Michellamine Alkaloids Inhibit Protein Kinase C

Archives of Biochemistry and Biophysics Vol. 365, No. 1, May 1, pp. 25–30, 1999 Article ID abbi.1999.1145, available online at http://www.idealibrary.com on

Michellamine Alkaloids Inhibit Protein Kinase C E. Lucile White,* ,1 Wan-ru Chao,† Larry J. Ross,* David W. Borhani,‡ Peter D. Hobbs,† Velaparthi Upender,† and Marcia I. Dawson† ,2 *Department of Biochemistry and ‡Department of Organic Chemistry, Southern Research Institute, Birmingham, Alabama 35205; and †Drug Discovery, SRI International, Menlo Park, California 94025

Received November 18, 1998, and in revised form January 29, 1999

Michellamines A, B, and C have shown antiviral activity against HIV-1 and HIV-2 in cell culture. They act in a complex manner by at least two reported antiviral mechanisms, inhibition of HIV reverse transcriptase and inhibition of HIV-induced cellular fusion. On the basis of their structural similarity to other protein kinase C (PKC) inhibitors, we have investigated another possible mechanism—inhibition of PKC. The michellamines were found to inhibit rat brain PKC with IC 50 values in the 15–35 mM range. Michellamine B was a noncompetitive PKC inhibitor with respect to ATP with a K i value of 4 – 6 mM, whereas mixed-type inhibition was observed when the peptide concentration was varied. Michellamine B inhibited the kinase domain of PKC similarly. These results indicate that the michellamines bind to the PKC kinase domain and not its regulatory domain. Molecular modeling showed that all three michellamines can bind in the active site cleft of the PKC kinase domain, to block both the ATP and the peptide substrate subsites. © 1999 Academic Press

Key Words: michellamine; protein kinase C.

The dimeric alkaloids michellamines A, B, and C (Fig. 1) from the liana Ancistrocladus korupensis were initially investigated as antiviral agents because of their activity in cell culture against HIV-1 and HIV-2 (2, 3). Michellamine B is reported to inhibit HIV reverse transcriptase and HIV-induced cellular fusion and syncytium formation (1). We recently discovered that michellamines behave as potent antioxidants (4).

The anti-HIV compound hypericin, which has a perylenequinone ring system, was found to inhibit protein kinase C (PKC) 3 (5). Two other perylenequinones, shiraiachrome-A and calphostin C, also inhibit PKC (6, 7), apparently by binding to the PKC regulatory domain. On the basis of the structural similarity between these compounds and the michellamines, we postulated that michellamine A, B, or C might inhibit PKC activity. Protein kinases are key components of the myriad signal transduction pathways that control many cellular processes and are characterized by their ability to transfer the g-phosphate of ATP or GTP to either an alcohol or phenol group on proteins. PKC is the major receptor for the tumor-promoting phorbol esters and is required for cell proliferation. Thus, effective inhibitors of PKC may have therapeutic potential in cancer treatment and prevention. MATERIALS AND METHODS Michellamine B was obtained from the Drug Synthesis & Chemistry Branch, Development Therapeutics Program, Division of Cancer Treatment, National Cancer Institute (Bethesda, MD). Michellamines A and C were synthesized as we previously reported (8). Rat brain protein kinase C, which contains primarily the a, b, and g isoforms, was from Promega (Madison, WI), and its kinase domain (PKCM) was from Calbiochem (San Diego, CA). Enzyme activity was determined by the transfer of the g-phosphate of ATP to an eight amino acid peptide (1098 Da) that is based on a sequence derived from a natural substrate specific for PKC (kit from Amersham Life Sciences, Arlington Heights, IL). The “mixed micelle” assay mixture contained 50 mM Tris–HCl, pH 7.5, 1.4 mM calcium acetate, 102 mM peptide substrate, 3.4 mM dithiothreitol, 0.11 mM ATP (0.2 mCi [a- 32P]ATP), 0.034 mg/ml phosphatidylserine, 2.73 mg/ml 12-O-tetradecanoylphorbol 13-acetate (TPA), 1 mM ethylenediaminetetraacetic acid, 2 mM ethylene glycol bis(b-aminoethyl ether) N,N,N9,N9-tetraacetic acid, 10 mg/ml phenylmethylsulfonyl

1

To whom correspondence should be addressed at Southern Research Institute, 2000 Ninth Ave. South, Birmingham, AL 35205. Fax: 11(205)5812877. E-mail: [email protected]. 2 Current address: Molecular Medicine Research Institute, 325 East Middlefield Rd., Mountain View, CA 94043. 0003-9861/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved.

3

Abbreviations used: PKC, protein kinase C; TPA, tetradecanoylphorbol 13-acetate; PKA, protein kinase A; H89, N-[2-(4-bromocinnamylamino)ethyl]-5-isoquinoline. 25

26

WHITE ET AL.

FIG. 1. Structures of michellamines A, B, and C.

fluoride, 2 mM benzamidine, and 3% glycerol. The lipid was provided by Amersham in a “detergent-dispersed solution” containing phosphatidylserine and TPA in buffer. The reaction mixture containing 10 ng PKC was incubated at 37°C for 30 min. Reactions were run in duplicate. The assay results were linear for at least 1 h at 37°C and were proportional to the amount of enzyme in the reaction up to 10 ng PKC. The K m values for both the peptide and the ATP substrates of PKC were determined using the nonlinear regression analysis program in the Winzyme software package (Biosoft, Cambridge, MA). The three-dimensional structure of the PKC kinase domain was modeled, using as a basis the 2.3-Å-resolution crystal structure of bovine cAMP-dependent protein kinase (protein kinase A; PKA) complexed to the inhibitor N-[2-(4-bromocinnamylamino)ethyl]-5-isoquinoline (H89; coordinates from Brookhaven protein data bank entry 1ydt) (9). We used this structure as a starting point, rather than that of PKA bound to ATP and a pseudosubstrate inhibitor (entry 1atp; 10) as did Srinivasan et al. (11) and Orr and Newton (12) in their models of PKC, because comparison of these two crystal structures shows that the glycine-rich flap (residues 343–354 in PKC) undergoes a significant shift (away from the ATP subsite) upon H89 binding to PKA. We thus thought the PKA z H89 crystal structure to be a more appropriate model for the conformation of PKC bound to inhibitors. The amino acid sequences of the bovine PKA a catalytic chain and rat PKC a (PIR entries okbo2c and kirtc, respectively) were aligned with GAP (40% identity). Using the graphics program O (13), the coordinates of PKA were transferred to the appropriate residues of PKC. Three loops in PKC (residues 393– 401, 484 – 496, and 617– 625) were placed using the lego_loop command of O, which searches a database of high-resolution, well-refined crystal structures for the best-fitting segment of amino acids. Side chains that differed between PKA and PKC were placed using the most commonly observed rotamer positions (lego_side_chain), with consideration of appropriate packing of hydrophobic side chains in the protein core. The model (PKC residues 327– 649) was refined using X-PLOR (Powell minimization, with a 500 kcal/(mol Å 2) harmonic restraint on the main chain and Cb atomic positions (14) to improve the side-chain positions. The final X-PLOR energy was 25521 kcal/mol, with van der Waals contacts contributing 2414 kcal/mol. The RMS deviations for bond lengths and angles were 0.007 Å and 1.84°. We also modeled the pseudosubstrate region of PKC (residues 9 –28), based on the structure of the inhibitory peptide PKI bound to PKA as found in the PKA z H89 crystal structure (9). Coordinates for michellamines A, B, and C were generated and energy minimized (MM3 force field) with MACROMODEL (15). The michellamines were docked into the PKC active site using O and

SYBYL (Tripos Assoc.). Solvent-accessible surface areas and cavity volumes were calculated with SYBYL and GRASP (16).

RESULTS AND DISCUSSION

The K m values for the peptide and ATP substrates of PKC were determined (Table I). The K m for ATP, approximately 36 mM, was determined at a peptide concentration of 102 mM, while the K m for the peptide of about 4 mM was determined at 110 mM ATP. K m values for protein kinases substrates appear to depend on such factors as assay type (plain or mixed micellar vesicle), substrate (peptide or protein), and enzyme source (recombinant or native) and so reported values vary (17). However, Seynaeve and co-workers (18) report similar K m values of 6.1– 6.6 mM for their peptide and 15–26 mM for ATP. Next, the IC 50 values (Table II) for michellamines A, B, and C were determined using nonsaturating concentrations of peptide (4 mM) and ATP (110 mM). Michellamines B and C have similar inhibitory potency (15–19 mM), whereas michellamine A had about half of this activity. Thus, the atropisomeric configuration of the michellamines about their 5– 89 and 899–5999 bonds appears to be of only minimal importance to their binding to PKC. PKC has both regulatory and catalytic (“kinase”) domains. In the absence of phospholipid plus calcium, PKC is inactive due to inhibition of the kinase domain TABLE I

Kinetic Constants for PKC Substrate Peptide (at 110 mM ATP) ATP (at 102 mM peptide) a

K m (mM) a 4.3 36.4

K m values were the average of two experiments in which the deviation between replicates was 0.5–1.0 mM.

27

PROTEIN KINASE C INHIBITION BY MICHELLAMINES TABLE II

TABLE IV

Inhibitory Effect of Michellamines on PKC

Kinetic Constants for Inhibition of Protein Kinase C by Michellamine B

Michellamine

IC 50 (mM) a

A B C

36 19 15

a IC 50, the concentration of michellamine that inhibited PKC activity by 50% in the assay containing lipid and calcium as described under Materials and Methods with ATP at 110 mM (three times K m value) and peptide at 4 mM (one times K m value). Values are the average of two experiments in which the deviation between replicates was 1–2 mM.

by a pseudosubstrate peptide sequence in its regulatory domain. Binding of phospholipid plus calcium to the PKC regulatory domain produces a conformational change that results in release of self-inhibition. Hypericin is reported to bind to the regulatory domain of PKC, thereby inhibiting the transformation of the kinase domain to its catalytically competent conformation (5). Thus, we investigated the inhibitory effects of michellamine B on the PKC kinase domain. Michellamine B inhibited the kinase domain in the presence and absence of phospholipid and calcium (Table III). The addition of detergent along with the phospholipid probably accounts for the higher IC 50 value in this assay. Michellamine, undoubtedly, binds to the detergent, reducing the effective concentration of compound. These results indicate that, unlike hypericin, michellamine B binds to the kinase domain rather than the regulatory domain of PKC. The K i values for inhibition of the intact PKC by michellamine B were determined against the peptide and ATP substrates (Table IV). Variation of either substrate did not produce classical inhibition patterns. (This failure could be due, in part, to using an enzyme stock containing a mixture of isozymes, predominantly a, b, and g. However, since the K m and V max values for these three isozymes are within a factor of two of each other and there is a high degree of conservation of TABLE III

Inhibitory Effects of Michellamine B on PKC Kinase Domain IC 50 (mM) a Enzyme

Ca 1 and lipid present

Ca 1 and lipid absent

PKC kinase domain

52

3

a The IC 50 value is the concentration of michellamine that inhibited PKC activity by 50% at 110 mM ATP (three times K m value) and 4 mM peptide (one times K m value).

Varied substrate

K i (mM) a

K 9i (mM)

Mode of inhibition

ATP (at 40 mM peptide) Peptide (at 330 mM ATP)

5 3.5

— 2.2

Noncompetitive Mixed

K i and K 9i values for ATP or peptide were the average of two or three experiments, respectively. a

residues involved in the catalytic process (17), this is probably a minor factor.) Nevertheless, inhibition of PKC enzymatic activity was judged to be noncompetitive with respect to ATP because michellamine B did not affect the K m for ATP, but did alter its V max (Fig. 2). Inhibition appeared to be mixed type with respect to peptide, because both the K m and the V max were affected (Fig. 3). A mixed inhibitor is defined as one that causes the double reciprocal plots (Lineweaver–Burke) to intersect anywhere to the left of the 1/V axis except on the 1/[S] axis. The K i values for michellamine B against both substrates were also determined for the PKC kinase domain. Again, inhibition was not classical, but was similar to that observed for native PKC. Inhibition by michellamine B with respect to ATP was judged to be best approximated with a noncompetitive model with a K i value of 20 mM. Inhibition of the kinase domain with respect to peptide was best approximated as “mixed.” The average K i value was 2.7 mM with a K9i of 3 mM. Because the michellamines are much larger than other PKC inhibitors (e.g., H89), we thought it would be informative to model their possible binding modes to the PKC kinase domain. We constructed a model of the PKC kinase domain based on the catalytic subunit of protein kinase A. All three michellamines could be satisfactorily docked into the active site cleft of this model. (Note that the C 2 symmetry of michellamines A and C restricts the possible binding modes.) A model of michellamine B bound to PKC (Fig. 4) has several interesting features. First, the large size and curved shape of michellamine B blocked the access of both ATP and peptide to their respective subsites in the PKC active site. This restricted access to both subsites may partly explain the observed kinetics of inhibition and their failure to follow classical patterns. Second, binding of michellamine B to PKC buried 420 Å 2 of solvent-accessible surface area of PKC and 564 Å 2 of the inhibitor (total, 984 Å 2). Third, the rigid geometry of the binaphthalene core of michellamine B limited its penetration deep into the PKC active site (particularly the adenine pocket of the ATP subsite). This feature of our model suggests that suitable modifications of the

28

WHITE ET AL.

FIG. 2. Inhibition of protein kinase C by michellamine B — ATP varied. ATP concentration was varied from 2 to 20 mM with the concentration of peptide fixed at 40 mM. Michellamine B concentrations were 0 (F), 4 (■), 20 (Œ), 30 (), and 40 (}) mM. The K i value was calculated from the intercepts of the Lineweaver–Burk plot (inset). An unweighted method of least-squares fit was used to fit the best straight line to the data.

michellamine tetrahydroisoquinoline ring(s) to permit filling of the active site may enhance inhibition. Models of michellamine A and C bound to the PKC kinase domain closely resembled the michellamine B model shown in Fig. 4. In particular, all three michellamines blocked both the ATP and peptide subsites and created solvent-inaccessible cavities in the ATP subsite. Explanation of the twofold difference in binding constants between the strongest inhibitor (michellamine C) and the weakest (michellamine A; see Table II), which corresponds to a DDG of only 0.4 kcal/mol, is beyond the predictive power of the present models.

PKC inhibitors are classified by their binding site (19). Inhibitors that bind to the PKC regulatory domain probably have specificity for PKC compared to other kinases. However, such inhibitors can also disrupt membrane architecture or bind to other phorbol ester receptors (n-chimerin). ATP-binding site inhibitors must be effective even in cells containing high ATP levels. Because other kinases and many other enzymes have ATP-binding sites, inhibitory selectivity against this site in PKC would be difficult to obtain. Although more specific than the ATP-binding site inhibitors, peptide-site inhibitors may have limited therapeutic

PROTEIN KINASE C INHIBITION BY MICHELLAMINES

29

FIG. 3. Inhibition of protein kinase C by michellamine B — Peptide varied. Peptide concentration was varied from 2 to 10 mM with the concentration of ATP fixed at 330 mM. Michellamine B concentrations were 0 (F), 5 (■), 10 (Œ), and 15 () mM. The K i and K 9i values were calculated from the slopes and intercepts of the Lineweaver–Burk plots (insets). An unweighted method of least-squares fit was used to fit the best straight line to the data.

usefulness because of their inability to penetrate cells. On the basis of their binding kinetics, the michellamine alkaloids may fall into an inhibitor class that interacts with both the peptide- and ATP-binding sites, although not competitively. In this regard, these alkaloids resemble staurosporine and UCN-01 (20, 18). Clearly, it will be of interest to determine the specificity of different protein kinases by the michellamines. All kinases appear to possess the bilobal structure appropriate for the binding of michellamines according to the model presented here. Nonetheless, these inhib-

itors appear to interact not just with the ATP-binding site (in which it has been proven difficult to obtain specificity), but also with the peptide-binding site. Interactions in the latter site would be expected to vary from kinase to kinase. This variation may allow the design of specific kinase inhibitors based upon the michellamine framework. Crystallization of PKC with michellamine B would more fully establish the nature of the interaction with both substrate binding sites. Our modeling and kinetic studies suggest that the michellamines will bind in such a way as to overlap

30

WHITE ET AL.

FIG. 4. Hypothetical model of michellamine B bound to the protein kinase C active site cleft. The solvent-accessible surface of the PKC kinase domain is shown as a light-blue surface. The solvent-accessible surface of michellamine B is shown as an orange surface. ATP (red ball-and-stick model) and the modeled PKC peptide pseudosubstrate (residues 9 –28 of PKC, purple stick model) are shown for reference. The N-terminal kinase subdomain is at the top, the C-terminal kinase subdomain is at the bottom, the activation loop is at the right, and the glycine-rich flap lies just above and behind the michellamine B.

both the ATP and peptide substrate subsites of PKC and possibly other protein kinases. ACKNOWLEDGMENT This work was supported by NIH Grant R01-AI36638 (M.I.D.).

REFERENCES 1. McMahon, J. B., Currens, M. J., Gulakowski, R. J., Buckheit, R. W., Jr., Lackman-Smith, C., Hallock, Y. F., and Boyd, M. R. (1995) Antimicrob. Agents Chemother. 39, 484 – 488. 2. Manfredi, K. P., Blunt, J. W., Cardellina, J. H., II, McMahon, J. B., Pannell, L. L., Cragg, G. M., and Boyd, M. R. (1991) J. Med. Chem. 34, 3402–3405. 3. Boyd, M. R., Hallock, Y. F., Cardellina, J. H., II, Manfredi, K. P., Blunt, J. W., McMahon, J. B., Buckheit, R. W., Jr., Bringmann, G., Scha¨ffer, M., Cragg, G. M., Thomas, D. W., and Jato, J. G. (1994) J. Med. Chem. 37, 1740 –1745. 4. White, E. L., Ross, L. J., Hobbs, P. D., Upender, V., and Dawson, M. I. (1999) Anticancer Res., in press. 5. Takahashi, I., Nakanishi, S., Kobayashi, E., Nakano, H., Suzuki, K., and Tamaoki, T. (1989) Biochem. Biophys. Res. Commun. 165, 1207–1212. 6. Wang, H.-K., Xie, J.-X., Chang, J.-J., Hwang, K.-M., Liu, S.-Y., Ballas, L. M., Jiang, J. B., and Lee, K.-H. (1992) J. Med. Chem. 35, 2717–2721. 7. Bruns, R. F., Miller, F. D., Merriman, R. L., Howbert, J. J., Heath, W. F., Kobayashi, E., Takahashi, I., Tamaoki, T., and Nakano, H. (1991) Biochem. Biophys. Res. Commun. 176, 288 –293.

8. Hobbs, P. D., Upender, V., and Dawson, M. I. (1997) Synlett 965–967. 9. Engh, R. A., Girod, A., Kinzel, V., Huber, R., and Bossemeyer, D. (1996) J. Biol. Chem. 271, 26157–26164. 10. Zheng, J., Trafny, E. A., Knighton, D. R., Xuong, N.-H., Taylor, S. S., Ten Eyck, L. F., and Sowadski, J. M. (1993) Acta Crystallogr. D49, 362–365. 11. Srinivasan, N., Bax, B., Blundell, T. L, and Parker, P. J. (1996) Proteins Struct. Funct. Genet. 26, 217–235. 12. Orr, J. W., and Newton, A. C. (1994) J. Biol. Chem. 269, 8383– 8397. 13. Jones, T. A., Zou, J. Y., Cowan, S. W., and Kjeldgaard, M. (1991) Acta Crystallogr. A47, 110 –119. 14. Bru¨nger, A. T., Kuriyan, J., and Karplus, M. (1987) Science 235, 458 – 460. 15. Mohamadi, F., Richards, N. G. J., Guida, W. C., Liskamp, R., Lipton, M., Caufield, C., Chang, G., Hendrickson, T., and Still, W. C. (1990) J. Comput. Chem. 11, 440 – 467. 16. Nicholls, A., Sharp, K. A., and Honig, B. (1991) Proteins Struct. Funct. Genet. 11, 281–296. 17. Dekker, L. V. (1997) in Protein Kinase C (Parker, P. J., and Dekker, L. V., Eds.), pp. 57–74, Landes, Austin, TX. 18. Seynaeve, C. M., Kazanietz, M. G., Blumberg, P. M., Sausville, E. A., and Worland, P. J. (1994) Mol. Pharmacol. 45, 1207–1214. 19. Nixon, J. S. (1997) in Protein Kinase C (Parker, P. J., and Dekker, L. V., Eds.), pp. 205–236, R. G. Landes Co., Austin, TX. 20. Ward, N. E., and O’Brian, C. A. (1992) Mol. Pharmacol. 41, 387–392.