binding stoichiometry. When p38 kinase was preincubated with FSBA (5Ј-[p-(fluorosulfonyl)benzoyl]adenosine), an ATP analogue that covalently modifies cAMP-dependent protein kinase at lysine 72 and precludes ATP binding (37,38), no binding of SB203580 was observed (Fig. 3, right panel).
Despite a lack of phosphorylation as detected by electrospray mass spectrometric analysis of tryptic peptide fragments (see "Experimental Procedures"; data not shown), E. coli-expressed p38 kinase nevertheless had an intrinsic ATPase activity in the absence of substrate. Using a spectrophotometric coupled-enzyme assay (39), the Michaelis-Menten k cat and K m for this ATPase activity were determined to be 0.006 s Ϫ1 and 9.6 mM ATP. The ATPase activity was also observed in the microcalorimeter as a steady-state production of heat that was inhibited by SB203580 (Fig. 4). The catalytic activity was eliminated when a molar ratio of about 1 SB203580/p38 kinase or more was added to the calorimeter.
Binding to Activated p38 Kinase-Because the above experiments dealt with the the binding of the pyridinyl imidazoles to the unactivated form of p38 kinase, it became important to determine whether the same features applied to the activated form of the enzyme. We have previously shown that p38 kinase expressed in yeast is constitutively active due to its phosphorylation by the yeast MAP kinase kinase, Pbs2p (30). Electrospray mass spectrometric analysis of p38 expressed in yeast using phosphopeptide-selective scanning and detection methods (33,34) showed that approximately 20% of the purified protein was phosphorylated on both Thr and Tyr in the activation loop, 40% was phosphorylated on Thr, and the rest was unphosphorylated (see "Experimental Procedures"). All of the kinase activity is assumed to be attributable to the doubly phosphorylated form as evidenced by mutagenesis of either or both of these residues (30).
A p38 kinase assay was developed using a peptide substrate with a sequence derived from the intracellular domain of the epidermal growth factor receptor. This peptide has been previously used to assay ERK activity (40). This peptide phosphorylation assay was used to compare the potency of pyridinyl imidazoles in inhibition of the yeast-expressed activated form of p38 kinase with their binding affinity to the unactivated form of the enzyme as measured in the cytosol of the THP.1 monocytic cell line. As shown in Fig. 5, the IC 50 values for binding to inactive p38 kinase correlated with those for kinase inhibition of the activated form of the enzyme. The correlation coefficient for a large number of pyridinyl imidazoles, ranging in potency over 3 orders of magnitude, was calculated from the data in Fig. 5 and found to be 0.911. Combined with the previously noted correlation of compound binding affinity with inhibition of proinflammatory cytokine production from lipopolysaccharide-activated monocytes (6), this indicates that one mechanism by which these pyridinyl imidazoles act is through inhibition of the catalytic activity of the activated form of p38 kinase.
To evaluate further the competition between pyridinyl imidazole binding and ATP observed for the unactivated p38 kinase above, the mode of inhibition of p38 kinase by SB203580 with respect to ATP was determined. The resulting Lineweaver-Burke plot (Fig. 6) shows that SB203580 inhibits p38 kinase in a manner competitive with ATP with a K i of 21 nM. The K m [ATP] was increased from 200 to 1392 M in the presence of 100 nM SB203580 (Fig. 6).
To determine potential sites important for compound binding, we examined the effect of selective mutations of p38 kinase expressed in yeast on the binding of two radiophotoaffinity cross-linkers, 125 I-SB206718 and a related cross-linker 125 I-SB227931, which has the azide group localized to a different region of the pyridinyl imidazole, closer to the active pharmacophore (27) (Fig. 1). Interestingly, mutation of the Thr 175 FIG. 1. Structures of inhibitor SB203580; radioligand SB202190 and radiophotoaffinity labels SB206718 and SB227931. Reported IC 50 values were obtained in the binding assay (6). Numbers in parentheses refer to values obtained in the kinase assay (27) .   FIG. 2. Cross-linking of 125 I-SB206718 to p38 kinase. The radioactivity associated with each amino acid residue in the isolated cyanogen bromide fragment prepared from p38 radioactively cross-linked with 125 I-SB206718 as described under "Experimental Procedures" is shown.
cross-linking site of 125 I-SB206718 to Ala did not eliminate cross-linking to either compound, suggesting that hydroxyl group of the Thr is not required for cross-linker attachment (Fig. 7). Mutation of other loop residues (Thr 180 and Tyr 182 ) also had no effect on cross-linking but did abrogate kinase activity. On the other hand, mutation of the catalytic residues Lys 53 and Asp 168 resulted in a reduction of both kinase activity and 125 I-SB206718 and 125 I-SB227931 cross-linking, indicating that these compounds bind in the catalytic site occupied by ATP. These mutations do not lead to gross misfolding because both mutants were phosphorylated on Tyr 182 by yeast Pbs2 as shown in Fig. 7 (see also Ref. 30), and D168A p38 kinase is still able to interact with its downstream substrate MAPKAP kinase-3 (12). Interestingly, the equivalent mutation to K53R in the closely related ERK (K52R) was shown not to affect ATP binding, but only the catalytic activity (41). Hence, this mutation may specifically affect the binding of the pyridinyl imidazoles and not ATP. In contrast, mutation of A34V, which results in a reduction in kinase activity (30), had no effect on cross-linking by either compound, once again suggesting differences in the binding of ATP and the pyridinyl imidazoles.
Finally, we wanted to know whether the binding of a pyridinyl imidazole to the unactivated form of p38 kinase prevented its activation by MAP kinase kinases in cells. HeLa cells were pretreated with increasing concentrations of SB203580 or SB202474 (a structurally related but inactive compound (6)) prior to addition of 0.4 mM sorbitol, which increases osmolarity. An in vitro kinase assay of immunoprecipitated p38 kinase isolated from these cells and washed free of compound showed that SB203580 had no effect on the p38 kinase activity stimulated by high osmolarity. Similarly, SB203580 has no effect on tyrosine phosphorylation of p38 kinase (Fig. 8). However, under these conditions, activation of the downstream in vivo substrate of p38 kinase, MAPKAP kinase-2, as measured by in vitro phosphorylation of hsp27, is inhibited by SB203580 (Fig.  8). In contrast, p38 kinase activity is inhibited if SB203580 is added to the in vitro reaction (data not shown (6,35)). It is unlikely that any activation is occurring in vitro by autophosphorylation because mutagenesis has shown that both tyrosine and threonine phosphorylation are required for activation of p38 kinase activity (30,42). These data suggest that SB203580 binding inhibits p38 kinase activity but does not prevent its activation by MAP kinase kinases. SB202474, a structurally Heat produced per second is proportional to the amount of p38 kinasecatalyzed ATP hydrolysis. The conditions used were 2.5 M p38 kinase in 5 mM HEPES, 150 mM NaCl, 20 mM MgCl 2 , 1 mM ATP, pH 7.5, 2% DMSO for 30°C. 8.3-l injections of 100 M SB203580 were made into a 1.4-ml solution of p38 kinase in the calorimeter. Curve is the best fit to a competitive inhibition Michaelis-Menton model with an upper limit to K i determined as K i Յ 100 nM.
FIG. 5. Potency of pyridinyl imidazole compounds in p38 kinase binding and kinase activity. Expression of p38 kinase in yeast, purification of the active form, and the peptide phosphorylation kinase assay were as detailed under "Experimental Procedures." Competition binding to p38 kinase in cytosol of THP.1 cells was performed as described previously (6). Kinase and binding IC 50 values were calculated from % inhibition data obtained using serial dilutions of each compound. related but inactive compound, has no effect on either the activity or the activation of p38 kinase or MAPKAP kinase-2 (6,35). DISCUSSION We have shown that the pyridinyl imidazoles typified by SB203580 inhibit the catalytic activity of p38 kinase by binding to the ATP site. This immediately raises questions about why the inhibitors are as selective in vitro and in vivo as they appear to be (20,27,35). Part of the explanation may be the high K m of p38 kinase for ATP (200 M), which is considerably higher than several other protein kinases, such as cAMP-dependent protein kinase (K m ϭ 10 M (43)). Hence p38 kinase will be more sensitive to inhibition in the presence of in vivo concentrations of ATP, which are estimated at 2-3 mM, and this may contribute to its in vivo protein kinase selectivity. However, this does not explain the in vitro selectivity, especially with respect to the highly related ERK2, which has a similar K m in vitro (K m ϭ 190 -350 M depending on substrate (41)).
The published x-ray crystal structures of the unactivated forms of ERK2 and p38 show an open conformation that does not create a competent active site (44,45). Comparison of these structures to the active structure of cAMP-dependent protein kinase (46,47) suggests that the phosphorylated and active form of p38 kinase will differ substantially in conformation from the unactivated form. However, a unique feature of SB203580 and related inhibitors is their ability to bind to both the unactivated and activated forms of p38 kinase. Because the binding affinity of pyridinyl imidazoles to unactivated p38 kinase correlates with inhibition of the catalytic activity of activated p38 kinase, this suggests that the inhibitors must bind similarly to both forms of the enzyme. Indeed, the binding affinities of individual inhibitors to the two forms of the enzyme appear comparable. It is thus possible that the inhibitors lock p38 kinase into the conformation seen in the unactivated, nonphosphorylated kinase and hence prevent it from adopting the active conformation.
It is not known whether binding to the unactivated form itself plays a role in the biological effects of the inhibitors. Based on evidence presented here, SB203580 does not prevent Wild type and mutant p38 kinases were expressed in ⌬hog1 S. cerevisiae, and the cells were lysed as described under "Experimental Procedures." CSBP1 is an internal splice variant of p38 kinase (CSBP2). All samples were analyzed by SDS-polyacrylamide gel electrophoresis. p38 kinase was immunoprecipitated with a rabbit polyclonal antiserum to p38 kinase and assayed for kinase activity in vitro using myelin basic protein as a substrate, whereas levels of phosphotyrosine or p38 kinase in immunoprecipitates were determined by Western blot using the monoclonal antibody PY20 or polyclonal rabbit anti-p38 kinase, respectively. Other portions of the same lysates were incubated with 125 I-SB206718 and 125 I-SB227931 and UV cross-linked as shown in the bottom two panels. The kinase activity and cross-linking were analyzed by autoradiography, whereas the Western blots were analyzed by ECL.

FIG. 8. Lack of inhibition of p38 kinase activation by SB203580.
A, HeLa cells were activated for 30 min with 0.4 M sorbitol prior to 30 min pretreatment with different concentration of SB203580 or SB202474 as indicated. Cells were lysed and an immune complex kinase assay for p38 kinase or MAPKAP kinase-2 was performed as described. Aliquots of the kinase assay samples were also immunoblotted with anti-phosphotyrosine and anti-p38 kinase antibody. IgG is the heavy chain of the anti-p38 kinase antibody used for immunoprecipitation. B, graphical representation of data in A showing the dosedependent inhibition of activation of MAPKAP Kinase-2 but not of p38 kinase by SB203580. activation by endogenous MAP kinase kinases because tyrosine phosphorylation of p38 kinase is unaffected, and the isolated enzyme, washed free of inhibitor, is fully active. This is consistent with the recent finding that both Thr 180 and Tyr 182 are surface-exposed in the x-ray crystal structure of unactivated p38 kinase (45). However, the present experiments do not address whether the observed preformed complexes of unactivated p38 with MAPKAP kinases 2 and 3 (12) might influence the way in which the inhibitors work in vivo. Our experiments have only addressed the behavior of free p38 kinase and peptide substrate, although qualitatively similar data have been obtained with myelin basic protein as substrate (data not shown). This will need to be investigated further.
One distinction between ERK2 and p38 kinase revealed by x-ray crystallography is the conformation of the activation loop that contains the Thr and Tyr that are phosphorylated by MAP kinase kinases (44,45). In p38 kinase, the shorter activation loop lies in the putative substrate peptide channel with the 125 I-SB206718 cross-linking site, Thr 175 , close to the ATP binding site (45). This suggests that the compound binds with the azide group pointing out through the open end of the ATP site and the pyridinyl and phenyl groups pointing inward. This is consistent with structure-activity data, which indicate that the azide group points away from the active pharmacophore and is in a region where several chemical groups are tolerated (27).
The present experiments only provide limited data regarding the critical residues for 125 I-SB206718 binding. Alteration of the highly conserved K53 and D168, both of which are normally required for catalytic activity, disrupt binding. However, without further structural data, it is not clear exactly how to interpret this result. The mutated residues might themselves interact directly with the inhibitors, they might occlude nearby interactions, or they might cause subtle local conformational changes. Certainly it is hard to see how these two residues could provide the selectivity of the inhibitors observed with respect to other protein kinases and especially MAP kinases (20,27,35). It is likely that other residues in or near the ATP pocket might be involved. For example, there are differences in several noncatalytic residues in the ATP binding site in addition to the different activation loop residues and conformation noted above. Some recent reports of the structures of other protein kinases complexed with inhibitors have provided direct evidence for the recognition of noncatalytic and non-ATP-binding residues by inhibitors (48 -50). The residues in p38 kinase responsible for binding the pyridinyl imidazoles are currently being investigated.
The recent discovery of several potent inhibitors of protein kinases that retain some measure of specificity in vitro and in vivo, despite competing at the ATP site (48 -50), bodes well for the development of therapeutic agents directed toward blocking protein kinases (e.g. see Ref. 11). Further understanding of the basis for this specificity will come from an understanding of the detailed interactions between enzyme and inhibitors through x-ray structural studies and further mutagenesis of p38 kinase.