cb2 new nj08903 - pnas · proc. natl. acad. sci. usa86 (1989) 3969 heated at a rate of 0.50c/min...

5
Proc. Nati. Acad. Sci. USA Vol. 86, pp. 3968-3972, June 1989 Biochemistry Binding of actinomycin D to DNA: Evidence for a nonclassical high-affinity binding mode that does not require GpC sites JAMES G. SNYDER*t§, NEIL G. HARTMAN*, BEATRICE LANGLOIS D'ESTANTOIT*, OLGA KENNARD*, DAVID P. REMETA*, AND KENNETH J. BRESLAUERt§ *University Chemical Laboratory, Lensfield Road, Cambridge CB2 lEW, United Kingdom; and tDepartment of Chemistry, Rutgers University, New Brunswick, NJ 08903 Communicated by Max Tishlerll, February 13, 1989 ABSTRACT We have employed a combination of temper- ature-dependent UV absorption spectroscopy, circular dichro- ism, and batch calorimetry to characterize the binding of actinomycin D to a series of oligomeric DNA duplexes. We find the duplex [d(CGTCGACG)12 to be unique in its ability to bind actinomycin D strongly despite the absence of a classic GpC site. We present evidence that this non-GpC-containing duplex binds two actinomycin D molecules in an apparently cooper- ative manner to form a complex that exhibits aberrant spec- troscopic and calorimetric behavior. We propose that these observations are consistent with actinomycin D exhibiting a high-affinity, sequence-dependent DNA-binding mode distinct from its classic binding to isolated GpC sites. The cytotoxicity of the antitumor drug actinomycin D (ActD; Fig. 1) is attributed to its ability to inhibit transcription by binding to double-stranded DNA. For this reason, the DNA binding properties of ActD have been studied extensively for many years (1-19). In general, these binding studies have shown that ActD complexes with duplex DNA by intercala- tion of its planar, aromatic rings between adjacent base pairs of the double helix. However, other binding modes also have been proposed (1, 20). In addition to its medical importance, ActD provides an important model for the study and the design of new se- quence-specific DNA-binding ligands. In this connection, ActD has been shown to exhibit a binding preference for the 3' side of guanine residues (10, 11). The dinucleotide site GpC exhibits an especially high binding affinity for ActD (11, 14, 15, 18, 19, 21). Other dinucleotide sites, such as GpT, GpA, GpG, and even CpG, also have been proposed as potential ActD binding sites, although most of the supporting evidence has been indirect and sometimes contradictory (9, 21-29). To investigate more directly the possibility of drug binding to non-GpC sites, we have conducted ActD binding studies using a series of oligomeric DNA host duplexes that contain systematically varied sequences. With one notable excep- tion, we found that oligomeric duplexes that lack GpC sites do not exhibit high affinities for ActD. However, ActD binds strongly to the non-GpC duplex [d(CGTCGACG)h2. Our data reveal the surprising result that two ActD molecules bind tightly to this duplex with apparent cooperativity and in a manner that exhibits atypical spectroscopic and thermody- namic properties. We propose that the mode of binding of ActD to [d(CGTCGACG)]z is distinct from its classic mode of binding to isolated GpC sites. MATERIALS AND METHODS Oligonucleotides. The oligodeoxynucleotides used in this study were synthesized with an Applied Biosystems 381A-00 0 11 H2C \/N\ CH3\ H2C C=O CH- HC CH3 /NH 0'C\ ___ CH- CH CH3 HN 0 CH, 11. I / CH- HC I \N N-CH2 CH3-N\ C =0\H HCCH~ I \ /N\ /CH2 0 =C CH2 /CH3 CH-HC HN CH3 CH3 CH3 FIG. 1. Structure of ActD. DNA synthesizer using phosphoramidite chemistry. The crude oligomers were purified by anion-exchange chroma- tography and reverse-phase HPLC. The extinction coeffi- cients of the oligonucleotides were calculated by using a nearest-neighbor approximation. ActD. The compound was purchased from Sigma and used without further purification. The concentration of ActD in solution was determined spectrophotometrically using an extinction coefficient (E44O) of 24,800 M-' cm-1 at 25°C. Buffer System. All solutions were prepared in a buffer containing 10 mM phosphate, 1 mM EDTA, and 45 mM Na+, adjusted to pH 7.0 at 25°C. UV Melting Curves. Absorbance-vs.-temperature profiles for each drug-free and drug-bound duplex were determined with a temperature-controlled, six-cuvette cell holder in- stalled in a programable Kontron Uvikon 810 spectropho- tometer interfaced to a BBC microcomputer. Samples were Abbreviations: ActD, actinomycin D; Tm and tim melting tempera- ture (absolute and Celsius). tPresent address: Department of Medicine, Baylor College of Med- icine, Texas Medical Center, Houston, TX 77030. §To whom reprint requests should be addressed. Il Deceased March 18, 1989. 3968 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Downloaded by guest on October 4, 2020

Upload: others

Post on 28-Jul-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: CB2 New NJ08903 - PNAS · Proc. Natl. Acad. Sci. USA86 (1989) 3969 heated at a rate of 0.50C/min while the temperature and absorbance at 260 nmwere recorded every 30 sec. Melting

Proc. Nati. Acad. Sci. USAVol. 86, pp. 3968-3972, June 1989Biochemistry

Binding of actinomycin D to DNA: Evidence for a nonclassicalhigh-affinity binding mode that does not require GpC sitesJAMES G. SNYDER*t§, NEIL G. HARTMAN*, BEATRICE LANGLOIS D'ESTANTOIT*, OLGA KENNARD*,DAVID P. REMETA*, AND KENNETH J. BRESLAUERt§*University Chemical Laboratory, Lensfield Road, Cambridge CB2 lEW, United Kingdom; and tDepartment of Chemistry, Rutgers University,New Brunswick, NJ 08903

Communicated by Max Tishlerll, February 13, 1989

ABSTRACT We have employed a combination of temper-ature-dependent UV absorption spectroscopy, circular dichro-ism, and batch calorimetry to characterize the binding ofactinomycin D to a series of oligomeric DNA duplexes. We findthe duplex [d(CGTCGACG)12 to be unique in its ability to bindactinomycin D strongly despite the absence of a classic GpCsite. We present evidence that this non-GpC-containing duplexbinds two actinomycin D molecules in an apparently cooper-ative manner to form a complex that exhibits aberrant spec-troscopic and calorimetric behavior. We propose that theseobservations are consistent with actinomycin D exhibiting ahigh-affinity, sequence-dependent DNA-binding mode distinctfrom its classic binding to isolated GpC sites.

The cytotoxicity of the antitumor drug actinomycin D (ActD;Fig. 1) is attributed to its ability to inhibit transcription bybinding to double-stranded DNA. For this reason, the DNAbinding properties of ActD have been studied extensively formany years (1-19). In general, these binding studies haveshown that ActD complexes with duplex DNA by intercala-tion of its planar, aromatic rings between adjacent base pairsof the double helix. However, other binding modes also havebeen proposed (1, 20).

In addition to its medical importance, ActD provides animportant model for the study and the design of new se-quence-specific DNA-binding ligands. In this connection,ActD has been shown to exhibit a binding preference for the3' side ofguanine residues (10, 11). The dinucleotide site GpCexhibits an especially high binding affinity for ActD (11, 14,15, 18, 19, 21). Other dinucleotide sites, such as GpT, GpA,GpG, and even CpG, also have been proposed as potentialActD binding sites, although most of the supporting evidencehas been indirect and sometimes contradictory (9, 21-29).To investigate more directly the possibility of drug binding

to non-GpC sites, we have conducted ActD binding studiesusing a series of oligomeric DNA host duplexes that containsystematically varied sequences. With one notable excep-tion, we found that oligomeric duplexes that lack GpC sitesdo not exhibit high affinities for ActD. However, ActD bindsstrongly to the non-GpC duplex [d(CGTCGACG)h2. Our datareveal the surprising result that two ActD molecules bindtightly to this duplex with apparent cooperativity and in amanner that exhibits atypical spectroscopic and thermody-namic properties. We propose that the mode of binding ofActD to [d(CGTCGACG)]z is distinct from its classic modeof binding to isolated GpC sites.

MATERIALS AND METHODSOligonucleotides. The oligodeoxynucleotides used in this

study were synthesized with an Applied Biosystems 381A-00

0

11

H2C\/N\CH3\ H2C C=O

CH- HC

CH3 /NH0'C\

___ CH- CH

CH3 HN

0

CH, 11.I /CH- HCI \N

N-CH2

CH3-N\

C =0\H

HCCH~I \

/N\ /CH20=C CH2 /CH3

CH-HC

HN CH3

CH3 CH3

FIG. 1. Structure of ActD.

DNA synthesizer using phosphoramidite chemistry. Thecrude oligomers were purified by anion-exchange chroma-tography and reverse-phase HPLC. The extinction coeffi-cients of the oligonucleotides were calculated by using anearest-neighbor approximation.

ActD. The compound was purchased from Sigma and usedwithout further purification. The concentration of ActD insolution was determined spectrophotometrically using anextinction coefficient (E44O) of 24,800 M-' cm-1 at 25°C.

Buffer System. All solutions were prepared in a buffercontaining 10 mM phosphate, 1 mM EDTA, and 45 mM Na+,adjusted to pH 7.0 at 25°C.UV Melting Curves. Absorbance-vs.-temperature profiles

for each drug-free and drug-bound duplex were determinedwith a temperature-controlled, six-cuvette cell holder in-stalled in a programable Kontron Uvikon 810 spectropho-tometer interfaced to a BBC microcomputer. Samples were

Abbreviations: ActD, actinomycin D; Tm and tim melting tempera-ture (absolute and Celsius).tPresent address: Department of Medicine, Baylor College of Med-icine, Texas Medical Center, Houston, TX 77030.§To whom reprint requests should be addressed.Il Deceased March 18, 1989.

3968

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

Oct

ober

4, 2

020

Page 2: CB2 New NJ08903 - PNAS · Proc. Natl. Acad. Sci. USA86 (1989) 3969 heated at a rate of 0.50C/min while the temperature and absorbance at 260 nmwere recorded every 30 sec. Melting

Proc. Natl. Acad. Sci. USA 86 (1989) 3969

heated at a rate of 0.50C/min while the temperature andabsorbance at 260 nm were recorded every 30 sec. Meltingtemperatures were obtained from these data as described(30).

Circular Dichroism. CD spectra were recorded by using aCary 60 instrument equipped with a Cary 6001 CD accessoryand a programable, thermoelectrically controlled cell holder(Aviv Associates, Lakewood, NJ).Job Plots. The ActD-DNA binding densities for [d(CGT-

CGACG)h2 and [d(CGATGCATG)]2 were determined fromJob plots (31) constructed from optical data obtained at 100C.

Helix-Coil Transition Enthalpies. Helix-to-coil transitionenthalpies for [d(CATGCATG)]2 and [d(CGTCGACG)]2were determined with a Microcal 2 differential scanningcalorimeter. The experimental protocols and analysis of thecalorimetric heat capacity curves have been described indetail (30). Transition enthalpies for the other duplex oligo-nucleotides used in this study were calculated by using thenearest-neighbor data base of Breslauer et al. (32).Batch Calorimetry. The batch calorimeter used for the

binding studies has been described in detail (33). The ActD-DNA binding enthalpies (AHb) were determined at 15'C bymixing solutions of ActD and the host DNA duplex. Thistemperature was selected to ensure the presence in solutionof a fully formed duplex. Significantly, however, bindingstudies over a range of temperatures revealed negligible heatcapacity changes (ACp). Consequently, the AHb values mea-sured at 15'C reasonably can be extrapolated to other tem-peratures. To correct for the heat associated with the binding-induced disruption of unbound ActD dimers, an ActD dimer-ization constant (Kd) of 1.4 x 103 M-1 and a dimerizationenthalpy (AHd) of +9.4 kcal/mol of dimer disrupted at 15'Cwere used (34). The relative values of the calculated AHb arenot dependent on the accuracy of this correction term, sinceit is applied equally to each measurement.Drug-DNA Binding Constants. Ligands that bind more

tightly to double-stranded DNA than to the single-strandedstate induce an increase in the melting temperature [Tm(absolute) or tm (Celsius)] of the host duplex. The theoreticalbasis for this phenomenon has been described by Crothers(35). Following Crothers' theoretical treatment, Snyder (34)has used the ligand-induced increase in duplex melting tem-perature (ATm) along with the ligand/DNA binding stoichi-ometry and the calorimetric transition enthalpy of the hostduplex to calculate ligand binding affinities (K) for the duplexstate by application of the equation

1- 1- nR In(1 + KaL),Tm Tm AHh

[1]

where TA and Tm are the melting temperatures of the oligo-meric DNA duplex in the presence and absence of addeddrug, respectively; n is the number of drug binding sites onthe duplex; AHh is the transition enthalpy for the duplex in theabsence of bound drug; aL is the activity of the free ligand atT = Tm; K is the drug-duplex association constant at T = Tm;and R is the gas constant. To extrapolate our calculatedbinding constants to reference temperatures of interest, weused the standard thermodynamic relationship a(ln K)!a(1/T) = AHb/R, where AHb is the drug-duplex bindingenthalpy measured directly by batch calorimetry. Use of thisequation assumes AHb to be independent oftemperature (i.e.,ACp = 0). Calorimetric measurements of AHb over a range oftemperatures have shown that this assumption introduceslittle error into the extrapolation of K (34). Note that Eq. 1was derived by assuming a model in which the ligand bindsto separate, noninteracting sites on the DNA lattice (35).Therefore, when n : 2, the calculated K should be consideredto be an apparent binding constant, Kapp. This "ATm"

method is especially useful for studying high-affinity drug-DNA binding events (36) for which reliable K values aredifficult to obtain by conventional optical titration methods.

RESULTSActD Binding to Oligomeric DNA Duplexes With and With-

out GpC. The self-complementary octanucleotide d(CATG-CATG) forms aDNA duplex with a single ActD intercalationsite at the central GpC. A Job plot analysis (not shown)revealed the formation of the expected 1:1 ActD/duplexcomplex. Fig. 2 shows the differential UV melting curves ofthe [d(CATGCATG)]2 duplex in its drug-free and drug-boundstates. The substantial thermal stabilization of the drug-bound vs. the drug-free duplex reflects a high ActD bindingaffinity for this oligomeric DNA. We calculated a ATm valueof 21 K from the two curves in Fig. 2 (see Table 1). We usedthis ATm value in conjunction with the calorimetrically mea-sured binding enthalpy (AHb, see below) to calculate an ActDbinding constant K(370C) = 1.5 x 107 M-1 (34-36). Themagnitude of this binding constant is consistent with previousActD studies that, corporately, indicated a high affinity ofActD for GpC intercalation sites in duplex DNA (7, 9-12, 14,15, 17-19, 21, 22).

Previous studies using polymeric host duplexes suggestedthat ActD also can exhibit a significant affinity for DNAduplexes containing GpA and GpT steps but devoid of GpCintercalation sites (9). To more directly investigate this pos-sibility, we extended our melting-curve studies on ActD-DNA complexes by employing a series of oligomeric hostduplexes containing guanine residues but devoid of GpCsites. The ATm data we determined from the melting curvesfor such non-GpC duplexes in the presence and absence ofActD (Table 1) show that for all but one of these non-GpCduplexes, ATm is small or equal to zero, thereby suggestinglittle, if any, ActD binding to this class of duplexes. In fact,small ATm values typify the melting behavior of all but one ofthe non-GpC oligomers we have examined to date (21, 34).Therefore, contrary to expectations based on earlier ActD-polymer studies (9), our results reveal that ActD binds onlyweakly, if at all, to oligomeric duplexes containing isolatedGpN sites but lacking GpC sites. In this regard, Chen (24) alsohas observed that ActD binds only weakly to the non-GpCDNA duplex [d(ATATACGTATAT)]2.

. [d(CATGCATG)]2

31C

t. = 32'C --

[d(CATGCATG)12S .: ~~+

ActD

lm 52 'C (1:1)

*I tm 52°C ;

, A:'0 20 40 60

Temperature, 0C80 100

FIG. 2. Differential UV melting curves of [d(CATGCATG)]2 inthe absence of ActD (Upper) and as a 1:1 ActD/duplex mixture(Lower).

Biochemistry: Snyder et al.

I-

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

Page 3: CB2 New NJ08903 - PNAS · Proc. Natl. Acad. Sci. USA86 (1989) 3969 heated at a rate of 0.50C/min while the temperature and absorbance at 260 nmwere recorded every 30 sec. Melting

Proc. Natl. Acad. Sci. USA 86 (1989)

Table 1. ActD-induced duplex thermal stabilization data (ATm)and calculated equilibrium binding constants (K) at 370C

ATm, Drug/duplexDuplex K ratio* K,t M-1

1. [d(CATGCATG)]2 21 1 1.4 x 1072. [d(CGTACG)]2 4 1 3.3 x 1053. [d(CTAGATCTAG)h2 0 1 <7.4 x 1044. [d(CATCGATG)]2 5 1 7.1 x 1055. [d(CGTCGACG)]2 33 2 1.4 x6. [d(CGTTAACG)]2 0 2 <3.1 x 1047. [d(TGTCGACA)]2 0 1 <5.4 x 104*Stoichiometries ofthe ActD-duplex complexes for sequences 1 and5 were determined directly by Job plot analysis (30). Other stoi-chiometries were assigned based on considerations of steric acces-sibility. Binding constants for sequences 2, 3, 4, 6, and 7 representupper limits. For example, if the drug/duplex ratio for sequence 4were 2 rather than 1, the calculated K value would be reduced by50%.

tCalorimetric binding enthalpies (see Table 2) were used to convertK(tm) to K(37C). For sequences for which the ActD binding enthalpywas not determined, we assigned a value of zero. Little error isintroduced by this assumption, since ATm is small for these se-quences.tWhen available, calorimetric transition enthalpies were used in thecalculation of K. Otherwise, transition enthalpies were calculatedby the method of Breslauer et al. (32).

A Non-GpC-Containing Duplex That Strongly Binds TwoActD Molecules. The oligomeric duplex [d(CGTCGACG)]2provides a notable exception to the observation that strongActD binding requires the presence of a GpC site. In fact, wefind that this duplex tightly binds not one but two ActDmolecules. This observation contrasts with the behavior of allother non-GpC duplexes we have studied to date (Table 1;refs. 21 and 34). Below, we present experimental evidence forthe strong binding of two ActD molecules to the [d(CGTC-GACG)h2 duplex.The differential melting profile for [d(CGTCGACG)]2 in

the presence of ActD (Fig. 3, solid curve) indicates a tm ofabout 43TC. After addition of one ActD molecule per duplexto the solution (Fig. 3, dashed curve), a new, higher-temperature transition appears, which undoubtedly corre-sponds to the melting ofan ActD-DNA complex. At the sametime, the intensity of the drug-free, lower-temperature tran-sition decreases to about half its original value. Upon addi-tion of a second equivalent of drug (Fig. 3, bold curve), thelow-temperature transition of the drug-free duplex is almosteradicated, while the area under the peak representing thehigher-temperature (750C) transition almost doubles withouta significant change in thermal stability. We propose that the750C transition in Fig. 3 corresponds to melting of the 2:1ActD-DNA complex. Taken together, these observationssuggest that the duplex form ofd(CGTCGACG) tightly bindstwo ActD molecules. A 2:1 ActD/duplex binding stoichiom-

E-4-.11-1'11.I

0 20 40 60Temperature, °C

80

etry also is supported by a Job plot analysis and our opticaltitration data. This 2:1 binding stoichiometry can be used inconjunction with the ATm and the calorimetrically measuredAHb (see below) to calculate an apparent binding constant of1.4 x 107 M-1 for the complexation of ActD with the duplex[d(CGTCGACG)]2 (Table 1). This association constant isnearly identical to the binding constant we measured forActD complexation to the duplex [d(CATGCATG)h2, whichcontains the classic high-affinity GpC binding site. Thus, ourresults indicate that ActD can bind strongly to a non-GpColigomeric DNA duplex. Further, the 2:1 drug/duplex stoi-chiometry suggests that strong non-GpC ActD binding mayrequire the presence of two binding sites that are properlypositioned for cooperative effects to be operative (see Dis-cussion).To further elucidate the sequence-dependent requirements

for this unusual ActD binding, we replaced the central CpGof [d(CGTCGACG)]2 with TpA to form the duplex[d(CGTTAACG)]2. UV melting curves (not shown) and ATmdata (Table 1) revealed that this altered duplex does not binda single ActD molecule despite the presence oftwo GpT sites.In an effort to make a more benign alteration, we changed thetwo terminal G-C base pairs of[d(CGTCGACG)]2 to A-T basepairs, thereby producing [d(TGTCGACA)]2. To our surprise,this altered duplex did not bind a single ActD moleculedespite the retention of all the original GpA and GpT sites.Thus, the ability ofthe non-GpC duplex [d(CGTCGACG)]2 tobind two ActD molecules is unique among the oligomericduplexes we have studied to date. Further studies are re-quired to elucidate precisely the sequence requirements forthe unique ActD binding properties of [d(CGTCGACG)]2

Spectroscopic Evidence for a Novel Mode ofActD Binding to[d(CATGCATG)h2. Previous investigators have used CD tomonitor and to characterize binding of ActD to DNA du-plexes (34-43). Our CD spectra suggest that ActD binds to[d(CGTCGACG)h2 in a manner that is different from its"classic" DNA-binding mode. The bold curve in Fig. 4a

x

4)

10

0-

-3-

-6-

-9-

_ 1 1

0-

-3-

-6-

-9-

- IL-T325

100

FIG. 3. Differential UV melting curves of [d(CGTCGACG)12 inthe absence of ActD (-), as a 1:1 ActD/duplex mixture (---), and asa 2:1 ActD/duplex mixture (-).

I I

375 425 475Wavelength, nm

525

FIG. 4. (a) CD spectra of free ActD (-) and ActD complexedwith salmon testis DNA (o), [d(CGTACG)h2 (A), [d(CATGCATG)12(-.-), and [d(CATGCATGCATG)]2 (-). (b) CD spectra of free ActD(-) and ActD complexed with salmon testis DNA (o) and [d(CGT-CGACG)]2 (-)-

a

--- --0

A/~~~~~~~

£ /,o0 000 0

.N *O

*u0* '00

0009 ~~~~~~~~~~0 0_9 ,,f, _0b 8 o°0 0°0 0o.o 000 00 0

3970 Biochemistry: Snyder et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

Page 4: CB2 New NJ08903 - PNAS · Proc. Natl. Acad. Sci. USA86 (1989) 3969 heated at a rate of 0.50C/min while the temperature and absorbance at 260 nmwere recorded every 30 sec. Melting

Proc. Natl. Acad. Sci. USA 86 (1989) 3971

shows that free ActD in solution yields a CD signal in thevisible range, characterized by a broad minimum at 375 nmand a smaller minimum at 450 nm. For comparative purposes,the CD spectra of ActD bound to salmon testis DNA and toseveral oligomeric DNA duplexes also are presented in Fig.4a. These spectra illustrate the change in the ActD CD signalwhich typically occurs upon binding to double-helical DNA.Specifically, ActD binding to polymeric or oligomeric DNAhost duplexes induces substantial increases in the magnitudesof both negative CD bands. These qualitative CD featuresgenerally are independent of the base sequence of the hostDNA. {Interestingly, Krugh and Young (43) found that ActDbinding to a daunomycin-poly[d(A-T)]2 complex also exhib-its a CD spectrum with two negative bands at 375 and 450nm.} Taken together, these results suggest that an increase inthe intensity ofthe negative bands at 375 and 450 nm providesa good CD fingerprint for classic ActD binding to DNA hostduplexes with GpN steps. The gross similarities in the visibleCD spectra reported to date for duplex-bound ActD may, infact, reflect structural features common to these ActD-DNAcomplexes.We have found only one exception to the CD pattern

described above. This exception is observed when ActDbinds to [d(CGTCGACG)h2. (Recall that this is the onenon-GpC host duplex for which ActD exhibits a high affinity,with a binding stoichiometry of two ActD molecules perduplex.) Inspection of the CD spectrum (dashed curve in Fig.4b) reveals that the binding of ActD to [d(CGTCGACG)]2induces the expected increase in the size of the negative bandat 375 nm. Surprisingly, however, the "normal" minimum at450 nm is replaced by a maximum in the same wavelengthrange. Although CD spectra of this type have been reportedfor aqueous solutions of ActD and mononucleotides (41), weknow of no previous report of a CD spectrum with theseunusual features for a complex between ActD and a DNAduplex.

Calorimetric Evidence for a Novel Mode of ActD Binding to[d(CATGCATG)h2. We used batch calorimetry to character-ize energetically the binding of ActD to the [d(CGTCGA-CG)]2 duplex. For comparative purposes, we performedparallel measurements on oligomeric duplexes that exhibitednormal CD profiles upon ActD binding. Table 2 lists theenthalpy changes (AHb) associated with the binding of ActDto each oligomeric duplex. Inspection of these data revealsthat the binding of ActD to the duplex [d(CGTCGACG)]2 ismore endothermic, by 4-5 kcal/mol of drug, than the bindingof ActD to the other oligomeric host duplexes. Thus, forma-tion ofan (ActD)2-[d(CGTCGACG)]2 complex not only givesrise to an anomalous CD spectrum but also is accompaniedby an atypical binding enthalpy. Tables 1 and 2 also revealthat strong ActD binding to GpC-containing duplexes pro-ceeds with little assistance from the binding enthalpy at 37TC.This feature contrasts with other intercalators such as ethid-ium bromide and daunomycin, which bind strongly to DNAduplexes with large enthalpic driving forces (44, 45). It hasbeen proposed (7, 14, 34, 46-50) that the large entropicdriving force observed for ActD binding to duplex DNA mayreflect solvent effects associated with binding-inducedchanges in the hydrations of the drug and/or the host duplex.

Table 2. Calorimetrically determined ActD-duplexbinding enthalpies

Duplex AHb,* kcal/mol

[d(CATGCATG)]2 -2.7[d(CATGCATGCATG)h2 -2.0[d(CGTACG)]2 -2.0[d(CGTCGACG)]2 +1.8

*Afb was determined at 15TCand includes a correction for disruptionof ActD dimers (see ref. 34).

DISCUSSIONOur UV melting curves reveal that ActD generally does notbind tightly to oligomeric DNA duplexes that lack the high-affinity GpC intercalation site. The one exception to thisgeneralization is provided by the [d(CGTCGACG)]2 duplex,which strongly binds two ActD molecules. The binding ofActD to this oligomeric duplex is accompanied by spectro-scopic and calorimetric changes that are distinct from thosepreviously reported, or observed in this study, for the bindingofActD to other oligomericDNA sequences. Taken together,these observations suggest that the high affinity of ActD for[d(CGTCGACG)]2 reflects a nonclassical mode of ActDbinding.

Cooperative ActD Binding to [d(CGTCGACG)h2. Obvi-ously, our melting curves, CD spectra, and AHb data aloneare insufficient to elucidate the specific molecular interac-tions that stabilize the (ActD)2-[d(CGTCGACG)]2 complex.A detailed molecular model must await the results of NMRand crystallographic analysis. However, the melting-curvedata suggest that formation of the (ActD)2-[d(CGTCGA-CG)]2 complex might involve a cooperative binding mecha-nism (i.e., the binding of the first ActD molecule mayfacilitate the binding of the second ActD molecule) as hasbeen observed by Krugh and coworkers (51, 52) for thebinding of ActD to polymeric DNA host duplexes at lowdrug/DNA ratios. In support of this interpretation, the dif-ferential melting profile of the equimolar ActD-DNA com-plex exhibits two, and only two, well-defined transitions (Fig.3). More importantly, the temperatures of the two transitionscorrespond to the tm values of the free duplex and the 2:1ActD-duplex complex, respectively. These observationssuggest that the two predominant species in an equimolarmixture of ActD and [d(CGTCGACG)]2 are the free duplexand the 2:1 complex (ActD)2-[d(CGTCGACG)]2. For a two-state process, the melting transition for a 1:1 ActD-duplexcomplex would be expected to appear at a temperaturemidway between the transitions of the free duplex and the(ActD)2-[d(CGTCGACG)]2 complex (34). Inspection of thedifferential melting curves in Fig. 3 reveals the absence ofsuch a transition. The absence of the 1:1 ActD-duplexspecies can be rationalized by postulating a highly coopera-tive drug-binding process. In other words, the 1:1 ActD-duplex state may not become significantly populated, be-cause binding of the first ActD molecule greatly facilitatesbinding of the second ActD molecule. Further, the absenceof ActD binding when either [d(CGTTAACG)]2 or [d(TG-TCGACA)]2 serves as the host duplex suggests that cooper-ative ActD binding to non-GpC oligomeric duplexes mayrequire the presence of multiple sites that not only areproperly positioned but also possess appropriate interveningor flanking sequences. These results suggest that the ActDligand "reads" a DNA recognition pattern that is uniquelydefined by an 8-base-pair domain. One should keep in mindthat sequence-dependent influences on local and global con-formations may cause one GpN site to be different fromanother GpN site in the same or a different duplex (53, 54).Such influences may well modulate the ability of a givendinucleotide site to bind a drug. Clearly, further studies arerequired before one can define the exact sequence/confor-mation-dependent requirements associated with the novelmode ofActD binding suggested by the studies reported here.

Special Case of Cooperative ActD Binding. A special case ofthe cooperative binding mechanism described above envi-sions the binding of two ActD molecules to adjacent sites ofthe oligomeric host duplex in a manner that permits drug-drug interactions. [The recent NMR studies of Scott et al.(18, 19) have shown that binding ofActD to two adjacent sitescan occur. However, the DNA duplex employed by Scott etal. contained two adjacent GpC sites, whereas the DNA

Biochemistry: Snyder et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020

Page 5: CB2 New NJ08903 - PNAS · Proc. Natl. Acad. Sci. USA86 (1989) 3969 heated at a rate of 0.50C/min while the temperature and absorbance at 260 nmwere recorded every 30 sec. Melting

Proc. Natl. Acad. Sci. USA 86 (1989)

duplex used in our study contained no GpC.] Obviously,binding ofActD to adjacent sites would be consistent with theobserved 2:1 drug/duplex stoichiometry. The apparent bind-ing cooperativity suggested by our melting-curve data mightresult from favorable drug-drug interactions between theadjacent ActD molecules [Scott et al. (18, 19) observedanticooperativity in their system]. In addition, the observedaberrant binding enthalpy of ActD for [d(CGTCGACG)]2may reflect differential solvation effects (7, 14, 47-50) as wellas differences in drug-drug and drug-DNA interactions whentwo ActD molecules bind to adjacent rather than nonadjacentsites (18, 19). In this connection, the x-ray studies of Bermanand coworkers (47, 49) should be noted. These investigatorsreported the crystal structure of an ActD dimer and showedthat the drug-drug interactions involved in dimerization"saturate" some of the drug sites that normally would beused by monomeric ActD to interact with duplex DNA. If thedimer crystal structure of ActD provides a partial model fordrug-drug interactions that may be expressed when twoActD molecules occupy adjacent sites on DNA, then it isreasonable to expect different thermodynamics and spectralchanges for adjacent-site vs. nonadjacent-site binding of twoActD molecules to duplex DNA.For the reasons discussed above, the calorimetric and

spectroscopic data reported here are consistent with a co-operative binding model that invokes either adjacent- ornonadjacent-site occupancy for formation of the (ActD)2-[d(CGTCGACG)]2 complex. Ultimately, NMR and/or crys-tallographic studies on the (ActD)2-[d(CGTCGACG)]2 com-plex should allow one to distinguish between these twostructural models, thereby providing insight into the molec-ular basis for the intriguing observation reported here-namely, the cooperative binding of two monomeric ActDmolecules to a non-GpC-containing DNA duplex.Concuding Remarks. In this paper, we have presented

evidence that two ActD molecules can bind strongly to anon-GpC-containing oligomeric DNA duplex to form a com-plex that exhibits aberrant spectroscopic and calorimetricbehavior. These observations are consistent with ActD ex-hibiting a sequence-dependent DNA-binding mode distinctfrom its classic binding to isolated GpC sites. Our results raisethe possibility that other naturally occurring DNA-bindingdrugs also may be capable of expressing alternative, nonclas-sical binding modes. Consideration of this possibility is im-portant because the ideas for the design of sequence-specificDNA-binding ligands often originate from the examination ofhow naturally occurring ligands interact with DNA (55, 56).

This work was supported by National Institutes of Health GrantsGM23509 and GM34469 (K.J.B.), the Johnson and Johnson Discov-ery Research Fund (K.J.B.), and the Medical Research Councilunder a Program Grant (O.K.).1. Goldberg, I. H., Rabinowitz, M. & Reich, E. (1962) Proc. Nati.

Acad. Sci. USA 48, 2094-2101.2. Goldberg, I. H. & Rabinowitz, M. (1962) Science 136, 315-316.3. Hamilton, L., Fuller, W. & Reich, E. (1963) Nature (London) 198,

538-540.4. Kahan, E., Kahan, R. M. & Hurwitz, J. (1963) J. Biol. Chem. 238,

2491-2497.5. Haselkorn, R. (1964) Science 143, 682-684.6. Cavalieri, L. & Nemchin, R. (1964) Biochim. Biophys. Acta 87,641.7. Gellert, M., Smithy, C. E., Neville, D. & Felsenfeld, G. (1965) J.

Mol. Biol. 11, 445-457.8. Muller, W. & Crothers, D. M. (1968) J. Mol. Biol. 35, 251-290.9. Wells, R. D. & Larson, J. E. (1970) J. Mol. Biol. 49, 319-342.

10. Sobell, H. M. & Jain, S. C. (1972) J. Mol. Biol. 68, 21-34.11. Krugh, T. R. & Neely, J. W. (1973) Biochemistry 12, 1775-1782.12. Patel, D. J. (1974) Biochemistry 13, 2396-2402.

13. Reid, D. G., Salisbury, S. A. & Williams, D. H. (1983) Biochem-istry 22, 1377-1385.

14. Marky, L. A., Snyder, J. G., Remeta, D. & Breslauer, K. J. (1983)J. Biomol. Struct. Dyn. 1, 487-507.

15. Snyder, J. G. (1983) Biophys. J. 41, 424a (abstr.).16. Brown, S. C., Mullis, K., Levenson, C. & Shafer, R. H. (1984)

Biochemistry 23, 403-408.17. Jones, R. L., Scott, E. V., Zon, G., Marzilli, L. G. & Wilson,

W. D. (1988) Biochemistry 27, 6021-6026.18. Scott, E. V., Jones, R. L., Banville, D. L., Zon, G., Marzilli, L. G.

& Wilson, W. D. (1988) Biochemistry 27, 915-923.19. Scott, E. V., Zon, G., Marzilli, L. G. & Wilson, W. D. (1988)

Biochemistry 27, 7940-7951.20. Takusagawa, F. & Berman, H. M. (1983) Cold Spring Harbor

Symp. Quant. Biol. 47, 315-321.21. Snyder, J. G., Hartman, N. G., d'Estantoit, B. L. & Kennard, 0.

(1987) in Fifth Conversation in Biomolecular Stereodynamics, ed.Sarma, R. H. (Adenine, Guilderland, NY), pp. 200-201 (abstr.).

22. Snyder, J. G., Remeta, D. P., Hartman, N. G., d'Estantoit, B. L.,Kennard, 0. & Breslauer, K. J. (1987) in Fifth Conversation inBiomolecular Stereodynamics, ed. Sarma, R. H. (Adenine, Guild-erland, NY), pp. 202-203 (abstr.).

23. Cantor, R. C. & Schimmel, P. R. (1980) in Biophysical Chemistry(Freeman, San Francisco), pp. 1258-1259.

24. Chen, F.-M. (1988) Biochemistry 27, 1843-1848.25. Allen, F. S. & Gray, D. M. (1984) Biopolymers 23, 2661-2668.26. Allen, F. S., Jones, M. B. & Hollstein, U. (1977) Biophys. J. 20,69-

78.27. Fox, K. R. & Waring, M. J. (1984) Nucleic Acids Res. 12, 9271-

9285.28. Scamrov, A. V. & Beabealashivilli, R. S. (1983) FEBS Lett. 164,

97-101.29. Sanger, W. (1984) Principles of Nucleic Acid Structure (Springer,

New York).30. Marky, L. A. & Breslauer, K. J. (1987) Biopolymers 26, 1601-1620.31. Cantor, R. C. & Schimmel, P. R. (1980) in Biophysical Chemistry

(Freeman, San Francisco), pp. 1135-1138.32. Breslauer, K. J., Frank, R., Blocker, H. & Marky, L. A. (1986)

Proc. Natl. Acad. Sci. USA 83, 3746-3750.33. Mudd, C. P., Berger, R. L., Hopkins, H. P., Friauf, W. S. &

Gibson, C. J. (1982) Biochem. Biophys. Methods 6, 179-203.34. Snyder, J. G. (1985) Ph.D. Thesis (Rutgers Univ., New Brunswick,

NJ).35. Crothers, D. M. (1971) Biopolymers 10, 2147-2160.36. Marky, L. A., Curry, J. & Breslauer, K. J. (1985) in Molecular

Basis of Cancer, ed. Rein, R. (Liss, New York), pp. 155-173.37. Hollstein, U. (1974) Chem. Rev. 74, 625-652.38. Krugh, T. J., Wittlin, F. N. & Cramer, S. P. (1975) Biopolymers 14,

197-210.39. Crothers, D. M., Sabol, S. L., Ratner, D. I. & Muller, W. (1968)

Biochemistry 7, 1817-1823.40. Yamoka, K. & Ziffer, H. (1968) Biochemistry 7, 1001-1008.41. Homer, R. B. (1969) Arch. Biochem. Biophys. 129, 405-407.42. Quadrifoglio, F. & Crescenzi, V. (1974) Biophys. Chem. 2, 64-69.43. Krugh, T. R. & Young, M. A. (1977) Nature (London) 269, 627-

628.44. Chou, W. Y., Marky, L. A., Zaunczkowski, D. & Breslauer, K. J.

(1987) J. Biomol. Struct. Dyn. 5, 345-359.45. Breslauer, K. J., Chou, W. Y., Ferrante, R., Zaunczkowski, D.,

Curry, J., Remeta, D. P., Snyder, J. G. & Marky, L. A. (1987)Proc. Natl. Acad. Sci. USA 84, 8922-8926.

46. Berman, H. M., Sowri, A., Ginell, S. & Beveridge, D. (1988) J.Biomol. Struct. Dyn. 5, 1101-1110.

47. Ginell, S., Lessinger, L. & Berman, H. M. (1988) Biopolymers 27,843-864.

48. Berman, H. M. (1986) Ann. N.Y. Acad. Sci. 482, 166-178.49. Berman, H. M., Ginell, S. L., Sawzik, P. & Lessinger, L. (1988) in

Structure and Expression, eds. Sarma, R. H. & Sarma, M. H.(Adenine, Guilderland, NY), Vol. 2, pp. 317-328.

50. Berman, H. M. (1986) Trans. Am. Cryst. Assoc. 22, 107-119.51. Winkle, S. A. & Krugh, T. R. (1981) Nucleic Acids Res. 9, 3175-

3186.52. Walker, G. T., Stone, M. D. & Krugh, T. R. (1985) Biochemistry

24, 7471-7479.53. Dickerson, R. E. & Drew, H. R. (1981) J. Mol. Biol. 149, 761-786.54. Chen, F.-M. (1988) Biochemistry 27, 6393-6397.55. Gale, E. F., Cundliffe, C., Reynolds, P. E., Richmond, M. H. &

Waring, M. J. (1981) The Molecular Basis of Antibiotic Action(Wiley, New York), pp. 258-401.

56. Dervan, P. B. (1986) Science 232, 464-471.

3972 Biochemistry: Snyder et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

4, 2

020