best binding sites; the clearest footprints are found around the dinucleotide TpG, especially at the sequence TGC. An in vitro transcription assay also suggested ...
Eur. J. Biochem. 209,31-36 (1992) 0FEBS 1992
Footprinting studies of DNA-sequence recognition by nogalamycin Keith R. FOX and Zafir ALAM Department of Physiology and Pharmacology, University of Southampton, England (Received June 12,1992)- EJB 92 0829
We have studied the DNA sequence binding preference of the antitumour antibiotic nogalamycin by DNase-I footprinting using a variety of DNA fragments. The DNA fragments were obtained by cloning synthetic oligonucleotides into longer DNA fragments and were designed to contain isolated ligand-binding sites surrounded by repetitive sequences such as (A),, . (T),, and (AT),. Within regions of (A), . (T),, clear footprints are observed with low concentrations of nogalamycin (< 5 pM), with apparent binding affinities for tetranucleotide sequences which decrease in the order TGCA > AGCT = ACGT > TCGA. In contrast, within regions of (AT),,, the ligand binds best to AGCT; binding to TCGA and TGCA is no stronger than to alternating AT. Within (ATT),, the preference is for ACGT > TCGA. Although each of these binding sites contains all four base pairs, there is no apparent consensus sequence, suggesting that the selectivity is affected by local DNA dynamic and structural effects. At higher drug concentrations ( > 25 pM), nogalamycin prevents DNAse-I cleavage of (AT), but shows no interaction with regions of (AC), . (GT),. Regions of (A), . (T)", which are poorly cut by DNase I, show enhanced rates of cleavage in the presence of low concentrations of nogalamycin, but are protected from cleavage at highcr concentrations. We suggest that this arises because drug binding to adjacent regions distorts the DNA to a structure which is more readily cut by the enzyme and which is better able to bind further ligand molecules.
The antitumour antibiotic nogalamycin (Fig. 1) is unusual in that it possesses bulky groups at both ends of its chromophore [l], yet still binds to DNA by intercalation. Nogalamycin appears to bind by insertion between the base pairs, positioning its sugar residues in both major and minor grooves. Although this model is sterically feasible, it presents considerable dynamic problems, since the minimum width of the antibiotic is 1.2 nm, and it is not possible to open up a potential intercalation site beyond about 1.0 nm. The DNA therefore needs to be disrupted before the drug can bind, and, as a result, the antibiotic may discriminate between sequences on the basis of their dynamic properties. It binds fastest to those sites that are easiest to disrupt [2], yet dissociates more slowly (half-life, 100-10000 s) from the most stable regions ~31. Previous footprinting studies have demonstrated that nogalamycin binds best to regions of alternating purines and pyrimidines, especially when these contain all four DNA bases [4, 51. Regions of alternating AT or GC do not present the best binding sites; the clearest footprints are found around the dinucleotide TpG, especially at the sequence TGC. An in vitro transcription assay also suggested TpG as the preferred binding site [6]. However this selectivity is not absolute, and at higher concentrations the drug binds to almost all sequences, except for poly(dA), in common with many other intercalating drugs. Various NMR [7 - 101 and X-ray crystallographic [ l l 141studies have been performed on complexes of nogalamycin Correspoizdcnce to K. R. Fox, Dept. Physiology and Pharmacology, University of Southampton, Bassett Crescent East, Southampton, England SOY 3TU
with short oligonucleotides. These have provided important information about the potential contacts between the drug and DNA, but in view of the very large ligand concentrations used in these experiments, they need not necessarily represent binding to the best sequences. Indeed, two different oligonucleotide families have been used in these studies, with the drug intercalated at C h (TG) of GCATGC and AGCATGCT [7-91 and at CG of CGTACG [lo-141 and GACGTC [15]. These studies place the nogalose sugar within the DNA minor groove, with the amino sugar in the major groove, although the structures differ in the orientation of the sugar residues with respect to the binding site. The DNA is distorted by the binding of nogalamycin; the base pairs are highly buckled with the chromophore asymmetrically disposed between TG rather than the complementary CA. The DNA is not unwound at the intercalation site, but at the preceding base-pair step [14]. Although some studies have suggested the formation of specific hydrogen bonds between the antibiotic and functional groups in the major and minor grooves [7, 8, 10-131, others claim that hydrogen bonds are not responsible for the sequence-specific interaction 1141.
We have therefore further examined the sequence selectivity of nogalamycin by performing DNase-I footprinting experiments on a series of synthetic DNA fragments. These fragments contain potential nogalamycin-binding sites, flanked by regions of (AT), or (A), .(T),,, which have been cloned into longer DNA fragments. We are also able to detect longer-range drug-induced changes in DNA structure by enhancements in the susceptibility of surrounding regions to
32 urea. Gels were run at 1.5 kV for about 2 h. The gel was then fixed in 10% acetic acid, transferred to Whatmann 3 MM paper, dried under vacuum at 80°C and subjected to autoradiography at - 70°C with an intensifying screen.
RESULTS
H r n O C H 3 OCH,
CH3
Fig. 1. The structure of nogalarnycin.
DNase-I clcavage, as previously described for other sequenceselective antibiotics [16- 181.
Previous footprinting studies, together with crystallographic and NMR data, suggest that nogalamycin binds best to regions of alternating pyrimidines and purines. This binding is strongest at sites containing all four bases and may involve the formation of specific hydrogen bonds to guanine. However, DNase-I footprinting patterns on natural DNA fragments of mixed sequence are often difficult to interpret, since they contain many overlapping binding sites. We have therefore examined the interaction of nogalamycin with several DNA fragments containing isolated potential binding sites which have been cloned into longer DNA fragments.
MATERIALS AND METHODS Drugs and enzymes
GC or CG sites surrounded by blocks of (A),, * (T),,
Nogalamycin was a gift from Dr. P. F. Wiley, Upjohn Company, Kalamazoo. The drug was prepared as a 2-mM stock solution in 10 mM Tris/HCl, pH 8.0, containing 10 mM NaCl, and stored at 4°C in the dark. DNase I was purchased from Sigma and stored as previously described [16- 191. All restriction enzymes were purchased from Pharmacia or Northumbria Biologicals.
We have begun these studies by examining the interaction of nogalamycin with DNA fragments containing potential binding sites which are surrounded by regions of (A),, and (T),,. These are especially useful since, in common with many other intercalating ligands, nogalamycin does not bind to poly(dA) . poly(dT). Fig. 2 presents DNase-I digestion patterns of a DNA fragment containing a dimeric insert of (T),GC(A), in the presence or absence of various concentrations of nogalamycin. The control cleavage pattern has been previous described [ 151 and shows five clear bands around each of the TGCA sites with very little cleavage within (A),, . (T)". In the presence of 1 pM ligand, the cleavage pattern is dramatically altered. Clear footprints can be seen around each of the GC sites with five or six bonds protected from enzyme attack. New cleavage products are also evident, both above and below each site. With 25 pM nogalamycin, the enhancements are much less evident, and there is widespread inhibition of enzyme activity; at 100 pM ligand, DNase-I cleavage is totally abolished. Very similar patterns are seen when the DNA is labelled at either the Hind111 or EcoRl end, as expected for a symmetrical insert. These data suggest that TGCA within a region of (A),, and (T),, constitutes a good nogalamycin-binding site. Fig. 3 shows DNase-I digestion patterns of a fragment containing the dimeric insert (A),GC(T), in the presence or absence of nogalamycin. Cleavage in the control is weak, even around AGCT, with best cutting evident at the centre of the dimer around the (T),(A)9 junction 1151. Due to this poor cutting, it is less easy to assess drug binding around the central AGCT, although a footprint is evident at concentrations above 5 pM. In addition, dramatic increases in DNase-I cleavage are evident in both (A)9 and (T)9, providing further evidence for some interaction between the drug and this DNA sequence. However the footprints are not as clear as those with (T),GC(A), suggesting that this constitutes a weaker binding site. It is also worth noting that enzyme cleavage is not abolished at the highest ligand concentrations. Similar footprinting patterns for fragments containing the inserts (T)15CG(A)15 and (A),,CG(T), are presented in Fig. 4. Control cleavage patterns of these fragments have previously been described [17]. A clear footprint can be seen in the centre of (A)15CC(T),,, evident at concentrations of 25 pM and above. This is accompanied by increases in enzyme
DNA sequences Plasmids containing the synthetic inserts (A)&C(T),, (T)9GC(A)9, (TI 1 5CG(A)15, (A) 1sCG(T) I 5 , (AT),GC(AT) 5. (TA)5GC(TA)5 , T(AT),CG(AT)i 5 , (ATT),CG(AAT),, (TAA),CG(TTA),, (AT)loCCCG(AT)l, and (AC),GC(GT), cloned into the SmaJ site of pUC19 were prepared as previously described [16, 18 - 191. Radiolabelled polylinker fragments containing the inserts were obtained by digesting with HindIII, labelling at the 3' end using reverse transcriptase, and [ E - ~ ~ P I ~ A and T Pcutting again with EcoRl. In some cases, the DNA was 32Plabelled at the opposite end by reversing the order of addition of HindIII and EcoR1. Since the fragment with the insert (A)9GC(T)9 contains a central Hind111 site, the 32P-labelledfragment was obtained by digesting with EcoRl and PstI. DNase-I footprinting DNase-I footprinting reactions were performed as previously described [15-181. 2 pl 32P-labelled DNA (about 5 pmol bp) was mixed with 3 p1 drug solution. Since the association and dissociation reactions of nogalamycin are known to be very long, this complex was left to equilibrate at 37 "C for at least 30 min before digesting with DNase I. Digestion was initiated by adding 3 p1 of a suitable concentration of DNase I (typically 0.03 Ujml), dissolved in 2 mM MgCI2, 2 mM MnClz and 20mM NaCI. 3-pl samples were then removed after 1 min and 5 min, and the reaction stopped by adding 3~180% formamide containing 10 mM EDTA. Samples were heated at 100°C for 3 niin before gel electrophoresis. Gel electrophoresis
The products of DNase-I digestion were separated on 8 12% polyacrylamide gels (40 cm x 0.3 mm) containing 8 M
33 T9GCA9
A
Hindm
B
EcoRl
t
c
Fig. 2. DNase-I digestion patterns of a fragment containing the insert (IJ9CC(A), in the presence or absence of varying concentrations of nogalamycin. The DNA is a HindITIIEcoRI fragment, 32Plabelled a t the 3’ end of the Hind111 site (A) or the EcoRl site (B). Each pair of lanes corresponds to digestion by the enzyme for 1 min and 5 min. con, control lanes; drug concentrations (pM) are shown at the top o f the drug-treated lanes. G, dimethylsulphate-piperidine markers specific for guanine. The markcd scction on the right of each gel shows thc position and length of the insert; the arrows indicate the positions of the GC sites.
cleavage in the surrounding AT regions. In contrast, no clear footprint can be seen with (T)15CG(A)15ralthough the enhanced cleavage within (A), . (T)nis suggestive of some interaction with the drug. On the basis of these results, it appears that nogalamycin binds best to TGCA, less well to AGCT and ACGT, and weaker still to TCGA. To examine this selectivity further, we have performed DNAse-I footprinting experiments on similar tetranucleotide sequences located with regions of alternating A and T residues. GC or CG sites surrounded by blocks of (AT),,
Fig. 5 presents DNase-I digestion patterns for fragments containing the sequences (TA),GC(TA),, (AT),GC(AT), and ?‘(AT),CG(AT),, in the presence or absence of nogalamycin. The only one of these fragments revealing a footprint at low nogalamycin concentrations is (TA)5GC(TA)5,which shows a reduction in DNase-I cleavage around the central AGCT in the presence of 5 pM ligand. At higher ligand concentrations (25 - 100 pM), DNase-I cleavage is almost totally abolished for all three fragments, suggesting that nogalamycin has
Fig. 3. DNase-I digestion patterns of a fragment containing the insert (A),GC(T), in the presence or absence of varying concentrations of nogalarnycin. The DNA is an EcoRI/PstI fragment, 32P labelled at the 3’ end of the EcoRl site. The other details are as for Fig. 2.
bound within (AT),. It therefore appears that, within regions of alternating AT, nogalamycin binds best to the sequence AGCT. Binding to the other sites is weaker and equivalent to that at alternating AT itself.
CG sites flanked by other AT-rich sequences Fig. 6 presents DNase-I digestion of fragments containing the inserts (TAA)4CG(TTA)4 and (ATT)4CG(AAT)4 in the presence or absence of nogalamycin. Both fragments show almost total protection from enzyme cleavage at the highest ligand concentration (100 pM), consistent with the ability of nogalamycin to bind to AT-rich DNA. However, at low concentrations there is a dramatic difference between these DNA fragments. Clear footprints can be seen with (TAA),CG(TTA)4 in the presence of 2 pM nogalamycin, centred around each of the ACGT sequences, extending over 6 bp. In contrast, no changes are apparent with (ATT)4CG(AAT)4. Other sequences
Fig. 7 presents DNase-I footprinting patterns for the sequences (AC),GC(GT), and (AT)loCCCG(AT)lo. We chose
34 DISCUSSION Sequence selectivity
Fig. 4. DNase-I digestion patterns of fragments containing the inserts (T)15CG(A)15and (A),,CG(T),, in the presence or absence of varying concentrations of nogalamycin. The DNA arc both HindIIllEcoRl fragments 32Plabelled at the 3' end of the Hind111 site. The other details are as for Fig. 2.
these fragments since they contain central regions of four adjacent CC, flanked by regions of alternating (AC) . (GT) or (AT),. Each fragment also contains a CG flanked on one side by AT and on the other side by GC. Looking first at (AC),GC(GT),, there is a drug-induced reduction in DNase I cleavage, extending over about 10 bp, centred around the middle of the insert in the sequence ACGCGT. This is generated in the presence of 25 pM ligand and above; no general inhibition of cleavage occurs at the highest concentration, suggesting that (AC), . (GT), does not represent a good nogalamycin-binding site. Although we can not be sure where nogalamycin is binding within this sequence, the large footprint suggests the presence of two or more bound ligands. Possible locations for these could be in the centre of GCGT (ACGC) or CGTG (CAGC). Since 25 pM nogalamycin is required to produce this footprint, it seems that binding to these sites is not as good as that noted for several other sequences, but is roughly equivalent to that at (AT),. N o binding is evident within (AC), . (GT),. With (AT)loCCCG(AT)lo, cleavage is totally abolished at nogalamycin concentrations of 25 pM and above, as noted for several other fragments containing blocks of alternating AT. At lower concentrations, there is a reduction in DNase-I cleavage of the two ApT bonds below the central guanine. This suggests that, within this sequence environment, nogalamycin binds better to CGAT or GATA than the surrounding regions of alternating AT.
The results presented above reveal footprints for certain fragments at low nogalamycin concentrations (< 5 pM) which are located within central GC-rich regions. DNase-I cleavage of the surrounding blocks of A and T is only inhibited at much higher ligand concentrations [25 pM for alternating AT and 50 - 100 pM for (A), . (T),]. These results therefore suggest a requirement for guanine in the nogalamycin-binding site. Within this series of DNA fragments, the clearest footprints are observed with (T),GC(A), and (TAA)4CG(TTA)4. Although these two fragments have almost identical base compositions, the central regions have no dinucleotides in common [(T),GC(A), contains TG(CA) and GC, (TAA),CG(TTA), contains AC (GT) and CG]. In each case, the footprint extends over 5 bp or 6 bp, suggesting that only one drug molccule has bound in the centre of the fragment. The simultaneous binding of two ligand molecules would produce a much larger footprint [similar to that seen with (AC)sGC(GT)5]. The apparent site size of 5 bp or 6 bp is similar to that previously determined from DNase-I footprinting with natural DNA fragments, and is close to the minimum footprint possible on account of the size of DNase I. If we assume that these footprints correspond to nogalamycin binding at the centre of each of these inserts, then the data confirm that nogalamycin does not have an absolute sequence binding requirement, even at its strongest sites. On the basis of previous NMR and crystallographic studies, it would seem likely that nogalamycin intercalates at CG of (TAA)4CG(TTA)4and at TG (CA) of (T),GC(A),. If this is the case then only one ligand binds at the centre of TGCA (at GC) but two drug molecules could bind to each (AC) . (GT) of (TAA)4CG(TTA)4. It is not clear whether the two ligand could be simultaneously bound or whether the observed footprint corresponds to a mixture of two bound forms. Although each of these footprints is centred around CG or TG, these dinucleotides alone cannot be a sufficient determinant of strong nogalamycin-binding since higher concentrations are required to footprint with (A),,CG(T),,. In addition, (A),GC(T), which contains no YG, also yields a footprint, albeit at higher ligand concentrations. The binding selectivity of nogalamycin therefore appears to be complex and is dictated by long-range sequence and/or structural effects. Within the context of surrounding (A), . (T),, the binding affinity of-nogalamycin to tetranucleotide sequences deceases in order TGCA > AGCT = ACGT > TCGA. In contrast, when these sequences are placed within (AT),, the only one yielding a specific footprint is AGCT (we have no data on ACGT). Within (ATT),, the preference is for ACGT > TCGA. The absence of clear footprints with (T), SCG(A)l and (ATT)4CG(AAT)4 may be partly related to some peculiar feature of their structure, since the CGselective ligand, echinomycin, also failed to produce footprints with either sequence [18]. These results further demonstrate that the local environment plays a strong role in determining the sequence-binding characteristics of nogalamycin. The observation that (TAA),CG(TTA), contains a much better binding site than (A)15CG(T)15,even though both contain the central tetranucleotide ACGT, confirms the effect of surrounding sequences and may be related to local DNA dynamics. It is known that poly[d(A-T)] melts at a lower temperature than poly(dA). poly(dT); this may be related to the high propeller twist found in (R), . (Y),, maximising basestacking interactions. Regions of alternating R-Y can not
,
35
Fig. 5. DNase-I digestion patterns of fragments containing the inserts (TA),GC(TA),, (AT),GC(AT), and T(AT)sCG(AT)ISin the presence or absence of varying concentrations of nogalamycin. The arrows indicate the positions of the GC and CG sites. The other details are as for Fig. 2.
Fig. 6. DNase-1 digestion patterns of fragments containing the inserts ('I'AA),CG(TTA), and (ATT),CG(AAT), in the presence or absence of varying concentrations of nogalamycin. The arrows show the position of the CG sites. The other details are as for Fig. 2.
Fig. 7. DNase-I digestion patterns of fragments containing the inserts (AC),CG(GT), and (AT)loCCCG(AT)loin the presence or absence of varying concentrations of nogalamycin. The details are as for Fig. 2.
36 undergo high propeller twist because of the R-R clash that would be generated in the minor grove. These factors must affect the DNA breathing parameters and thereby modulate the selective binding of nogalamycin,
Structural changes Several of the fragments studied revealed enhanced DNase-I digestion in regions surrounding the nogalamycin footprints. This is especially noticeable for the surrounding blocks of (A)" (TL in (T)15CG(A)is, (A)15CG(T)15,(ThGC(A), and (A),GC(T),, and is evident even in sequences which do not yield clear footprints. These have previously been interpreted as arising from drug-induced changes in DNA structure, propagated away from the actual ligandbinding site, which render the DNA more susceptible to enzyme attack. This is frequently found in blocks of (A),, . (T), adjacent to intercalation sites, and results from increases in the local minor groove width caused by unwinding of the DNA helix together with flattening the propeller-twisted base pairs caused by insertion of the rigid, flat chromophore. Studies with other drugs have also shown the presence of enhancements surrounding poor binding sites, which do not yield clear footprints. These confirm that some interaction with the ligand has occurred and suggest that the changes in DNA structure persist for longer than the residence time of the ligand on DNA. Despite the fact that nogalamycin binds poorly to poly(dA) . poly(dT), DNase-I cleavage of several (A),, . (T), blocks is completely inhibited in the presence of elevated nogalamycin concentrations. This may also be related to the structural changes induced by nogalamycin binding to adjacent sites which alter the structure and dynamics of (A), * (T), making them better binding sites. At low nogalamycin concentration, the ligand binds adjacent to these regions, alten'ng their structure and thereby rendering them more susceptible to DNase-I attack; at higher concentrations, this altered structure is able to accept further drug molecules. N o enhancements are detected in adjacent regions of (AT), and (AC), . (GT),, for which cleavage of R-Y is still much better than the corresponding Y-R. '
This work was supported by grants from the Scienceand Engineering Research Council and the Cancer Research Campaign. KRF is a Lister Institute Research Fellow.
REFERENCES 1. Arora, S . K. (1983) J . Am. Chem. Soc. 105,1328-1332. 2. Fox, K. R. & Waring, M. J. (1984) Biochim. Biophyys. Acta 802, 162-168. 3. Fox, K. R., Brassett, C. & Waring, M. J. (1985) Biochim. Biophys. Act0 840, 383 -392. 4. Fox, K. R. & Waring, M. J. (1986) Biochemistry 25, 43494356. 5. Fox, K. R. (1988) Anti-cancer Drug Des. 3, 157-168. 6. White, R. J. & Phillips, I). R. (1989) Biochemistry 28, 42774293. 7. Searle, M. S., Hall, J. G., Denny, W. A. & Wakelin, L. P. G. (1988) Biochemistry 27, 4340 - 4349. 8. Zhang, X. & Patel, D. J. (1990) Biochemistry 29, 9451 -9466. 9. Searle, M. &Lane, A. N. (1992) FEBS Letr. 297, 292-296. 10. Robinson, H., Liaw, Y.-C., van der Marel, G. A., van Boom, J. H. & Wang, A. H.-J. (1990) Nucleic Acids Res. IS, 4851 4858. 11. Liaw, Y.-C. Gao, Y.-G., Robinson, H., van der Marel, G. A., van Boom, J. H. & Wang, A. H.-J. (1989) Biochemistry 28, 9913 - 9918. 12. Williams, L. D., Egli, M., Gao, Q., Bash, P., van der Marel, G. A., van Boom, J. H., Rich, A. & Frederick, C. A. (1990) Proc. Natl Acad. Sci. USA 87,2225 - 2229. 13. Gao, Y.-G., Liaw, Y.-C., Robinson, €1. & Wang, A. H.-J. (1990) Biochemistry 29, 10 307-10 316. 14. Egli, M., Williams, L. D., Frederick, C. A. & Rich, A. (1991) Biochemistry 30, 1364- 1372 15. Searle, M. S. & Bicknell, W. (1992) Eur. J. Biochem. 205, 4558. 16. Waterloh, K. & Fox. K . R. (1991) J . B i d . Chon. 266, 63816388. 17. Cons, B. M. G. & Fox, K. R. (1991) Biochemistry 30, 63146321. 18. Waterloh, K. & Fox, K. R. (2991) Nucleic Acids Res. 19, 67196724. 19. Fox, K. R., Marks, J. N. & Waterloh, K. (1991) Nucleic Acids Res. 19,6725-6730.
