Recognition of (dG)n.(dC)n sequences by endonuclease G. Characterization of the calf thymus nuclease

We report the purification of endonuclease G (Ruiz-Carrillo, A., and Renaud, J. (1987) EMBO J. 6, 401-407) from calf thymus nuclei and whole tissue. The enzyme has been enriched 29,000-fold, and the activity was unambiguously identified with a 26-kDa protein after renaturation following sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The native nuclease behaves as a 50-kDa species by gel filtration, suggesting that it is composed of two subunits, presumably identical. In terms of absolute amounts, endonuclease G (endo G) is a nuclear enzyme although it was also detected in purified mitochondria. Endo G is highly specific for (dG)n.(dC)n tracts in DNA, nicking either strand of relaxed substrates with similar kinetics. The sensitivity of the homopolymer tracts is proportional to their length (from n = 8 to 29), insofar as the flanking sequences are constant. However, the overall rate of cleavage is influenced by the composition of the flanking DNA. Minor cleavage sites contain shorter (dG)n.(dC)n clusters (n = 3-7). Endo G efficiently cleaves (dC)n but not (dG)n runs in single-stranded DNA, suggesting that it may recognize an asymmetric strand conformation of the homopolymer tracts. Endo G does not recognize other homo(co)-polymer sequences or cruciform structures in DNA.     

Monoclonal antibodies specific for poly(dG) X poly(dC) and poly(dG) X poly(dm5C)

Most duplex DNAs that are in the "B" conformation are not immunogenic. One important exception is poly(dG) X poly(dC), which produces a good immune response even though, by many criteria, it adopts a conventional right-handed helix. In order to investigate what features are being recognized, monoclonal antibodies were prepared against poly(dG) X poly(dC) and the related polymer poly(dG) X poly(dm5C). Jel 72, which is an immunoglobulin G, binds only to poly(dG) X poly(dC), while Jel 68, which is an immunoglobulin M, binds approximately 10-fold more strongly to poly(dG) X poly(dm5C) than to poly(dG) X poly(dC). For both antibodies, no significant interaction could be detected with any other synthetic DNA duplexes including poly[d(Gm5C)] X poly[d(Gm5C)] in both the "B" and "Z" forms, poly[d(Tm5Cm5C)] X poly[d(GGA)], and poly[d(TCC)] X poly[d(GGA)], poly(dI) X poly(dC), or poly(dI) X poly(dm5C). The binding to poly(dG) X poly(dC) was inhibited by ethidium and by disruption of the DNA duplex, confirming that the antibodies were not recognizing single-stranded or multistranded structures. Furthermore, Jel 68 binds significantly to phage XP-12 DNA, which contains only m5C residues and will precipitate this DNA in the absence of a second antibody. The results suggest that (dG)n X (dm5C)n sequences in natural DNA exist in recognizably distinct conformations.     

Detection of triple-helix related structures adopted by poly(dG)-poly(dC) sequences in supercoiled plasmid DNA.

Negative superhelical strain induces the poly(dG)-poly(dC) sequence to adopt two totally different types of triple-helices, either a dG.dG.dC triplex in the presence of Mg(+)+ at both neutral and acidic pHs or a protonated dC+.dG.dC triplex in the absence of Mg(+)+ ions at acidic pH (1). To examine whether there are still other types of non-B DNA structures formed by the same sequence, we constructed supercoiled plasmid DNAs harboring varying lengths of the poly(dG) tract, and the structures adopted by each supercoiled plasmid DNA were studied with a chemical probe, chloroacetaldehyde. The potential of a poly(dG)-poly(dC) sequence to adopt non-B DNA structures depends critically on the length of the tract. Furthermore, in the presence of Mg(+)+ and at a mildly acidic pH, in addition to the expected dG.dG.dC triplex detected for the poly(dG) tracts of 14 to 30 base pairs (bp), new structures were also detected for the tracts longer than 35 bp. The structure formed by a poly(dG) tract of 45 bp revealed chemical reaction patterns consistent with a dG.dG.dC triplex and protonated dC+.dG.dC triple-helices fused together. This structure lacks single-stranded stretches typical of intramolecular triplexes.     

Poly(dG)-poly(dC) sequences, under torsional stress, induce an altered DNA conformation upon neighboring DNA sequences

Supercoiled plasmid DNAs (at bacterial superhelical density) harboring the homopurine-homopyrimidine sequence, poly(dG)-poly(dC), were reacted with bromoacetaldehyde (BAA), a reagent that reacts with unpaired DNA bases. Not only did the poly(dG)-poly(dC) sequence react with BAA but, surprisingly, neighboring sequences located 3' to the contiguous G sequences also reacted. The altered conformation in the poly(dG)-poly(dC) sequence and in the neighboring sequence occurred in the same supercoiled plasmid DNA molecule. Furthermore, the occurrence of an "unpaired" conformation in the neighboring sequence is strictly due to a positional effect, since it is observed when the poly(dG)-poly(dC) segment is adjacent to a variety of neighboring sequences.     

Endonuclease G from mammalian nuclei is identical to the major endonuclease of mitochondria.

Two Mg(2+)-dependent DNA endonucleases have been isolated from mammalian cells which have a strong preference to nick within long tracts of guanine residues in vitro. One endonuclease activity is mitochondrial (mt). The other endonuclease, called Endonuclease G, is associated with isolated nuclei, and is released when the nuclear chromatin is exposed to moderate ionic strength. Our laboratory has previously purified the mt endonuclease to near homogeneity from mitochondria of bovine heart and reported the enzyme to be a homodimer of a approximately 29 kDa polypeptide [Cummings, O. W. et al. (1987) J. Biol. Chem., 262, 2005-2015]. Although the purified mt endonuclease will extensively fragment M13 viral ssDNA and plasmid dsDNAs in vitro, the enzyme displays an unusually strong preference to nick within a (dG)12:(dC)12 sequence tract which resides just upstream from the origin of DNA replication in the mitochondrial genome. The nuclear Endonuclease G first identified from its selective targeting of several (dG)n:(dC)n tracts in vitro (where N = 3-29), was subsequently purified from calf thymus nuclei and shown to be a homodimer of a approximately 26-kDa polypeptide [Côté, J. et al. (1989) J. Biol. Chem., 264, 3301-3310]. In the present study, we find that Endonuclease G partially purified from calf thymus nuclei will extensively degrade both viral ss- and dsDNAs in vitro, and that the enzyme possesses biochemical properties and specificity for nucleotide sequences in DNA that are strongly related or identical to those of the mt endonuclease. These findings and the discovery of sequence identity between the proteins strengthen the conclusion that the nuclear Endonuclease G is the same enzyme as the mt endonuclease.     

Substrate specificity of HeLa endonuclease R. A G-specific mammalian endonuclease

We examined the substrate specificity of endonuclease R (endo R) a mammalian endonuclease that cleaves G.C-rich DNA sequences. The best substrates for double-stranded cleavage were homopolymeric stretches of poly(dG).poly(dC). Plasmids which contain other G-rich sequences were also cleaved but at a reduced frequency. These included the telomeric sequences, d(G4T2) and d(G2-6A), which were cleaved at approximately one-third the frequency of d(G)n.d(C)n. The alternating copolymer d(GA) and the terminal sequences of adeno-associated virus d(G1-3T/A) were also cut. Poly(dA).poly(dT) and the alternating copolymer d(GC)n were not detectably cleaved. Although endo R has a nicking activity which converts supercoiled plasmids to nicked circular DNA, the nicking activity is random with respect to plasmid sequences. Specific cleavage of G-rich sequences appears to occur by a concerted double-stranded mechanism. The cleavage pattern within the G-rich runs suggests that cleavage can occur anywhere within the G-rich region. Product ligation experiments indicate that a limited number of cleavage events (1-2) occur/molecule. Inasmuch as the best substrates for endo R are d(G)n.d(C)n and telomeric sequences, we suggest that endo R may directly recognize and cleave DNA that contains G.G base pairing.     

Chain length-dependent association of distamycin-type oligopeptides with A X T and G X C pairs in polydeoxynucleotide duplexes

Different binding affinities of various distamycin analogs including the deformylated derivative with poly(dA-dC) X poly(dG-dT) were investigated using CD measurements. The inhibitory effect of distamycins on the DNAase I cleavage activity of DNA duplexes strongly supports the binding data. The base specificity of the ligand interaction with duplex DNA depends on the chain length of distamycin analogs. Netropsin, distamycin-2 and the deformylated distamycin-3 show no binding to dG X dC containing sequences at moderate ionic strength and are classified as highly dA X dT specific. In contrast distamycin having three, four or five methylpyrrolecarboxamide groups also forms more or less stable complexes with dG X dC-containing duplexes. These ligands possess a lower basepair specificity. The correlation between binding behavior and oligopeptide structure shows that presence of the number of hydrogen acceptor and donor sites determines the basepair and sequence specificity. The additional interaction with dG X dC pairs becomes essential when the number of hydrogen acceptor sites exceeds n = 3.     

The isolation of duplex DNA fragments containing (dG.dC) clusters by chromatography on poly(rC)-Sephadex.

Duplex DNA containing oligo(dG.dC)-rich clusters can be isolated by specific binding to poly(rC)-Sephadex. This binding, probably mediated by the formation of an oligo(dG.dC)rC+ triple helix, is optimal at pH 5 in 50% formamide, 2 M LiCl; the bound DNA is recovered by elution at pH 7.5. Using this method we find that the viral DNAs PM2, lambda and SV40 contain at least 1, 1 and 2 sites for binding to poly(rC)-Sephadex, respectively. These binding sites have been mapped in the case of SV40; the binding sites can in turn be used for physical mapping studies of DNAs containing (dG.dC) clusters. Inspection of the sequence of the bound fragments of SV40 DNA shows that a (dG.dC)6-7 tract is required for the binding of duplex DNA to poly(rC)-Sephadex. Although about 60% of rabbit DNA cleaved with restriction endonuclease KpnI binds to poly(rC)-Sephadex, no binding is observed for the 5.1 kb DNA fragment generated by KpnI digestion, which contains the rabbit beta-globin gene. This indicates that oligo(dG.dC) clusters are not found close to the rabbit beta-globin gene.     

Intramolecular dG.dG.dC triplex detected in Escherichia coli cells.

The formation of an intramolecular dG.dG.dC triplex in Escherichia coli cells is demonstrated at single-base resolution. The intramolecular dG.dG.dC triplex structure was probed in situ for E. coli cells containing plasmid DNAs with varying lengths of poly(dG).poly(dC) tracts employing chloroacetaldehyde. This chemical probe reacts specifically with unpaired DNA bases. The triplex structure formed with the poly(dG).poly(dC) tracts of 35 and 44 base-pairs, but not with 25 base-pairs. The triplex was detected only one to two hours after the chloramphenicol treatment: the period at which the extracted plasmid DNA revealed the maximal superhelical density.     

Protonated polynucleotides structures - 22.CD study of the acid-base titration of poly(dG).poly(dC)

The acid-base titration (pH 8 --> pH 2.5 --> pH 8) of eleven mixing curve samples of the poly(dG) plus poly(dC) system has been performed in 0.15 M NaCl. Upon protonation, poly(dG).poly(dC) gives rise to an acid complex, in various amounts according to the origin of the sample. We have established that the hysteresis of the acid-base titration is due to the non-reversible formation of an acid complex, and the liberation of the homopolymers at the end of the acid titration and during the base titration: the homopolymer mixtures remain stable up to pH 7. A 1G:1C stoichiometry appears to be the most probable for the acid complex, a 1G:2C stoichiometry, as found in poly(C(+)).poly(I).poly(C) or poly(C(+)).poly(G).poly(C), cannot be rejected. In the course of this study, evidence has been found that the structural consequences of protonation could be similar for both double stranded poly(dG).poly(dC) and G-C rich DNA's: 1) protonation starts near pH 6, dissociation of the acid complex of poly(dG).poly(dC) and of protonated DNA take place at pH 3; 2) the CD spectrum computed for the acid polymer complex displays a positive peak at 255 nm as found in the acid spectra of DNA's; 3) double stranded poly(dG).poly(dC) embedded in triple-stranded poly(dG).poly(dG).poly(dC) should be in the A-form and appears to be prevented from the proton induced conformational change. The neutral triple stranded poly(dG).poly(dG).poly(dC) appears therefore responsible, although indirectly, for the complexity and variability of the acid titration of poly(dG).poly(dC) samples.     

Homopolymer tract length dependent enrichments in functional regions of 27 eukaryotes and their novel dependence on the organism DNA (G+C)% composition

Background  

DNA homopolymer tracts, poly(dA).poly(dT) and poly(dG).poly(dC), are the simplest of simple sequence repeats. Homopolymer tracts have been systematically examined in the coding, intron and flanking regions of a limited number of eukaryotes. As the number of DNA sequences publicly available increases, the representation (over and under) of homopolymer tracts of different lengths in these regions of different genomes can be compared.     

Factor C from rabbit liver. A new poly(dC) and poly[d(G-C)] template-selective stimulatory protein of DNA polymerases

We have undertaken a search for mammalian DNA-binding proteins that enhance the activity of DNA polymerases in a template sequence-specific fashion. In this paper, we report the extensive purification and characterization of a new DNA-binding protein from rabbit liver that selectively stimulates DNA polymerases to copy synthetic poly[d(G-C)] and the poly(dC) strand of poly(dC).poly(dG) as well as single-stranded natural DNA that contains stretches of oligo(dC). The enhancing protein, a polypeptide of 65 kDa designated factor C, stimulates the copying of the two synthetic templates by Escherichia coli DNA polymerase I, Micrococcus luteus polymerase, and eukaryotic DNA polymerases alpha and beta, but not by avian myeloblastosis virus polymerase. Factor C, however, does not affect utilization by these polymerases of the poly(dG) strand of poly(dC).poly(dG), of poly(dC) primed by oligo(dG), or of poly(dA).poly(dT) and poly[d(A-T)]. With polymerase I, Michaelis constants (Km) of poly[d(G-C)] and of the poly(dC) strand of poly(dC).poly(dG) are decreased by factor C 37- and 4.7-fold, respectively, whereas maximum velocity (Vmax) remains unchanged. By contrast, neither the Km value of the poly(dG) strand of poly(dC).poly(dG) nor the Vmax value with this template is altered by factor C. Rates of copying of activated DNA, denatured DNA, or singly primed M13 DNA are not affected significantly by factor C. However, primer extension analysis of the copying of recombinant M13N4 DNA that contains runs of oligo(dC) within an inserted thymidine kinase gene shows that factor C increases processivity by specifically augmenting the efficiency at which polymerase I traverses the oligo(dC) stretches. Direct binding of factor C to denatured DNA is indicated by retention of the protein-DNA complex on columns of DEAE-cellulose. Binding of factor C to poly[d(G-C)] is demonstrated by the specific adsorption of the enhancing protein to columns of poly[d(G-C)]-Sepharose. We propose that by binding to poly[d(G-C)] and to poly(dC).poly(dG), factor C enables tighter binding of some DNA polymerases to these templates and facilitates enzymatic activity.     

Magnesium-dependent supercoiling-induced transition in (dG)n.(dC)n stretches and formation of a new G-structure by (dG)n strand.

Plasmids containing (dG)27.(dC)27 inserts (pPG27), (dG)37.(dC)37 inserts (pPG37), and (dG)24C(dG)21.(dC)24G(dC)21 inserts (pPG46C) were constructed for the study of structural transitions within (dG)n.(dC)n stretches. Two-dimensional gel electrophoresis has shown that a Mg2+-dependent supercoiling-induced structural transition takes place at pH 8 in plasmid pPG46C. The transition occurs at -0=0.06 and involves a supercoiling release corresponding to 5 superhelical turns. After denaturation of the restriction fragments containing (dG)n.(dC)n inserts, the strands do not renature completely and (dG)n-containing strand migrates in PAGE much faster than the (dC)n-containing one. Chemical modification experiments with the (dG)n-strand have revealed the periodic nature of the protection of guanines against dimethyl sulfate methylation. The (dG)n strand in the presence of Mg2+ forms complexes with the complementary (dC)n strand, which differ from the native duplex in mobility. We believe these effects to be due to the formation of an intrastrand structure within the (dG)n strand stabilized by G.G interactions (we called it G-structure), which in the presence of Mg2+ forms an interstrand complex. with the (dC)n strand.     

500-MHz 1H NMR study of poly(dG).poly(dC) in solution using one-dimensional nuclear Overhauser effect

Secondary structures of poly(dG).poly(dC) and poly(dG).poly(dm5C) in solution are determined by nuclear Overhauser effect (NOE) measurements on GH8-deuterated and -nondeuterated DNAs with low presaturation pulse lengths (10-25 ms) and low-power and prolonged accumulations in the range of 50,000-72,000 scans. Under these conditions, the NOE difference spectra were free from diffusion. Primary NOEs between base protons GH8/CH6 and sugar protons H1', H2'/H2', and H3' suggest that in poly(dG).poly(dC) both guanine and cytosine nucleotides adopt a C3'-endo, low anti X = 200-220 degrees conformation. Computer modeling of the NOE data enable identification for the first time, in terms of the geometry of the nucleotide repeat, handedness, and helix geometry, of the structure of poly(dG).poly(dC) to be the A form, and the derived structure for the polymer duplex is very close to the single crystal structure of the double-helical d-GGGGCCCC [McCall, M., Brown, T., & Kennard, O. (1985) J. Mol. Biol. 183, 385-396]. Similar nuclear Overhauser effect data on poly(dG).poly(dm5C) revealed that G and m5C adopt a C2'endo, anti X = 240-260 degrees conformation, which indicates that this DNA exhibits the B form in solution. In summary, the results presented in this paper demonstrate that methylation of cytosines in poly(dG).poly(dC) causes A----B transition in the molecule.     

Factor D is a selective single-stranded oligodeoxythymidine binding protein.

Factor D, a protein purified from rabbit liver that selectively enhances traversal of template oligodeoxythymidine tracts by diverse DNA polymerases, was examined for the sequence specificity of its binding to DNA. Terminally [32P]-labeled oligomers with the sequence 5'-d[AATTC(N)16G]-3', N being dT, dA, dG, or dC, were interacted with purified factor D and examined for the formation of protein-DNA complexes that exhibit retarded electrophoretic mobility under nondenaturing conditions. Whereas significant binding of factor D to 5'-d[AATTC(T)16G]-3' is detected, there is no discernable association between this protein and oligomers that contain 16 contiguous moieties of dG, dA, or dC. Furthermore, factor D does not form detectable complexes with the duplexes oligo(dA).oligo(dT) or poly(dA).poly(dT). The preferential interaction of factor D with single-stranded poly(dT) is confirmed by experiments in which the polymerase-enhancing activity of this protein is protected by poly(dT) against heat inactivation two- and four-fold more efficiently than by poly(dA) or poly(dA).poly(dT), respectively.     

Pisum sativum endonuclease. Studies on substrate specificity and possible use as a biochemical tool

An endonuclease purified from germinating pea (Pisum sativum) seeds has been shown to catalyze the hydrolysis of heat-denatured single-stranded DNA. Since P. sativum endonuclease shows appreciable activity in the presence of DNA destabilizing agents and, unlike many similar endonucleases, significant activity at neutral pH, it is a potentially valuable tool for studies of the secondary structure of nucleic acids. The residual hydrolysis of duplex DNA is directed towards partially denatured, A,T-rich areas in native DNA. The rate of hydrolysis of deoxypolynucleotides was in the order poly(dT) greater than denatured DNA greater than poly(dA) greater than poly(dA-dT) = native DNA. Neither poly(dC), poly(dG) nor poly(dC).poly(dG) were attacked by the enzyme. Supercoiled, covalently closed circular phage PM2 form I DNA is converted to singly hit nicked circular form II and doubly hit linear from III duplexes. Prolonged treatment with enzyme does not further cleave the linear form III DNA. Addition of increasing concentrations of NaCl in the incubation mixture suppresses the conversion of form I to form II, but not the conversion of form II to form III, which is enhanced with the increasing ionic strength. The enzymatically relaxed circular form, I degree, obtained by unwinding of supercoiled DNA with a DNA-relaxing protein, is resistant to the action of the enzyme. Molecules with intermediate superhelix densities do not serve as substrates. The sites of cleavage of P. sativum endonuclease in PM2 DNA occur within regions that are readily denaturable in a topologically constrained superhelical molecule.     

Poly(dG)-poly(dC) DNA appears shorter than poly(dA)-poly(dT) and possibly adopts an A-related conformation on a mica surface under ambient conditions

Three types of DNA: approximately 2700 bp polydeoxyguanylic olydeoxycytidylic acid [poly(dG)-poly(dC)], approximately 2700 bp polydeoxyadenylic polydeoxythymidylic acid [poly(dA)-poly(dT)] and 2686 bp linear plasmid pUC19 were deposited on a mica surface and imaged by atomic force microscopy. Contour length measurements show that the average length of poly(dG)-poly(dC) is approximately 30% shorter than that of poly(dA)-poly(dT) and the plasmid. This led us to suggest that individual poly(dG)-poly(dC) molecules are immobilized on mica under ambient conditions in a form which is likely related to the A-form of DNA in contrast to poly(dA)-poly(dT) and random sequence DNA which are immobilized in a form that is related to the DNA B-form.     

S1 nuclease sensitivity to cis- and trans-diamminedichloroplatinum(II) modified DNAS: influence of (G+C) content and nucleotide sequence

The sensitivity of S1 nuclease to cis- and trans-(NH3)2PtCl2 modified DNAs is examined as a function of the level of cis- and trans-(NH3)2PtCl2 bound, the % (G+C) content in DNA from different sources and the sequence dependence in poly(dG).poly(dC) and poly(dG-dC).poly(dG-dC). The extent of DNA digested increases with increasing levels of either isomer and is inversely influenced by the % (G+C) content of the DNA. However, the difference in the extent of digestion between the cis-and trans-(NH3)2PtCl2 modified DNAs at equivalent levels of bound isomer follows the order, calf-thymus greater than M. lysodeikticus greater than poly(dG-dC).poly(dG-dC). While there is virtually no difference in the digestion profiles for poly(dG-dC).poly(dG-dC) modified with the two isomers, there is a striking difference in the extent of digestion between cis- and trans-(NH3)2PtCl2 modified poly(dG).poly(dC). These results are discussed in light of the possible modes of binding for cis-(NH3)2PtCl2, previously reported findings on modified DNA and possible implications for modifications in cellular chromatin.     

Comparison between poly(dG-dC).poly(dG-dC) and DNA modified by cis-diamminedichloroplatinum (II): immunological and spectroscopic studies

The importance of the base composition and of the conformation of nucleic acids in the reaction with the drug cis-diamminedichloroplatinum(II) has been studied by competition experiments between the drug and several double-stranded polydeoxyribonucleotides. Binding to poly(dG).poly(dC) is larger than to poly (dG-dC).poly(dG-dC). There is no preferential binding in the competition between poly(dG-dC).poly(dG-dC), poly(dA-dC).poly(dG-dT) and poly(dA-dG).poly(dC-dT). In the competition between poly(dG-dC).poly (dG-dC) (B conformation) and poly(dG-br5dC).poly(dG-br5dC) (Z conformation), the drug binds equally well to both polynucleotides. In natural DNA, modification of guanine residues in (GC)n.(GC)n sequences by the drug has been revealed by the inhibition of cleavage of these sequences by the restriction enzyme BssHII. By means of antibodies to platinated poly(dG-dC), it is shown that some of the adducts formed in platinated poly(dG-dC) are also formed in platinated pBR322 DNA. The type of adducts recognized the antibodies is not known. Thin layer chromatography of the products after chemical and enzymatic hydrolysis of platinated poly(dG-dC) suggests that interstrand cross-links are formed. Finally, the conformations of poly(dG-dC) modified either by cis-diamminedichloroplatinum(II) or by trans-diamminedichloroplatinum (II) have been compared by circular dichroism. Both the cis-isomer and the trans-isomer stabilize the Z conformation when they bind to poly(dG-m5dC) in the Z conformation. When they bind to poly(dG-m5dC) in the B conformation, the conformations of poly(dG-m5dC) modified by the cis or the trans-isomer are different. Moreover, the cis-isomer facilitates the B form-Z form transition of the unplatinated regions while the trans-isomer makes it more difficult.     

Characterization of a multisubunit human protein which selectively binds single stranded d(GA)n and d(GT)n sequence repeats in DNA.

A protein which selectively binds d(GA)n and d(GT)n sequence repeats in single stranded DNA has been identified in human fibroblasts. This protein, designated PGB, has been purified at least 500-fold by ammonium sulfate precipitation followed by DEAE-Sepharose column chromatography and affinity chromatography in a column of d(GA)-Sepharose. Electrophoretic mobility shift assays revealed that the PGB protein bound most avidly d(GA)n and d(GT)n tracts of n > 5. It also bound other G-rich DNA sequence repeats, including dGn tracts, with lower affinities. It did not manifest significant binding affinities to single stranded M13 DNA, or to the homopolynucleotides poly dA, poly dC and poly dT, or to various DNA sequence repeats which do not contain G residues, such as d(A-C)n and d(TC)n. It did not bind double stranded d(T-C)n.d(GA)n tracts or other double stranded DNA sequences. In glycerol gradient centrifugation assays the d(GA)n- and the d(GT)n-binding activities cosedimented as a homogeneous protein species having an S20,w = 9.4 +/- 0.7 and an estimated native molecular weight of 190,000 +/- 7,000. UV crosslinking assays revealed that the protein contains 33.6 +/- 2.1 kd subunits which bind d(GA)n and d(GT)n sequences. However, SDS-polyacrylamide gel electrophoresis of the purified protein followed by silver staining indicated that it may also contain other subunits that do not contact the DNA. It is proposed that binding of the PGB protein to single stranded d(GA)n or d(GT)n tracts in double stranded topologically restricted DNA may stimulate strand separation and formation of triple helices or other unusual DNA structures.