Proc. Nati. Acad. Sci. USAVol. 88, pp. 8835-8839, October 1991Chemistry

Studies on DNA-cleaving agents: Computer modeling analysis ofthe mechanism of activation and cleavage of dynemicin-oligonucleotide complexesPAUL A. WENDER*, ROBERT C. KELLYt, SUZANNE BECKHAM, AND BENJAMIN L. MILLERDepartment of Chemistry, Stanford University, Stanford, CA 94305

Communicated by John I. Brauman, July 15, 1991 (received for review May 13, 1991)

ABSTRACT Dynemicin A is a recently identified antitu-mor antibiotic. Upon activation, dynemicin is reported to causedouble-stranded cleavage of DNA, putatively through theintermediacy of a diradical. Computer modeling of this acti-vation and cleavage process is described herein as part of aneffort to establish a structural hypothesis for this mechanisticsequence and for the design of simple analogues. Intercalationcomplexes of duplex dodecamers [d(CGCGAATTCGCG)J2and [d(GC)6]2 with both enantiomers of dynemicin and of allrelated mechanistic intermediates are evaluated. Examinationof these structures shows that cycloaromatization of dynemicinto a diradical intermediate results in the rotation of thediradical-forming subunit with respect to the intercalationplane that is of an opposite sense for the two dynemicinenantiomers. In addition, the activation of the (2S) enantiomerof dynemicin occurs by a less restricted approach trajectorythan the corresponding (2R) enantiomer. In all complexes, the5'-3' strand is at least 1 A closer than the 3'-5' strand to thediyl intermediate. As a result, complexes are produced in whichthe diyl moiety is aligned along [(2S)J or across [(2R)] the minorgroove, leading to different predictions for the selectivity ofradical-initiated, oxidative lesion ofDNA. Molecular dynamicssimulations are found to support these predictions, includingthe 3-base-pair offset cleavage reported for dynemicin.

The cleavage of DNA is a key process in the transfer ofgenetic information, the mode of action of certain chemo-therapeutic agents, and the function of reagents designed forDNA modification and structure determination. DNA cleav-age can be effected with a variety of agents ranging from thesimple hydroxyl radical to relatively complex restrictionenzymes. Within the past 5 years, the antitumor antibioticscalicheamicin (1), esperamicin (2), and neocarzinostatin (3)have emerged as a new structural and mechanistic class ofDNA-cleaving agents that are proposed to operate throughthe inducible generation ofan arenyl or indenyl diradical. Themost recently identified member of this class is dynemicin A(structure 1 in Scheme I), a compound that exhibits potentcytotoxicity and in vivo antitumor activity (4, 5). Dynemicinhas been shown to interact with the minor groove of DNAand, upon activation, to cause double-strand breaks 3 basepairs (bp) apart (6). Examination of the structure of dynemi-cin suggests that it could be activated for DNA cleavagethrough reduction of its anthraquinone subunit, resulting inheterolysis of the adjacent epoxide ring. Addition of a nu-cleophile to, or protonation of, the resultant anthraquinonemethide (structure 2 in Scheme I) would provide an activatedderivative 3 which, in the absence of the constraints imposedby the original epoxide ring, would undergo facile cycloar-omatization to diradical 4 (Scheme I). [Heats of formationwere determined by using the AMPAC program (QCPE no.

506, version 2.1) with the AM1 Hamiltonian (7).] Abstractionof proximate deoxyribosyl hydrogens by this diradical wouldinitiate oxidative cleavage on opposing DNA strands. Con-version of diradical 4 to the alternative ene-diyne structure 7is not observed.

Semiempirical (8) and molecular mechanics (9) studies ondynemicin itself have provided valuable information in sup-port of the above mechanism. Thus far, however, computa-tional methods have not been used to evaluate the role ofDNA in the mode of action of dynemicin, although they havebeen applied to calicheamicin (10) and neocarzinostatin (11),yielding models that are consistent with known DNA cleav-age patterns. We describe herein computer modeling studiesdesigned to delineate at the molecular level the interaction ofdynemicin and dynemicin-derived intermediates with oligo-nucleotides selected to emulate native DNA. These studiesaddress several fundamental issues that are crucial to thedevelopment ofa structural hypothesis for the mode of actionof dynemicin and its analogues, including (i) the mechanisticfate of the two possible enantiomers of dynemicin [structures1-(2R) or 1-(2S)], a point of much interest since the absolutestereochemistry of dynemicin is as yet unknown, (ii) theeffect of nucleophile size and approach trajectory in theactivation step and the dynamics of this activation process,(iii) the influence of oligonucleotide sequence and length ondynemicin intercalation and activation, and (iv) the relation-ship of intercalation sites to cleavage sites. The answers tothese questions provide a structural basis for evaluatingmechanistic proposals, for predicting DNA cleavage pat-terns, and for designing new cleaving agents based on thedynemicin lead.

METHODSA DNA octamer corresponding to [d(CGAATTCG)L2 anddodecamers corresponding to [d(CGCGAATTCGCG)]2 and[d(GC)62 were constructed in B-DNA form by using theprogram MACROMODEL (versions 2.0 and 3.0; W. C. Still,Columbia University) running on a MicroVax 3900 and Evansand Sutherland PS340 system. All sequences were minimizedto a gradient of <0.100 kJ/mol per A under the AMBER forcefield (12) before intercalation experiments were begun. Bothenantiomers corresponding to dynemicin A 1, the putativequinone methide intermediate 2, the proposed quinone me-thide addition product 3, and a surrogate for the cyclized,diradical intermediate 5 were minimized to a gradient of<0.200 kJ/mol per A by using the MM2 force field (13, 14).A surrogate for diradical intermediate 4 was necessary be-cause current molecular mechanics force fields are not pa-rameterized for diradicals; the pyrazine ring was chosen as asurrogate because of its geometric similarity to the putativediradical species. As the current parameter set available in

*To whom reprint requests should be addressed.tSenior Scientist on leave from the Upjohn Company, Kalamazoo,MI 49001.

8835

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.

Dow

nloa

ded

by g

uest

on

July

15,

202

0

8836 Chemistry: Wender et al.25

-89.261 wlro

24

OH 0 HN O,H

NOHB C 3

-89.261 kcaVmolOH 0 OH

0OH

OH OH HN OH

pH-147.107 kcaVmol

OH OH OH

3

0

OH OH HNHO OH

NCOCHH 3

-1 55.887 kcaVmolOH OH OH

7

AMBER did not extend to acetylenes, it was necessary tointroduce standard parameters from MM2 into AMBER. In thecase of stretching interactions, this was done by multiplyingthe MM2 constant by a proportionality factor; with otherparameters it was possible to use them without modification,or to use values for available substructures in AMBER whichhad force constants in MM2 identical to those of acetylenes.The resulting modified AMBER force field gave structures thatwere identical to those obtained with MM2 and, for dyne-micin A itself, produced a structure consistent with the x-raycrystallographic structure (5). (Copies of the modified pa-rameters used in these calculations are available from theauthors upon request.)

Initial intercalation spaces in the duplex oligonucleotideswere formed by docking only the anthraquinone portion ofdynemicin into position between base pairs and minimizingfor 1000 iterations under AMBER. The intercalator was thenremoved, and the molecule of interest was placed into thegap. The entire structure was then minimized to a gradient of<0.200 kJ/mol per A by using the modified force field.Structures used to study the effect of nucleophile size onDNA structure and strain energy were generated by addinga hydrogen at C8 of 2 in the minimized complex, rearrangingthe bond orders to give an unminimized form of 3, andcarrying out a substructure minimization on C8 to permit thatatom to pyramidalize. This yielded a complex designed tomimic the conditions that would exist immediately afteraddition of a nucleophile. The hydrogen at C8 was thenreplaced by a methyl, ethyl, isopropyl, or tert-butyl group.Each complex was then minimized as described above, andthe strain energy was calculated.Minor groove widths were determined by measuring the

perpendicular distance between the centers of mass of phos-phorus atoms on complementary strands and subtracting 5.8A, the Van der Waals diameter of a phosphate group (15).Angles for hydrogen abstraction (N-lone pair-H angles) forthe pyrazine diyl surrogate complexes were calculated byassuming the pyrazine ring to be planar and measuring the

Proc. Natl. Acad. Sci. USA 88 (1991)

0

OH 0 HN OH

'- - OCH3 -

-125.574 kcaVrmolOH OH OH

2

N N0 0

OH OH HNHOP OH OH OH HNHO OH

- N N E CH3 -, N N OCH3H > H

-131.156 kcaVmolOHOH OH OH OH OH

4 5

0

OH OHHNHO OH

I H H

-255.815 kcal/molOH OH OH

6

Scheme I

H-N-N angle. The H-lone pair-N angle was obtained fromthis value, and the measured N-H distance by triangulation,where the N-lone pair distance was set to 0.600 A, the valuegiven in the MM2 force field (13, 14).

Molecular dynamics simulations were carried out by usingthe minimized complexes of 5-(2R) and 5-(2S) with [d(CGC-GAATTCGCG)h2 as starting structures. Bonds to hydrogenwere constrained with the SHAKE algorithm (16), and thethermal stability of each system was maintained by couplingto a 300 K external bath (17). Following a 15-ps preequili-bration, the complexes were observed for 30 ps. A timestepof 1 fs was used during both the preequilibration and obser-vation periods, and the nonbonded interaction array wasupdated every 0.5 ps.

RESULTSBecause of the absence of DNA cleavage data at the outsetof this study, G+C- and A+T-rich oligonucleotide sequenceswere selected to explore two generic intercalation sites.Sequence selection around these sites was guided by sym-metry considerations (to simplify calculations) and/or by theavailability of solid state (18) or solution phase (19-24)structural information (for calibration). Sequence length wasinitially set at 8 bp to minimize computational time. How-ever, an early finding of this study was that the intercalationcomplexes ofthe resultant duplex octamers exhibit disruptedbase pairing at terminal residues when full, rather thansubstructure, minimizations were conducted. Since this mod-eling suggested, in agreement with solution studies of DNAintercalation complexes, that the effects of intercalation aretransmitted more than four residues in both directions fromthe site of intercalation, the oligonucleotide length was ex-tended to duplex dodecamers. The resultant duplex structurefor uncomplexed [d(CGCGAATTCGCG)h2 was found to dif-fer only slightly from the x-ray crystal structure (18), thelatter being slightly more tightly wound as a result of crystal-packing forces. Intercalation complexes ofboth duplex dode-

Dow

nloa

ded

by g

uest

on

July

15,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991) 8837

camers produced a DNA model that displayed normal endgroup base pairing.

Since the absolute stereochemistry of dynemicin has notbeen established, intercalation complexes were constructed ofthe aforementioned duplex dodecamers with both possibleenantiomers of dynemicin and of the resultant quinone meth-ide, cyclization precursor, and pyrazine surrogate for thediradical. All minimized duplexes were found to be in theright-handed, Watson-Crick B form, with most of the sugarring puckers C2'-endo. The overall structure of the oligonu-cleotides in the complexes was little changed from the uncom-plexed form. Separation of the base pairs immediately aboveand below the intercalation site increased from 3.4 A to6.4-6.6 A upon intercalation ofdynemicin, and Watson-Crickbase pairing remained intact. The intercalation site was wedge-shaped, similar to that observed for a nogalamycin-oligonucleotide complex (25) and a daunomycin-oligonucleo-tide complex (26). Base pairs directly adjacent to the interca-lation site exhibited a slight propellor twist and were buckled.The anthraquinone moiety was puckered and not planar.Although this puckering is not observed in the daunomycincomplex, it is seen in nogalamycin, which, like dynemicin, hasan aliphatic ring rigidly attached to the intercalator subunit(25).Two general energy minima were found for the complexes,

one with the anthraquinone intercalated more deeply and theene-diyne moiety closer to the double-helical axis than in theother. The proposed explanation for this is that H5' and/orH4' of base 8 cause unfavorable Van der Waals interactionswhen they are in the same plane as the ene-diyne (or diradicalsurrogate) portion. This results in a local minimum when theene-diyne/diradical surrogate is inserted beyond these hy-drogens. Although the size of the barrier between these twominima is unknown, experimental data available for thestructurally related daunomycin-oligonucleotide complex, inwhich the amino sugar lies on the floor of the minor grooveand the D ring protrudes into solution from the major groove,suggests that deeply intercalated species are energeticallyaccessible. This also suggests that intercalation depth may bean important structural parameter in the design of analoguesfor abstraction of specific hydrogens.Comparison ofthese complexes provides an understanding

of the structural changes that attend the activation andcyclization steps, including the differing roles of the twopossible dynemicin enantiomers. Specifically, for the con-version of the (2S)-dynemicin complex (Fig. 1 Left) to the

FIG. 1. (2S)-dynemicin (Left) and (2R)-dynemicin (Right) in[d(CGCGAATrCGCG)]2.

FIG. 2. (2S)-diyl surrogate (Left) and (2R)-diyl surrogate (Right)in [d(CGCGAA1TCGCG)]2-

corresponding (2S)-diyl surrogate complex (Figs. 2 Left and3 Left), atoms C23-C28 move in a clockwise sense withrespect to the pseudo-plane ofthe intercalated anthraquinonesubunit, resulting in a structure in which these atoms arealigned along the minor groove. The initial alignment of the(2R)-dynemicin complex (Fig. 1 Right) is similar to that foundfor the (2S) complex. However, as expected from the con-trasting handedness of the (2R)-dynemicin isomer, its con-version to the (2R)-diyl surrogate proceeds with the coun-terclockwise movement of atoms C23-C28 with respect tothe intercalated anthraquinone subunit, producing a structurein which these atoms align across the minor groove (Figs. 2Right and 3 Right).

Although alignment of the intercalating species along oracross the minor groove does not by itself determine whetherthe resultant diyl could cleave DNA, it does determine accessto and steric encumbrance, at least for nucleophilic attack, atthe site of attachment for the incoming nucleophile or protonsource. This is clearly seen for the (2S) derivatives, where thetrajectory of nucleophilic attack or protonation would beeasily accommodated along the minor groove, while the pathfor the (2R) derivatives would encounter severe steric en-

FIG. 3. Corey-Pauling-Koltun models of the (2S)-diyl surrogate(Left) and (2R)-diyl surrogate (Right) in [d(CGCGAATTCGCG)]2.The closest hydrogen for each static complex is highlighted in red.

Chemistry: Wender et al.

Dow

nloa

ded

by g

uest

on

July

15,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991)

30-.

25-

20

15

10

E OHB

.1 5'%

(2S) bases 1-12

666 6

000.|l.aBasseS.HI

I 9Angstm a

.1 B2 B3 B4 B5 B6 B7 B8 B9 B10 Bll B12-

31

30

2520

15

10

5I 5'0

sU (4) oases 13-24

zwA1ill;.itI

.5 6!

5 i il.Hi'

i7

Angstrolillu§§ins .-.ul28@@|§ell.0B13 B14 B15 B16 B17 B18 B19 B20 B21 B22 B23 B245' O3'

B13 B14 B15 B16 B17 B18 B19 B20 B21 B22 B23 B24St - 3'

FIG. 4. Maximal (x), minimal (o), and average (-) N-H5', N-H4', N-H3', and N-Hi' distances for 5 in [d(CGCGAATTCGCG)h2 frommolecular dynamics simulations.

cumbrance from the 3'-5' strand of the oligonucleotide. Thistrend is further supported by the relative stabilities of theactivated complexes: in all cases, the (2S) complex is morestable than the (2R) complex, with energy differences rangingfrom 8.82 kcal/mol for nucleophiles similar to a methyl groupin size to 14.72 kcal/mol for a tert-butyl group. For anactivation process involving protonation ofan anthraquinonemethide, the product complexes are of comparable energy,suggesting that either could be accommodated in the minorgroove. However, protonation of the (2R)-anthraquinonesubunit would preferentially occur prior to association withDNA, since direct protonation of the anthraquinone subunitwhen complexed with DNA would be sterically disfavored.Alternatively, a relay proton transfer mechanism could beinvolved, in which case a small proton transfer agent (e.g.,water) could associate with DNA prior to the intercalation ofdynemicin and might remain present in the minor groove soas to allow direct contact with dynemicin.An analysis of distances between abstractable hydrogens

on the sugar-phosphate backbone and radical centers on 4

toE2

U

._O

_ [d(GC)6J2

provides an indication of the manner in which cleavage sitesare related to intercalation sites and suggests where the initialabstraction might occur. It was found that the 5'-3' strandwas, on average, greater than 1 A closer to the "diyl" thanthe 3'-5' strand in all of the static cases examined. For both(2S)-diyl complexes, the hydrogens on base 8 are closest.(The optimum transition-state geometry for abstraction ofhydrogen from methane by methyl radical has been deter-mined by ab initio methods; see ref. 27.)Molecular dynamics simulations ofthe complexes of5-(2R)

and 5-(2S) with [d(CGCGAATTCGCG)J2 at 300 K confirmthe above general observations on the static complexes andprovide additional information with respect to the differinginteraction of the two enantiomers with DNA. While base 7is once again closest to the "diyl" moiety of 5-(2R) through-out the simulation, the contrasting handedness of 5-(2S)coupled with motion within the intercalation site causes base8 to be closest in the 5-(2S)-DNA complex (Fig. 4). None ofthe bases ofthe 3'-5' strand come within abstraction distanceof the 5-(2R)-"diyl" during the simulation, the closest being

.3 y-2S-1+DNA ia(C)P- -C(G) 13

-- 2R-1+DNA Pair#121Pairc#,4m-2S-+DNA etc. i'!!P

,'~~~~~ ~~~~~~a) 11,..~~..2R-5+DNA (C! Pair#1

s\, ~~~~6(cQy---(G) If

M4(ERG- -C)

-y 2 (C)G'--(G)23_ ~ ~~~~~~~~~1-(Gc-_C X

6 e0 7 3 G 2

FIG. 5. Groove widths for [d(CGCGAATTCGCG)h2 and [d(GC)6]2 complexes. 2S-1, 2R-1, 2S-5, and 2R-5, (2S) and (2R) enantiomers ofstructures 1 and 5 of Scheme I.

Cud._ 2

.I

1 2 3 4 5 6 7 8 1 2 3 4

Pair Number

8838 Chemistry: Wender et al.

MONaL --- s .2e % A

Dow

nloa

ded

by g

uest

on

July

15,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991) 8839

base 18 at an approximate minimum distance of 4 A. Bases18 and 19 approach within 2 and 3.5 A, respectively, of the5-(2S)-"diyl".

Plots ofgroove widths measured for each dynemicin modelintercalated into [d(CGCGAATTCGCG)12 and for the freedodecamer are shown in Fig. 5. As expected from the"wrong-handed" twist observed on cyclization of 3-(2R),intercalation of 5-(2R) and its diyl model caused greaterbroadening of the minor groove than intercalation of 5-(2S).The larger groove widths observed for the (2R) enantiomerrelative to the (2S) enantiomer in [d(CGCGAATTCGCG)]2systems are not observed for the GC dodecamer [(GC)612(Fig. 5). It has been observed (15) in x-ray diffraction studieson DNA fibers that the minor groove width of B-DNAincreases with the proportion of G-C base pairs, and it isprobable that the larger minor groove width of the GCdodecamer permits intercalation with less distortion.

CONCLUSIONSixteen intercalation complexes ofboth possible enantiomersof dynemicin and of dynemicin-derived intermediates withduplex octamers and dodecamers were constructed andminimized. While the duplex octamer complexes sufferedfrom disrupted base pairing at the termini, complexes withduplex dodecamers exhibited normal end group base-pairing.This suggests that duplex octamers may not be a reliablemodel for DNA in dynemicin-oligonucleotide complexes,and longer oligonucleotides should be used. Both enantio-mers of dynemicin and putative precyclization intermediatesyield intercalation complexes with similar orientations of theene-diyne within the minor groove ofDNA. Upon examina-tion of the trajectory of attack ofa nucleophile to the quinonemethide intermediate, however, differences between the twoenantiomers become apparent. The approach of a nucleo-phile to either enantiomer is hindered by the 3'-5' strand ofDNA, and the effect is greater for the (2R) enantiomer. Inaddition to this kinetic effect on activation, it is also foundthat the activated (25) complex is lower in energy than the(2R) complex for various nucleophiles. Cyclization to the diylintermediate results in a rotation of the diyl moiety withrespect to the intercalation plane, improving the alignment ofthe (2S) enantiomer within the minor groove, while causingthe (2R) enantiomer to become aligned across the minorgroove. Although it is not yet known whether dynemicinintercalates before or after activation occurs, the aboveevidence suggests that if intercalation occurs first, the (2S)enantiomer of dynemicin is better suited for nucleophilicactivation within the complex and also indicates that anynucleophile attacking at C8 of the quinone methide interme-diate must be small.

In all complexes, the diyl is much closer to the 5'-3' strandof the oligonucleotide than to the 3'-5' strand. Even so,significant motion ofthe intermediate within the intercalationcomplex would be expected to occur, permitting abstractionfrom the 3'-5' strand. Molecular dynamics simulations de-signed to address this issue suggest that, although motionwithin the intercalation site brings the 5-(2S)-diyl surrogatewithin abstraction distance of both strands ofDNA, abstrac-

tion from the 3'-5' strand would necessarily follow abstrac-tion from the 5'-3' strand for the complex with 4-(2R).

Support of this research by the National Cancer Institute throughGrant CA31845 is gratefully acknowledged. S.B. is a NationalScience Foundation Research Fellow, and B.L.M. is a W. R. GraceGraduate Fellow.

1. Zein, N., Sinha, A. M., McGahren, W. J. & Ellestad, G. A.(1988) Science 240, 1198-1201.

2. Long, B. H., Golik, J., Forenza, S., Ward, B., Rehfuss, R.,Dabrowiak, J. C., Catino, J. J., Musial, S. T., Brookshire,K. W. & Doyle, T. W. (1989) Proc. Nati. Acad. Sci. USA 86,2-6.

3. Napier, M. A., Holmquist, B., Strydom, D. J. & Goldberg,I. H. (1979) Biochem. Biophys. Res. Commun. 89, 635-642.

4. Konishi, M., Ohkuma, H., Matsumoto, K., Tsuno, T., Kamei,H., Miyaki, T., Oki, T., Kawaguchi, H., VanDuyne, G. D. &Clardy, J. (1989) J. Antibiot. 42, 1449-1452.

5. Konishi, M., Ohkuma, H., Tsuno, T., Oki, T., VanDuyne,G. D. & Clardy, J. (1990) J. Am. Chem. Soc. 112, 3715-3716.

6. Sugiura, Y., Shiraki, T., Konishi, M. & Oki, T. (1990) Proc.Nati. Acad. Sci. USA 87, 3831-3855.

7. Dewar, M. J. S., Zoebisch, E. G., Healy, E. F. & Stewart,J. J. P. (1985) J. Am. Chem. Soc. 107, 3902-3909.

8. Snyder, J. P. & Tipsword, G. E. (1990) J. Am. Chem. Soc. 112,4040-4042.

9. Semmelhack, M. F., Gallagher, J. & Cohen, D. (1990) Tetra-hedron Lett. 31, 1521-1522.

10. Hawley, R. C., Kiessling, L. L. & Schreiber, S. L. (1989)Proc. Nati. Acad. Sci. USA 86, 1105-1109.

11. Galat, A. & Goldberg, I. H. (1990) Nucleic Acids Res. 18,2093-2099.

12. Weiner, S. J., Kollman, P. A., Case, D., Singh, U. C., Ala-gona, G., Profeta, S. & Weiner, P. (1984) J. Am. Chem. Soc.106, 765-784.

13. Allinger, N. L. (1977) J. Am. Chem. Soc. 99, 8127-8134.14. Burkert, U. & Allinger, N. L. (1982) Molecular Mechanics

(Am. Chem. Soc., Washington).15. Arnott, S. (1981) in Topics in Nucleic Acid Structure, ed.

Neidle, S. (Halsted, New York), pp. 65-82.16. Ryckaert, J.-P., Ciccotti, G. & Berendsen, H. J. C. (1977) J.

Comput. Phys. 23, 327.17. Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F.,

DiNola, A. & Haak, J. R. (1984) J. Chem. Phys. 81, 3684-3690.18. Drew, H. R., Wing, R. M., Takano, T., Broka, C., Tanaka, S.,

Itakura, K. & Dickerson, R. E. (1981) Proc. Natl. Acad. Sci.USA 78, 2179-2183.

19. Patel, D. J. (1976) Biopolymers 15, 533-558.20. Uesugi, S., Shida, T. & Ikehara, M. (1981) Chem. Pharm. Bull.

29, 3573-3585.21. Cheng, D. M., Kan, L. S., Frechet, D., Ts'O, P. 0. P., Ue-

sugi, S., Shida, T. & Ikehara, M. (1984) Biopolymers 23,775-795.

22. Patel, D. J., Kozlowski, S. A., Markey, L. A., Broka, C.,Rice, J. A., Itakura, K. & Breslauer, K. J. (1982) Biochemistry21, 428-436.

23. Patel, D. J., Pardi, A. & Itakura, K. (1982) Science 216,581-590.

24. Pardi, A., Morden, K. M., Patel, D. J. & Tinoco, I., Jr. (1982)Biochemistry 21, 6567-6574.

25. Searle, M. S., Hall, J. G., Denny, W. A. & Wakelin, L. P. G.(1988) Biochemistry 27, 4340-4349.

26. Wang, A. H.-J., Ughetto, G., Quigley, G. J. & Rich, A. (1987)Biochemistry 26, 1152-1163.

27. Wildman, T. A. (1986) Chem. Phys. Lett. 126, 325-329.

Chemistry: Wender et al.

Dow

nloa

ded

by g

uest

on

July

15,

202

0