Interaction of Cationic Porphyrins with DNA

Utilizing linear dichroism (LD), circular dichroism (CD), and fluorescence energy transfer, the binding geometries of a series of Co3+-porphyrins and their free ligands were examined. The compounds studied were Co-meso-tetrakis(N-methylpyridinium-4-yl)porphyin (CoTMPyP) and its free ligand (H2TMPyP), Co-meso-tetrakis(N-n-butylpyridinium-4-yl)porphyrin (CoTBPyP) and its free ligand (H2TBPyP), and Co-meso-tetrakis(N-n-octylpyridinium-4-yl)porphyrin (CoTOPyP). The two non-metalloporphyrins exhibit negative LD, having angles of roughly 75' relative to the DNA helix axis. They also display negative CD and a significant contact energy transfer from the DNA bases. On the other hand, the three metalloporphyrins display orientation angles of roughly 45' between the porphyrin plane and the helix axis of DNA. Furthermore, they exhibit positive CD and no contact energy transfer from DNA bases. These observations show that the metalloporphyrins are not intercalated whereas non-metalloporphyrins having four freely rotating meso-aryl groups intercalate between the base pairs of DNA. In the presence of KHSOs, the cobalt porphyrins cleave closed circular PM2 DNA in a single strand manner, i.e., a single activation event on the porphyrin leads to a break in one of the DNA strands. A kinetic analysis of the cleavage data revealed that cleavage rates are in the order CoTMPyP > CoTBPyP > CoTOPyP with the difference being due to different DNA affinities rather than differences in cleavage rate-constants. Based on these and earlier observations, the metalloporphyrins appear bound to a partially melted region of DNA. Water-soluble cationic porphyrins bind to DNA. The pioneering work of Fiel (1989) and Pasternack and their coworkers (Pasternack & Gibbs, 1989) established that the binding mechanism could be modulated by the nature of the metal ion and the size and location of the substituent groups on the periphery of the porphyrin. Generally, the free bases and square planar complexes such as those with Ni2+ and Cu2+ intercalate between the base pairs of DNA. For complexes having axially bound ligands such as those with Mn3+, Fe3+, and Co3+ or those with bulky substituents located on the periphery of the structure, intercalation is blocked and "outside" binding occurs. Recent studies show that intercalation versus outside binding may also be influenced by the charge on the porphyrin core (Marzilli et al., 1992; Kuroda et al., 1990). The most extensively studied DNA binding porphyrin is meso-tetrakis(N-methylpyridinium-4-yl)porphyrin, H2TMPyP (Figure 1). As was originally demonstrated by Fiel and coworkers (Fiel et al., 1982), FeTMPyP can be chemically activated to produce breaks in DNA. Additional work by the groups of Dabrowiak and co-workers (Raner et al., 1989; Ward et al., 1986a,b) and of Bernadou and Meunier and their co-workers (Fouguet et al., 1987; Pratviel et al., 1989; Bernadou et al., 1989) established that M-TMPyP where M is Fe3+, Mn3+, and Co3+ can be activated with a variety of oxidizing agents to produce breaks in DNA. Analysis of Support to P.C. from the National Science Foundation, Research Experience for Undergraduates [CHEM-92005011, is greatly acknowledged. * To whom correspondence should be addressed. 11 Chalmers University of Technology. t Yeungnam University. 1 Syracuse University. 1 Johnson-Matthey, Inc. * Abstract published in Advance ACS Abstracts, December 15,1993. R R

metalloporphyrins display orientation angles of roughly 45' between the porphyrin plane and the helix axis of DNA. Furthermore, they exhibit positive CD and no contact energy transfer from DNA bases. These observations show that the metalloporphyrins are not intercalated whereas non-metalloporphyrins having four freely rotating meso-aryl groups intercalate between the base pairs of DNA. In the presence of KHSOs, the cobalt porphyrins cleave closed circular PM2 DNA in a single strand manner, i.e., a single activation event on the porphyrin leads to a break in one of the DNA strands. A kinetic analysis of the cleavage data revealed that cleavage rates are in the order CoTMPyP > CoTBPyP > CoTOPyP with the difference being due to different DNA affinities rather than differences in cleavage rate-constants. Based on these and earlier observations, the metalloporphyrins appear bound to a partially melted region of DNA.
Water-soluble cationic porphyrins bind to DNA. The pioneering work of Fiel (1989) and Pasternack and their coworkers ) established that the binding mechanism could be modulated by the nature of the metal ion and the size and location of the substituent groups on the periphery of the porphyrin. Generally, the free bases and square planar complexes such as those with Ni2+ and Cu2+ intercalate between the base pairs of DNA. For complexes having axially bound ligands such as those with Mn3+, Fe3+, and Co3+ or those with bulky substituents located on the periphery of the structure, intercalation is blocked and "outside" binding occurs. Recent studies show that intercalation versus outside binding may also be influenced by the charge on the porphyrin core (Marzilli et al., 1992;Kuroda et al., 1990).
The most extensively studied DNA binding porphyrin is meso-tetrakis(N-methylpyridinium-4-yl)porphyrin, H2TMPyP (Figure 1). As was originally demonstrated by Fiel and coworkers (Fiel et al., 1982), FeTMPyP can be chemically activated to produce breaks in DNA. Additional work by the groups of Dabrowiak and co-workers (Raner et   The cobalt porphyrins have axial ligands ((21-, -OH, H20) above and below the porphyrin plane.
cleavage products using the sequencing methods showed that the preferred binding site is a triplet containing only adenine and thymine. In the presence of the oxidizing agent KHS05, MnTMPyPcan be used as a footprinting agent for AT-specific drugs . Also, the temperature dependence of the compound's cleavage specificity suggests that it may be a probe for low-melting regions of DNA (Raner et al., 1988(Raner et al., , 1989. A detailed study by Bernadou and Meunier and their coworkers (Pratviel et al., 1991;Gasmi et al., 1991;Piti6 et al., 1992) showed that MnTMPyP cleaves DNA from its minor grove. For calf thymus DNA, the activated oxo species attacks Cl'-H, which is located deep within the minor groove. For oligonucleotides, cleavage selectivity occurs on the 3' sides of three contiguous AT base pairs. The oxo species primarily hydroxylates the Scarbon of deoxyribose, located in theminor groove, ultimately causing a break which leaves a 3'-phosphate and a 5'-aldehyde at the scission site.
In addition to chemical activation, porphyrins can be photochemically activated to attack biological molecules. For example, photodynamic therapy (Berg et al., 1978;Diamond et al., 1977) very likely involves damage to membranes through the photoinduced production of singlet oxygen. Although the details of the cleavage mechanisms are not known, studies with plasmid DNA indicate that certain metalloporphyrins can cause both single-and double-strand breaks in DNA (Praseuth et al., 1986;Fiel et al., 1981).
The structure of the porphyrin-DNA complex has been characterized using a variety of physical techniques. Extensive NMR, equilibrium dialysis, flow dichroism, and viscometry measurements on oligomeric and polymeric DNA's by Marzilli, Wilson, and their co-workers [for an excellent review see Marzilli (1990);Strickland et al., 19901 have supported the original proposal that porphyrins intercalate into GC-rich regions and that they bind in an outside manner at AT sites (Fiel, 1989). Geacintov et al. (1987) found that H2TMPyP, which intercalates, is aligned perpendicualr to the DNA helix axis, while ZnTMPyP, which binds outside to AT sites (Fiel, 1989;Ward et al., 1986a,b), lies at an angleof 62-67O relative to the helix axis. These structural features are consistent with the results of molecular modeling and energy minimization studies on dinucleotide (Ford et al., 1987) and oligonucleotide (Hui et al., 1990) complexes of H2TMPyP.
In an attempt to learn more about the manner in which cationic porphyrin complexes interact with DNA, especially those which bind outside in the minor groove, we characterized the DNA binding and cleavage properties of the cobalt porphyrins shown in Figure 1. The porphyrins possess four alkylated pyridine residues having methyl (HzTMPyP), n-butyl (HzTBPyP), and n-octyl (H2TOPyP) groups. We reasoned that if the DNA helix wereintact at thesiteof binding, bulky substituent groups on the periphery of the porphyrin could alter the structure of the porphyrin-DNA complex. Changes in structure could in turn be detected by alterations in the angle that the porphyrin plane makes with the helix axis and/or changes in rates associated with porphyrin cleavage of DNA.
When bound to calf thymus DNA, the cobalt porphyrins exhibit induced CD effects that are similar in sign and magnitude, suggesting a common binding mode for all compounds. Furthermore, linear dichroism (LD) shows that the porphyrin plane is oriented at an angle of 42-45' relative to the helix axis, indicating that the compounds are "end-on" bound. In the range 220-300 nm, LD also shows that DNA molecules having the bound cobalt porphyrins are difficult to orient along the flow lines. These observations sharply contrast with those of the non-metal porphyrins H2TMPyP and H2-TBPyP, which have induced CD effects and orientation angles consistent with intercalation into DNA. In the presence of KHS05, the coblat porphyrins cleave the closed circular form of PM2 DNA in a single-strand manner, i.e., a single activation event on the porphyrin leads to a break in one of the DNA strands. A kinetic analysis shows that the apparent cleavage Sehlstedt et al. rates for conversion of form I DNA to form I1 DNA are in the order CoTMPyP > CoTBPyP > CoTOPyP. This is likely caused by differences in DNA affinities rather than by differences in cleavage rate constants. These results, and those of earlier work, support a model having the metalloporphyrin in a partially melted region of DNA. The unmetallated analogues, on the other hand, intercalate into DNA.

Materials.
The PM2 DNA was used as obtained from Boehringer Mannheim Biochemical. Oxone, KHS05, was purchased from Aldrich Chemical Co. Calf thymus DNA, CT DNA (source Sigma Chemical Co.), was dissolved in 1 mM EDTA, 100 mM NaCl, and 5 mM cacodylic buffer at pH 7.0 by exhaustive stirring at 4 "C. The DNA solution was dialyzed against 5 mM cacodylic buffer with 20 mM NaCl at 4 OC, and the buffer was changed six times with 5-h intervals. Afterwards, the hypochromicity of DNA at 260 nm was found to be 40%. All of the synthesized compounds had acceptable combustion analytical data.

( C O T M P~P ) C~~.~N~C~.~~H Z~.
To 11 g of H2TMPyP-(TOS)4 (Strem Chemical Co.) in 225 mL of water was added 10 g of Dowex 1 anion-exchange resin in the chloride form. After being stirred for 1 h, the resin was removed by filtration, and 18 g of CoC1~6H20 was added to the filtrate. The solution was refluxed for 14 h and cooled to room temperature, and its volume was reduced in vacuo. The addition of excess aqueous NaC104 resulted in the precipitation of the perchlorate salt of the complex. The c104-counter-ion was exchanged for C1-using Dowex 1 in the chloride form, and the product was recovered by freeze drying. The yield was 7.2 g.
(CoTBPyP)BFy6H20. To 8 g of H2TBPyP(BF4)4 (Strem Chemical Co.) in 500 mL of water was added 15 g of Dowex 1 anion-exchange resin in the chloride form. After 1 h the resin was removed by filtration, and 18g of CoCly6H20 was added to the filtrate. After being refluxed for 6 h, the solution was cooled and excess aqueous NaBF4 added. The product, as BF4-salt, was recovered by filtration and was air-dried. The yield was 7.2 g.
HzTOPyPC14. A total of 158 g (2 mol) of pyridine was added to 0.5 mol of octanol. The reaction mixture was placed in an ice bath, and 105 g (0.55 mol) of p-toluenesulfonyl chloride was slowly added so that the temperature never exceeded 12 OC. The reaction mixture was then stirred for 3 h at a temperature around 15 OC, after which time it was diluted with 300 mL of HCl in 1 L of ice water. Then the product was extracted from the reaction with ether. The ether was then removed in vacuo, and 90 g of octyl tosylate was obtained.
A total of 50 g of octyl tosylate was added to 500 mL of DMF followed by 6 g of tetra(4-pyridy1)porphine. The reaction mixture was refluxed for 10 h. The volume of DMF was reduced by evaporation to 80 mL, and 200 mL of water was added. After being filtered, the semisolid was stirred in a cold water/5% acetone mixture, filtered again, and airdried. Then the counter-ion was changed to chloride using a Dowex 1 ion-exchange resin. The yield was 8 g.
( CoTOPyP)Cly8Hz0. The free ligand H2TOPyPC14,4g, in a small amount of acetone was added to 500 mL of water containing 15 g of CoCl2.6H20. The reaction was refluxed for 6 h. After the mixture cooled, excess aqueous NaC104 was added to induce crystallization. The ClO4-counter-ion was converted to the chloride form in the manner described for the previous compounds.
Cleavage Reactions. The cleavage reactions involving DNA werecarried out in a totalvolume of 9 pL at room temperature. To 4 p L of the appropriate porphyrin solution in 10 mM Tris-HCl, pH 8.0, was added 1 pL of PM2 DNA (0.25 pg/pL) in water. Following the equilibration of each porphyrin-DNA solution, 4 p L of freshly prepared aqueous 10 mM oxone (potassium monopersulfate) was added to activate the metalloporphyrin. The final concentrations of various components in the reaction mixture were as follows: PM2 DNA, 46 pM base pairs; oxone, 4.4 mM; Tris-HC1, 4.4 mM. The final concentration of CoTMPyP in the various reactions, which is also the order in the wells of the gel from left to right, was, 0,9.2 X 10-8,0,9.2 X 1.15, 1.54, 2.31,3.08,4.63,6.17, 9.26 X lo-'; 1.85, 4.63 X lo4; and 4.63 X M. The concentrations of CoTBPyP and CoTOPyP were 0, 0, 9.2 X 9.2 X lo-*; 1.15,1.54,2.31,3.08,4.63,6.17,9.26 X lo-' M; 1.85,4.63, and 9.27 X 10" M. The reaction was allowed to proceed for 15 min, at which time 4 pL of 10 mM sodium dodecyl sulfate (SDS) was added to terminate the reaction. Finally, 2 pL of a nondenaturing gel-loading buffer containing 0.25% bromophenol blue, 0.25% xylene cyanol, and 50% glycerol was added to each solution. The solutions were then loaded onto a 1% (w/vol) horizontal agarose gel and electrophoresed at a constant voltage using tris-acetate (TAE) buffer.
After electrophoresis, the gel was soaked in a solution of ethidium bromide (10 mg/mL) in 400 mL of water for approximately 30 min. While resting on a UV transilluminator, the gel was photographed, and both a positive and a negative were obtained. The negative was "fixed" in an aqueous solution containing sodium sulfite and then rinsed several times in water. The negatives were then scanned by a Molecular Dynamics microdensitometer, Model 300A, in order to obtain the intensity of each spot appearing on the gel.
Analysis of DNA Cleavage. Due to non-uniform intensity of the transilluminator, the scanned intensities were corrected in the previously described manner (Kishikawa, 1991). Because of its superhelical nature, the ability of form I DNA to stain under saturated ethidium bromide conditions was less than that of forms I1 and I11 DNA: the earlier determined factor of 0.8 was used to multiply calculated intensities of form I DNA (Bauer & Vinograd, 1968).
The kinetic scheme assumed, with I, 11, and I11 representing supercoiled, open circular, and linear PM2 DNA, respectively, was as follows: Assuming all reactions to be first order in the concentration of DNA, expressions for the concentrations of the three forms as a function of time are easily derived: as pseudo-first-order because the concentration of porphyrin does not change with timeduring a digest. Each rate constant, ki, in these pseudo-first-order processes is actually a product of a second-order rate constant ki' and the concentration of porphyrin. Because rate constants always appear multiplied by time and porphyrin concentration, changing time t is equivalent to changing porphyrin concentration.
In the expressions for [I], [11], and [111], e-kl' is replaced by exp(-kl'c,r), where cp is the porphyrin concentration and the value of t is fixed in the experiments at 7 , the digest time, which is always 10 min. Other exponential terms are replaced similarly. The coefficients of the exponential terms involve ratios of rate constants, so they remain unchanged except that k / replaces ki. Spot intensities on the gels are proportional to concentrations of DNA. By fitting the intensities for forms I, 11, and I11 as a function of porphyrin concentration to the above expressions, one can extract values for the six parameters The value of the constant was used to normalize the data, i.e., all intensities for each concentration of, for example, CoTMPyP were multiplied by the average value of S divided by the corresponding value of S for that concentration. This compensates for dilution errors, loading errors, and unequal illumination of different gel lanes by the transilluminator. The resulting correcting intensity data were used in the kinetic analysis.
Linear Dichroism. Linear dichroism (LD) on flow-aligned DNA was measured in a couette cell of Wada type (Wada & Kozawa, 1964) on a Jasco J500A spectropolarimeter as described earlier (Norden & Tjerneld, 1976;Nordtn & Seth, 1985;Nordtn et al., 1992). The reduced linear dichroism (LD'), i.e., the ratio between LD and isotropic absorbance, Aimo, at a given wavelength, A, is related to the angle, a, between the light-absorbing transition dipole and the local helix axis of DNA according to eq 6, where S is an orientation function such that S = 1 for DNA perfectly aligned parallel to the flow lines and S = 0 for isotropic orientation (Nordtn et al., 1992): An effective value of a was calculated from the maximum LDr observed for the in-plane polarized aa* transitions in the Soret and visible porphyrin absorption bands by comparison with the LD' observed for the DNA bases using an effective a value of 86' at 260 nm (Matsuoka & Nordtn, 1983).
Large a values (80-90') are indicative of intercalation whereas small values, such as 40-50°, are rather consistent with groove binding and, in any case, exclude intercalation binding geometries. In the present case, as a result of degenerate in-plane polarization of 7ra* transitions of the porphyrin chromophores, the apparent a! value corresponding to a tilt of the porphyrin plane by an angle / 3 will be larger than the tilt (i.e., a > p). The tilt is obtained by replacing COS& in eq 6 by (1/2)c0s2/3 (Hard & Nordtn, 1986).
Here the initial concentrations of I, 11, and I11 are CO, NO, and &, respectively, and The processes are actually second-order, but may be treated Energy Transfer. Fluorescence energy transfer is the transfer of electronic excitation energy from a donor to an acceptor. Since the first report of contact energy transfer from DNA bases to bound ligands by Le Pecq and Paoletti in 1967, this technique has been used frequently in DNAligand interaction studies. In our work, the technique was applied to assess the interaction between the porphyrin chromophore and the DNA bases.
where Z denotes the fluorescence intensity and e denotes absorption. The indices band f refer to bound and free ligands, respectively. X = 310 nm was chosen as the normalization wavelength because of the very low absorption of DNA in this region of the spectrum.
The fluorescence spectra were recorded on an Aminco SPF-500 spectrofluorometer in the quantum corrected mode. In order to avoid second-order spectra, a 375-nm cutoff filter was utilized. The excitation spectra were recorded for an emission wavelength of 660 nm and band width of 40 nm. All spectra were corrected for inner filter effects.

RESULTS AND ANALYSIS
Absorption Spectra. The interaction of CoTMPyP, CoT-BPyP, CoTOPyP, H2TMPyP, and H2TBPyP with C T DNA was monitored with visible absorption spectroscopy. The absorption spectra of the free porphyrins and the porphyrins bound to C T DNA are depicted in Figure 2. The increase in absorbance in the 220-280-nm region upon addition of the porphyrins is attributable to the porphyrin molecules (Geacintov et al., 1987). For all the porphyrins, binding to DNA leads to a substantial red-shift of the Soret maximum. For the free ligands, H2TMPyP and H2TBPyP hypochromicity in the Soret band is 50 and 35% respectively, while CoTMPyP, CoTBPyP, and CoTOPyP show smaller hyperchromicities in the Soret band of 7, 14, and 6% respectively. Induced CD. The porphyrins and metalloporphyrin derivatives considered here do not display circular dichroism (CD) spectra in the absence of nucleic acids, but CD spectra are induced for the various porphyrins with DNA. A number of representative spectra are shown in Figure 3. All of the metalloporphyrin derivatives examined display positive CD spectra when bound to CT DNA. This profile is similar to that of di-tert-butylproflavin (Dalgleish et al., 1972) which does not intercalate and has a specificity for AT-rich DNA (Muller et al., 1973). Conversely, HzTMPyP displays a negative CD spectrum, and the induced CD spectrum of H2-TBPyP is composed of both positive and negative components, where the positive band has a smaller ellipticity than the negative band. This spectrum is similar to the visible CD spectrum observed for the proflavin complex with DNA (Dalgleish et al., 1969). Proflavin is a known intercalator in both GC and AT sites.
LD and LD'. The LD spectra of DNA complexes of all the porphyrins examined are depicted in Figure 4a-e. The LD signals at long wavelength (300-650 nm) for all the DNAporphyrin complexes are negative, which indicates that the transition moments that are oriented within the planes of the molecules (Weiss et al., 1972) are on the average tilted more perpendicular than parallel to the flow lines (Wada & Kozawa, 1964; Nord6n & Tjerneld, 1976). However, as an effect of degeneracy or near-degeneracy of electronic states in the effectively 4-fold symmetric metal porphyrins, the apparent angle a, as obtained from eq 6, does not correspond to the angle the plane of the chromophore makes to the helix axis. The true angle, j3, is instead obtained as the solution to cos2 j3 = 2 cos2 a (Materials and Methods). In Table 1 the corresponding tilts, j3, are given. Clearly, whereas the metalfree porphyrins show a perpendicular binding geometry, consistent with intercalation, the metal porphyrins bind to DNA at angles close to the pitch of the minor groove, a ..  geometry that excludes intercalation. A significant decrease in LDr values in the 220-300-nm region upon addition of the cobalt porphyrins (Figure 5a-c) suggests that the ability of the DNA molecules to orient along the flow lines is decreased upon binding to these porphyrins. Fluorescence Energy Transfer. Contact energy transfer from nucleic acid bases to bound porphyrins was measured from fluorescence excitation spectra recorded between 220 and 3 10 nm. Since energy transfer can only occur if a ligand closely contacts the DNA bases, the technique is often used as a criterion for intercalation (Reinhardt et al., 1982). Such energy-transfer experiments were performed with all the porphyrins studied. The three cobalt porphyrins did not show  energy transfer from DNA bases (curve A in Figure 6), while HzTMPyP and HzTBPyP increased the fluroescence quantum yield by a factor of 1.6 and 2.2, respectively (curves B and C in Figure 6), in the DNA absorption region, indicating that significant energy is transferred from excited DNA bases to the ligands. Cleavage Kinetics of the Cobalt Porphyrins. The corrected spot intensities as described in the Materials and Methods section were fitted to eq 1 and 2 by searching for the values of the four parameters which minimized the sum of the squared deviations between the normalized intensities and the theo- .. retical expressions. The resulting values for the parameters, including the root mean square deviation between theoretical and experimental for all three cobalt porphyrins, are given in Table 2. The normalized spot intensities are shown in Figure  8a-q together with the values of the theoretical expressions using these parameters.

DISCUSSION
CD and Absorption Spectra. The free bases, H2TMPyP and H2TBPyP, which have no axial ligands show a substantial red-shift at the Soret maximum ( Figure 2). Upon binding to CT DNA they also exhibit negative CD bands in the visible region ( Figure 3). These observations indicated that the unmetallated porphyrins have strong interactions with the bases of DNA. For the cobalt porphyrins, which have axial ligands attached to the metal ion, red-shifts of 5-8 nm of the Soret maxima and rather strong, positive induced CD bands are observed in the visible region when they are complexed with CT DNA. In addition, the cobalt porphyrins display hyperchromicity or very small hypochromicity with CT DNA rather than the usual substantial hypochromism associated with intercalating molecules. By way of analogy, when proflavin intercalates into DNA, significant hypochromicity of its visible absorption band occurs, unlike the nonintercalating derivative 2,7-di-tert-butylproflavin which displays significant hyperchromicity (Muller et al., 1973). Thediaxially liganded metalloporphyrins have a ligand-metal-ligand distance of approximately 7-10 A. ZnTMPyP, which is 5 coordinate and does not intercalate (Pasternack et al., 1983), has an approximate "thickness" of 5 A (Hoard, 1975). Since the spacing between base pairs adjacent to an intercalated ligand is -6.8 A, cationic metalloporphyrin possessing even oneaxial group is too thick for intercalation into DNA.  (Figure 4d). Therefore, this porphryin does not contribute significantly to the linear dichroism of the complex at -260 nm; suggesting that the pyridyl groups, which absorb significantly in this wavelength range, are characterized by either a low degree of orientation or orientation angles of -55' (the angle at which LD is -0, eq 6). The LD for the H2TBPyP complex is somewhat lower than for DNA alone in the 260-nm region (Figure 4), indicating some positive contribution of the DNA helix, probably from the bound porphyrin.
For the H2TMPyP-DNA and H2TBPyP-DNA complexes, the linear dichroism maxima within the 440-nm Soret band and within the 525-530-nm absorption band are slightly redshifted (by 4 nm) with respect to the absorption maxima. Because of this, the reduced linear dichroism is not constant as a function of the wavelength is the visible region of the spectrum. Assuming that the average angle between the base pairs and the DNA helix axis is 86' (Matsuoka & Norden, 1982, the average angles of the transition moment -0.02 .. around the Soret maximum can be calculated according to eq 6 as 85 f 1' and 75 f 1' for H2TMPyP-DNA and Hz-TBPyP-DNA, respectively. The angles calculated from all ratios of porphyrin to DNA fall into the same range, suggesting homogeneous binding mode for the two unmetallated porphyrins with CT DNA. Also, thevalues of the angles strongly indicate that both H2TMPyP and HzTBPyP bind by intercalation to DNA. A clear difference in the LD' values for x 2.5 1 Biochemistry, Vol. 33, No. 2, 1994 423 In summary, the planes of the metalloporphyrin derivatives studied are oriented at angles of roughly 45O with respect to the axis of DNA. This conclusion is reinforced by the LDr values obtained for the y-polarized transitions in the 550-570-nm range; the reduced linear dichroism in this wavelength range is similar in magnitudes to the values determined within the Soret band. The calculated orientation angles of the complexed metalloporphyrin molecules are clearly inconsistent with those of intercalation complexes and in agreement with earlier conclusions that similar molecules are bound to DNA in an externally binding mode (Fiel et al., 1979;Carvlin & Fiel, 1983;Carvlin et al., 1983;Pasternack et al., 1983;Geacintov et al., 1987). In light of the increased flexibility indicated by the difficulty in orienting DNA (decreased S value), it is possible that the porphyrins are end-on bound in a partially melted region of DNA. The orientation of the chromophore is also consistent with groove binding (Nordbn et al., 1992).
Energy Transfer. Porphyrins that contact closely with DNA bases are characterized by clear increases of their fluorescence quantum yields for an excitation around 260 nm corresponding to an energy transfer from DNA bases to the porphyrins (Sari et al., 1990). Such an energy transfer was obtained for the two unmetallated porphyrins examined here. The fluorescence quantum yield was increased by a factor of 1.6 for HzTMPyP and 2.2 for H2TBPyP ( Figure 6), while none of the metalloporphyrin derivatives gave any increase in the quantum yield. From this observation, the conclusion can be drawn that both H2TMPyP and H2TBPyP are in close contact with the DNA bases whereas neither CoTMPyP, CoTBPyP, nor CoTOPyP show such a behavior.
As is shown in Figure 7, the cobalt porphyrins, in the presence of KHSOs, can cleave PM2 DNA. The mechanism by which this occurs is not clear, but extensive work by Groves and co-workers (Groves, 1979;Groves et al., 1983a,b) has shown that, in non-aqueous media, iron and manganese porphyrins can be oxidized with iodosobenzene or superoxide to higher valent oxo species having a ligated oxygen atom. In the presence of alkanes or olefins, the coordinated oxygen atom can be inserted into C-H and C=C bonds to form alcohols and epoxides, respectively. Little is known about the activation chemistry of cationic cobalt porphyrins, but CoT-MPyP has been reported to exhibit superoxide dismutase activity in water (Peretz et aZ., 1982).
The inspection of Figure 7 and the kinetic analysis of the cleavage data reveal that all three cobalt porphyrins are singlestrand cleavage agents, i.e., a single activation event on the porphyrin leads to a break in one of the DNA strands. In this respect, the compounds are similar to the DNA hydrolytic enzyme DNase I and the anticancer agent esperamicin A1 (Kishikawa, 1991). Analogywith MnTMPyP/KHSOS (Pitib etal., 1992;PrativeletaZ., 199l),suggests thatDNAbreakage by the cobalt porphyrins probably involves oxygen insertion from a high valent oxo species, ultimately resulting in a strand break. No evidence was found that the cobalt porphyrins participate in a double-strand cleavage process. This would require an initial break on one strand followed quickly by a second break on the opposing strand. The overall process, effectively a double-strand or "simultaneous" cleavage event, would cause form I DNA to be converted directly to form 111 DNA. This is in fact the case for the anticancer drug calicheamicin, which proceeds through a phenylene diradical intermediate in causing a double-strand break in DNA (Kishikawa, 1991). denoting the absorbance spectra of DNA and bound dye. The CT DNA concentration is 10 pM, and the concentration of the porphyrins is 1 pM. Emission wavelength was 660 nm (bandpass 40 nm). Excitation bandpass was 4 nm, and the optical path length was 10 nm. and y transition moments, therefore, indicates that these transition moments have somewhat different angles relative to the DNA helix axis.
The absorption band above 500 nm is oriented along the y direction of the ring system, while the Soret band is characterized by a mixed x and y polarization (Weiss, 1972). The generally high values of the reduced linear dichroism in the 40@-600-nm range support the previous conclusion, based on the LDr values observed within the Soret band, that the orientation of the planes of the two porphyrin ring systems are parallel or nearly parallel to the planes of the DNA bases.
LD and LLY of the Cobalt Porphyrins. The magnitudes of the LD signals are only slightly smaller than that of DNA alone in the 260-nm region, except for CoTOPyP which decreases the LD signal significantly compared to that of free DNA (Figure 4a-c). The decrease in magnitude can be explained by external binding in such a way so as to increase the flexibility of DNA. In the Soret band region, in contrast to H2TMPyP-DNA and H*TBPyP-DNA, the reduced linear dichroism is about 70% smaller than that of the DNA region (Figure 4a-c). Furthermore, the LDr values are not constant across this absorption band, suggesting that the x and y directions of the porphyrin ring systems with respect to the planes of the DNA bases are somewhat different. On the basis of eq 6, and with methods outlined elsewhere (Nordbn & Tjerneld, 1976;Nord6n et al. 1992)  products. Units of rate constants are pM-I min-I. The quantity D is the root mean square deviation between experimental and calculated data. The apparent cleavage rates for the conversion of form I DNA to form I1 DNA, kl', in Table 2 are about an order of magnitude larger than those observed for esperamicin AI, a kl' of 1.95 (Kishikawa, 1991). Without knowledge of the binding constants for the drug and the porphyrins, it is difficult to determine if the rate differences are due to differences in cleavage rate constants or differences in affinities of the agents for DNA. Despite literature reports that binding constants for porphyrin-DNA interactions are measurable (Strickland et al., 1990), efforts to determine the DNA affinities of the cobalt porphyrins were unsuccessful. Centrifugation of solutions containing 5 pM CoTMPyP caused the absorbance of the solution to decrease by -20%. This observation, which is consistent with aggregation, complicated attempts to analyze the optical data via Scatchard methods. The LD studies show that all three porphyrins have the same orientation relative to the DNA helix axis. This observation strongly suggests that the disposition between the activated porphyrin and its DNA target is similar for all three compounds and that the observed differences in cleavage rates ( Table 2) are due to differences in affinities rather than differences in cleavage rate constants.
The binding model which emerges from these studies confirms earlier views of the porphyrin DNA interaction. The fact that the unmetallated porphyrins, H2TMPyP and H2-TBPyP, intercalate shows that the length of the alkyl chain does not affect the insertion of the porphyrin plane between the base pairs of DNA. The mechanism by which this occurs is not understood, but the size of the substituent groups and the dimensions of the porphyrin strongly suggest that hydrogen-bonding and stacking forces involving the bases of DNA are disrupted as insertion occurs. The resulting complex is stabilized by favorable electrostatic interactions between the alkylated pyridine nitrogen atoms and the negatively charged phosphate groups of DNA. Additional stabilization is provided by van der Waals contacts (stacking interactions) between the porphyrin core and the base pairs at the intercalation site. The role of water in stabilizing the complex is also important since intercalation effectively "hides" the hydrophobic core of the porphyrin from solvent.
Outside-binding porphyrins are believed to interact in the minor groove of DNA (Pratviel et al., 199 1 ;Dabrowiak et al., 1989;Hui et al., 1990;Fiel, 1989;Marzilli, 1990;. Although many structural details are lacking, a number of factors indicate that DNA may be distorted at the porphyrin binding site. As was earlier mentioned, the DNA molecules are difficult to orient when bound to the cobalt complexes shown in Figure 1. Difficulty in orienting was also noted by Geacintov et al. (1 987) in studies with the outside-binding porphyrin ZnTMPyP. These workers concluded that the compound is bound to a site which is bent, kinked, or possesses some type of open structure. Similar observations have been made for ARu(phen)3*+ which kinks DNA (Satyanarayana et al., 1992(Satyanarayana et al., , 1993 and reduces the ability of DNA to orient in flow dichroism experiments (Hiort et al., 1990). Studies involving rod-like DNA and ZnTMPyP, a porphyrin which binds in an outside manner to DNA (Marzilli, 1990), show that the viscosity of solutions decrease as ligand binding occurs (Banville et al., 1983). This effect is the same as that observed for several aromatic substituted diammonium cations (Kopicak & Gabby, 1975), ARu-(phen)32+ (Satyanarayana et al., 1992(Satyanarayana et al., ,1993) and the propellershaped dye, crystal violet (Wakelin et al., 198 l), all of which are believed to kink DNA. The kink is caused by partial insertion of a ligand group between the base pairs of DNA. In view of the structures of the non-intercalating porphyrins and the angle that they make with the helix axis (-45O), it is difficult to see how they could kink DNA through partial intercalation. Axial ligands on the metal ion block intercalation of the porphyrin core while the substituents on the aoo Biochemistry, Vol. 33, No. 2, 1994 425 on the pyridine groups would be greatly reduced. This would allow "inclusionn of the porphyrin into a partially melted region of DNA. This model, which was earlier used to explain the identical DNA cleavage patterns of cationic manganese porphyrins having different substituents on the alkylated pyridine groups (Raner et al., 1988), is also consistent with the elegant mechanistic work of Bernadou and Meunier and their co-workers (PitiB, 1992) on porphyrin cleavage of DNA. The latter work showed that the activated oxo intermediate of MnTMPyP cleaves single-stranded polynucleotides and double-stranded calf thymus DNA by hydroxylating the anomeric carbon atom C1' (Pratviel et al., 1991;Bernadou et al., 1989). For double-stranded DNA, this carbon atom is located deep within the minor groove, and without a significant distortion in DNA it is difficult to see how this site can be hydroxylated by the activated porphyrin.
While intercalation of porphyrins is analogous to the "threading" which takes place with the anticancer drug nogalamycin (Williams et al., 1990), the model proposed for the outside-binding porphyrins has no precedence in the ligand-DNA literature. However, AT-rich regions are easier to melt than GC sites, and it is possible that the binding energy may be sufficient to overcome the energy loss due to the separation of Watson-Crick base pairing. In addition to favorable electrostatic interaction, the unusual mode of binding may be enhanced by the formation of ?r complexes between unpaired adenine and thymine residues with planar surface on the porphyrin core and pyridine groups attached to the methine carbon atoms. Clearly, additional work will be necessary to ascertain if this model is in fact correct. However, in view of the measured angle between the porphyrin plane and helix axis and the aforementioned observations, an "inclusionn type model is an attractive possibility.