Pim-1 Ligand-bound Structures Reveal the Mechanism of Serine ...

THE JOURNAL OF BIOLOGICAL CHEMISTRY © 2005 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 280, No. 14, Issue of April 8, pp. 13728 –13734, 2005 Printed in U.S.A.

Pim-1 Ligand-bound Structures Reveal the Mechanism of Serine/Threonine Kinase Inhibition by LY294002* Received for publication, November 22, 2004, and in revised form, January 5, 2005 Published, JBC Papers in Press, January 17, 2005, DOI 10.1074/jbc.M413155200

Marc D. Jacobs‡, James Black, Olga Futer, Lora Swenson, Brian Hare, Mark Fleming, and Kumkum Saxena From Vertex Pharmaceuticals Incorporated, Cambridge, Massachusetts 02139

Pim-1 is an oncogene-encoded serine/threonine kinase primarily expressed in hematopoietic and germ cell lines. Pim-1 kinase was originally identified in Maloney murine leukemia virus-induced T-cell lymphomas and is associated with multiple cellular functions such as proliferation, survival, differentiation, apoptosis, and tumorigenesis (Wang, Z., Bhattacharya, N., Weaver, M., Petersen, K., Meyer, M., Gapter, L., and Magnuson, N. S. (2001) J. Vet. Sci. 2, 167–179). The crystal structures of Pim-1 complexed with staurosporine and adenosine were determined. Although a typical two-domain serine/ threonine protein kinase fold is observed, the interdomain hinge region is unusual in both sequence and conformation; a two-residue insertion causes the hinge to bulge away from the ATP-binding pocket, and a proline residue in the hinge removes a conserved main chain hydrogen bond donor. Without this hydrogen bond, van der Waals interactions with the hinge serve to position the ligand. The hinge region of Pim-1 resembles that of phosphatidylinositol 3-kinase more closely than it does other protein kinases. Although the phosphatidylinositol 3-kinase inhibitor LY294002 also inhibits Pim-1, the structure of the LY294002䡠Pim-1 complex reveals a new binding mode that may be general for Ser/ Thr kinases.

The Pim-1 oncogene was first identified as the preferred site for integration of the slow transforming Maloney murine leukemia virus in lymphoblastic T-cells (1). Direct evidence for the oncogene potential of the Pim-1 gene comes from the study of transgenic mice in which overexpression of Pim-1 produces a low but spontaneous rate of tumor incidence (2). These mice are highly susceptible to chemical carcinogens, x-ray radiation, and Maloney murine leukemia virus-induced lymphomagenesis. Pim-1 knock-out mice did not show any obvious phenotype, suggesting in vivo functional redundancy of this highly conserved oncogene (3). Since the initial report of the cloning of mouse Pim-1 gene (4), Pim-1 has been cloned from human, rat, bovine, and zebrafish cDNA libraries (1). In humans, the Pim-1 gene is expressed mainly in the developing fetal liver and spleen (5) and in hematopoietic malignancies (6, 7). Two homologs of the Pim-1 gene, pim-2 (8) and pim-3/kid-1 (9), have also been identified. * This work was supported in part by the United States Department of Energy under Contract DE-AC03-76SF00098 at Lawrence Berkeley National Laboratory. 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. ‡ To whom correspondence should be addressed: 130 Waverly St., Cambridge, MA 02139. Tel.: 617-444-6451; Fax: 617-444-6566; E-mail: marc_ [email protected].

The human Pim-1 gene encodes a 313-amino acid serine/ threonine kinase (10) and is associated with multiple cellular functions such as proliferation, differentiation, apoptosis, and tumorigenesis (1). Several cellular substrates of Pim-1 have been identified, including the transcription factors cMyb (11) and NFATc1 (12), the transcriptional co-activator of cMyb p100 (13), phosphatases Cdc25A (14) and PTPU2S (15), Pim-1-associated protein 1 (16), cell cycle inhibitor p21/WAF1 (17), heterochromatin protein 1 (18), TRAF2/SNX6 (19), and nuclear mitotic apparatus (20). The consensus sequence for Pim-1 substrate recognition is Lys/Arg-Lys/Arg-Arg-Lys/Arg-Leu-Ser/ Thr-Xaa, where Xaa is an amino acid with a small side chain (21). The expression of Pim-1 is induced by a number cytokines, mitogens, and hormones (reviewed in Ref. 1). The JAK/STAT (22, 23), AKT (24), mitogen-activated protein kinase, and phosphatidylinositol 3-kinase (PI3K)1 (23) pathways may all mediate Pim-1 expression. The x-ray structure of Pim-1 was pursued, in part, to determine how an unusual sequence feature in the active site affects ligand binding. In protein kinases, one hydrogen bond between the ATP (N1 atom) and a protein main chain NH is highly conserved. In the Pim-1 sequence, however, a proline (Pro123) occupies this position, so the main chain amide nitrogen is not available to participate in a hydrogen bond. A proline at this position is extremely rare; in fact, no other kinases with known structure have a similarly placed proline. Thus, the hydrogen bond to N1 of ATP is not necessary for substrate binding or catalysis in these kinases, and other interactions are sufficient to correctly position ATP. Qian et al. (25) have determined the structures of both unliganded Pim-1 and the Pim-1䡠AMP-PNP complex. This work showed the mechanism by which Pim-1 is constitutively active and the manner in which an ATP analog is bound in the absence of the one of the main chain hydrogen bond donors. In our study, two kinase inhibitors, staurosporine and adenosine, both of which accept a hydrogen bond from this main chain NH in other kinase structures, were chosen for co-crystallization with Pim-1. The conformations of these ligands further elucidate the interactions needed for ligand positioning in the absence of a conserved hydrogen bond. In the course of determining the Pim-1 structure, similarities with the active site of PI3K were observed. The PI3K inhibitor, LY294002, was also found to inhibit Pim-1; we determined this additional co-complex structure to better understand the mechanisms by which LY294002 inhibits protein kinases. EXPERIMENTAL PROCEDURES

Cloning and Expression of Pim-1—Full-length Pim-1 (residues Met1– Lys313) was cloned in two parts by PCR from a human IMAGE Consor1 The abbreviations used are: PI3K, phosphatidylinositol 3-kinase; DTT, dithiothreitol; GST, glutathione S-transferase; PKA, cAMP-dependent kinase; AMP-PNP, adenosine 5⬘-(␤,␥-iminotriphosphate); CK2, casein kinase 2.

13728

This paper is available on line at http://www.jbc.org

Structure of Pim-1 Kinase

13729

TABLE I IC50 values of kinase inhibitors determined for Pim-1

FIG. 1. Purification of Pim-1. The ion exchange chromatography (MonoQ) elution profile is shown as a function of fraction number, with the eluate conductance drawn as an overlay. tium clone (GenBankTM accession number GI 1845036) and from a human bone marrow cDNA library (BD Biosciences, Clontech, Palo Alto, CA). The pieces were fused by PCR and inserted into the NdeI and EcoRI sites of the dual promoter vector pBEV1, encoding a protein with an N-terminal His6 tag and thrombin cleavage site (26). The amino acid sequence of this Pim-1 clone is identical to SwissProt entry P11309. BL21/DE3 pLysS Escherichia coli cells were transformed with the construct encoding full-length human Pim-1 kinase, using a standard transformation protocol (Stratagene, La Jolla, CA). Freshly transformed cells were grown at 37 °C in brain heart infusion medium (DIFCO Laboratories, Detroit, MI) supplemented with 100 ␮g/ml carbenicillin and 35 ␮g/ml chloramphenicol. The cells were grown at 37 °C up to an optical density of 0.75 at 600 nm, and expression was induced at 28 °C with 1 mM isopropyl ␤-D-thiogalactopyranoside. The cells were harvested via centrifugation 4 h post-induction and stored at ⫺80 °C prior to purification. Protein Purification—Frozen cell pellets (⬃30 g) were thawed in 7 volumes of Buffer A (50 mM HEPES, pH 7.8, 300 mM NaCl, 10% (v/v) glycerol, 3 mM ␤-mercaptoethanol) containing 0.1% (v/v) Tween 20, 50 ␮M diisopropyl fluorophosphate, 1 ␮g/ml E-64, 1 ␮g/ml leupeptin, and 10 ␮g/ml pepstatin (Roche Applied Science) and lysed in a microfluidizer (Microfluidics, Newton, MA). The lysate was centrifuged at 54,000 ⫻ g for 45 min, and the supernatant was incubated with 1 ml of Talon metal affinity resin (BD Biosciences, Clontech)/5 mg of protein overnight at 4 °C. The resin was washed with 20 column volumes of Buffer A and eluted with Buffer A containing 100 mM imidazole. Fractions containing Pim-1 were pooled and concentrated by ultrafiltration using a 30-kDa molecular mass cut-off membrane in an Amicon stirred cell concentrator (Millipore, Billerica, MA). The Pim-1 fraction was loaded onto a Superdex 200 column (90 ⫻ 2.6 cm; Amersham Biosciences) equilibrated in Buffer B (50 mM HEPES, pH 7.8, 200 mM NaCl, 10% (v/v) glycerol, and 5 mM ␤-mercaptoethanol). The fractions were pooled based on SDS-PAGE, diluted to 25 mM NaCl with 50 mM HEPES, pH 7.8, 10% (v/v) glycerol, and 5 mM dithiothreitol (DTT), and loaded onto a 8-ml prepacked MonoQ (HR 10/10) anion exchange column (Amersham Biosciences) equilibrated in Buffer C (50 mM HEPES, pH 7.8, 20 mM NaCl, 10% (v/v) glycerol, 5 mM DTT). Pim-1 was eluted using a gradient of 0 – 40% Buffer D (buffer C plus 1 M NaCl) over 60 column volumes. Peak fractions were collected as four separate pools (I–IV) based on the elution chromatogram. The Pim-1 fraction was dialyzed into 20 mM Tris, pH 8.0 (25 °C), 100 mM NaCl, 5 mM DTT and concentrated to 10 mg/ml using a 10-kDa molecular mass cut-off Vivaspin concentrator (Vivascience, Hanover, Germany). The identity of the purified Pim-1 was confirmed by N-terminal amino acid sequencing. Typically, preparations contained a mixture of species with 0 –5 phosphoryl groups, which were partially resolved by anion exchange chromatography (Fig. 1). Phosphoamino acid analysis showed that Pim-1 purified from E. coli was extensively phosphorylated in the His6 tag (MGSSHHHHHHSSGLVPRGSH), and the four MonoQ pools differed mainly in the degree of phosphorylation in this region. Dephosphorylation of Pim-1 with Lambda phosphatase (New England Biolabs) followed by autophosphorylation showed that Pim-1 readily autophosphorylates in the His6 tag region (data not shown). Ser261 was the major

phosphorylation site observed in Pools III and IV. Other minor sites, Ser8, Thr23, and Ser98, were phosphorylated to varying degrees in each pool. Pim-1 crystallized from different MonoQ pools yielded similar crystal forms. The phosphorylation state of each of the MonoQ-purified pools I-IV of Pim-1 was determined by electrospray mass spectrometry of thrombincleaved Pim-1. Spectra were collected using a Micromass Quattro II triple quadrupole mass spectrometer (Waters Corp., Milford, MA) (27). Phosphorylation sites were identified from tryptic digests of the MonoQpurified pools I–IV to liquid chromatography-tandem mass spectrometry on a QSTAR Pulsar quadrupole time-of-flight tandem mass spectrometer (AB/MDS-Sciex, Toronto, Canada) equipped with a nanoelectrospray ion source (MDS Protana, Odese, Denmark). The data were analyzed using the Mascot search engine (Matrix Science, London, UK).

13730

Structure of Pim-1 Kinase TABLE II Data collection and refinement statistics Data set

Data collection X-ray source Space group Unit cell parameters (Å) Resolution (Å) Unique reflections Redundancy Completeness (%)a Rmergea,b 具I/␴a典 Refinement Reflections used Test reflections R factor Free R factor (% data) Root mean square deviation Bond lengths (Å) Bond angles (°) Dihedral angles (°) Protein atoms Solvent atoms

Staurosporine

Adenosine

LY294002

Rigaku RU-H3R P65 a ⫽ b ⫽ 97.7 c ⫽ 80.5 20–2.15 22615 3.6 94.9 (74.8) 0.050 (0.250) 10.6 (2.3)

ALS 5.0.2 P65 a ⫽ b ⫽ 98.3 c ⫽ 80.4 20–2.4 16430 5.2 94.3 (96.1) 0.060 (0.361) 14.7 (3.9)

Rigaku RU-H3R P65 a ⫽ b ⫽ 97.7 c ⫽ 80.7 20–2.5 14445 3.1 94.9 (87.6) 0.072 (0.336) 12.0 (2.6)

22526 1706 0.205 0.233 (7.6)

16152 1268 0.210 0.246 (7.9)

14206 1097 0.208 0.259 (7.7)

0.015 1.7 23.1 2202 142

0.007 1.3 22.8 2202 81

0.009 1.2 22.2 2202 136

a

The values for the highest resolution shell are shown in parentheses. Rmerge ⫽ ⌺hkl ⌺i 兩I(hkl)i ⫺ 具I(hkl典兩/⌺hkl ⌺i具I(hkl)i典 over i observations of reflection hkl. R factor ⫽ ⌺储Fobs兩 ⫺ 兩Fcalc储/⌺兩Fobs兩, where Fobs and Fcalc are the observed and calculated structure factors, respectively. Free R factor is calculated from a randomly chosen subset of reflections not used for refinement. b

Kinase Assays—A coupled-enzyme assay (28) was used to quantify the ADP generated in the kinase reaction with S6 peptide (RRRLSSLRA) (American Peptide Co, Sunnyvale, CA) as the phosphoacceptor substrate. The assay was carried out in a total volume of 100 ␮l in 0.1 M HEPES, pH 7.6, containing 10 mM MgCl2, 2.5 mM phosphoenolpyruvate, 0.2 mM NADH, 30 ␮g/ml pyruvate kinase, 10 ␮g/ml lactate dehydrogenase (Roche Applied Science), and 2 mM DTT in a 96-well plate and read at 340 nm at 30 °C on a Spectramax spectrophotometer (Molecular Devices, Sunnyvale, CA). Pim-1 concentration was 25 nM in all assays. The reaction was started by the addition of ATP after 10 min of preincubation of the reaction mixture at 30 °C. Substrate concentrations were 1 mM S6 peptide, 2 mM ATP for activity assays and 40 ␮M S6 peptide, 100 ␮M ATP for IC50 determinations. Inhibitors were dissolved in Me2SO and added to the reaction to 2.5% Me2SO final at the beginning of preincubation period. Kinetic analysis was performed by nonlinear regression fitting using the program Prism (GraphPad Software, San Diego, CA) (Table I). Crystallization and X-ray Analysis—Pim-1 crystals were grown by the vapor diffusion method at 22 °C. Equal volumes of protein (12 mg/ml protein, 20 mM HEPES, pH 8.0, 100 mM NaCl, 5 mM DTT) and well solution (1 M (NH4)2HPO4, 100 mM citrate buffer pH 5.5, 200 mM NaCl) were mixed and suspended over 1 ml of well solution. Over 4 days, the crystals reached a final size of ⬃250 ⫻ 40 ⫻ 40 ␮m. Crystals were harvested and flash-frozen in a solution composed of the well solution with 30% (v/v) glycerol. A complex of Pim-1 with either staurosporine (Sigma-Aldrich) or the inhibitor LY294002 (Calbiochem, La Jolla, CA) was made by soaking unliganded crystals (grown as above) with 500 ␮M compound and 5% Me2SO (final concentration) for 24 h at room temperature. The adenosine䡠Pim-1 complex was made by adding adenosine (2 mM) to the protein prior to crystallization. For the staurosporine and LY294002 complexes, x-ray diffraction data were recorded using a RU-200 x-ray generator and RaxisV⫹⫹ detector (Rigaku, The Woodlands, TX), and the intensities were integrated and scaled using CrystalClear (29). Diffraction data for the adenosine complex crystals were recorded at Beamline 5.0.2 at the Advanced Light Source (Lawrence Berkeley Laboratories, Berkeley, CA). The intensities were integrated and scaled using the programs DENZO and SCALEPACK (30) and CrystalClear. The structure was determined by molecular replacement using homology models based upon phosphorylase kinase (Protein Data Bank code 1PHK) (31) and death-associated protein kinase (Protein Data Bank code 1JKK) (32). The molecular replacement solution was determined using AMORE (Navaza, CCP4 distribution) (33). The crystals belong to the space group P65, and a single protein monomer comprises the asymmetric unit. The protein model was built using QUANTA (Accelrys, San Diego, CA) and refined with both CNX (Accelrys) (34) and BUSTER (Global Phasing Inc., Cambridge, UK) (35) (Table II).

The refined model consists of the protein kinase catalytic domain. Although full-length protein was used for crystallization (313 residues), 32 residues at the N terminus, 8 residues at the C terminus, and 4 residues in one loop (80 – 83) were not sufficiently ordered to be built into the electron density. Phosphorylation of Ser261 is clearly visible in the electron density map. The phosphoserine side chain participates in both intra- and intermolecular interactions and may be important in formation of the crystal packing interactions. Also, the electron density map reveals additional density adjacent to the sulfur of Cys161, indicating an adduct at this residue. The electron density was large enough to accommodate four nonhydrogen atoms; it was modeled as a ␤-mercaptoethanol adduct; however, it is also consistent with a partially ordered DTT adduct. Both DTT and ␤-mercaptoethanol were used in the purification. RESULTS AND DISCUSSION

Enzymatic Activity—Kinase activity assays of the MonoQ pools I–IV were performed to determine whether the observed phosphorylation affected the catalytic activity. All four pools showed very similar kinetic parameters (kcat ⫽ 4 ⫾ 0.4 s⫺1, peptide Km ⫽ 51 ⫾ 2 ␮M, and ATP Km ⫽ 120 ⫾ 16 ␮M). Pim-1 purified from E. coli was phosphorylated at Ser261 as well as multiple sites in the His tag region. Palaty et al. (36) have identified Ser190 in Xenopus Pim-3 as the major autophosphorylation site and showed that S190A and S190E mutants are 7-fold less active than the wild type Pim-3. The equivalent residue in human Pim-1, Ser189, was not phosphorylated in the E. coli purified preparations. The fact that all four MonoQ Pim-1 pools exhibit very similar kinetic parameters indicates that the enzyme is constitutively active and that the phosphorylation state does not affect enzymatic activity. The specific activity (5 ⫾ 0.2 ␮mol/min/mg) observed here is much higher than previously reported (21, 36 –38). It is 60-fold greater than that reported by Friedman et al. (21) for human GST-Pim-1 using a histone H1 peptide (KRRASGP) and over 104-fold greater than that reported by Palaty et al. (38) for GST fusions of human Pim-1 using S6 peptide (AKRRRLSSLRA). Because both studies utilized GST fusions for expression and purification, it is possible that this large protein tag had a detrimental effect on enzyme activity, either by interfering with substrate access to the active site or with overall protein folding.

Structure of Pim-1 Kinase A panel of kinase inhibitors was evaluated for their ability to inhibit Pim-1. Staurosporine and structurally similar compounds, such as K-252a and bisindolyl-maleimides I and IX, were found to inhibit Pim-1 with submicromolar potency (Table I). These compounds are nonspecific inhibitors of Ser/Thr and Tyr kinases (39). Surprisingly, LY294002 was found to be an inhibitor of Pim-1 with IC50 ⫽ 4 ␮M. This compound was originally described as a specific inhibitor of PI3K with 1.4 ␮M IC50 (40). Later, Davies et al. (39) reported that LY294002 inhibits PI3K and casein kinase 2 with a similar potency (10 and 6.9 ␮M, respectively). Protein Structure—The structure of Pim-1 has a global fold typical of protein serine/threonine kinases, consisting of two domains linked by a hinge region (Fig. 2). The smaller, Nterminal domain (residues 33–121) consists primarily of ␤-strands with one ␣-helix, and the C-terminal domain (residues 128 –305) is largely ␣-helical. The active site is formed by a groove at the interface between these two domains and is enclosed by the hinge region (residues 122–127), the glycine rich loop (residues 44 –52), and the activation loop (residues 186 –210). The Pim-1 structure was compared with several other protein kinases with high sequence homology such as cAMP-dependent kinase (PKA) and phosphorylase kinase. When secondary structural elements are aligned, a root mean square deviation of 1.3 Å for C-␣ atom positions is observed between Pim-1 and both PKA or phosphorylase kinase (using 213 residues from phosphorylase kinase (Protein Data Bank code 1PHK) (31) and 220 residues from cAMP-dependent kinase (Protein Data Bank code 1ATP) (41), respectively). The Pim-1 protein structures described here are similar to unliganded Pim-1 (25), with a root mean square deviation of 0.4 Å for C-␣ atoms. The conformation of the glycine-rich loop (residues 45–52) in this structure differs from that of the PKA structures and most closely resembles that of unliganded Pim-1 (25). The Pim-1 glycine-rich loop moves toward the C-terminal domain, and Phe49 adopts a rotamer in which the side chain points toward the hinge region, thereby filling the space usually occupied by ATP phosphates (Fig. 2). In the Pim-1䡠AMP-PNP complex structure, the ligand phosphates displace Phe49, and the loop adopts a more open conformation. This is similar to GSK-3␤, for instance, where the corresponding phenylalanine residue is observed both within the active site pointing toward the hinge (peptide complex) (42) and, in another structure (unliganded), outside the active site, pointing away from the hinge (43). The Pim-1 hinge sequence is unusual because of a tworesidue insertion relative to kinases such as CDK-2 (44) and JNK-3 (45) and a single residue insertion relative to PKA. A comparison of the hinge regions of Pim-1 and PKA is shown in Fig. 3. Residues before and after the insertion superimpose well; however, at the point of insertion (Pro125), the hinge bulges away from the ATP-binding site by up to 4 Å. Some of the additional space created by the change in main chain position is occupied by the Val126 side chain, which is oriented toward the ATP-binding pocket and interacts with Pro123. This unique hinge conformation could be utilized for the design of specific Pim-1 inhibitors and creates a space for substitution at the position corresponding to C-2 of ATP. For instance, polar interactions with the carbonyl oxygen of Pro123 or hydrophobic contacts with the side chain of Val126 would be unique to PIM. One kinase sharing this two-residue insertion is PI3K (46). Although there is little sequence homology in the hinge region between PI3K and Pim-1, the main chain conformations are remarkably similar (0.86 Å root mean square deviation over 13 C-␣ positions). The PI3K and Pim-1 hinge conformation differ most at Pro125 (Asp884 in PI3K) (Fig. 4).

13731

FIG. 2. Overall structure of the Pim-1䡠staurosporine complex. The structure is shown with ␤-sheets as arrows and the ␣-helices as cylinders. The N-terminal domain (dark blue) is shown with the glycinerich loop drawn in green. The hinge connecting the two domains is colored orange. The C-terminal domain is shown in light blue with the activation loop shown in yellow. Staurosporine (red) is shown in the active site, bound between the Phe49 side chain (colored gray in the glycine-rich loop) and the hinge region. The salt bridge stabilizing the conformation of the activation loop is formed by residues Asp200 and Arg166 side chains, drawn with gray carbon atoms. The site of phosphorylation, Ser261, is shown. All of the structural figures were prepared with Pymol (54).

Adenosine and Staurosporine Ligand Structures—The positions of staurosporine and adenosine bound to Pim-1 are similar to those observed in other protein kinase complex structures. Both compounds are sandwiched between hydrophobic residues from the glycine rich loop, the C-terminal domain, and the hinge. In each case, only one hydrogen bond between the ligand and the hinge is observed. In the Pim-1䡠adenosine complex, a hydrogen bond is observed between the Glu121 main chain carbonyl and the N-6 amino group. The pyrrolidinone nitrogen of staurosporine forms a similar hydrogen bond. The presence of proline at position 123 prevents each of the two compounds from forming the second hydrogen bond to the hinge. The Pim-1 ligand structures were aligned and compared with PKA. Relative to the PKA-adenosine complex (Protein Data Bank code 1FMO) (47), the adenosine in Pim-1 rotates by ⬃20° toward the hinge into the additional space formed by the proline insertion (rotation axis perpendicular to the plane of the adenine ring; see Fig. 3B). The purine rings of adenosine in the Pim-1 and PKA structures are approximately co-planar. The relative position of the adenosine in the two structures is, in part, fixed by the length of the side chain to which the sugar moiety forms a hydrogen bond (Asp128 in Pim-1 and Glu128 in PKA). Likewise, the aromatic rings of staurosporine in Pim-1 and PKA (Protein Data Bank code 1STC) (48) are approximately co-planar. Staurosporine in Pim-1 rotates toward the hinge by about 10° and also forms a hydrogen bond to Asp128 (Fig. 3A). A comparison of the ligand positions in the Pim-1 and the PI3K complexes (Protein Data Bank codes 1E8Z and 1E8X) (46, 49) reveals a shift and a rotation. In PI3K, two hydrogen bonds are made between the main chain (Val882 NH and Glu880 O)

13732


FIG. 3. Ligand interactions with the hinge region: comparison with PKA. Structures of Pim-1 and PKA bound to staurosporine and adenosine were aligned to optimize the superposition of residues adjacent to the hinge regions. The Pim-1 structure is colored by atom type and labeled in black, and PKA is drawn and labeled in pink. Hydrogen bonds are depicted as green dotted lines. The view is rotated ⬃90° from Fig. 2. In this orientation, the glycine-rich loop lies above and in the plane of the page. A, the structures of the PKA䡠staurosporine (Protein Data Bank code 1STC) and Pim-1䡠staurosporine complexes. B, Pim-1䡠adenosine and PKA complexes (Protein Data Bank code 1FMO). C, sequence alignments of hinge regions. Residues that accept and donate hydrogen bonds to the adenine ring of ATP are highlighted in red and blue, respectively.

and the ligands (adenosine N1 and N6 and staurosporine pyrrolidinone N and O). Adenosine and staurosporine bound to PI3K are shifted toward the outermost edge of the hinge by about 2.5 Å relative to the Pim-1 structure (Fig. 4). Also, the staurosporine is tilted about 30° about an axis parallel to the main chain of the hinge (between Ile879 and Val882), such that the pyrrolidinone ring lies below (toward the C-terminal domain) the same ring in the Pim-1 structure (Fig. 4A). Although the hinge conformations between Pim-1 and PI3K are very similar, the positions of adenosine and staurosporine differ. In fact, the orientations of the ligands in Pim-1 more closely resemble those found in other protein kinases. The PI3K binding mode, characterized by the shift toward the hinge and tilt toward the C-terminal domain, is sterically hindered in Pim-1. The presence of the C-␦ atom of Pro123 and the larger side chain at position 126 (valine instead of alanine) prevent the shift toward the hinge. The tilt toward the C-terminal domain is hindered by the side chain of Ile104 in Pim-1. In the absence of the conserved pair of hydrogen bonds to the hinge, a number of van der Waals contacts constrain the position of the ligands. LY294002 Complex—The structure of PI3K inhibitor LY294002 bound to Pim-1 (Fig. 5A) was pursued based upon the observation of the inhibitory activity of the compound in the Pim-1 in vitro assay, as well as the conformational similarity between the Pim-1 and PI3K hinges. When bound to PI3K, the morpholine oxygen of LY294002 accepts a hydrogen bond from the amide nitrogen of Val882, making the same interaction as seen with N1 of ATP (Protein Data Bank code 1E7V) (Fig. 5B).

The structure of the Pim-1䡠LY294002 complex reveals that the compound orientation is very different. Relative to the PI3K structure, the LY294002 compound rotates about 180° about the bond common to the two rings in the chromone group. In this case, the only interaction with the hinge is between the main chain carbonyl of Glu121 and two aromatic hydrogens of the chromone rings. The chromone carbonyl oxygen makes a hydrogen bond to a water molecule, which in turn interacts with the main chain amide of Asp186. The phenyl group of LY294002 packs against the side chains of Arg122, Val126, and Leu174, whereas the morpholine group interacts with Phe49 in the glycine-rich loop. The contacts between the Pim-1 hinge and LY294002 are highly unusual. Typically, protein kinase ligands interact with the hinge via hydrogen bonds, where the donor hydrogen is bonded to either oxygen or nitrogen. In this case, only hydrogens bonded to aromatic carbon atoms interact with the hinge. If indeed these interactions are important for LY294002 binding, we would expect the arrangement of the atoms to be favorable for hydrogen bonding. The ideal (C)H to O distance is ⬃2.6 –2.7 Å, and the distance between the Glu121 carbonyl and the LY294002 hydrogens is 2.6 and 2.9 Å (50). The ideal O-CH angle is between 90 and 180°. The angles observed with LY294002 in Pim-1 are 140 and 130°. Further, the hydrogen and the amide bond should be co-planar, which is the case in the Pim-1-LY294002 complex. It is likely, therefore, that a pair of aromatic CH hydrogen bonds are formed between LY294002 and the Pim-1 hinge.


13733

FIG. 4. Ligand interactions with the hinge region: comparison with PI3K. Structures of Pim-1 and PI3K bound to staurosporine, adenosine, and ATP were aligned to optimize the superposition of residues adjacent to the hinge regions. The Pim-1 structure is colored by atom type and labeled in black type, and PI3K is drawn and labeled in blue. A, Pim1䡠staurosporine and PI3K䡠staurosporine complexes (Protein Data Bank code 1E8Z). B, Pim-1䡠adenosine and PI3K䡠ATP complexes (Protein Data Bank code 1E8X).

FIG. 5. Binding mode of LY294002. The structures are drawn such that the glycine-rich loop would lie above and in the plane of the page. A, the structure of the Pim-1䡠LY294002 complex is shown. The Fo ⫺ Fc electron density map is drawn around the compound at the 2.5 ␴ level. A water molecule is drawn as a red sphere with hydrogen bonds to the chromene oxygen and the Asp186 amide drawn in purple. B, the PI3K䡠LY294002 complex (Protein Data Bank code 1E7V).

The fact that LY294002 binds to Pim-1, a protein with a PI3K-like hinge, appears to be coincidental, because the compound orientation is dramatically different in the two proteins. Although the proteins have structural similarities, none of the features in common contribute to the binding of LY294002. In fact, superposition of the two complexes reveals that the PI3K binding mode is sterically hindered by Pro123 in Pim-1. Also, the Pim-1 binding mode is incompatible with the PI3K structure; Trp182 in PI3K packs against the phenyl and the morpholine rings of LY294002 but would collide with the phenyl ring if the compound bound in the Pim-1 orientation. The compound LY294002 is commonly used to assess the role of PI3K in cell signaling and does not significantly inhibit most protein kinases (39). For instance, PKA activity is reduced by only 9% (⫾5%) in the presence of 50 ␮M LY294002, so we would not expect the structure of PKA to easily accommodate LY294002 binding. Indeed, both the Pim-1 and PI3K binding modes are sterically hindered by Thr183 and Val123, respectively, in PKA. One kinase inhibited by LY294002 is casein kinase 2 (CK2) (IC50 6.9 ␮M). The structures of CK2 and Pim-1 were aligned to predict how LY294002 might bind to CK2. The

PI3K binding mode is blocked by the side chain of Val116 in CK2. However, the CK2 active site will accommodate LY294002 in the Pim-1 binding mode, with a only 0.5 Å translation to avoid a close contact with Ile66. Interestingly, in addition to kinases, LY294002 has also been observed to bind to proteins with unrelated sequences and functions. In these cases, inhibition by LY294002, but not by wortmannin (another PI3K inhibitor), was observed and indicated a PI3K-independent mechanism. For instance, LY294002 has been shown to alter intracellular calcium concentrations in bronchial smooth muscle cells (51), block the Kv2.1 and Kv1.4 channels (52), and also bind to and inhibit estrogen receptor (53). This promiscuity may be due to the fact that LY294002 is a relatively small, planar, and unelaborated molecule with several hydrogen bonding opportunities. It is possible that there are other as yet unidentified targets of this compound, and therefore LY294002 should be used with caution in cellular assays. Conclusions—In protein kinases, the hinge conformation and the hydrogen bonds to ATP are highly conserved. The Pim-1 structure reveals how a catalytically competent sub-

13734


strate complex can form even when the hinge is unusual in both sequence and conformation. The structures of the adenosine and staurosporine complexes show van der Waals contacts playing the same role as a conserved hydrogen bond to position the substrate. Although the Pim-1 hinge closely resembles the analogous region in the active site of PI3K, the compound LY294002 interacts with the hinges of the two proteins in very different ways. This new binding mode provides a model to show how LY294002 might inhibit other protein kinases, and this structure can be used to aid in the design of specific inhibitors, utilizing unique features of the Pim-1 active site. Acknowledgments—We thank H. Hsiao for preparation of DNA constructs, J. Fulghum for protein expression, M. Pitta for mass spectrometry, X. Liu for analytical ultracentrifugation, and G. McDermott and the staff at Beamline 5.0.2 at the Advanced Light Source for assistance in diffraction data collection. We also thank J. Moore, S. Raybuck, and E. Fox for helpful discussions. REFERENCES 1. Wang, Z., Bhattacharya, N., Weaver, M., Petersen, K., Meyer, M., Gapter, L., and Magnuson, N. S. (2001) J. Vet. Sci. 2, 167–179 2. Domen, J., van der Lugt, N. M., Laird, P. W., Saris, C. J., and Berns, A. (1993) Leukemia 7, (Suppl. 2) S108 –S112 3. Domen, J., van der Lugt, N. M., Acton, D., Laird, P. W., Linders, K., and Berns, A. (1993) J. Exp. Med. 178, 1665–1673 4. Selten, G., Cuypers, H. T., Boelens, W., Robanus-Maandag, E., Verbeek, J., Domen, J., van Beveren, C., and Berns, A. (1986) Cell 46, 603– 611 5. Amson, R., Sigaux, F., Przedborski, S., Flandrin, G., Givol, D., and Telerman, A. (1989) Proc. Natl. Acad. Sci. U. S. A. 86, 8857– 8861 6. Meeker, T. C., Nagarajan, L., ar-Rushdi, A., Rovera, G., Huebner, K., and Croce, C. M. (1987) Oncogene Res. 1, 87–101 7. Nagarajan, L., Louie, E., Tsujimoto, Y., ar-Rushdi, A., Huebner, K., and Croce, C. M. (1986) Proc. Natl. Acad. Sci. U. S. A. 83, 2556 –2560 8. van der Lugt, N. M., Domen, J., Verhoeven, E., Linders, K., van der Gulden, H., Allen, J., and Berns, A. (1995) EMBO J. 14, 2536 –2544 9. Feldman, J. D., Vician, L., Crispino, M., Tocco, G., Marcheselli, V. L., Bazan, N. G., Baudry, M., and Herschman, H. R. (1998) J. Biol. Chem. 273, 16535–16543 10. Padma, R., and Nagarajan, L. (1991) Cancer Res. 51, 2486 –2489 11. Winn, L. M., Lei, W., and Ness, S. A. (2003) Cell Cycle 2, 258 –262 12. Rainio, E. M., Sandholm, J., and Koskinen, P. J. (2002) J. Immunol. 168, 1524 –1527 13. Leverson, J. D., Koskinen, P. J., Orrico, F. C., Rainio, E. M., Jalkanen, K. J., Dash, A. B., Eisenman, R. N., and Ness, S. A. (1998) Mol. Cell 2, 417– 425 14. Mochizuki, T., Kitanaka, C., Noguchi, K., Muramatsu, T., Asai, A., and Kuchino, Y. (1999) J. Biol. Chem. 274, 18659 –18666 15. Wang, Z., Bhattacharya, N., Meyer, M. K., Seimiya, H., Tsuruo, T., Tonani, J. A., and Magnuson, N. S. (2001) Arch. Biochem. Biophys. 390, 9 –18 16. Maita, H., Harada, Y., Nagakubo, D., Kitaura, H., Ikeda, M., Tamai, K., Takahashi, K., Ariga, H., and Iguchi-Ariga, S. M. (2000) Eur. J. Biochem. 267, 5168 –5178 17. Wang, Z., Bhattacharya, N., Mixter, P. F., Wei, W., Sedivy, J., and Magnuson, N. S. (2002) Biochim. Biophys. Acta. 1593, 45–55 18. Koike, N., Maita, H., Taira, T., Ariga, H., and Iguchi-Ariga, S. M. (2000) FEBS Lett. 467, 17–21 19. Ishibashi, Y., Maita, H., Yano, M., Koike, N., Tamai, K., Ariga, H., and Iguchi-Ariga, S. M. (2001) FEBS Lett. 506, 33–38 20. Bhattacharya, N., Wang, Z., Davitt, C., McKenzie, I. F., Xing, P. X., and Magnuson, N. S. (2002) Chromosoma 111, 80 –95 21. Friedmann, M., Nissen, M. S., Hoover, D. S., Reeves, R., and Magnuson, N. S. (1992) Arch. Biochem. Biophys. 298, 594 – 601 22. Nagata, Y., Nagahisa, H., Nagasawa, T., and Todokoro, K. (1997) Leukemia 11, (Suppl. 3) 435– 438 23. Krumenacker, J. S., Narang, V. S., Buckley, D. J., and Buckley, A. R. (2001) J. Neuroimmunol. 113, 249 –259 24. Krishnan, N., Pan, H., Buckley, D. J., and Buckley, A. (2003) Endocrine 20, 123–130

25. Qian, K. C., Wang, L., Hickey, E. R., Studts, J., Barringer, K., Penge, C., Kronkaitis, A., Li, J., White, A., Mische, S., and Farmer, B. (November 3, 2004) J. Biol. Chem. 280, 6130 – 6137 26. Chambers, S. P., Austen, D. A., Fulghum, J. R., and Kim, W. M. (2004) Protein Expression Purif. 36, 40 – 47 27. Fox, T., Fitzgibbon, M. J., Fleming, M. A., Hsiao, H. M., Brummel, C. L., and Su, M. S. (1999) FEBS Lett. 461, 323–328 28. Fox, T., Coll, J. T., Xie, X., Ford, P. J., Germann, U. A., Porter, M. D., Pazhanisamy, S., Fleming, M. A., Galullo, V., Su, M. S., and Wilson, K. P. (1998) Protein Sci. 7, 2249 –2255 29. CrystalClear (1997–2002) An Integrated Program for the Collection and Processing of Area Detector Data, Rigaku/MSC, The Woodlands, TX 30. Otwinoski, Z. (1993) Oscillation Data Reduction Program: Data Collection and Processing (L. Sawyer, N. W. I., and S. Bailey, eds) SERC Daresbury Laboratory, Warrington, UK 31. Owen, D. J., Noble, M. E., Garman, E. F., Papageorgiou, A. C., and Johnson, L. N. (1995) Structure 3, 467– 482 32. Tereshko, V., Teplova, M., Brunzelle, J., Watterson, D. M., and Egli, M. (2001) Nat. Struct. Biol. 8, 899 –907 33. CCP4 (1994) Acta Crystallogr. Sect. D Biol. Crystallogr. 50, 760 –763 34. Rice, L. M., and Brunger, A. T. (1994) Proteins 19, 277–290 35. Roversi, P., Blanc, E., Vonrhein, C., Evans, G., and Bricogne, G. (2000) Acta Crystallogr. Sect. D Biol. Crystallogr. 56, 1316 –1323 36. Palaty, C. K., Kalmar, G., Tai, G., Oh, S., Amankawa, L., Affolter, M., Aebersold, R., and Pelech, S. L. (1997) J. Biol. Chem. 272, 10514 –10521 37. Hoover, D., Friedmann, M., Reeves, R., and Magnuson, N. S. (1991) J. Biol. Chem. 266, 14018 –14023 38. Palaty, C. K., Clark-Lewis, I., Leung, D., and Pelech, S. L. (1997) Biochem. Cell Biol. 75, 153–162 39. Davies, S. P., Reddy, H., Caivano, M., and Cohen, P. (2000) Biochem. J. 351, 95–105 40. Vlahos, C. J., Matter, W. F., Hui, K. Y., and Brown, R. F. (1994) J. Biol. Chem. 269, 5241–5248 41. Zheng, J., Knighton, D. R., ten Eyck, L. F., Karlsson, R., Xuong, N., Taylor, S. S., and Sowadski, J. M. (1993) Biochemistry 32, 2154 –2161 42. Bax, B., Carter, P. S., Lewis, C., Guy, A. R., Bridges, A., Tanner, R., Pettman, G., Mannix, C., Culbert, A. A., Brown, M. J., Smith, D. G., and Reith, A. D. (2001) Structure (Camb.) 9, 1143–1152 43. ter Haar, E., Coll, J. T., Austen, D. A., Hsiao, H. M., Swenson, L., and Jain, J. (2001) Nat. Struct. Biol. 8, 593–596 44. De Bondt, H. L., Rosenblatt, J., Jancarik, J., Jones, H. D., Morgan, D. O., and Kim, S. H. (1993) Nature 363, 595– 602 45. Xie, X., Gu, Y., Fox, T., Coll, J. T., Fleming, M. A., Markland, W., Caron, P. R., Wilson, K. P., and Su, M. S. (1998) Structure 6, 983–991 46. Walker, E. H., Perisic, O., Ried, C., Stephens, L., and Williams, R. L. (1999) Nature 402, 313–320 47. Narayana, N., Cox, S., Shaltiel, S., Taylor, S. S., and Xuong, N. (1997) Biochemistry 36, 4438 – 4448 48. Prade, L., Engh, R. A., Girod, A., Kinzel, V., Huber, R., and Bossemeyer, D. (1997) Structure 5, 1627–1637 49. Pacold, M. E., Suire, S., Perisic, O., Lara-Gonzalez, S., Davis, C. T., Walker, E. H., Hawkins, P. T., Stephens, L., Eccleston, J. F., and Williams, R. L. (2000) Cell 103, 931–943 50. Pierce, A. C., Sandretto, K. L., and Bemis, G. W. (2002) Proteins 49, 567–576 51. Ethier, M. F., and Madison, J. M. (2002) Cell Calcium 32, 31–38 52. El-Kholy, W., Macdonald, P. E., Lin, J. H., Wang, J., Fox, J. M., Light, P. E., Wang, Q., Tsushima, R. G., and Wheeler, M. B. (2003) FASEB J. 17, 720 –722 53. Pasapera Limon, A. M., Herrera-Munoz, J., Gutierrez-Sagal, R., and UlloaAguirre, A. (2003) Mol. Cell. Endocrinol. 200, 199 –202 54. DeLano, W. L. (2002) The PyMOL Molecular Graphics System, DeLano Scientific, San Carlos, CA 55. Cohen, P. (2002) Nat. Rev. Drug. Discov. 1, 309 –315 56. Dumas, J. (2001) J. Exp. Opin. Ther. Patents 11, 405– 429 57. Hashimoto, Y., Nakayama, T., Teramoto, T., Kato, H., Watanabe, T., Kinoshita, M., Tsukamoto, K., Tokunaga, K., Kurokawa, K., and Nakanishi, S. (1991) Biochem. Biophys. Res. Commun. 181, 423– 429 58. Berg, M. M., Sternberg, D. W., Parada, L. F., and Chao, M. V. (1992) J. Biol. Chem. 267, 13–16 59. Mizuno, K., Saido, T. C., Ohno, S., Tamaoki, T., and Suzuki, K. (1993) FEBS Lett. 330, 114 –116 60. Harris, T. E., Persaud, S. J., and Jones, P. M. (1996) Biochem. Biophys. Res. Commun. 227, 672– 676