DNA sequence determinants for binding of the Escherichia coli catabolite gene activator protein.

The consensus DNA site for binding of the Escherichia coli catabolite gene activator protein (CAP) is 22 base pairs in length and is 2-fold symmetric: 5'-AAATGTGATCTAGATCACATTT-3'. Positions 4 to 8 of each half of the consensus DNA half-site are the most strongly conserved. In this report, we analyze the effects of substitution of DNA base pairs at positions 4 to 8, the effects of substitution of thymine by uracil and by 5-methylcytosine at positions 4, 6, and 8, and the effect of dam methylation of the 5'-GATC-3' sequence at positions 7 to 10. All DNA sites having substitutions of DNA base pairs at positions 4 to 8 exhibit lower affinities for CAP than does the consensus DNA site, consistent with the proposal that the consensus DNA site is the ideal DNA site for CAP. Specificity for T:A at position 4 appears to be determined solely by the thymine 5-methyl group. Specificity for T:A at position 6 and specificity for A:T at position 8 appear to be determined in part, but not solely, by the thymine 5-methyl group. dam methylation has little effect on CAP.DNA complex formation. The thermodynamically defined consensus DNA site spans 28 base pairs. All, or nearly all, DNA determinants required for maximal affinity for CAP and for maximal thermodynamically defined CAP.DNA ion pair formation are contained within a 28-base pair DNA fragment that has the 22-base pair consensus DNA site at its center. The quantitative data in this report provide base-line thermodynamic data required for detailed investigations of amino acid-base pair and amino acid-phosphate contacts in this protein-DNA complex.     

Identification of C/EBP basic region residues involved in DNA sequence recognition and half-site spacing preference.

C/EBP and GCN4 are basic region-leucine zipper (bZIP) DNA-binding proteins that recognize the dyad-symmetric sequences ATTGCGCAAT and ATGAGTCAT, respectively. The sequence specificities of these and other bZIP proteins are determined by their alpha-helical basic regions, which are related at the primary sequence level. To identify amino acids that are responsible for the different DNA sequence specificities of C/EBP and GCN4, two kinds of hybrid proteins were constructed: GCN4-C/EBP chimeras fused at various positions in the basic region and substitution mutants in which GCN4 basic region amino acids were replaced by the corresponding residues from C/EBP. On the basis of the DNA-binding characteristics of these hybrid proteins, three residues that contribute significantly to the differences in C/EBP and GCN4 binding specificity were defined. These residues are clustered along one face of the basic region alpha helix. Two of these specificity residues were not identified as DNA-contacting amino acids in a recently reported crystal structure of a GCN4-DNA complex, suggesting that the residues used by C/EBP and GCN4 to make base contacts are not identical. A random binding site selection procedure also was used to define the optimal recognition sequences for three of the GCN4-C/EBP fusion proteins. These experiments identify an element spanning the hinge region between the basic region and leucine zipper domains that dictates optimal half-site spacing (either directly abutted for C/EBP or overlapping by one base pair for GCN4) in high-affinity binding sites for these two proteins.     

DNA sequence preferences of GAL4 and PPR1: how a subset of Zn2 Cys6 binuclear cluster proteins recognizes DNA.

Biophysical and genetic experiments have defined how the Saccharomyces cerevisiae protein GAL4 and a subset of related proteins recognize specific DNA sequences. We assessed DNA sequence preferences of GAL4 and a related protein, PPR1, in an in vitro DNA binding assay. For GAL4, the palindromic CGG triplets at the ends of the 17-bp recognition site are essential for tight binding, whereas the identities of the internal 11 bp are much less important, results consistent with the GAL4-DNA crystal structure. Small reductions in affinity due to mutations at the center-most 5 bp are consistent with the idea that an observed constriction in the minor groove in the crystalline GAL4-DNA complex is sequence dependent. The crystal structure suggests that this sequence dependence is due to phosphate contacts mediated by arginine 51, as part of a network of hydrogen bonds. Here we show that the mutant protein GAL4(1-100)R51A fails to discriminate sites with alterations in the center of the site from the wild-type site. PPR1, a relative of GAL4, also recognizes palindromic CGG triplets at the ends of its 12-bp recognition sequence. The identities of the internal 6 bp do not influence the binding of PPR1. We also show that the PPR1 site consists of a 12-bp duplex rather than 16 bp as reported previously: the two T residues immediately 5' to the CGG sequence in each half site, although highly conserved, are not important for binding by PPR1. Thus, GAL4 and PPR1 share common CGG half sites, but they prefer DNA sequences with the palindromic CGG separated by the appropriate number of base pairs, 11 for GAL4 and 6 for PPR1.     

The M.EcoRV DNA-(adenine N6)-methyltransferase uses DNA bending for recognition of an expanded EcoDam recognition site

The M.EcoRV DNA methyltransferase recognizes GATATC sites. It is related to EcoDam, which methylates GATC sites. The DNA binding domain of M.EcoRV is similar to that of EcoDam suggesting a similar mechanism of DNA recognition. We show that amino acid residue Lys11 of M.EcoRV is involved in recognition of Gua1 and Arg128 contacts the Gua in base pair 6. These residues correspond to Lys9 and Arg124 in EcoDam, which recognize the Gua residues in both strands of the Dam recognition sequence, indicating that M.EcoRV and EcoDam make similar contacts to outermost base pairs of their recognition sequences and M.EcoRV recognizes its target site as an expanded GATC site. In contrast to EcoDam, M.EcoRV considerably bends the DNA (59+/-4 degrees) suggesting indirect readout of the AT-rich inner sequence. Recognition of an expanded target site by DNA bending is a new principle for changing DNA recognition specificity of proteins during molecular evolution. R128A is inefficient in DNA bending and binding, whereas K11A bends DNA with relaxed sequence specificity. These results suggest a temporal order of the formation of protein-DNA contacts in which the Gua6-Arg128 contact forms early followed by DNA bending and, finally, the formation of the Lys11-Gua1 contact.     

Interactions between the trp repressor and its operator sequence as studied by base analogue substitution.

A series of modified trp operator sequences has been prepared by the incorporation of seven different base analogues. Four of the analogues allow the site-specific deletion of functional groups present on the dA-dT and dT-dA base pairs at positions -4/+4 and -5/+5 in the trp operator. The remaining three analogues permit the incorporation of structural analogues of the native dA-dT or dG-dC base pairs. The duplex operator sequences all exhibit Tm values well above ambient temperature (48-70 degrees C), and these values generally correlate very well with the number of interstrand hydrogen bonds present. The affinity between the trp repressor and 14 modified operator sequences was examined using a recently developed alkaline phosphatase protection assay. The results from the analogue sequences used in this study suggest that the structure of the dA-dT or dT-dA base pairs at positions -4/+4 and -5/+5, respectively, has relatively little effect upon the solution binding by the trp repressor, but the protein is very sensitive to the orientation of the amino and carbonyl functional groups at the -4/+4 positions, which are involved in the formation of an interbase hydrogen bond present in the major groove. (The term structure in this case refers to the hydrogen bonding structure of the base pairs. We recognize that the introduction of conservative functional group deletions or reversals may affect other structural criteria such as hydration.) The deletion of individual functional groups from the operator sequence suggests that the carbonyl at dT+4 is critical for formation of the high-affinity sequence-specific complex. Additionally, the thymine methyl group at dT+4 and the N7 nitrogen of dA+5 appear to be critical contacts necessary for high-affinity binding by the repressor. The thymine carbonyl and the adenine N7 nitrogen are each responsible for approximately -1.5 kcal/mol of apparent free energy of binding. The thymine methyl provides a somewhat smaller contribution of -0.7 kcal/mol. Deletion of either of the adenine amino groups at dA-4 or dA+5 results in a sequence that binds to the repressor with a higher affinity than observed with the native sequence; this can be explained in that the functional groups lost are not critical for binding, and the resulting increased flexibility of the operator, or the creation of a more hydrophobic surface at these sites, enhances van der Waals contacts between the protein and the nucleic acid.     

The role of the minor groove substituents in indirect readout of DNA sequence by 434 repressor

The sequence of non-contacted bases at the center of the 434 repressor binding site affects the strength of the repressor-DNA complex by influencing the structure and flexibility of DNA (Koudelka, G. B., and Carlson, P. (1992) Nature 355, 89-91). We synthesized 434 repressor binding sites that differ in their central sequence base composition to test the importance of minor groove substituents and/or the number of base pair hydrogen bonds between these base pairs on DNA structure and strength of the repressor-DNA complex. We show here that the number of base pair H-bonds between the central bases apparently has no role in determining the relative affinity of a DNA site for repressor. Instead we find that the affinity of DNA for repressor depends on the absence or presence the N2-NH(2) group on the purine bases at the binding site center. The N2-NH(2) group on bases at the center of the 434 binding site appears to destabilize 434 repressor-DNA complexes by decreasing the intimacy of the specific repressor-DNA contacts, while increasing the reliance on protein contacts to the DNA phosphate backbone. Thus, the presence of an N2-NH(2) group on the purines at the center of a binding site globally alters the precise conformation of the protein-DNA interface.     

Mutations in 5S DNA and 5S RNA have different effects on the binding of Xenopus transcription factor IIIA

The effects on TFIIIA binding affinity of a series of substitution mutations in the Xenopus laevis oocyte 5S RNA gene were quantified. These data indicate that TFIIIA binds specifically to 5S DNA by forming sequence-specific contacts with three discrete sites located within the classical A and C boxes and the intermediate element of the internal control region. Substitution of the nucleotide sequence at any of the three sites significantly reduces TFIIIA binding affinity, with a 100-fold reduction observed for substitutions in the box C subregion. These results are consistent with a direct interaction of TFIIIA with specific base pairs within the major groove of the DNA. A comparison of the TFIIIA binding data for the same mutations expressed in 5S RNA indicates that the protein does not make any strong sequence-specific contacts with the RNA. Although the protein footprinting sites on the 5S DNA and 5S RNA are coincident, nucleotide substitutions in 5S RNA which moderately reduce TFIIIA binding affinity do not correspond at all to the three specific TFIIIA interaction sites within the gene. The implications of these results for models which attempt to reconcile the DNA and RNA binding activities of TFIIIA by proposing a common structural motif for the two nucleic acids are discussed.     

Alternative heterocycles for DNA recognition: an N-methylpyrazole/N-methylpyrrole pair specifies for A.T/T.A base pairs

Side-by-side pairs of three five-membered rings, N-methylpyrrole (Py), N-methylimidazole (Im), and N-methylhydroxy-pyrrole (Hp), have been demonstrated to distinguish each of the four Watson Crick base pairs in the minor groove of DNA. However, not all DNA sequences targeted by these pairing rules achieve affinities and specificities comparable to DNA binding proteins. We have initiated a search for new heterocycles which can expand the sequence repetoire currently available. Two heterocyclic aromatic amino acids. N-methylpyrazole (Pz) and 4-methylthiazole (Th), were incorporated into a single position of an eight-ring polyamide of sequence ImImXPy-gamma-lmPyPyPy-beta-Dp to examine the modulation of affinity and specificity for DNA binding by a Pz/Py pair and or a Th/Py pair. The X/Py pairings Pz/Py and Th/Py were evaluated by quantitative DNase I footprint titrations on a DNA fragment with the four sites 5'-TGGNCA-3' (N=T, A, G, C). The Pz/Py pair binds T.A and A.T with similar affinity to a Py/Py pair but with improved specificity. disfavoring both G.C and C.G by about 100-fold. The Th/Py pair binds poorly to all four Watson Crick base pairs. These results demonstrate that in some instances new heterocyclic aromatic amino acid pairs can be incorporated into imidazole-pyrrole polyamides to mimic the DNA specificity of Py/Py pairs which may be relevant as biological criteria in animal studies become important.     

A comparison of the different DNA binding specificities of the bZip proteins C/EBP and GCN4.

The bZip proteins GCN4 and C/EBP differ in their DNA binding specificities: GCN4 binds well to the pseudopalindromic AP1 site 5'-A4T3G2A1C0T1C2'A3'T4'-3' and to the palindromic ATF/CREB sequence 5'-A4T3G2A1-C0*G0'T1'C2'A3'T4'-3'; C/EBP preferentially recognizes the palindromic sequence 5'-A4T3T2G1C0*G0'C1'A2'-A3'T4'-3'. According to the X-ray structures of GCN4-DNA complexes, five residues of the basic region of GCN4 are involved in specific base contacts: asparagine -18, alanine -15, alanine -14, serine -11 and arginine -10 (numbered relative to the start point of the leucine zipper, which we define as +1). In the basic region of C/EBP position -14 is occupied by valine instead of alanine, the other four residues being identical. Here we analyse the role of valine -14 in C/EBP-DNA complex formation. Starting from a C/EBP-GCN4 chimeric bZip peptide which displays C/EBP specificity, we systematically mutated position -14 of its basic region and characterized the DNA binding specificities of the 20 possible different peptides by gel mobility shift assays with various target sites. We present evidence that valine -14 of C/EBP interacts more strongly with thymine 2 than with cytosine 1' of the C/EBP binding site, unlike the corresponding alanine -14 of GCN4, which exclusively contacts thymine 1' of the GCN4 binding sites.     

Combinatorial determination of sequence specificity for nanomolar DNA-binding hairpin polyamides

Development of sequence-specific DNA-binding drugs is an important pharmacological goal, given the fact that numerous existing DNA-directed chemotherapeutic drugs rely on the strength and selectivity of their DNA interactions for therapeutic activity. Among the DNA-binding antibiotics, hairpin polyamides represent the only class of small molecules that can practically bind any predetermined DNA sequence. DNA recognition by these ligands depends on their side-by-side amino acid pairings in the DNA minor groove. Extensive studies have revealed that these molecules show extremely high affinity for sequence-directed, minor groove interaction. However, the specificity of such interactions in the presence of a large selection of sequences such as the human genome is not known. We used the combinatorial selection method restriction endonuclease protection, selection, and amplification (REPSA) to determine the DNA binding specificity of two hairpin polyamides, ImPyPyPy-gamma-PyPyPyPy-beta-Dp and ImPyPyPy-gamma-ImPyPyPy-beta-Dp, in the presence of more than 134 million different sequences. These were verified by restriction endonuclease protection assays and DNase I footprinting analysis. Our data showed that both hairpin polyamides preferentially selected DNA sequences having consensus recognition sites as defined by the Dervan pairing rules. These consensus sequences were rather degenerate, as expected, given that the stacked pyrrole-pyrrole amino acid pairs present in both polyamides are unable to discriminate between A.T and T.A base pairs. However, no individual sequence within these degenerate consensus sequences was preferentially selected by REPSA, indicating that these hairpin polyamides are truly consensus-specific DNA-binding ligands. We also discovered a preference for overlapping consensus binding sites among the sequences selected by the hairpin polyamide ImPyPyPy-gamma-PyPyPyPy-beta-Dp, and confirmed by DNase I footprinting that these complex sites provide higher binding affinity. These data suggest that multiple hairpin polyamides can cooperatively bind to their highest-affinity sites.     

Indirect Readout of DNA Sequence by P22 Repressor: Roles of DNA and Protein Functional Groups in Modulating DNA Conformation

The repressor of bacteriophage P22 (P22R) discriminates between its various DNA binding sites by sensing the identity of non-contacted base pairs at the center of its binding site. The “indirect readout” of these non-contacted bases is apparently based on DNA's sequence-dependent conformational preferences. The structures of P22R–DNA complexes indicate that the non-contacted base pairs at the center of the binding site are in the B′ state. This finding suggests that indirect readout and therefore binding site discrimination depend on P22R's ability to either sense and/or impose the B′ state on the non-contacted bases of its binding sites. We show here that the affinity of binding sites for P22R depends on the tendency of the central bases to assume the B′-DNA state. Furthermore, we identify functional groups in the minor groove of the non-contacted bases as the essential modulators of indirect readout by P22R. In P22R–DNA complexes, the negatively charged E44 and E48 residues are provocatively positioned near the negatively charged DNA phosphates of the non-contacted nucleotides. The close proximity of the negatively charged groups on protein and DNA suggests that electrostatics may play a key role in the indirect readout process. Changing either of two negatively charged residues to uncharged residues eliminates the ability of P22R to impose structural changes on DNA and to recognize non-contacted base sequence. These findings suggest that these negatively charged amino acids function to force the P22R-bound DNA into the B′ state and therefore play a key role in indirect readout by P22R.     

Contacts between the factor TUF and RPG sequences

The yeast TUF factor binds specifically to RPG-like sequences involved in multiple functions at enhancers, silencers, and telomeres. We have characterized the interaction of TUF with its optimal binding sequence, rpg-1 (1-ACACCCATACATTT-14), using a gel DNA-binding assay in combination with methylation protection and mutagenesis experiments. As many as 10 base pairs appear to be engaged in factor binding. Analysis of a collection of 30 different RPG mutants demonstrated the importance of 8 base pairs at position 2, 3, 4, 5, 6, 7, 10, and 12 and the critical role of the central GC pair at position 5. Methylation protection data on four different natural sites confirmed a close contact at positions 4, 5, 6, and 10 and suggested additional contacts at base pairs 8, 12, and 13. The derived consensus sequence was RCAAYCCRYNCAYY. A quantitative band shift analysis was used to determine the equilibrium dissociation constant for the complex of TUF and its optimal binding site rpg-1. The specific dissociation constant (K8) was found to be 1.3 x 10(-11) M. The comparison of the K8 value with the dissociation constant obtained for nonspecific DNA sites (Kn8 = 8.7 x 10(-6) M) shows the high binding selectivity of TUF for its specific RPG target.     

Triple-helix formation is compatible with an adjacent DNA-protein complex.

The effect of oligonucleotide-directed triple-helix formation on the binding of a protein to an immediately adjacent sequence has been examined. A double-stranded oligonucleotide was designed with a target site for the binding of a pyrimidine oligonucleotide located immediately adjacent to the recognition sequence for the herpes simplex virus type 1 (HSV-1) origin of replication binding protein, which is encoded by the UL9 gene of HSV-1. Since the optimal conditions for the binding of the UL9 protein and the pyrimidine oligonucleotide to the duplex DNA are markedly different, a pyrimidine oligonucleotide was designed to optimize binding affinity and specificity for the target duplex oligonucleotide. Consideration was given to length and sequence composition in an effort to maximize triple-strand formation under conditions amenable to the formation of the UL9-DNA complex. Using gel mobility shift assays, a trimolecular complex composed of duplex DNA bound to both a third oligonucleotide strand and the UL9 protein was detected, indicating that the UL9-DNA complex is compatible with the presence of a triple helix in the immediately adjacent sequences.     

Sequence specificity of the binding of 9-aminoacridine- and amsacrine-4-carboxamides to DNA studied by DNase I footprinting.

DNase I footprinting has been used to probe the sequence selectivity of binding of a series of intercalating amsacrine-4-carboxamides and a related 9-aminoacridine-4-carboxamide to three DNA restriction fragments. These ligands have good experimental antileukemic activity, and for those members of the series that gave evaluable footprints, our principal finding is that they bind preferentially to GC-rich regions in agreement with the conclusion of equilibrium and kinetic measurements. The highest affinity sites generally occur in clusters of GC base pairs with runs of AT pairs being excluded from binding. It is important to appreciate that the 9-aminoacridine- and amsacrine-4-carboxamides exhibit a very high degree of selectivity for GC sites which, to our knowledge, has not been previously matched by acridine derivatives in footprinting experiments. The principal determinant of specificity appears to be the 4-carboxamide group itself since neither variations in the terminal funtionality of the 4-carboxamide sidechain nor the presence of the 9-anilino substituent modifies sequence preferences. The molecular origins of selectivity may be discerned in terms of potential hydrogen bonding interactions between the 4-carboxamide moiety and carbonyl oxygen and amino groups of GC base pairs in the DNA minor groove at CG dinucleotide sites. The related therapeutic agent amsacrine failed to inhibit cleavage by DNase I, so no conclusion can be drawn concerning its binding selectivity, save to note that amsacrine does not possess the 4-carboxamide group which appears to be the crucial determinant of GC specificity. Whether selectivity for binding to GC-rich sequences is an important element in the antitumor activity of both the 9-aminoacridine- and amsacrine-4-carboxamides remains to be determined.