Proc. Natl. Acad. Sci. USAVol. 85, pp. 7197-7201, October 1988Biochemistry

Modeling of the bryostatins to the phorbol ester pharmacophore onprotein kinase C

(tumor promotion/phorbol ester receptor/aplysiatoxin/teleocidin/structure-activity analysis)

PAUL A. WENDER*, CYNTHIA M. CRIBBS*, KONRAD F. KOEHLER*, NANCY A. SHARKEYt,CHERRY L. HERALDt, YOSHIAKI KAMANOt, GEORGE R. PETTITf, AND PETER M. BLUMBERGt*Department of Chemistry, Stanford University, Stanford, CA 94305; tMolecular Mechanism of Tumor Promotion Section, Laboratory of CellularCarcinogenesis and Tumor Promotion, National Cancer Institute, Bethesda, MD 20892; and tCancer Research Institute and Department ofChemistry, Arizona State University, Tempe, AZ 85287

Communicated by John I. Brauman, June 27, 1988 (received for review March 21, 1988)

ABSTRACT The bryostatins are macrocyclic lactones thatrepresent an additional structural class of potent activators ofprotein kinase C. These marine animal biosynthetic productsare of unusual interest because they induce only a subset of thebiological responses induced by the phorbol esters. We havenow determined the binding affinities of naturally occurringand semisynthetic bryostatins for protein kinase C by compe-tition analysis with [26-3Hlbryostatin 4 as the radioactiveligand. Esterification of the hydroxyl group at C26 causeddramatic loss of activity as did inversion of the asymmetriccenter at this position. In contrast, neither of the ester groupsat C7 and C20 had a major influence on activity. Computermodeling of the phorbol esters, related diterpenes, and indolealkaloids suggested that the C20, C9, and C4 oxygens ofphorbol represented critical elements of the phorbol esterpharmacophore. The C26 oxygen of the bryostatins, togetherwith the C1 and C19 oxygens, gave an excellent spatialcorrelation with this model, with a root-mean-square deviationof 0.16 A (compared to 0.10-0.35 A among phorbol-relatedditerpenes). The extension of the phorbol ester pharmacophoremodel to the bryostatins and its agreement with the structure-activity relations for the bryostatin class of compounds provideadditional support for the validity of the model.

Protein kinase C mediates one arm of the signal-transductionpathway proceeding through inositol phospholipid break-down (1, 2). This pathway is directly involved in the actionofa broad range of cellular effectors, including growth factorsand oncogenes, and indirectly affects other transductionpathways such as that of the cyclic AMP second-messengersystem. Protein kinase C is proposed to be activated by1,2-diacyl-sn-glycerol compounds and by certain exogenousanalogs such as the phorbol ester tumor promoters (3, 4). Inaddition to the phorbol esters and related diterpenes, anumber of structurally distinct natural products have beenidentified that also bind to protein kinase C with high affinityand that resemble the phorbol esters in their actions atnanomolar concentrations in biological systems. These com-pounds include the indole alkaloids such as teleocidin and thepolyacetates such as aplysiatoxin (5).

Recently, considerable attention has focused on an unusualclass of potent activators of protein kinase C, the bryostatins.These compounds are macrocyclic lactones isolated fromBugula neritina and other marine bryozoans on the basis ofantineoplastic activity against the P388 leukemia cell system(6). The bryostatins at nanomolar concentrations inhibitphorbol ester binding to protein kinase C and stimulateenzymatic activity in vitro to a comparable degree as do thephorbol esters (7-9). Biologically, however, the bryostatins

induce only a subset of the typical phorbol ester responses.Moreover, the bryostatins block, in an apparently noncom-petitive fashion, the action of the phorbol esters on thoseresponses that they themselves do not induce. Examplesinclude differentiation in HL-60 promyelocytic leukemiacells (7, 10), in Friend erythroleukemia cells (11), and inprimary mouse epidermal cells (12); arachidonic acid releasein C3H1OT1/2 cells (13); and tumor promotion in mouse skin(14). An understanding of the mechanism for the differencesin bryostatin and phorbol ester action may permit selectiveintervention in the protein kinase C pathway.The bryostatins demonstrate that structurally different

activators of protein kinase C need not be functionallyequivalent. For full exploration of the structure-functionrelations of the protein kinase C system, synthetically moreaccessible compounds than the potent naturally occurringactivators would be of great value. As a guide to the designof such compounds, we have used computer comparison ofthe x-ray and/or calculated structures ofthe diterpene, indolealkaloid, and 1,2-diacyl-sn-glycerol classes of protein kinaseC activators to identify the isospatial functional groups andhydrophobic regions that constitute the putative phorbolester pharmacophore (15). This approach suggested a criticalrole for the C4, C9, and C20 oxygens of phorbol. Based onthis modeling, an initial series of benzyl alcohol and hydroxy-methylindole derivatives were prepared that inhibited phor-bol ester binding to protein kinase C and that showed effectson epidermal growth factor binding and phosphorylation inplatelets similar to those of the phorbol esters (15).We report here on the structure-activity analysis for

binding of natural and semisynthetic bryostatins to proteinkinase C. The bryostatins yield an excellent fit to thepharmacophore model derived from the phorbol esters andindole alkaloids, and the experimental structure-activityrelations are largely consistent with this model.

MATERIALS AND METHODSProtein kinase C was purified from the brains of female CD-1mice (Charles River Breeding Laboratories) through theconcentration step after DE-52 chromatography as describedby Jeng et al. (16). The enzyme preparation was frozen in thepresence of 10% (wt/vol) glycerol and stored at - 70°C untiluse.

Bryostatins were isolated as described: bryostatin 1 (6),bryostatin 2 (17), bryostatin 3 (18), bryostatin 4 (19), bryosta-tins 5-7 (20), bryostatin 8 (21), bryostatin 9 (22), andbryostatin 10 (23). Bryostatin 4 13,30-epoxide and bryostatin4 26-acetate were synthesized from bryostatin 4 as reported(23). Preparation and characterization of [26-3H]bryostatin 4and the epimer of [26-3H]bryostatin 4 (specific activity, 6.33Ci/mmol; 1 Ci = 37 GBq) will be described elsewhere

7197

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

7198 Biochemistry: Wender et al.

(N.A.S., D. J. de Vries, Y.K., G.R.P., and P.M.B., unpub-lished data).

Dissociation constants for binding of bryostatins to proteinkinase C were determined by inhibition of [3H]bryostatin 4binding. Partially purified protein kinase C (7.5,ag/ml, 230 +40 pmol/mg), 0.1 mM CaC12, 0.02 M Tris chloride (pH 7.4),immunoglobulin G (Sigma) at 1 mg/ml, Triton X-100 at 1.5mg/ml, phosphatidylserine (Sigma) at 300 /xg/ml, approxi-mately 3.0 nM [3H]bryostatin 4, and nonradioactive bryosta-tin were incubated for 5 min at 370C in a volume of250 1.l. Thesamples were then chilled and two aliquots of 50 1ul wereremoved for determination of total radioactivity. Two otheraliquots were applied to 2.5-cm Whatman DE-81 ion-exchange paper disks and allowed to adsorb for 15-30 sec.The disks were washed with 10 ml of 55% (vol/vol)methanol/20 mM Tris chloride, pH 7.4/0.1 mM Ca2+ andchilled to 00C, and the radioactivity was measured. Nonspe-cific binding was measured in the absence of added enzyme.Free ligand was calculated from the difference between totalligand and that bound to the filter. Because the bryostatinsare highly lipophilic ligands, it should be noted that "freeligand" will be present almost entirely in the detergentmicelle and that nominal molar concentrations for dissocia-tion constants will thus be inversely proportional to thedetergent concentration. The bryostatins absorb avidly tosurfaces when in aqueous solution. However, they are stableunder our assay conditions in the presence of Triton X-100/phosphatidylserine. Inhibition of [3H]bryostatin 4 bind-ing was determined for concentrations of nonradioactivebryostatin from 0.3 to 300 nM. The 50% inhibitory concen-trations were determined from a computer fit of the bindingdata, and the dissociation constants were calculated from the50% inhibitory concentrations as described (24). Directbinding of [26-3H]bryostatin 4 and epi[26-3H]bryostatin 4were measured as above except that the concentration ofradioactive bryostatin was varied.

RESULTSInhibition of [3H]Bryostatin 4 Binding by Nonradioactive

Bryostatins. The naturally occurring bryostatin derivativesexamined here (Fig. 1) display marked variation in the lengthof the ester side chains. Moreover, individual derivatives areentirely lacking the ester group at 07 (bryostatin 2), arelacking the oxygen at C20 together with the associated estergroup (bryostatin 10), or are esterified on the hydroxyl groupsat C19 (bryostatin 3) and at C26. The relative bindingaffinities of the derivatives for protein kinase C were deter-mined by competition with [26-3H]bryostatin 4. Under theassay conditions used, binding was reversible and the kinet-ics were competitive. For bryostatin 4, reasonable but notperfect consistency was found between the Kd determinedhere by competition (1.3 x 10-9 M) and that measured bydirect binding (6.8 x 10-8 M)1 (N.A.S., D. J. de Vries, Y.K.,G.R.P., and P.M.B., unpublished data).

Esterification of bryostatin 4 with acetate at C26 causeddramatic loss of activity (Table 1). Similar results wereobtained for bryostatin 4 26-m-bromobenzoate and bryosta-tin 1 26-acetate or 26-m-bromobenzoate (measured for inhi-bition of [20-3H]phorbol 12,13-dibutyrate binding under non-equilibrium conditions; data not shown). In contrast, thestructural differences among the other derivatives causedquite limited variation in binding activity, suggesting that thesestructural differences did not disrupt receptor recognition.

Binding of 26-Epi[26-3H]bryostatin 4 to Protein Kinase C.The [26-3H]bryostatin 4 was prepared by oxidation of thehydroxyl group at C26, followed by reduction with sodiumborotritiide. This reduction generated both the natural ste-reoisomer of bryostatin 4 (R configuration at C26) as well asthe epimer (S configuration at C26), which were separated bynormal-phase HPLC. In light ofthe potent deactivating effectof esterification at C26, we compared the direct bindingactivity of 26-epi[26-3H]bryostatin with that of the naturalstereoisomer. Under our assay conditions, no specific bind-

CH3 CH3

R,

COCH3HCOCH2CH(CH3)2COCH2CH (CH3)2COCH2CH2CH3COCH3COCH2CH2CH3COCH3COCH2CH(CH3)2

BRYOSTATIN 3

oco~oco\1 N1

OCOCH2CH2CH3OCOCH3OCOCH3OCOCH3OCoCH2CH2CH3OCoCH2CH2CH3H

FIG. 1. Structures of the bryostatins.

BRYOSTATIN

245678910

Proc. Natl. Acad. Sci. USA 85 (1988)

Proc. Natl. Acad. Sci. USA 85 (1988) 7199

Table 1. Comparison of binding affinities to protein kinase C ofbryostatins (Bryo) and their derivatives

CompoundBryo 1Bryo 2Bryo 3Bryo 4Bryo 5Bryo 6Bryo 7Bryo 8Bryo 9Bryo 10Bryo 4 13,30-epoxideBryo 4 26-acetatePhorbol 12,13-dibutyrate

K, x 109*

1.35 ± 0.175.86 ± 1.132.75 ± 0.051.30 ± 0.191.04 ± 0.101.18 ± 0.290.84 ± 0.071.72 ± 0.101.31 ± 0.003.36 ± 0.060.54 ± 0.07>>100

12.3 ± 1.6t

n

5322222222327

*Mean ± range (n = 2) or + SEM (n > 2).tKd, measured by direct binding.

ing was observed at concentrations up to 23 nM 26-epi[26-3H]bryostatin 4, indicating marked loss of activity (Fig. 2).

Structural Comparison of Bryostatins with Phorbol. Struc-ture-activity studies of phorbol derivatives indicate that along-chain acid esterified to either the C12 or C13 hydroxylgroup, the hydroxyl group at C20, and the C4-hydroxyland/or C3-keto functionalities are necessary for tumor-promoting activity (discussed in greater detail in ref. 15).Except for the C20 oxygen, the heteroatoms of phorbol areattached directly to rings with relatively little conformationalmobility and, consequently, their relative spatial coordinatesshould be similar, if not identical, in both the enzyme-boundand the crystal-structure conformation(s). In contrast, therelative spatial coordinates for the C20 oxygen will vary withrotation about the C6-C20 bond. Calculations using theMMP2 force field program (ref. 30; with 1985 parametersavailable from N. L. Allinger, University of Georgia, Athens,GA) indicate that torsion about this bond produces threeconformational minima, differing by at most 1.1 kcal/mol.Comparisons of the spatial coordinates of the heteroatoms inthese three phorbol rotamers with those of (S)-1,2-diacyl-sn-glycerol conformers show that a particularly good correlation(root-mean-square deviation = 0.03 A) exists between theC4, C9, and C20 oxygens of one phorbol rotamer with thecarbonyl and hydroxyl oxygens of a low-energy conformer of(S)-1,2-diacyl-sn-glycerol.

200

0,E 1600Eo 120z0

z

80

0cc 40

6 8 10 12

FREE BRYOSTATIN, nM

FIG. 2. Binding of [26-3H]bryostatin 4 or 26-epi[26-3H]bryostatin4 to protein kinase C. Points represent the average of duplicatedeterminations in single experiments. The combined results from twoindependent experiments with each ligand are presented.

In contrast to phorbol, bryostatin 1 is characterized by aconformationally mobile 26-membered macrolide ring. Thetwo compounds possess, however, a comparable arrange-ment of pharmacophoric groups. In particular, the hydroxylgroup at C26 of bryostatin 1 represents a reasonable surro-gate for the corresponding C20 hydroxyl of phorbol. Therelationship of the C4 and C9 oxygens to the reference C20oxygen in phorbol is then seen to be approximately that foundfor the C1 and C19 oxygens, respectively, to the C26 oxygenin bryostatin 1. Furthermore, comparison of bryostatin 1 and(S)-1,2-diacyl-sn-glycerol reveals a more obvious correspon-dence of subunits, with both possessing vicinal hydroxyalkyl,and ester groups (see Fig. 3).

Quantification of the structural relationships among bry-ostatin 1, phorbol, and (S)-1,2-diacyl-sn-glycerol was accom-plished by computer-assisted comparison of the respectivepharmacophoric oxygens of the three types of protein kinaseC activators. The coordinates of these oxygens in the phorbolesters and in 1,2-diacyl-sn-glycerol were obtained as de-scribed above from crystal structure data and/or calcula-tions. Coordinates for all but the hydroxyethyl atoms ofbryostatin 1 were derived from its crystal structure. Since therelative coordinates of the rotationally mobile hydroxyethylatoms can vary, the relative energies of all conformationsresulting from rotation about the C25-C26 bond werecalculated (MMP2). Three low-energy conformers were pro-duced corresponding to C24-C25-C26-O dihedral anglesof 350, 1900, and 285°, the latter two corresponding closely torotamers that coexist in the crystal structure. The coordi-nates of the pharmacophoric oxygens in these and otherlow-energy conformations of bryostatin 1 were then com-pared with the coordinates of the respective pharmacophoricoxygens in all three C6-C20 rotamers of phorbol. The bestcorrelation was found between phorbol with a C7-C6-C20-O dihedral angle of 138°, previously referred to asphorbol model I, and bryostatin 1 with a C24-C25-C26-O dihedral angle of 3100, the latter corresponding to aconformation that is only 4.68 kcal/mol above the calculatedglobal energy minimum. This correlation gave a root-mean-square deviation of fitted atoms of 0.16 A, which is wellwithin the range of 0.10-0.35 A found in previous correla-tions of phorbol esters with other naturally occurring tumorpromoters (15). Furthermore, in addition to an excellentcorrelation of centers of mass of the pharmacophoric oxy-gens, a good alignment of orbitals was observed as isnecessary for similar binding activity.Table 2 presents the correlation between bryostatin 1 and

phorbol models I and II (differing by a rotation of 1200 aboutthe C6-C20 bond). A significantly better correlation isobtained for phorbol model I, indicating that model I betterdescribes the active conformation. Furthermore, it is expectedthat those structures previously shown to correlate well withphorbol model I should also correlate with bryostatin 1,although not necessarily as well. In fact, good correlations areobtained for (S)-1,2-diacyl-sn-glycerol, ingenol, gnidimacrin,dihydroteleocidin, and aplysiatoxin with bryostatin 1.An additional requirement for biological activity according

to our model is the presence of a lipophilic group in a regionwhere it presumably affects binding orientation and parti-tioning of the molecule into the membrane (15). In phorbolderivatives, a fatty acid is esterified to the C12 hydroxyl; iningenol, an A-ring ester occupies the same region. Whilebryostatin 1 contains no long-chain ester in this area, thelargely lipophilic body of the molecule does occupy the sameregion, thereby fulfilling this requirement.

DISCUSSION

The hydroxyl group at C26 of bryostatin corresponds in thepharmacophore model to the hydroxyl groups at C20 of

Biochemistry: Wender et al.

Proc. Natl. Acad. Sci. USA 85 (1988)

Bryostatin 1

n-C,3H27C00

OH

Phorbol Myristate Acetate I

R 0 O R

OH

(S)-1 ,2-Diacyl-sn-glycerol

FIG. 3. Stereoplot comparisons (taken from the GRMOL option ofCHEMLAB-II). Dashed lines connect heteroatoms whose relative spatialcoordinates are the same in each structure. Certain nonpharmacophoric atoms are deleted from the stereoplot of bryostatin for clarity.

phorbol and C24 of teleocidin. However, it differs in being asecondary rather than a primary hydroxyl. Activity requiresthe R configuration at this position, and we do not knowwhether substituents larger than methyl group will be toler-ated. The exciting possibility exists that substitution at thisposition can be used to lock the freely rotating hydroxyl in theoptimal orientation for receptor interaction, enhancing bind-ing potency, or that orientation might differentially affectbinding to different enzymes. Since the S configuration issterically restricted for binding, the methyl group may itselfhelp limit rotation at this position when the bryostatin isbound to the enzyme and thereby contribute to the apparenthigh binding affinity observed for the bryostatins.The aplysiatoxins, a structurally distinct class of protein

kinase C activators, also possess a single exocyclic, secondaryhydroxyl group. As in the bryostatins, this hydroxyl group (atC30) is in the R configuration (25). Thus, the contribution ofthemethyl group to activity can potentially be assessed, sinceoscillatoxin A represents 20-desmethyldebromoaplysiatoxin.Consistent with expectations, oscillatoxinA was less potent forinducing mouse ear reddening and ornithine decarboxylase andfor inhibiting [20-3H]phorbol 12-myristate 13-acetate binding toan epidermal particulate fraction (26). Unfortunately, inade-quate information was provided to assess the validity of thequantitation for binding. In contrast to the results with thebryostatins, acetylation at C30 of debromoaplysiatoxin 20-acetate caused a loss of activity by only a factor of 6.

Previous efforts to identify the binding site of the phorbol

esters on their receptor by affinity labeling have proven onlypartially successful. Photoactivatable substituents on theester side chains label the lipid moiety of the receptorcomplex (27). The current results suggest that location of thephotoaffinity group proximal to the exocyclic hydroxyl groupmight be an alternative with a better chance of interactingwith the protein portion of the complex.

In accord with the similar activity exhibited by the naturalbryostatins, the critical pharmacophoric functionalities inbryostatin 1 are conserved in all but one other member of thisseries, bryostatin 3. In bryostatin 3, the pharmacophoric C19hydroxyl oxygen common to other bryostatins is replaced bya C19 lactone oxygen. However, since comparisons withother protein kinase C activators suggest that the C19 oxygenof the bryostatins is acting as a hydrogen-bond acceptor, thelactone and hydroxyl groups should be functionally equiva-lent if their spatial orientations are similar. MMP2 calcula-tions indicate that this is indeed the case. Specifically,comparison of the position of the pharmacophoric oxygens ofbryostatins 1 and 3 gave a good fit, with a root-mean-squaredeviation of0.25 A. Furthermore, comparison ofthe C1, C19,and C26 oxygens of bryostatin 3 (dihedral angle C24-C25C26-O = 53) with the C4, C9, and C20 oxygens of phorbol(phorbol model I) provided a good correspondence withroot-mean-square deviations of 0.36 A and 0.37 A for the Rand S configurations at C34, respectively.

Analysis of structure-activity relations in C3H1OT1/2 cellsindicated that different bryostatin derivatives varied in the

7200 Biochemistry: Wender et al.

Proc. Natl. Acad. Sci. USA 85 (1988) 7201

Table 2. Distances between the pharmacophoric heteroatoms in bryostatin, phorbol, and othernaturally occurring protein kinase C ligands

Dihedral angle,* Distance,t A rms deviationiStructure degrees C1C19 C1-C26 C19-C26 A

Bryostatin 1 310 4.61 4.69 6.34Phorbol model 1 138 4.31 4.92 6.38 0.16Phorbol model II 258 4.31 5.39 5.29 0.63Bryostatin 3 53 4.93 4.48 6.80 0.25(S)-DAG 158 4.19 5.18 6.32 0.27Ingenol 127 4.45 4.56 5.79 0.22Gnidimacrin 216 4.25 5.04 6.15 0.23Dihydroteleocidin 178 4.74 4.81 6.67 0.13Aplysiatoxin 292 4.22 5.38 6.48 0.33

Coordinates used in distance calculations were taken from x-ray crystal structures or MMP2-optimized geometries for bryostatin 3 and (S)-1,2-diacyl-sn-glycerol [(S)-DAG]. In (S)-DAG, themovement of the pharmacophoric heteroatoms was restricted.*The dihedral angles are defined by the following atoms: bryostatins, C24-C25-C26-O; diterpenes(phorbol, ingenol, and gnidimacrin), C7-C6-C20--O; (S)-DAG, C1-C2--C3-O; dihydroteleoci-din, C8-C9-C24C-O; and aplysiatoxin, C28-C29-C30-.tDistances are measured between the centers of mass of the oxygens attached to the indicated carbonsof bryostatin 1 and the corresponding heteroatoms of the other structures.tRoot-mean-square (rms) deviations ofthe C1, C19, and C26 oxygens ofbryostatin 1 and the correspondingheteroatoms of the other structures. For aplysiatoxin, the proposed pharmacophoric oxygens are thoseat C27, C3, and C30 (corresponding to the C4, C9, and C20 oxygens, respectively, of phorbol).

degree to which they could cause arachidonic acid release or,reciprocally, block release induced by the phorbol esters (13).Bryostatins 1 and 4 induced no more than 17% of the levelinduced by phorbol ester, whereas bryostatin 3 induced 61%,a finding again consistent with a role for the C19 hydroxyl inrecognition.The bryostatin binding to protein kinase C showed rela-

tively little dependence on the length of the side chains at C7and C20 under our assay conditions, where virtually all of thebryostatin will have partitioned into the lipid phase. Theinflammatory potencies of ingenol 3-monoesters show similarbehavior, with comparable activities for ingenol 3-butyrateand ingenol 3-hexadecanoate and 8-fold higher activity foringenol 3-tetradecanoate (28).The extension of the phorbol ester pharmacophore model

to the bryostatins provides further support for its generalvalidity. However, it does not clarify why the bryostatinsdisplay a mixture of phorbol-ester agonistic and phorbol-ester antagonistic properties. One mechanism currently be-ing evaluated is that the bryostatins both interact with proteinkinase C at the high-affinity phorbol ester binding site toinduce phorbol ester-like responses and also interact at adistinct target, not efficiently recognized by the phorbol esters,to suppress protein kinase C responses. Indirect evidence forthis model is that bryostatins inhibit phorbol ester actionnoncompetitively in Friend erythroleukemia cells (11) andinduce in HL-60 promyelocytic leukemia cells unique phos-phoproteins as well as all those induced by phorbol ester (29).If this second target exists, it might represent either a distinctbryostatin receptor or, alternatively, a form of protein kinaseC modified to recognize phorbol esters only weakly. Thestructural features requisite for the unique bryostatin effectsare only beginning to be explored.

We thank Dr. Stuart H. Yuspa for critical reading of the manu-script. For other assistance we thank Dr. John E. Leet. We thankMolecular Design, Ltd., for a grant to purchase CHEMLAB-II.Financial support to P.A.W. was provided by Grant CA 31841-05from the National Institutes of Health. The Arizona State UniversityCancer Research Institute Laboratory received support from theFannie E. Rippel Foundation, the Arizona Disease Control ResearchCommission, the Robert B. Dalton Endowment Fund, and Grant CA16049-07-11 from the National Institutes of Health.

1. Nishizuka, Y. (1984) Nature (London) 308, 693-698.2. Nishizuka, Y. (1986) Science 233, 305-312.

3. Blumberg, P. M. (1988) Cancer Res. 48, 1-8.4. Ashendel, C. L. (1985) Biochim. Biophys. Acta 822, 219-242.5. Sugimura, T. & Fujiki, H. (1983) in Gann Monograph on Cancer

Research, eds. Makita, A., Tsuiki, S., Fujiki, H. & Warren, L. (Univ. ofTokyo Press, Tokyo), Vol. 29, pp. 3-15.

6. Pettit, G. R., Herald, C. L., Doubek, D. L., Herald, D. L., Arnold, E.& Clardy, J. (1982) J. Am. Chem. Soc. 104, 6846-6848.

7. Kraft, A. S., Smith, J. B. & Berkow, R. L. (1986) Proc. Natl. Acad. Sci.USA 83, 1334-1338.

8. Berkow, R. L. & Kraft, A. S. (1985) Biochem. Biophys. Res. Commun.131, 1109-1116.

9. Smith, J. A., Smith, L. & Pettit, G. R. (1985) Biochem. Biophys. Res.Commun. 132, 939-945.

10. Stone, R. M., Sariban, E., Pettit, G. R. & Kufe, D. W. (1986) Blood 68,193a (abstr.).

11. Dell'Aquila, M. L., Nguyen, H. T., Herald, C. L., Pettit, G. R. &Blumberg, P. M. (1987) Cancer Res. 47, 6006-6009.

12. Sako, T., Yuspa, S. H., Herald, C. L., Pettit, G. R. & Blumberg, P. M.(1987) Cancer Res. 47, 5445-5450.

13. Dell'Aquila, M. L., Herald, C. L., Kamano, Y., Pettit, G. R. & Blum-berg, P. M. (1988) Cancer Res. 48, 3702-3708.

14. Hennings, H., Blumberg, P. M., Pettit, G. R., Herald, C. L., Shores, R.& Yuspa, S. H. (1987) Carcinogenesis 8, 1343-1346.

15. Wender, P. A., Koehler, K. F., Sharkey, N. A., Dell'Aquila, M. L. &Blumberg, P. M. (1986) Proc. Natl. Acad. Sci. USA 83, 4214-4218.

16. Jeng, A. Y., Sharkey, N. A. & Blumberg, P. M. (1986) Cancer Res. 46,1966-1971.

17. Pettit, G. R., Herald, C. L., Kamano, Y., Gust, D. & Aoyagi, R. (1983)J. Natural Prod. 46, 528-531.

18. Pettit, G. R., Herald, C. L. & Kamano, Y. (1983) J. Org. Chem. 48,5354-5356.

19. Pettit, G. R., Kamano, Y., Herald, C. L. & Tozawa, M. (1984) J. Am.Chem. Soc. 106, 6768-6771.

20. Pettit, G. R., Kamano, Y., Herald, C. L. & Tozawa, M. (1985) Can. J.Chem. 63, 1204-1208.

21. Pettit, G. R., Kamano, Y., Aoyagi, R., Herald, C. L., Doubek, D. L.,Schmidt, J. M. & Rudloe, J. J. (1985) Tetrahedron 41, 985-994.

22. Pettit, G. R., Kamano, Y. & Herald, C. L. (1986) J. Natural Prod. 49,661-664.

23. Pettit, G. R., Kamano, Y. & Herald, C. L. (1987) J. Org. Chem. 52,2848-2854.

24. Cheng, Y. C. & Prusoff, W. H. (1973) Biochem. Pharmacol. 22, 3099-3108.25. Moore, R. E., Blackman, A. J., Cheuk, C. E., Mynderse, J. S., Matsu-

moto, G. K., Clardy, J., Woodward, R. W. & Craig, J. C. (1987) J. Org.Chem. 52, 5749.

26. Suganuma, M., Fujiki, H., Tahira, T., Cheuk, C., Moore, R. E. &Sugimura, T. (1984) Carcinogenesis 5, 315-318.

27. Delclos, K. B., Yeh, E. & Blumberg, P. M. (1983) Proc. Natl. Acad. Sci.USA 80, 3054-3058.

28. Sorg, B. & Hecker, E. (1982) Z. Naturforsch. Teil B 37, 748-756.29. Warren, B. S., Kamano, Y., Pettit, G. R. & Blumberg, P. M. (1988)

Cancer Res., in press.30. Ulrich, B. & Allinger, N. L. (1982) Molecular Dynamics (American

Chemical Society, Washington, DC).

Biochemistry: Wender et al.