Bisintercalation of Homodimeric Thiazole Orange Dyes in DNA: Effect

Oct 15, 1997 - Dan Stærk, Atef A. Hamed, Erik B. Pedersen, and Jens Peter Jacobsen*. Department of Chemistry, Odense University, Odense M DK-5230 ...
10 downloads 0 Views 416KB Size
Bioconjugate Chem. 1997, 8, 869−877

869

Bisintercalation of Homodimeric Thiazole Orange Dyes in DNA: Effect of Modifying the Linker Dan Stærk, Atef A. Hamed, Erik B. Pedersen, and Jens Peter Jacobsen* Department of Chemistry, Odense University, Odense M DK-5230 Denmark. Received April 30, 1997X

The thiazole orange dye 1,1′-(4,4,8,8-tetramethyl-4,8-diazaundecamethylene)-bis[4-[3-methyl-2,3dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO) binds to double-stranded DNA (dsDNA) in a sequence selective bisintercalation. Each chromophore is sandwiched between two base pairs in a (5′-CpT-3′):(5′-ApG-3′) site, and the linker spans over two base pairs in the minor groove. The binding of analogs of TOTO in which the linker has been modified is examined. The aim of the study is to utilize the sequence selectivity of the TOTO chromophores to enhance and/or alter the overall selectivity of the binding. One- and two-dimensional 1H-NMR investigations of complexes between TOTO analogs and various dsDNA oligonucleotides are reported. The following analogs were synthesized and used: 1,1′-(4,4,8,8-tetramethyl-4,8-diazadodecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO10), 1,1′-(5,5,9,9-tetramethyl-5,9diazatridecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO11), and 1,1′-(6,6,10,10-tetramethyl-6,10-diazapentadecamethylene)-bis[4-[3-methyl2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO13). The results show that with a longer linker the dyes can bisintercalate into two (5′-CpT-3′):(5′-ApG-3′) sites separated by one or two base pairs. Bisintercalation in two such “isolated” binding sites yields non-nearestneighbor bisintercalation in which the linker spans over more than two base pairs. The investigations also showed that an exact length of the linker is not crucial for the site selectivity since TOTO, TOTO10, and TOTO11 are almost equally suitable in binding selectively to the (5′-CTAG-3′)2 sequence. Fluorescence measurements show that TOTO10, TOTO11, and TOTO13 have higher fluorescence quantum yields than TOTO when bound to d(CGCTAGCG)2. This indicates that the length of the linker in TOTO may not be the optimum one in terms of using the dye as a fluorescence marker.

INTRODUCTION

The search for nonradioactive DNA stains that are stable under gel electrophoretic conditions has led to the synthesis and characterization of a family of homo- and heterodimeric DNA-binding dyes (Benson et al., 1993a,b; Rye et al., 1992). These dyes form highly fluorescent and stable noncovalent complexes with double-stranded DNA (dsDNA1). Prominent among these are 1,1′-(4,4,8,8-tetramethyl-4,8-diazaundecamethylene)-bis[4-[3-methyl-2,3dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO, Scheme 1). The enhancement of the fluorescence quantum yield of TOTO upon complex formation with dsDNA is >1000 (Rye et al., 1992), and t1/2 for the dissociation of the dsDNA-TOTO complex * Author to whom correspondence should be addressed (email [email protected]). X Abstract published in Advance ACS Abstracts, October 15, 1997. 1 Abbreviations: AMBER, assisted model building with energy refinement; DNA, deoxyribonucleic acid; DMSO, dimethyl sulfoxide; dsDNA, double-stranded DNA; DSS, 2,2-dimethyl-2silapentane-5-sulfonate; EDTA, ethylenediaminetetraacetic acid; HSQC, heteronuclear single quantum coherence; NMR, nuclear magnetic resonance; NOESY, nuclear Overhauser effect spectroscopy; RMD, restrained molecular dynamics; TOCSY, total correlation spectroscopy; TOTO, 1,1′-(4,4,8,8-tetramethyl-4,8diazaundecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo-1,3thiazole)-2-methylidene]]quinolinium tetraiodide; TOTO10, 1,1′(4,4,8,8-tetramethyl-4,8-diazadodecamethylene)-bis[4-[3-methyl2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide; TOTO11, 1,1′-(5,5,9,9-tetramethyl-5,9-diazatridecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2methylidene]]quinolinium tetraiodide; TOTO13, 1,1′-(6,6,10,10tetramethyl-6,10-diazapentadecamethylene)-bis[4-[3-methyl-2,3dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide; TPPI, time proportional phase incrementation.

S1043-1802(97)00067-0 CCC: $14.00

under gel electrophoretic conditions is approximately 11 h (Benson et al., 1993). This makes TOTO an excellent choice as a marker with a detection limit of ∼4 pg of DNA (Benson et al., 1993a,b). Recent work in our laboratory has focused on 1H-NMR studies of the binding mode and sequence selectivity of TOTO with various dsDNA oligonucleotides. We have shown that TOTO binds via bisintercalation preferentially to oligonucleotides containing the (5′-CTAG-3′)2 or the (5′-CCGG-3′)2 sequences (Hansen et al., 1996; Jacobsen et al., 1995). In studies of complex formation in oligonucleotides containing both of these binding sites, we showed that the binding to the (5′-CTAG-3′)2 sequence is favored compared to the (5′-CCGG-3′)2 sequence. The solution structure of the complex of TOTO bisintercalated in the (5′-CTAG-3′)2 binding site of the d(5′CGCTAGCG-3′)2 oligonucleotide shows that the sequence selectivity of TOTO is caused by the ability of the TOTO chromophores to adapt to the propeller twist of the nucleobases (Spielmann et al., 1995). This is possible since the benzothiazole ring can twist relative to the quinolinium ring due to the flexible cyanine methine bond between. The benzothiazole is intercalated between two pyrimidine bases and the quinolinium ring between the two purine bases. In contrast to other intercalators, the sequence selectivity of TOTO arises directly from the interaction of the intercalated chromophores with the nucleobases. Although the linker of TOTO adds significantly to the binding strength, it is only a spectator with respect to the sequence selectivity. With the chromophore responsible for the sequence selectivity, several opportunities for linker modifications are possible to enhance or alter the overall selectivity of TOTO analogs. The bisintercalation of TOTO obeys the nearest-neighbor exclusion principle with the linker © 1997 American Chemical Society

870 Bioconjugate Chem., Vol. 8, No. 6, 1997 Scheme 1. TOTO Numbering Scheme

Stærk et al. Scheme 2. Numbering Duplexes Used

Scheme

for

the

dsDNA

Scheme 3. Synthetic Path for the TOTO Analogs

spanning over two bases in the minor groove. A priori, there seems to be no structure-based justification that the linker length in TOTO is the optimal one. A slightly longer or shorter linker may enhance the binding strength. Furthermore, a significant longer linker may allow TOTO analogs to bisintercalate in “isolated” (5′-CpT-3′):(5′-ApG3′) sites, i.e. such sites separated by one or two base pairs. Similar considerations on the length of the linker in other bisintercalating agents with completely different chromophores have been reviewed earlier (Wakelin, 1986). In this paper, we present the results of one- and twodimensional 1H-NMR studies of complexes between dsDNA and TOTO analogs with various linker lengths. We have synthesized 1,1′-(4,4,8,8-tetramethyl-4,8-diazadodecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo1,3-thiazole)-2-methylidene]quinolinium tetraiodide (TOTO10, Scheme 1), 1,1′-(5,5,9,9-tetramethyl-5,9-diazatridecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO11, Scheme 1), and 1,1′-(6,6,10,10-tetramethyl6,10-diazapentadecamethylene)-bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium tetraiodide (TOTO13, Scheme 1) and investigated their complexes with various oligonucleotides. These oligonucleotides contain the (5′-CpT-3′):(5′-ApG-3′) and the (5′-ApG3′):(5′-CpT-3′) sequences separated by 0, 1, or 2 base pairs (Scheme 2). We found that with a longer linker, TOTO analogs bind into isolated (5′-CpT-3′):(5′-ApG-3′) base pairs by bisintercalation. To our knowledge this is the first time that non-nearest-neighbor bisintercalation has been observed.

SYNTHESIS

So far, nobody seems to have synthesized TOTO derivatives with extended methylene bridges between the quinolinium ring and the quaternary ammonium nitrogen. In fact, two structures with 4-sulfobutyl are the only examples reported with a chain of more than three carbons attached to the quinolinium nitrogen of the corresponding monomeric cyanine dye (Kagawa and Kawashima, 1989). For the synthesis of the TOTO analogs TOTO10, TOTO11, and TOTO13, we selected the strategy of Brooker et al. (1941, 1942) (Scheme 3) to condense the benzothiazole derivative 1 with the quinolinium compound 2 in the presence of triethylamine in absolute ethanol to give the intermediate cyanine dye 3, which subsequently was reacted with 0.5 equiv of N,N,N′,N′-tetramethyl-1,3-diaminopropane in refluxing anhydrous methanol according to the procedure of Glazer and o-workers (Benson et al., 1993; Rye et al., 1992) to give the symmetrical derivatives TOTO11 and TOTO13. The starting quinolinium compound 2 was synthesized by treating lepidine with 5 equiv of the appropriate diiodoalkane in refluxing dioxane followed by recrystallization from acetone. The unsymmetrical TOTO10 was synthesized by treatment of 3a (Brooker et al., 1941, 1942) with N,N,N′,N′-tetramethyl-1,3-diaminopropane in excess followed by reaction with 3b. RESULTS

The d(CGCTAGCG)2 (CTAG) Oligonucleotide. The CTAG + TOTO11 Mixture. The line width observed in

Bisintercalation of Homodimeric Thiazole Orange Dyes

Figure 1. Aromatic to H1′ region of a NOESY spectrum (mixing time of 150 ms) of the CTAG-TOTO11 complex at 25 °C. The sequential H1′(n-1)-H6/H8(n)-H1′(n) connectivity pattern is shown with solid lines. Interruption in the 5′-C3pT4-3′ and the 5′-A5pG6-3′ binding sites are indicated with broken lines. A few pairs of cross peaks between TOTO11 and the DNA are marked with solid arrows. Intramolecular connectivities between protons from TOTO11 are shown with broken arrows.

the 1H spectrum of CTAG + TOTO11 is ∼6-7 Hz, indicating essentially no dynamic exchange between two or more complexes. Furthermore, only one resonance frequency is observed for the protons on each strand of the dsDNA and for each chromophore. This shows that the CTAG + TOTO11 mixture forms a single complex with dyad symmetry. The CTAG-TOTO11 complex arises from bisintercalation in the 5′-CTAG-3′ binding site. The amount of this complex is at least 100 times higher than that of any other complex as seen from inspection of the spectra. Figure 1 shows the aromatic to H1′ part of the NOESY spectrum of the complex. The sequential assignment of the H1′(n-1)-H6/H8(n)-H1′(n) resonances is shown with solid lines. Interruption due to intercalation in the 5′C3pT4-3′ and the 5′-A5pG6-3′ intercalation sites is indicated with broken lines. The large up-field shifts of C3H6 (1.02 ppm), C3H5 (0.85 ppm), T4H1′ (0.58 ppm), and T4H6 (0.44 ppm) relative to the free oligonucleotide are caused by the intercalation of the aromatic chromophore. A few cross peaks between the dsDNA and TOTO11 are indicated with unbroken arrows. Combined with the interruption of the connectivity pattern, this establishes binding in the 5′-CTAG-3′ site and shows the proximity of the benzothiazole protons H1-H4 and the quinolinium protons H13 and H14 to protons in the 5′C3pT4-3′ and the 5′-A5pG6-3′ binding sites, respectively. Intramolecular connectivities between protons from TOTO11 are shown with broken arrows. A characteristic strong cross peak between H8 and H16 on TOTO11 confirms the relative conformation of the benzothiazole and the quinolinium ring as drawn in Scheme 1. Figure 2 (left) shows the aromatic to H2′/H2′′ area of a NOESY spectrum. The aromatic protons yield cross peaks to H2′/H2′′ of their own and the previous oligonucleotide except at the 5′-C3pT4-3′ and 5′-A5pG6-3′ binding sites. Thus, this region is a further strong tool for confirming the bisintercalation in the 5′-C3pT4-3′ and 5′-A5pG6-3′ sites. H13, H14, and H15 on TOTO11 yield cross peaks to A5 H2′/H2′′, while H1 and H3 have cross

Bioconjugate Chem., Vol. 8, No. 6, 1997 871

peaks to C3 H2′/H2′′. The above connectivities show that the benzothiazole is sandwiched between the C3 and T4, while the quinolinium ring is sandwiched between A5 and G6. The methyl group protons of T4 experience an up-field shift of 0.88 ppm and yield cross peaks to H1H4 of TOTO11. This implies that the methyl group is positioned above the benzothiazole. The relative assignment of H2′ and H2′′ is based on analysis of the buildup rates of intramolecular aromatic to H2′/H2′′ cross peaks (Scheek et al., 1984) (Figure 2, right). The cross peaks with the fastest buildup rate are assigned to H2′ and vice versa. This analysis was not possible for A5 and G6 due to overlap of the H2′/H2′′ pairs. The 1H-31P HSQC spectrum is shown in Figure 3. The 1H-31P connectivities (Searle and Lane, 1992) found are in full agreement with the assignment of H3′, H4′, and H5′/H5′′ through NOESY and TOCSY spectra. Normally the dispersion of 31P chemical shift values lies within a range of ∼0.7 ppm (Roongta et al., 1990). The large dispersion of the 31P chemical shift values of C3, T4, and A5 is due to unwinding of the oligonucleotide upon bisintercalation . A complete assignment of 1H and 31P chemical shift values of the complex and the free oligonucleotide is given in Table 1. Intermolecular contacts between the N-methyl groups and A5/G6 confirm that the linker crosses the minor groove. The cross peak pattern observed for the CTAGTOTO11 complex is very similar to that observed for the CTAG-TOTO complex (Spielmann et al., 1995). Restrained molecular dynamics calculations using cross peak intensities in NOESY buildup experiments yielded a structure of the CTAG-TOTO11 complex similar to the structure of the CTAG-TOTO complex (results to be published). The CTAG + TOTO10 Mixture. Similar to the CTAG + TOTO11 mixture, the mixture of CTAG + TOTO10 forms only one complex. Neither the 1D nor the 2D spectrum shows any sign of more than one compound. The CTAG-TOTO10 complex arises from (5′-CTAG-3′)2 bisintercalation. Only small chemical shift differences are observed between the proton resonances of the two strands, and the observed differences are restricted to the (5′-CTAG-3′)2 binding site. Chemical shift differences of the proton resonances of the two chromophores are also observed and are consistent with the asymmetric bisintercalation. A complete assignment of the chemical shift values of the CTAG-TOTO10 complex compared to those of the free oligonucleotide is given in the Supporting Information (Table 1S). The CTAG + TOTO13 Mixture. Unlike TOTO11 and TOTO10, the dye TOTO13 forms more than one complex with the CTAG oligonucleotide. The major form (50%) has been identified as the complex with TOTO13 binding in the (5′-CTAG-3′)2 site. In addition to that, two minor forms exist. Severe overlap in the spectra prevents conclusive evidence that would allow identification of the binding sites of TOTO13 in these two minor forms. The d(CGCTGAGCG):d(CGCTCAGCG) (CTGAG) Oligonucleotide. The CTGAG + TOTO13 Mixture. The CTGAG-TOTO13 mixture yields one major form (90%) and one minor form (10%). Dynamic exchange between the major and the minor form yields extensive line broadening and exchange cross peaks. This makes the interpretation of the spectra difficult. The major form is asymmetric and arises from nearest-neighbor bisintercalation in the 5′-CTGAG-3′ binding site. Figure 4 shows the aromatic to H2′/H2′′ part of a NOESY spectrum. Interruption in the sequential H2′/ H2′′(n-1)-H6/H8(n)-H2′/H2′′(n) pattern of the 5′-C3pT4-3′, 5′-G5pA6-3′, 5′-T13pC14-3′, and 5′-A15pG16-3′ sequences

872 Bioconjugate Chem., Vol. 8, No. 6, 1997

Stærk et al.

Figure 2. (Left) Aromatic to H2′/H2′′ region of a 100 ms NOESY spectrum of the CTAG-TOTO11 complex at 25 °C. Intraresidue connectivities between aromatic protons and H2′/H2′′ are shown with solid arrows, and inter-residue connectivities to H2′/H2′′ on the previous deoxyribose are shown with broken arrows. A few paisr of cross peaks between TOTO11 and the DNA are also marked. (Right) Buildup curves used for the relative assignment of intrraesidue connectivities between the aromatic protons and the H2′/ H2′′ pair.

Figure 3. 1H-31P HSQC spectrum of the CTAG-TOTO11 complex. The sequential connectivities establish the intrasugar connectivities found through NOESY and TOCSY spectra. 31P chemical shift values are relative to 85% phosphoric acid. The arrows shown indicate the following 1H-31P backbone connectivities (Searle and Lane, 1992): intraresidue, 31P(n) f H3′(n); and inter-residue, 31P(n) f H4′(n+1) and 31P(n) f H5′/H5′′(n+1).

is clear evidence of 5′-CTGAG-3′ bisintercalation. Cross peaks between A15H2′/H2′′ and H13/H14/H15 of one of the chromophores are also consistent with intercalation in the (5′-C3pT4-3′):(5′-A15pG16-3′) sequence. Similarly, we observed cross peaks between G5H2′/H2′′ and H13/

H14/H15 of the other chromophore consistent with intercalation in the (5′-G5pA6-3′):(5′-T13pC14-3′) sequence. A complete assignment of chemical shift values of the major complex and the free oligonucleotide is given in the Supporting Information (Table 2S) . A model of the major form of the CTGAG-TOTO13 complex is shown in Figure 5. The model was obtained from restrained molecular dynamics calculations using 20 intermolecular restraints obtained from cross peak intensities of a NOESY spectrum with a mixing time of 200 ms as described under Experimental Procedures. It is obvious that the linker is positioned in the minor groove crossing from one side of the groove to the other. A complete assignment of the minor forms is not possible, but exchange cross peaks and exchangetransferred NOEs show that TOTO13 bisintercalates in the 5′-CTGAG-3′ binding site. This is demonstrated in the upper part of Figure 4, which is a part of the NOESY spectrum obtained at 25 °C corresponding to the lower part of the figure, which is a part of the NOESY spectrum obtained at 10 °C. Transferred NOEs at 25 °C correlate the resonances of T13 H6 and T13 CH3 of the major form (6.55 and 0.91 ppm, respectively) with those of the minor form (7.02 and 0.80 ppm, respectively). This is indicated with broken lines. The corresponding transferred NOEs show that H6 and CH3 of T4minor resonate at 6.54 and 0.92 ppm, respectively. The CTGAG + TOTO11 Mixture. TOTO11 yields one major form with CTGAG (80%) and one minor form (20%). Both the major and minor forms are asymmetric complexes with nearest-neighbor bisintercalation in the 5′-CTGAG-3′ and 5′-CTGAG-3′ binding sites, respectively. These complexes are similar to those observed in the mixture of CTGAG + TOTO13. The assignments of the resonances from the major form show many features similar to the spectra of CTGAG + TOTO13 with bisintercalation in the 5′-CTGAG-3′ binding site. A complete assignment of chemical shift values of the major

Bioconjugate Chem., Vol. 8, No. 6, 1997 873

Bisintercalation of Homodimeric Thiazole Orange Dyes

Table 1. 1H Chemical Shift Values (in Parts per Million) for the CTAG-TOTO11 Complex Compared with Those of the Free Oligonucleotide in Parenthesesa H8/H6 H5/Me/H2 H1′ H2′ H2′′ H3′ H4′ H5′ H5′′ H1/H3 H4/H6 H4/H6

C1

G2

C3

T4

A5

G6

C7

G8

7.54 (7.67) 5.80 (5.95) 5.68 (5.82) 1.86 (2.03) 2.29 (2.47) 4.63 (4.74) 4.02 (4.11) 3.67 (3.75) 3.67 (3.75)

7.75 (8.01)

6.44 (7.46) 4.56 (5.41) 6.00 (6.00) 1.93 (2.08) 2.53 (2.53) 4.63 (4.78) 4.06 (4.27) 4.15 (4.27) 4.04 (4.21)

6.98 (7.42) 0.83 (1.71) 5.05 (5.63) 1.97 (2.15) 2.14 (2.46) 4.67 (4.90) 4.13 (4.17) (4.11) (4.11) 13.51 (13.79)

8.18 (8.24) 6.83 (7.40) 5.60 (6.05) 2.75 (2.76) 2.75 (2.90) 5.12 (5.06) 4.36 (4.42) 4.14 (4.17) 4.04 (4.09)

7.80 (7.71)

7.28 (7.31) 5.30 (5.35) 5.72 (5.78) 1.85 (1.89) 2.28 (2.35) 4.78 (4.80) 4.17 (4.16) (4.12) (4.21)

7.90 (7.95)

7.95 (8.30) 6.59 (6.59)

TOTO11 H1 H2 H3 H4 CH3(6) H8 H9

7.59 7.47 7.59 7.43 3.58 7.17 6.46

a

H10 H13 H14 H15 H16

7.99 6.96 6.88 6.64 7.46

C3 31P T4 31P

The values are relative to DSS at 25 °C.

31P

H19′ H19′′ H20′ H20′′ H21′ H21′′ H22′ H22′′

TOTO11 4.04 4.55 2.02 2.02 1.87 1.87 3.52 3.52

phosphate 1.45 1.02

A5 31P G6 31P

6.10 (6.17) 2.59 (2.62) 2.34 (2.39) 4.67 (4.68) 4.16 (4.18) 4.05 (4.08) 4.15 (4.08)

8.22 (8.36) 6.43 (6.49)

TOTO11

phosphate 0.46 0.66

5.81 (5.73) 2.61 (2.52) 2.67 (2.62) 4.85 (4.97) 4.49 (4.38) 4.02 (4.21) 4.35 (4.21) 12.00 (12.78)

7.24 (-) 6.54 (-)

TOTO11

phosphate C1 31P G2 31P

5.67 (5.98) 2.43 (2.74) 2.35 (2.79) 4.84 (5.02) 4.25 (4.40) 4.02 (4.14) 3.87 (4.04) 12.77 (13.05)

H23′ H23′′ H24′ H24′′ Η25′ H25′′

3.22 3.26 3.52 3.52 2.37 2.37

phosphate 2.73 0.52

C7 31P

0.70

chemical shift values are relative to 85% phosphoric acid.

form and the free oligonucleotide is given in the Supporting Information (Table 3S). A similar complete assignment of the minor form is not possible, but transferred NOEs and exchange cross peaks prove that the minor form of CTGAG + TOTO11 arises from bisintercalation in the 5′-CTGAG-3′ binding site. The d(CGCTAAGCG):d(CGCTTAGCG) (CTAAG) Oligonucleotide. The CTAAG + TOTO13 Mixture. The mixture of CTAAG + TOTO13 yields rather broad lines in the spectrum due to exchange between three complexes. The major form (∼75%) arises from non-nearestneighbor bisintercalation in the 5′-CTAAG-3′ binding site. This major form constitutes the first complex that has been shown to contain a non-nearest-neighbor bisintercalation of a ligand in dsDNA. Figure 6 shows the aromatic to H2′/H2′′ part of a NOESY spectrum of the CTAAG-TOTO13 complex. Interruption in the sequential H2′/H2′′(n-1)-H6/H8(n)H2′/H2′′(n) pattern is observed in the 5′-C3pT4-3′, 5′C12pT13-3′, 5′-A6pG7-3′, and 5′-A15pG16-3′ sequences. More importantly, sequential connectivities for the 5′T4A5A6-3′ and 5′-T13T14A15-3′ sequences show that no intercalation takes place in this part of the oligonucleotide, proving that the linker spans over 3 base pairs. Interruption in the sequential H1′(n-1)-H6/H8(n)-H1′(n) pattern confirms the binding site. The aromatic protons on A6 and G7 have cross peaks to H15 and/or H14 on one of the chromophores, while the aromatic protons on A15 an G16 have cross peaks to H15 and H14 on the other chromophore. Furthermore, an up-field shift of ∼0.8 ppm of the methyl group protons of T4 and T13 is observed. Such an up-field shift is normally observed in binding sites containing a thymidine (Jacobsen et al., 1995; Hansen et al., 1996). All together, this establishes bissintercalation in the 5′-CTAAG-3′ site. A complete assignment of chemical shift values of the major form and the free oligonucleotide is given in the Supporting Information (Table 4S). A similar complete assignment of the minor forms is not possible, but exchange cross peaks and transferred

NOEs show that the two minor forms arise from nearestneighbor 5′-CTAAG-3′ and 5′-CTAAG-3′ bisintercalation, respectively. One aromatic thymidine proton from each of the two minor forms experiences an up-field shift of ∼0.5 ppm (not shown). The methyl group protons from these thymidines have an up-field shift of ∼0.8 ppm. This large up-field shift is typical when a TOTO chromophore intercalates on the 3′ site of a nucleobase. Transferred NOEs from the T4/T13 methyl group protons on the major form to the up-field-shifted thymidine protons on the minor forms are indicated with broken lines in Figure 6. Combined with exchange cross peaks, this makes it possible to assign the binding site of the minor forms. The CTAAG + TOTO11 Mixture. The mixture of CTAAG + TOTO11 yields complexes similar to those in the CTAAG-TOTO13 mixture, but the ratio between the complexes varies. There is no pronounced major form observed. The d(CGCTGCAGCG)2 (CTGCAG) Oligonucleotide. The CTGCAG + TOTO13 Mixture. Dynamic exchange between different forms yields very broad lines in the spectrum of the CTGCAG + TOTO13 mixture. It is not possible to establish the binding sites conclusively. However, it was observed that the thymidine methyl groups in all cases were shifted up-field. A mixture of a symmetric form with intercalation between the 5′-C3pT43′ and the 5′-A7pG8-3′ in the 5′-CTGCAG-3′ binding sites and an asymmetric form with intercalation in the 5′C3pT4-3′ and the 5′-C6pA7-3′ in the 5′-CTGCAG-3′ binding sites is a possible explanation of the features in the spectra. In both cases, this gives a non-nearestneighbor bisintercalation where the linker spans over more than two base pairs in the minor groove. Complete assignment of the spectra was not possible. Absorbance and Fluorescence Emission Spectroscopy. Data from absorbance and fluorescence emission measurements of the free dye and the dye-dsDNA complexes are given in Table 2. A red shift of ∼35 nm of the absorbance maximum and a blue shift of 38-49 nm of the fluorescence emission maximum are observed upon complex formation. Fluorescence enhancement

874 Bioconjugate Chem., Vol. 8, No. 6, 1997

Figure 4. Aromatic to H2′/H2′′/methyl area of a NOESY spectra (mixing time of 200 ms) of CTGAG + TOTO13. (Lower part) Spectrum obtained at 10 °C. The H8/H6(n) f H2′/H2′′(n) connectivities of the major form are shown with unbroken double arrows, while the H8/H6(n) f H2′/H2′′(n-1) connectivities are indicated with broken double arrows. (Top part) Part of the spectrum acquired at 25 °C. Transferred NOEs used for assignment of the minor form are shown with broken lines.

Stærk et al.

Figure 6. Aromatic to H2′/H2′′ part of a NOESY spectrum (mixing time of 200 ms) of the CTAAG-TOTO13 complex at 25 °C. (Lower part) The H8/H6(n) f H2′/H2′′(n) connectivities of the major form are shown with unbroken double arrows, while the H8/H6(n) f H2′/H2′′(n-1) connectivities are indicated with broken double arrows. (Upper part) Transferred NOEs showing the correlation between the major form and the two minor forms.

the enhancement reported from binding of TOTO to calf thymus DNA (Rye et al., 1992). The uncertainties on the values of the fluorescence enhancement (Fbound/Ffree) are expected to be very large because the free dye is nearly nonfluorescent and it has not been possible to ensure that the dye was absolutely free of fluorescent impurities. Consequently, we have determined the relative fluorescence intensities at equal normalized absorbance intensities of the dye-dsDNA complexes. The values are given in the last column in Table 2 showing that TOTO10, TOTO11, and TOTO13 have higher fluorescence quantum yields than TOTO. DISCUSSION

Figure 5. Stereoview of a stickplot of the CTGAG-TOTO13 complex looking into the minor groove. Nearest-neighbor bisintercalation in the 5′-CTGAG-3′ binding site forces the linker between the N-methyls to cross the minor groove nearly perpendicularly. Deoxyribose protons have been omitted for clarity.

upon binding to the 5′-CTAG-3′ sequence is several thousand for all four dyes, which is much larger than

The sequence selectivity and the binding mode of the TOTO analogs to various dsDNA oligonucleotides examined in this work are very similar to those of TOTO. In the CTAG oligonucleotide both TOTO10 and TOTO11 bind exclusively to the 5′-CTAG-3′ site. Also, TOTO13 has this sequence as a preferred binding site, but due to the longer linker other binding sites are possible as well. On the basis of the NMR spectra it is somewhat difficult to establish the relative binding strengths of TOTO, TOTO10, and TOTO11 to the CTAG oligonucleotide. However, a comparison of the line width in the NMR spectra indicates that there is no reason to believe that either TOTO10 or TOTO11 binds more weakly or less preferentially to the (5′-CTAG-3′)2 site than TOTO. The results of the fluorescence spectroscopy experiments using the CTAG oligonucleotide show that the fluorescence quantum yield of TOTO and analogs (Table 2) increases with the length of the linker. Thus, TOTO13

Bioconjugate Chem., Vol. 8, No. 6, 1997 875

Bisintercalation of Homodimeric Thiazole Orange Dyes

Table 2. Data from Absorbance and Fluorescence Emission Spectra of TOTO Analogs Bound to the CTAG Oligonucleotide

TOTO TOTO10 TOTO11 TOTO13

λA (nm) free

bound

479.5 479 474 475.5

513.5 512.5 510 509.5

 (M-1 cm-1) free bound 1.18 × 105 1.16 × 105 0.90 × 105 1.37 × 105

0.81 × 105 0.98 × 105 0.91 × 105 0.91 × 105

λF (nm) free bound 565 565 575 575

527 528 526 527

Fbound/Ffree

rel fluorescence intensity

∼5000 ∼7400 ∼2400 ∼6000

1 1.2 1.3 1.4

a A red shift of ∼35 nm of the absorbance maximum (λA) and a blue shift of 38-49 nm of the fluorescence emission maximum (λF) upon complex formation is observed. The fluorescence enhancement upon complex formation, Fbound/Ffree, is the ratio of the intensities at the fluorescence emission peaks. Measurements were performed in the same buffer with normalized absorbance intensities at the excitation wavelengths. The relative fluorescence emission intensities show that TOTO10, TOTO11, and TOTO13 have higher fluorescence quantum yields than TOTO.

gives the largest enhancement of the fluorescence upon binding to the CTAG oligonucleotide. This effect is not at all correlated to the binding selectivity. One may argue that the most useful fluorescence marker is the one with the lowest selectivity and the highest quantum yield. This would make TOTO13 a better choice as a marker than TOTO. Both TOTO11 and TOTO13 were observed to give nonnearest-neighbor bisintercalation in oligonucleotides with isolated (5′-CpT-3′):(5′-ApG-3′) base pairs. On the basis of the results of TOTO11 and molecular modeling, we refrained from examining the binding of TOTO and TOTO10 to such oligonucleotides since it seems obvious that the length of the linker is too short in these two dyes to accommodate the binding into isolated (5′-CpT-3′):(5′ApG-3′) binding sites. The CTGAG sequence contains 5′-CpT-3′ and 5′-ApG3′ sites separated by one base pair. This is anticipated to lead to non-nearest-neighbor bisintercalation in the 5′-CTGAG-3′ binding site of TOTO11 and TOTO13. However, in both cases we observed predominantly nearest-neighbor bisintercalation in the 5′-CTGAG-3′ binding site. Earlier results have shown the 5′-CTGA3′ is a good binding site for TOTO, although less favorable than the 5′-CTAG-3′ binding site (Jacobsen et al., 1995). The results in the present work demonstrate that this is also the case for TOTO11 and TOTO13. Consequently, it was not possible to observe non-nearest-neighbor bisintercalation in the CTGAG oligonucleotide since the competition from four base pair binding sites is too strong. In the CTAAG oligonucleotide it is more favorable to bisintercalate in a non-nearest-neighbor fashion. The mixture of TOTO13 and CTAAG gives as the major form a complex in which the linker spans over three base pairs. The linker of TOTO11 seems to be slightly too short to really favor the non-nearest-neighbor bisintercalation, although a minor form of that kind was observed in the binding to the CTAAG oligonucleotide. In case of the CTGCAG oligonucleotide, the linker of TOTO13 was observed to be able to span over four base pairs in the binding to the 5′-CTGCAG-3′ site. This is probably initiated by the presence of the 5′-GpC-3′ step, which is known to be a very poor TOTO binding sequence (Jacobsen et al., 1995; Spielmann et al., 1995; Hansen et al., 1996). There seems to be a somewhat rational trend in the ability of TOTO analogs to bind in a non-nearest-neighbor fashion with difficulties in the case of TOTO11, whereas such binding is favored by TOTO13 in some oligonucleotides. However, this behavior probably hides a much more complicated situation than just a simple relation to the length of the linker. The position of the linker in the minor groove is stabilized by electrostatic interactions between the negatively charged phosphate groups on the DNA backbone and the positively charged nitrogen atoms on the linker. It was previously shown

(Spielmann et al., 1995) that the positively charged nitrogens are positioned near the negatively charged phosphate groups between T4 and A5 in the CTAGTOTO complex with the linker crossing the minor groove. Three CH2 groups are probably needed to make this crossing. This leaves the first three CH2 groups of the 4,4,8,8-tetramethyl-4,8-diazaundecamethylene linker with the task of reaching from the intercalating chromophore to the position of the nearest phosphate group. In TOTO11 this can be achieved by the first four CH2 groups in the 5,5,9,9-tetramethyl-5,9-diazatridecamethylene linker, since it may not be very essential whether there are three or four CH2 groups in this part of the linker. On the other hand, it may be more crucial for the ability to cross in the minor groove that there are three CH2 groups between the two positively charged nitrogen atoms in the linker. Preliminary molecular modeling studies have supported these types of considerations on the positions of CH2 groups in the linker. In our opinion, the results in this work are best understood if the stability of the complexes is determined not just by the length of the linker but also by the ability of the linker to cross the minor groove so as to place the positively charged nitrogen atoms close to backbone phosphate groups. This makes it understandable that TOTO11 does not bisintercalate in a non-nearestneighbor fashion, while TOTO13 is able to do so in even the CTGCAG oligonucleotide where the linker has to span over four base pairs. TOTO11 has the optimal length of the linker for nearest-neighbor bisintercalation, while TOTO13 with two extra CH2 groups in the linker can span much longer in the minor groove. The non-nearest-neighbor bisintercalation complex of TOTO13 with the CTGCAG was not observed to be kinetically very stable. The linker in the minor groove spans over four base pair, but only two of the four possible phosphate groups can be close to a positively charged nitrogen atom in the linker. This situation is obviously a very dynamic one, giving broad lines in the NMR spectra. The linker may be slightly too short in TOTO13 to span over four base pairs, but more importantly the linker should contain just as many positively charged nitrogen atoms as the number of phosphate groups it meets in the minor groove. A TOTO analog with a 4,4,8,8,12,12,16,16-octamethyl-4,8,12,16-tetraazanonadecamethylene linker would probably be optimal for TOTO bisintercalation in the CTGCAG oligonucleotide. Likewise, a 4,4,8,8,12,12-hexamethyl-4,8,12-triazapentadecamethylene linker should be optimal for binding in two isolated 5′-CT-3′ sites separated by one base pair. Recently, Glazer and co-workers (Benson et al., 1995; Zeng et al., 1995) have dealt with the problem of modifying the length of the linker in bisintercalating fluorescent dimeric dyes. In the case of the heterodimeric dyes TOTAB and TOTIN, which are closely related to TOTO, they found that the linker length strongly affected

876 Bioconjugate Chem., Vol. 8, No. 6, 1997

the emission spectra of the dsDNA-bound dye. In agreement with our work they observed that the energy transfer from donor to acceptor chromophores was optimized in the analog with a linker slightly longer than the 4,4,8,8-tetramethyl-4,8-diazaundecamethylene linker in TOTO. EXPERIMENTAL PROCEDURES

Synthesis. 1-(4-Iodoalkyl)-4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]quinolinium Iodide (3). Equimolar amounts of 1 and 2 were dissolved in absolute ethanol by heating. One equivalent of triethylamine was added, and the mixture was stirred at room temperature for 30 min. The title compound precipitated on addition of ether and was recrystallized from acetone/ether to give a red powder. 3b: yield 75%; mp 216-219 °C; 1H NMR [(CD3)2SO] δ 7.00-8.75 (m, 10H, Ar), 6.78 (s, 1H, CH), 4.54 (m, 2H, NCH2), 3.93 (s, 3H, NCH3), 3.33 (m, 2H, CH2I), 1,89 (m, 4H, 2 × CH2); FAB MS m/z 473 (M+). 3c: yield 75%; mp 206-209 °C; 1H NMR [(CD3)2SO] δ 7.26-8.77 (m, 10H, Ar), 6.85 (s, 1H, CH), 4.58 (m, 2H, NCH2), 3.99 (s, 3H, NCH3), 3.29 (m, 2H, CH2I), 1.83 (m, 4H, 2 × CH2), 1.46 (m, CH2); FAB MS m/z 487 (M+). 1,1′-(4,4,8,8-Tetramethyl-4,8-diazadodecamethylene)bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium Tetraiodide, TOTO10. A suspension of 3a in anhydrous methanol and 8 equiv of N,N,N′,N′-tetramethyl-1,3-propanediamine were refluxed for 12 h. Evaporation of the solvent and subsequent recrystallization from acetone/ether afforded 4 in 80% yield, mp 230-233 °C. Equivalent amounts of 4 and 3b in anhydrous methanol were refluxed for 12 h. TOTO10 was precipitated by addition of ether and recrystallized from acetone/ether: yield 85%; mp 238-240 °C; 1H NMR [(CD3)2SO] δ 7.26-8.77 (m, 20H, Ar), 6.85 (s, 1H, CH), 6.88 (s, 1H, CH) 4.68 (m, 4H, NCH2), 4.00 (s, 3H, NCH3), 3.99 (s, 3H, NCH3), 3.5-3.9 (m, 8H, 4xNCH2), 3.21 (s, 12H, NCH3), 2.40 (m, 4H, 2 × CH2), 1.96 (m, 4H, 2 × CH2); FAB MS m/z 1189 (M+ - 127). Anal. Calcd for C50H60I4N6S2‚4H2O: C, 43.24; H, 4.94; N, 6.05. Found: C, 42.80; H, 4.86; N, 5.70. 1,1′-(5,5,9,9-Tetramethyl-5,9-diazatridecamethylene)bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2-methylidene]]quinolinium Tetraiodide, TOTO11. A suspension of 3b and 0.5 equiv of N,N,N′,N′-tetramethyl-1,3propanediamine in anhydrous methanol was heated at reflux for 24 h. Recrystallization from DMF/MeOH afforded the title compound as an orange-red solid: yield 80%; mp 258-261 °C; 1H NMR [(CD3)2SO] δ 7.15-8.70 (m, 20H, Ar), 6.76 (s, 2H, CH) 4.68 (m, 4H, NCH2), 3.94 (s, 6H, NCH3), 3.61 and 3.49 (2m, 8H, 4 × NCH2), 3.20 (s, 12H, NCH3), 2.34 (m, 2H, CH2), 1.97 (m, 8H, 4 × CH2); FAB MS m/z 1203 (M+ - 127). Anal. Calcd for C51H62I4N6S2‚2H2O: C, 44.82; H, 4.87; N, 6.15. Found: C, 44.73; H, 4.90; N, 5.78. 1,1′-(6,6,10,10-Tetramethyl-6,10-diazapentadecamethylene)bis[4-[3-methyl-2,3-dihydro(benzo-1,3-thiazole)-2methylidene]]quinolinium Tetraiodide, TOTO13. The same procedure as for TOTO11 was used: yield 90%; mp 253-255 °C; 1H NMR [(CD3)2SO] δ 7.23-8.76 (m, 20H, Ar), 6.82 (s, 2H, CH) 4.62 (m, 4H, NCH2), 3.98 (s, 6H, NCH3), 3.40 (m, 8H, 4 × NCH2), 3.13 (s, 12H, NCH3), 2.26 (m, 2H, CH2), 1.91-1.94 (m, 8H, 4 × CH2), 1.46 (m, 4H, 2 × CH2); FAB MS m/z 1231 (M+ - 127). Anal. Calcd for C53H66I4N6S2‚3H2O: C, 45.04; H, 5.18; N, 5.95. Found: C, 44.67; H, 5.18; N, 5.69. Materials. Purified DNA oligonucleotides were purchased from DNA Technology (Aarhus, Denmark) and

Stærk et al.

used without further purification. The non-self-complementary single-stranded DNA oligomers were added to an equivalent amount of the complementary strand and duplexes formed by annealing from 80 °C over 2 h. In this work four different oligonucleotides have been used (Scheme 2). Stock solutions of TOTO10, TOTO11, and TOTO13 in DMSO-d6 were used for complex formation with the dsDNA oligonucleotides according to the procedure described earlier (Jacobsen et al., 1995). A phosphate buffer containing ethylenediaminetetraacetic acid (EDTA), 2,2dimethyl-2-silapentane-5-sulfonate (DSS), and NaN3 was added, giving a final concentration in the NMR sample of 10 mM Pi (pD 7.0), 0.025 mM EDTA, 0.1 mM DSS, and 0.01 mM NaN3. The complexes were redissolved in 500 µL of 99.96% D2O (from Cambridge Isotope Laboratories). For experiments in H2O a mixture of 90% H2O/ 10% D2O (500 µL) was used. The samples were kept under N2 and remained stable for months. Absorbance and Fluorescence Emission Spectra. For determination of the absorbance and fluorescence emission spectra of the dye-dsDNA complexes, aqueous solutions of 0.02 mM were used. A phosphate buffer was added, giving a final concentration in a 3 mL sample of 10 mM Pi (pD 7.0), 0.02 mM EDTA, 0.1 mM DSS, and 0.02 mM NaN3. After recording of the absorption spectrum, the excitation wavelength for the fluorescence emission was selected to the wavelength of maximum absorbance. A similar aqueous solution of the free dye was prepared and diluted until the absorbance was identical to that of the dye-dsDNA. Once again, the excitation wavelength was selected to the wavelength of the maximum absorbance. The intensities at the fluorescence emission maximum were used to calculate the fluorescence enhancement defined as Fbound/Ffree. Absorbance spectra were recorded on a Shimadzu UV-160 spectrophotometer, and fluorescence measurements were performed with a Perkin-Elmer MPF-3 fluorescence spectrophotometer. All spectra were recorded at room temperature. NMR Experiments. The NMR spectra of the free dyes were obtained on a Bruker AC250 NMR spectrometer. All NMR experiments on the DNA-dye complexes were performed at 500 MHz on a Varian Unity 500 NMR spectrometer. NOESY spectra in D2O were acquired with a mixing time of 200 ms using 1024 complex points in t2 and a spectral width of 5000 Hz. A total of 512 t1 experiments with 64 scans each were acquired using the States-Haberkorn-Ruben phase cycle. Furthermore, experiments with mixing times of 25, 50, 100, and 150 ms were acquired for the mixture of CTAG + TOTO11. These experiments were acquired successively without removing the sample from the magnet. NOESY spectra in H2O were acquired using 2048 complex points and a spectral width of 10 000 Hz using a NOESY pulse sequence in which the last 90° pulse was replaced by a pulse containing a notch to suppress the water signal (Stein et al., 1995). This ensured suppression of the water signal together with a linear excitation profile over the whole spectral width. TOCSY experiments in D2O were acquired with mixing times of 30 and 90 ms using 1024 complex points in t2 and a spectral width of 5000 Hz. A total of 512 t1 experiments with 64 scans each were acquired using the TPPI phase cycle. The 1H-31P HSQC experiments were recorded with 1024 complex points and a spectral width of 1000 Hz in both dimensions. A total of 512 t1 experiments with 288 scans each were acquired using the States-HaberkornRuben phase cycle.

Bioconjugate Chem., Vol. 8, No. 6, 1997 877

Bisintercalation of Homodimeric Thiazole Orange Dyes

One-dimensional spectra in D2O were acquired separately. All spectra were recorded at 10 or 25 °C using DSS as reference. The acquired data were processed using FELIX (version 95.0, Biosym/MSI, San Diego, CA). The TOCSY and NOESY spectra were assigned in the conventional way as described in the literature (Boelens et al., 1985; Feigon et al., 1983; Hare et al., 1983; Scheek et al., 1984, 1983). Molecular Modeling. The model structure of the CTGAG-TOTO13 complex was made using an isolated spin pair approximation. The assigned cross peaks from a 200 ms NOESY spectrum were integrated and transformed into a restraint file. Twenty structures were calculated using the following restrained molecular dynamics (RMD) protocol: initially 100 and 10 000 cycles of steepest descent and conjugate gradient energy minimization, respectively. This was followed by 28 ps of RMD (steps of 1 fs) with the following temperature profile: 425 K for 4 ps and then cooled to 250 K in 25 K steps of 3 or 4 ps each. Force constants of 50 kJ/mol were used. Finally, the structures were energy minimized using conjugate gradient until the derivative was