InI. J. Rodmion Oncology Bml. Phvs , Vol. 23, PP. 579-584 Printed in the U.S.A. All rights reserved.

0360-3016/92 $5.00 + II0 Copyright 0 1992 Pergamon Press Ltd.

0 Biology Original Contribution DNA LIGANDS

AS RADIOMODIFIERS: STUDIES WITH MINOR-GROOVE BINDING BIBENZIMIDAZOLES R. F. MARTIN,

PH.D.

AND L. DENISON,

PH.D.

Molecular Sciences Group, Peter MacCaIlum Cancer Institute, 48 1 Lt. Lonsdale St., Melbourne, 3000, Australia An iodinated bibenzimidazole, iodoHoecbst 33258, was previously reported to markedly sensitize DNA and cells to UV-A, exemplifying the potential of iodinated DNA llgands as radlosensitizers, a rational extension of sensitization by halogenated pyrimidines. However, unlike the latter sensitizers, iodoHoechst 33258 is not a sensitizer of ionizing radiation, presumably due to the innate radioprotective properties of the uniodinated ligand. Experiments with purified DNA show that both Hoechst 33258 and Hoechst 33342 decrease the yield the radiation-induced DNA strand breakage. The ligands bind at discrete sites in the minor groove of DNA, and analysis on DNA sequencing gels show pronounced protection at the ligand binding sites, as well as more generalized protection. The extent of protection of strand breakage on plasmid DNA and the fact that it persists in the presence of 0.5 M NaCl (which prevents low aflinity ionic binding between the high affinity sites) suggests that the protective effects of bound ligand are not confined to the high affinity binding sites in the minor groove. The mechanisms of this generalized protection is unknown, but there is some evidence indicating that the H-atom donation from the ligand may account for the site-specific protection. The extent of protection is much diminished, but still evident, in the presence of 100 mM mannitol, a known hydroxyl radical scavenger, indicating that some of the protective effects might relate to DNA damage mediated by direct action. Further evaluation of the mechanisms of protection should enable development of both more active radioprotectors and, by elimination of the radioprotectlve features from halogenated DNA ligands, more effective radiosensitizers. Radioprotectors, donation.

DNA ligands, Hoechst

33342, Hoechst

33258, Plasmid DNA, DNA strand breaks, H-atom

livery of sensitizers to tumors and of radioprotectors to critical normal cell populations. We have directed our attention to the first of these challenges, although it seems reasonable to expect that the second will not be insurmountable. Two features of the requirements for preferential delivery of radiomodifiers suggest grounds for optimism. One is that the extent of preferential delivery needs only to be modest to provide useful therapeutic gain, and the second that the requirement for preferential delivery is restricted to the treatment field. This latter point implies that the radiomodifiers are non-toxic per se, and this consideration has influenced our choice of DNA ligands. The use of bibenzimidazoles such as Hoechst 33342 as a basis for fluorescence activated cell sorting according to DNA content, at concentrations which achieve saturation binding with little impact on cell survival is notable (3), but there is considerable variation in sensitivity to these ligands between different cell lines ( 17). The marked UV-sensitization of DNA and cells by iodoHoechst 33258 illustrates the potential of DNA li-

INTRODUCTION

The complexity of the response of cells to ionizing radiation provides many possible approaches to the design of radiomodifiers. Those approaches that have attracted most attention are based on modulating redox state, DNA repair, or levels of endogenous thiols. We have decided to focus on trying to influence the level of initial damage in irradiated DNA, particularly DNA strand breaks, using DNA ligands. The idea of exploiting DNA ligands is an obvious one and by no means novel, especially for radiosensitization ( 13- 15) and there are reports of a correlation between DNA binding and radioprotective activity of some aminothiols ( 18, 19). The development of radiomodifiers that might be useful in improving cancer radiotherapy invokes two distinct challenges. The first of these is the identification of agents which substantially influence the radiation response at concentrations that can be achieved physiologically. The second is the development of systems for preferential dePresented at the Seventh International Conference in Chemical Modifiers of Cancer Treatment, 2-5 February 1991, Clearwater, FL. Reprint requests to: Dr. R. F. Martin.

The research was supported in part, by a grant from the Australian Research Council. Accepted for publication 3 1 December 1991. 579

580

I. J. Radiation Oncology 0 Biology 0 Physics

gands as radiomodifiers (11). Sensitization by iodoHoechst 33258 can be rationalized as an extension of the wellestablished sensitization of DNA ceils by covalent incorporation of iodine or bromine, principally as the halogenated uracil (6, 7). Whereas the halogen atom is covalently incorporated into DNA in the original case, the iodinated DNA ligand enables non-covalent association of the halogen atom (or more precisely the carbon-halogen bond) with DNA. Although the sensitization to UV-B by halouracils is much more pronounced than that to ionizing radiation, both effects are thought to be mediated by the uracilyl radical, formed by photolytic disassociation of the carbon-halogen bond, or by indirect action, respectively. The carbon-centered radical on the uracil induces a strand break in the nucleotide immediately 5’to the modified nucleotide, presumably by abstraction of a H-atom from one of the sugar carbons. On the basis of the structure of the B-DNA helix, the 2’-deoxyribosyl carbon has been suggested as the most likely candidate for attack (6,7). In the case of iodoHoechst 33258, where the halogen is covalently attached to a ligand non-covalently associated with DNA (in contrast to covalent attachment to a DNA base), the carbon-centered radical produced by photolysis is located on the ligand bound in the minor groove. Apparently the stereochemistry of the iodoHoechst/DNA complex is such that the radical is located close to deoxyribosyl hydrogens, resulting in H-atom abstraction and strand cleavage. Indeed the sites of cleavage reflect the discrete binding sites of the ligand in DNA and our results suggest that H-atom abstraction is from the 5’-deoxyribosyl carbon (11). Disappointingly, although iodoHoechst markedly sensitizes DNA and cells to UV, we have failed to demonstrate even modest sensitization to ionizing radiation. We suggest that this failure is due to the innate radioprotective properties of the non-halogenated bibenzimidazoles. Radioprotection of cells by Hoechst 33342 has already been described (16, 2 1). This report focuses on experiments with purified DNA, aimed at determining the molecular basis of the radioprotection. The basic observation of inhibition of strand-breakage is detailed in a separate study (2), whereas this report emphasizes later experiments which address two specific mechanistic issues. The first of these concerns the possibility that the protective effect of the DNA ligand is due to hydroxyl radical scavenging, and the relative contributions of bound vs (excess) unbound ligand. Furthermore, the use of 0.5 M NaCl provides the possibility of distinguishing between the effects of high affinity site specific minor groove binding vs low affinity ionic binding (8, 9). The second issue addressed involves investigation of a free radical scavenger (mannitol) in experiments with purified DNA, in an attempt to ascertain whether all the Hoechst 33342 protection ob-

* Eppendorf, Hamburg, Germany. +Aldrich Chemical Company, Milwaukee,

Volume 23, Number 3, 1992

served in cell culture experiments is mediated by indirect action. Although elucidation of the mechanism of radioprotection by Hoechst 33342 may provide approaches to the design of even better radioprotectors, it would also have an important impact on the design of novel radiosensitizers. If the radioprotective features can be “designed-out” of iodoHoechst analogues, then the marked sensitization by these ligands of UV-irradiation, may extend to ionizing radiation. METHODS

AND MATERIALS

Samples of pBR322 DNA (Boehringer Mannheim) were irradiated in 1.5 ml microfuge* tubes in 2.5 mM phosphate buffer (pH 6.8), using either 6oCo or ‘37Cs gamma sources. Hoechst 33258 and Hoechst 33342 were obtained from Aldrich,+ and used without further purification. Strand breakage in pBR322 DNA was assayed by agarose gel electrophoresis as described (1 l), using laser densitometryt of Polaroid Type 665 negatives to quantitate the relative amounts of ethidium bromide-stained supercoiled, nicked, and linear species. After correcting for the small level of breakage in the unit-radiated control, the densitometric data was fitted using a non-linear regression program to obtain the value for the constant a in the expression y = e+. Dose modifying factors were then obtained from ratios of a values. Radiation-induced strand breakage in 32P-end labeled DNA (the HindIII/DdeI fragment from pBR322) was analyzed on DNA sequencing gels as described previously (10-12). RESULTS Figure 1 illustrates the protective effect of Hoechst binding on strand-breakage induced in plasmid DNA by 60Co-gamma irradiation. The exponential decrease in the fraction of intact plasmid molecules with increasing radiation dose, reflects the linear dose-response for induction of single-strand breaks. The presence of 25 uM Hoechst 33258 suppresses strand breakage; the dose modifying factor being 7.5. The results of a similar experiment are shown in Figure 2, however, in this case 0.5 M NaCl is included with the 2.5 mM phosphate buffer. The addition of salt results in some protection per se (dose modifying factor of 1.6), but does not compromise the extent of protection by 25 uM Hoechst 33258. The dose modifying factors, relative to the respective controls, are 9.9 and 9.6 for with and without salt, respectively. Further experiments of the sort shown in Figure 1 were done at different concentrations of Hoechst 33258, for both 2.5 mM phosphate buffer alone and for phosphate buffer plus 0.5 M NaCI. The D37 (or Do) was obtained

* Biomed Instruments, Fullerton, CA. WI.

DNA

1

,

u .-

I

I

1

E

cn 0

ligands as radiomodifiers 0

581

R. F. MARTIN AND L. DENISON

I

A

ii + : +-

-c

A

0.3

c

.-0 4

4: t

A

A

\ 0.1 0

I

I

I

I

10

20

30

40

50

-0

5

10

15

20

30

35

40

45

50

[H33258]

Dose(Gy) Fig. 1. Protection by Hoechst 33258 of strand breakage in pBR322 DNA by 60Co-gamma irradiation. Plasmid DNA (37.5 PM in bp) in 2.5 mM phosphate buffer was irradiated at the indicated doses, and the fraction of the DNA remaining as the intact supercoiled species determined as described in Methods and Materials. Some samples included Hoechst 33258 (A) at a concentration of 25 PM, compared to the controls (0) without added ligand.

from the gradients of log [intact plasmid] vs radiation dose, and plotted against ligand concentration, as shown in Figure 3. The extent of protection clearly increases as more and more l&and is added, even up to 50 uM Hoechst (compared to 75 uM DNA bp). The inclusion of mannitol, a hydroxyl radical scavenger, during irradiation markedly protects against strand

0

25

I

I

I

I

,

5

10

15

20

25

30

Fig. 3. Effect on the concentration of Hoechst 33258 on protection of pBR322 DNA. Experiments of the sort described for Figure 2 were conducted at various ligands concentrations, the results were plotted as shown for Figures 1 and 2, and the Dj7 value obtained from the gradient for each ligand concentration. The abscissa shows the ligand concentration in micromolar. There were two series of experiments, one in phosphate buffer alone (0) and the other with NaCl added to 0.5 M (+).

breakage. The protection is concentration-dependent but maximal at 100 mM, for which the dose-modifying factor is 95 (results not shown). When both 100 mM mannitol and 25 uM Hoechst 33258 are included, the DNA ligand confers additional protection-a dose modifying factor of 1.4 relative to mannitol alone (Fig. 4). Analysis of radiation-induced strand breakage in 32P-end-labeled DNA by DNA sequencing techniques shows that protection by Hoechst 33342 and Hoechst 33258 is enhanced at the

0.1 0

I

I

,

I

I

100

200

300

400

500

600

Dose(Gy) Dose(Gy) Fig. 2. Effect of 0.5 M NaCl on protection by Hoechst 33258. The experiment is similar to that in Figure 1 except that the radiation source was “‘Cs. All samples contained pBR322 DNA (75 uM bp) and 2.5 mM phosphate buffer. Some samples (hlled symbols) also contained 0.5 M NaCl. Two sets of samples included Hoechst 33258 (Cl, ?? ) and two sets had DNA only (0, 0).

Fig. 4. Effect of 100 mM mannitol on protection by Hoechst 33258. The experiment is similar to that in Figure 2 except that none of the samples contained NaCl. Instead, all samples contained 100 mM mannitol in 2.5 mM phosphate buffer. Hoechst 33258 (0) was added to some samples, compared the controls (H) with pBR322 (75 uM) without added ligand.

582

I. J. Radiation Oncology 0 Biology 0 Physics

ligand binding sites. For example, at the EcoRl site GAATTC (also a ligand binding site) in a 100 bp restriction fragment of pBR322, the amount of radiation-induced fragments resulting from breakage at the underlined Ts: GAATYJC, is substantially reduced in the presence of Hoechst 33258 or Hoechst 33342 (2). This specific depression of cleavage by bound ligand is also evident in the presence of 100 mM mannitol (Fig. 5).

DISCUSSION

Hoechst 33342 and Hoechst 33258 are known to bind in the minor groove of DNA at discrete sites (5, lo- 12), and the results of studies involving irradiation in 2.5 mM phosphate suggest that although protection was focussed at the binding sites, there also appeared to be protection between binding sites. In view of the possibility that the global effect could be mediated by low affinity ligand binding between the major (high affinity) sites, we investigated protection at 0.5 M NaCl-conditions which suppress the low affinity ionic binding (8, 9). The protection observed in Figure 2 is therefore presumably mediated by ligand bound at the high affinity sites. Moreover, the extent of protection seems too high to be accounted for only in terms of site specific effects. This intuitive assessment is borne-out by simple calculations. If the DNA is considered as a heterogeneous target with two regions with different radiosensitivities distinguished by a protection factor x, the probability of a single radiation-induced break in the entire target can be calculated as a function of x and of the relative portions of protected and normal regions. The protected region represents the (combined) ligand binding sites. In such simulations, even when x is 10, and the proportion of the protected region is 20%, the overall protection of the total target relative

Fig. 5. Site-specific protection by Hoechst ligands in the presence of 100 mM mannitol. The 100 bp Dde I/Hind III fragment from pBR322 was 3’end labeled with 32P,mixed with carrier pBR322 DNA to 75 uMbp in 2.5 mM phosphate buffer containing 100 mM mannitol, and irradiated with r3’Cs-gammas (1500 Gy). Further details are: lane 1, unirradiated control; lane 2, irradiated control; lane 3, 25 uM Hoechst 33258; lane 4, 25 uM Hoechst 33342; lanes 5 & 6, Maxam-Gilbert sequencing markers, G + A and C + T, respectively. A Hoechst binding site at the EcGRl site (GAATTC).

Volume 23, Number 3, 1992

to a homogeneous (unprotected) target of the same size is 1.22. The results of Figure 2 therefore indicate that the protective effect of the ligand is not confined to the binding sites. Such global protection implies some sort of energy or lesion transfer along DNA, as well as H-atom, electron, or energy transfer between the ligand and DNA. In the above analysis, the 20% figure for the protected region is based on the occurrence of Hoechst binding sites, which are comprised of three or more consecutive AT base pairs (5, lo- 12), in pBR322. There is approximately one such site per 20 bp in the pBR322 sequence, and the ligand occupies about 4 bp. Thus in the Hoechst 33258 titration experiment (Fig. 3), one might have expected that in the presence of 0.5 M NaCl, the high affinity sites would be saturated after adding about 4-5 uM ligand (per 75 uM DNA bp). Thus assuming that DNA-bound ligand would be a much more effective protector than unbound ligand, the expected results for the Figure 3 experiment would be a decreasing protective effect of additional ligand at the higher ligand concentrations, rather than the linear response that was observed. This could be explained by the dye-mediated binding described by Loontiens et al. (9), in which multiple ligand molecules are bound at each site on DNA, stabilized by l&and-ligand interactions. It could be argued that the complexity of the Hoechst/ DNA interaction raises some uncertainty about the binding model for 0.5 M NaCl in which about 80% of the DNA is ligand-free, and hence about the idea of “protection at a distance” by bound ligand. However, it does seem likely that it is DNA-bound, rather than free ligand that imparts any protective effect that is due to hydroxyl radical scavenging. Blazek and Peak (1) have studied a number of different hydroxyl radical scavengers and presented the results as in Figure 3. The gradients of the Do vs [protector] plots vary substantially, directly reflecting the rate constants of the hydroxyl radical/protector reaction. The direct relationship between the gradient of these plots and the rate constants for reaction with hydroxyl radical is taken as evidence that protection is due to hydroxyl radical scavenging alone. Thus if we assume that Hoechst 33258 protection is due to hydroxyl radical scavenging, and compare the Figure 3 data with that of Blazek and Peak, then we can estimate the apparent rate constant. Although we have corrected for the difference in size between pBR322 and the plasmid DNA used by Blazek and Peak, the comparison can only provide an estimate; other factors such as the buffer, and radiation quality will also have some impact. Nevertheless, the comparison yields a rate constant of 2.7 X 10” mole dmp3 set-’ for hydroxyl radical scavenging for Hoechst 33258. This is much higher than the expected value of 109- 10” mole dmm3 see-‘. The high value of the apparent rate constant is consistent with the idea that if protection by the DNA ligand is due to hydroxyl radical scavenging, it is DNA-bound ligands that are involved. A recent study comparing radioprotection by aminothiols reported marked differences in rates of DNA repair depending on

DNA ligands as radiomodifiers 0 R. F. MARTINAND L. DENISON

the charge on the aminothiol, and hence the concentration of the thiol close to DNA (4). Finally, it should be noted that all the protectors evaluated by Blazek and Peak were non-DNA binders. Although, there is some uncertainty about the mode of protection involved for DNA between high-affinity ligand binding sites, it is clear from the DNA sequencing gel analysis that the protective effect is more pronounced at the ligand binding sites. This site-specific protection could be viewed simply in terms of the ligand physically restricting access of radical species to the regions of DNA comprising the ligand binding sites. However, the fact that some site-specific protection is observed in the presence of mannitol (Fig. 5) suggests that not all site specific protection is due to restriction of access of hydroxyl radicals to the DNA at the ligand binding sites. On the basis of detailed investigation of the pattern of site-specific protection at the EcoR 1 site (GAATTC), and comparison of the published X ray crystal structure (20) of a complex of Hoechst 33258 and a dodecamer containing the GAATTC sequence, we suggested (2) that H-atom donation from the ligand to radiation-induced radicals on DNA could account for site-specific protection. More particularly, it was suggested that H-atom donation involves benzimidazole NHs and radiation-induced radicals on 4’~deoxyribosyl carbons. Such a mechanism provides an explanation for protection by bound ligand of DNA damage mediated by direct action, and thus for the results of the experiment in Figure 4 where Hoechst 33258 conferred (additional) protection against the induction of strand breaks in plasmid DNA in the presence of 100 mM mannitol. Likewise, the observed site-specific protection in the presence of 100 mM mannitol (Fig. 5) might suggest part (albeit a minor part) of the protection by the DNA ligands might be mediated by direct action, which could be important in the observed radioprotection of cells by

583

Hoechst 33342. However, this interpretation assumes that 100 mM mannitol is completely effective in scavenging hydroxyl radicals, and ignores the possible involvement of mannitol-derived radical species in DNA damage. Accordingly, we are now investigating other scavengers such as TRIS and formate, and the preliminary results do indicate some differences. In summary we conclude that: 1. the observed protection

2.

3.

4.

5.

6.

of radiation-induced DNA strand-breakage by Hoechst 33342 and Hoechst 33258 might account for the earlier observations (1, 13, 17) of radioprotection of cells by Hoechst 33342, the protection is comprised of both focused protection at the ligand binding sites, superimposed on a generalized protection, H-atom donation from the ligand to radiation-induced radicals in DNA may account for some of the sitespecific protection, part of the global protection could be due to DNAligands bound at high-affinity sites exerting some protective effect on the DNA between the binding sites, by an unknown mechanism, if hydroxyl radical scavenging is an important component of protection by the ligands, then it is largely mediated by DNA-bound species, rather than free ligand, and a portion of the protective effects might relate to damage mediated by direct action.

Presumably, the further elucidation of the molecular mechanisms of protection will help guide the development of more potent radioprotectors. Moreover, elimination of the radioprotective features from the halogenated ligands, will presumably lead to the development of more effective radiosensitizers.

REFERENCES I.

2.

3. 4. 5. 6.

Blazek, E. R.; Peak, M. J. The role of hydroxyl radical quenching in the protection by acetate and ethylenediamine tetracetate of supercoded plasmid DNA from ionising radiation-induced strand breakage. Int. J. Radiat. Biol. 53: 237-247; 1988. Denison, L.; Haigh, A.; D’Cunha, G.; Martin, R. F. DNA ligands as radioprotectors: Molecular studies with Hoechst 33342 and Hoechst 33258. Int. J. Radiat. Biol. 61: 69-81; 1992. Durand, R. E. Use of Hoechst 33342 for cell selection from multicell systems. J. Histochem. Cytochem. 30: 117-122; 1982. Fahey, R. C.; Prise, K. M.; Stratford, M. R. L.; Watfa, R. R.; Michael, B. D. Int. J. Radiat. Biol. 59: 901-9 17; 1991. Harshman, K. D.; Dervan, P. B. Molecular recognition of B-DNA by Hoechst 33258. Nucl. Acids Res. 13: 4825-4835; 1985. Hutchinson, F. The lesions produced by ultra-violet light in DNA containing 5-bromouracil. Quart. Rev. Biophys. 6: 201-246; 1973.

7. Hutchinson, F.; Kohnlein, W. The photochemistry of 5bromouracil and 5-iodouracil in DNA. Progr. Mol. Subcell. Biol. 7: l-42; 1980. 8. Latt, S. A.; Stetten, G. Spectral studies on 33258 Hoechst and related bisbenzimidazoles useful for fluorescent detection of deoxyribonucleic acid synthesis. J. Histochem. Cytochem. 24: 24-33; 1976. 9. Loontiens, F. G.; Regenfuss, P.; Zechel, A.; Dumortier, L.; Clegg, R. M. Binding characteristics of Hoechst 33258 with calf thymus DNA, Poly [d(A-T)] and d (CCGGAATTCCGG): multiple stoichiometries and determination of tight binding with a wide spectrum of site affinities. Biochemistry 29: 9029-9039; 1990. 10. Martin, R. F.; Holmes, N. Use of an ‘2sI-labelJed DNA ligand to probe DNA structure. Nature 302: 452-454; 1983. 11. Martin, R. F.; Murray, V.; D’Cunha, G.; Pardee, M.; Kampouris, E.; Haigh, A.; Kelly, D. P.; Hodgson, G. S. Radiation sensitization by an iodine-labelled DNA ligand. Int. J. Radiat. Biol. 57: 939-946; 1990. 12. Murray, V.; Martin, R. F. The sequence specificity of ‘251-

584

13. 14.

15.

16.

17.

I. J. Radiation Oncology 0 Biology 0 Physics

labelled Hoechst 33258 in intact human cells. J. Mol. Biol. 201: 437-442; 1988. Nias, A. H. W. Radiation and platinum-drug interactionreview. Int. J. Radiat. Biol. 48: 297-3 14; 1987. Roberts, P. B.; Anderson R. F.; Wilson, W. R. Hypoxiaselective radiosensitization of mammalian cells by nitracrine, an electron-afhnic DNA intercalator. Int. J. Radiat. Biol. 51: 641-654; 1987. Skov, K. A. Modification of radiation by metal complexes: a review with emphasis of nonplatinum Studies. Radiat. Res. 112: 217-242; 1987. Smith, P. J.; Anderson, C. 0. Modification of the radiation sensitivity of human tumour cells by a bis-benzimidazole derivative. Int. J. Radiat. Biol. 46: 331-344; 1984. Smith, P. J.; Lacy, M.; Debenham, P. G.; Watson, J. V. A mammalian cell mutant with enhanced capacity to disso-

Volume 23, Number 3, 1992

18.

19.

20.

2 1.

ciate a bisbenzimidazole dye-DNA complex. Carcinogenesis 9: 485-490; 1988. Smoluk, G. D.; Fahey, R. C.; Ward, J. F. 1986 Equilibrium dialysis of the binding of radioprotector compounds to DNA. Radiat. Res. 107: 194-204; 1986. Smoluk, G. D.; Fahey, R. C.; Ward, J. F. Interaction of glutathione and other low-molecular-weight thiols with DNA: evidence for counterion condensation and co-ion depletion near DNA. Radiat. Res. 114: 3-10; 1988. Teng, M.; Usman, N.; Frederick, C. A.; Wang, A. H. J. The Molecular Structure of the complex of Hoechst 33258 and the DNA dodecamer d(CGCGAATTCGCG)2. Nucl. Acids Res. 16: 2671-2690; 1988. Young, S. D.; Hill, R. P. Radiation sensitivity of tumour cells strand in vitro or in vivo with the benzimidazole fluorescence Hoechst 33342. Brit. J. Cancer 60: 7 15-72 1; 1989.