QUAD, a protein from hepatocyte chromatin that binds selectively to guanine-rich quadruplex DNA.

The single-stranded oligomer Q, whose nucleotide sequence 5'-d(TACAGGGGAGCTGGGGTAGA)-3' corresponds to the IgG switch region, forms in concentrated solutions and in the presence of alkali metal cation parallel four-stranded complexes termed G4 DNA (Sen, D., and Gilbert, W. (1988) Nature 334, 364-366). We show that G4 DNA was also formed during storage of dried oligomer Q. This quadruplex complex migrated more slowly than mono-strand oligomer Q during nondenaturing gel electrophoresis, the rate of its formation depended on the mass of stored oligomer Q, and N7 positions of guanine residues were involved in its stabilization. Here we report the purification of a protein designated QUAD that binds specifically to the G4 form of oligomer Q, from non-histone protein extracts of rabbit hepatocytes. QUAD was 80-90% purified by sequential steps of column chromatography on Sepharose 6B, DEAE-cellulose, phosphocellulose, and phenyl-Sepharose. Purified QUAD migrated on SDS-polyacrylamide gel electrophoresis as a 58 +/- 2.6-kDa polypeptide and had a native molecular mass of 57 +/- 2.5 kDa as determined by Sepharose 6B gel filtration. The dissociation constant of G4 DNA binding to QUAD was in the range of 2.5 to 7.0 x 10(-9) M/liter. Excess unlabeled monostranded oligomer Q did not compete with 5'-32P-labeled G4 DNA on its binding to QUAD. Further, that QUAD recognized the G4 DNA structure rather than a DNA sequence was also demonstrated by the inefficient competition on the binding of 5'-[32P]G4 DNA to QUAD by excess unlabeled single- or double-stranded DNA molecules that contained guanine clusters of different length or various other nucleotide sequences.


QUAD, a Protein from Hepatocyte Chromatin That Binds Selectively
to Guanine-rich Quadruplex DNA* (Received for publication, July 28, 1992) Pnina Weisman-Shomer and Michael Fry From the Unit of Biochemistry, The Bruce Rappaport Faculty of Medicine, Technion-Israel Institute of Technology, Haifa 31096, Israel " -The single-stranded oligomer Q , whose nucleotide sequence 5'-d(TACAGGGGAGCTGGGGTAGA)-3' corresponds to the IgG switch region, forms in concentrated solutions and in the presence of alkali metal cation parallel four-stranded complexes termed G4 DNA (Sen, D., and Gilbert, W. (1988) Nature 334,[364][365][366]. We show that G4 DNA was also formed during storage of dried oligomer Q. This quadruplex complex migrated more slowly than mono-strand oligomer Q during nondenaturing gel electrophoresis, the rate of its formation depended on the mass of stored oligomer Q , and positions of guanine residues were involved in its stabilization. Here we report the purification of a protein designated QUAD that binds specifically to the G4 form of oligomer Q , from nonhistone protein extracts of rabbit hepatocytes. QUAD was 80-90% purified by sequential steps of column chromatography on Sepharose 6B, DEAE-cellulose, phosphocellulose, and phenyl-Sepharose. Purified QUAD migrated on SDS-polyacrylamide gel electrophoresis as a 58 2 2.6-kDa polypeptide and had a native molecular mass of 57 f 2.5 kDa as determined by Sepharose 6B gel filtration. The dissociation constant of G4 DNA binding to QUAD was in the range of 2.5 to 7.0 X lo-' M/liter. Excess unlabeled monostranded oligomer Q did not compete with 5'-32P-labeled G4 DNA on its binding to QUAD. Further, that QUAD recognized the G4 DNA structure rather than a DNA sequence was also demonstrated by the inefficient competition on the binding of 5'-[32P]G4 DNA to QUAD by excess unlabeled single-or double-stranded DNA molecules that contained guanine clusters of different length or various other nucleotide sequences.
A growing body of experimental evidence indicates that unusual DNA structures such as cruciform DNA, alternating B-Z regions, bent DNA, and triplex and quadruplex DNA may be involved in specific cellular processes (Yagil, 1991;Palecek, 1991). Clusters of guanine residues appear in many chromosomal locations such as telomers (Zakian, 1989;Blackburn, 1990Blackburn, , 1991, in gene promoters (Evans et al., 1984;Kilpatrick et al., 1986;Clark et al., 1990), and in the immunoglobulin switch region (Shimizu and Honjo, 1984). Single-*This work was supported by grants (to M. F.) from the U S -Israel Binational Science Fund, the Israel Cancer Association through the late Judith Klienberger Fund, Munich, the Fund for Basic Research administered by the Israel Academy of Sciences and Humanities, and the E. and J. Bishop Research Fund. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. stranded DNA fragments from such regions are capable of aggregating into four stranded structures that are stabilized by Hoogsteen-bonded guanine quartets (Henderson et al., 1987;Sen and Gilbert, 1988, 1990Williamson et al., 1989;Sundquist and Klug, 1989;Panyutin et ai., 1989Panyutin et ai., , 1990Hardin et al., 1991;Lu et al., 1992). Oligomeric fragments of telomers can form in the presence of low concentrations of KC1 or NaCl fold-back structures which migrate in a nondenaturing gel more rapidly than the mono-stranded input oligomer (Henderson et al., 1987;Williamson et al., 1989). Two such hairpin fold-back telomeric fragments of Oxytricha and Tetrahymena can combine to generate guanine quartetstabilized dimeric quadruplex structure (Williamson et al., 1989;Sundquist and Klug, 1989;Hardin et al., 1991). Further, stretches of 27-37 guanine residues in denatured plasmid DNA are able to fold into antiparallel tetrahelical structures (Panyutin et al., 1989(Panyutin et al., , 1990. Lastly, single-stranded oligomers that contain short runs of guanine form in highly concentrated solutions, and in the presence of sodium ions, a fourstranded structure, termed G4 DNA, in which the strands run in a parallel orientation (Sen and Gilbert, 1988, 1990Lu et al., 1992).
Formation of four-stranded DNA under physiological conditions invites suggestions as to the potential in uiuo role of this unusual structure. Generation of parallel, Hoogsteenbonded quadruplex DNA during the interaction of four double helices has been proposed to tie together the four chromatids at meiosis (Sen and Gilbert, 1988). If involved in the formation of tight synapses between homologous chromosomes at their telomeric ends and elsewhere in the chromosome (Sen and Gilbert, 1988;Lu et al., 1992), such tetrahelical structure at guanine-rich regions in DNA may be instrumental in promoting recombination. In view of the relative fragility of Hoogsteen-bonded quadruplex DNA, it is conceivable that specific DNA binding proteins may promote its formation or stabilization. In this paper, we describe the purification of a protein, designated QUAD, from chromatin of hepatocytes that specifically binds a quadruplex form of oligomer Q, whose nucleotide sequence
Isolation of G4 DNA-Quadruplex G4 DNA was formed by interstrand association of oligomer Q. Firstly, aliquots (1.0 pgllane) of HPLC-purified oligomer Q were further purified by electrophoresis at room temperature under -30 V/cm for 1.5 h through a denaturing 8% polyacrylamide (acryl:Bis 19:l)-8.0 M urea gel in TBE buffer (0.45 M Tris borate buffer, pH 8.3, 1 mM EDTA). The 20-mer DNA band was visualized by autoradiography and excised from the gel, and 75-90% of the DNA was extracted from the minced gel slices during an overnight shaking at 4 "C in 100 pl of T E buffer (20 mM Tris-HCI buffer, pH 8.0, 1.0 mM EDTA). Salts were removed by centrifugation through a Sephadex G-50 mini-column (Sambrook et al., 19891, and the oligomer was desiccated a t room temperature by Speed-Vac (Savant Instruments) centrifugation. Purified oligomer (2.0-5.0 pg, spectrophotometrically determined) was resuspended in 10 p1 of polynucleotide kinase reaction mixture and was 5'-end-labeled with "P (Sambrook et al., 1989). The reaction was terminated by the addition of EDTA to a final concentration of 20 mM, and the volume was brought to 100 p1 with water. The mixture was passed through a Sephadex G-50 mini-column to remove salts and unincorporated [-y-"'PIATP, and the specific activity of the DNA was determined. The purified oligomer was denatured a t 100 "C for 5 min, dried, and stored at -20 "C for 24-72 h under desiccation. In the course of its storage, 15-70% of oligomer Q formed tetrahelical G4 DNA. To separate quadruplex and mono-strand oligomer Q, 0.6 pg of DNA/lane was electrophoresed a t room temperature under 10 V/cm for 2.0-2.5-h through a nondenaturing 12% polyacrylamide gel (acryl:Bis 30:1.2) in 0.5 X TBE. Mono-strand oligomer Q and tetrahelical G4 DNA, which had a lower electrophoretic mobility, were visualized by autoradiography. The slowly migrating G4 DNA band was excised from the gel, and DNA was extracted into TE buffer, passed through a Sephadex G-50 mini-column, and dried in batches of 0.1-0.2 pg of DNA and stored until used. Immediately prior to use, the G4 DNA was resuspended in TE buffer to a final concentration of 2.0-5.0 ng/ pl and run through a nondenaturing gel to resolve mono-strand oligomer Q from G4 DNA and in order to monitor the formation of a protein-G4 DNA complex (see mobility shift electrophoresis below). T h e gel was dried on DE81 filter paper, and amounts of tetrahelical and mono-strand DNA were determined by counting Cerenkov radioactivity in their respective cut bands. Typically, 30 to 70% of the purified G4 form dissociated into mono-strand oligomer Q in the course of storage.
A slowly migrating G4 form of the oligomer (dG),, was isolated by a similar procedure except that tetrahelical DNA was formed by incubating at 4 "C for 72 h, 9.2 pg of (dG)3s per p1 of 0.1 M Tris-HCI buffer, pH 8.0, 0.5 M NaCl, 1 mM EDTA.
Mobility Shift Electrophoresis-Complex formation between G4 DNA and QUAD was monitored by the electrophoretic retardation of G4 DNA upon its association with the protein. Protein fractions in a final volume of 5-15 p1 of DE buffer (25 mM Tris-HC1 buffer, pH 7.5, 0.5 mM DTT, 1 mM EDTA, 20% glycerol) were incubated a t 4 "C for 20 min with 0.5-3.0 ng of 5'-"P-labeled G4 DNA. The protein-DNA binding mixtures were electrophoresed a t room temperature under 10 V/cm for 1.5 h through a 6% polyacrylamide gel (acrykBis 301.2) in 0.5 X TBE. Complex formation was quantified by measuring Cerenkov radioactivity in cut bands of the gel that was dried on a DE81 filter paper. Amounts of free and protein-bound DNA were calculated from the known specific activity of the DNA. One unit of G4 binding activity was defined as the activity that binds 0.1 ng of G4 DNA under the described standard conditions. Preparation of Nuclear Extract-Salt extracts of non-histone nuclear proteins were prepared from isolated nuclei of hepatocytes of female New Zealand White rabbits as described by Sharf et al. (1988). Preparation of protein extracts and all the subsequent chromato-graphic purification steps were conducted at 4 "C.
Sepharose 6B Gel Filtration-Aliquots of 6.0 ml of non-histone nuclear protein extracts in DE buffer were filtered through a Sepharose GB-CL column (2 X 100 cm, 314-ml packed column volume). Fractions of 6.0 ml were collected, and aliquots were assayed by mobility shift electrophoresis for the presence of G4 DNA binding QUAD protein.
DEAE-cellulose Chromatography-Tetrahelical DNA binding activity that was pooled from 4-6 independent Sepharose 6B column chromatographies was loaded onto a DE buffer-equilibrated DE52 column at a ratio of 4 mg of protein per ml of packed resin. The loaded column was washed with a single column volume of DE buffer, and proteins were eluted by a 10-column volume gradient of 0 to 500 mM KC1 in DE buffer. Forty fractions were collected, 0.2 mg/ml ST1 protein stabilizer was added, and each fraction was dialyzed against DE buffer. Electrophoretic mobility shift analysis detected QUAD protein in fractions that were eluted by 145-245 mM KCI. Pooled fractions of the binding protein were dialyzed overnight against P buffer (50 mM KPi buffer, pH 8.5, 0.5 mM DTT, 1 mM EDTA, 20% glycerol).
Phosphocellulose Chromatography-DEAE-cellulose-resolved QUAD protein was loaded onto a P buffer-equilibrated P-11 column a t a ratio of 2 mg of protein/ml packed resin. The loaded column was washed with a single column volume of P buffer, and proteins were eluted by a 10-column volume gradient of P buffer that contained 50 to 400 mM KPi. Forty fractions were collected, 0.2 mg/ml ST1 was added, and, after dialysis of each fraction against DE buffer, the binding of G4 DNA was assayed. QUAD protein was eluted by 140-215 mM KP, and the pooled active fractions were dialyzed overnight against PS buffer (25 mM Tris-HCI buffer, pH 7.5, 0.5 mM DTT, 1 mM EDTA) that contained 4.0 M NaC1.
Phenyl-Sepharose Chromatography-Phosphocellulose-resolved QUAD protein was loaded onto a phenyl-Sepharose column equilibrated in P S buffer that contained 4.0 M NaCI, at a ratio of 1.0 mg of protein/ml of packed column volume. The column was washed with 2 column volumes of the equilibration buffer, and proteins were eluted by a 10-column volume gradient of 4.0 to 0.0 M NaCl in PS buffer followed by a final 2-column volume wash with PS buffer. Forty fractions were collected, 0.2 mg/ml ST1 and 0.05% Nonidet P-40 were added, and each fraction was dialyzed against DE buffer. G4 DNA binding activity of QUAD protein was detected in fractions that were eluted in the final PS buffer wash. Purified QUAD protein remained active for at least 2 months when stored a t -70 "C in DE buffer that contained 0.2 mg/ml STI, 0.05% Nonidet P-40.
Determination of the Amount of Protein-The Bio-Rad protein assay kit was used to determine the amount of protein.

RESULTS
Generation and Properties of Quadruplex G4 DNA-Oligomer Q, whose nucleotide sequence 5'-d(TACAGGGGAGC-TGGGGTAGA)-3' corresponds to the IgG switch region (Sen and Gilbert, 1988), has been shown to form by an interstrand association of its two (dG), clusters, parallel four-stranded complexes designated G4 DNA Gilbert, 1988,1990). In agreement with results reported by these authors, we also found that, when stored in solution, oligomer Q formed in a concentration-and NaCl (0.4-1.0 M)-dependent fashion a heat-labile G4 DNA species that migrated in a nondenaturing gel more slowly than mono-strand oligomer Q (Fig. 1, lane  B ) . We have observed, however, that when HPLC-and denaturing gel-purified oligomer Q was stored dried at -20 "C, a form was generated whose diminished electrophoretic mo- bility was indistinguishable from that of G4 DNA that was formed in solution (Fig. 1, lune A ) . Generally, when batches of 2.5-5.0 pg of oligomer Q were stored desiccated a t -20 "C for 24 to 48 h, 15-6096 of the DNA appeared in a major electrophoretically retarded form. In addition, a minor fraction that migrated more slowly than the major G4 DNA band was discerned in some preparations (Fig. 1, lune A, and Fig.  2, lune A ) . Both the slowly migrating major and minor forms that were generated during storage of dried oligomer Q were entirely converted into mono-strand DNA upon heating a t 100 "C for 5 min (Fig. 2, lune B ) as did the slowly migrating forms that were generated in solution (Fig. 1, lune C). The amount of G4 DNA that was generated under desiccation reached a maximum in 24-48 h, and formation of the G4 species was independent of the presence of salt (data not shown).
The rate of G4 DNA formation under desiccation depended on the mass of stored oligomer Q. Variable amounts (0.025 to 50 pg) of denatured and desiccated oligomer Q were stored a t -20 "C for 24 h, and the dried DNA was suspended in TE buffer and was electrophoresed through a nondenaturing gel. As seen in Fig. 3A, the relative proportion of G4 complex that was formed depended on the mass of stored oligomer, giving a slope of 1.5 for a log-log plot. This dependence is close but not equal to the quadratic dependence on DNA concentration of the formation of tetrahelical DNA in solution (Sen and Gilbert, 1990). This discrepancy probably reflects a different kinetics of G4 DNA formation in solution or under desiccation. To assess the concentration dependence of its stability, were dried and stored a t -20 "C for 24 h. The DNA was resuspended in T E buffer to a concentration of 5.0 pg of DNA/ml, and its G4 form was resolved electrophoretically from mono-stranded oligomer Q as described in the legend to Fig. 1. Radioactivity was determined in cut hands of mono-strand and G4 DNA as described under "Experimental Procedures." Results were plotted as the log of the amount of G4 formed as a function of the log of the total amount of stored desiccated DNA. R, stability of G4 DNA in solution. A desiccated mixture of mono-strand and the G4 form of oligomer Q was suspended in T E buffer and diluted to DNA concentrations ranging between 1.4 and 0.14 pg/ml. Mono-stranded oligomer Q and its G4 form were resolved electrophoretically, and their amounts were determined as described in A ahove. A value of 100% G4 DNA represents 52% of the total mono-strand and G4 oligomer Q mixture in the initial 1.4 pg/ml solution.
G4 DNA that was formed under desiccation was suspended in T E buffer to concentrations ranging between 1.4 and 0.14 pg of DNA/ml and electrophoresed through a nondenaturing gel. As shown in Fig. 3B, the slowly migrating G4 DNA remained stable in solutions of 1.4 to 0.35 pg of DNA/ml, but, upon dilution to 0.14 pg/ml, about 50% of the G4 DNA was dissociated to mono-strand oligomer Q. The lower mobility of G4 DNA relative to oligomer Q, its heat denaturation, and lack of rapid self-renaturation as well as the dependence of G4 formation and stability on DNA concentration suggest that it is formed by interaction between multiple DNA strands (Sen and Gilbert, 1988, 1990.
T o demonstrate the involvement of positions of guanine residues in the formation of G4 DNA, methylation-protection was compared for mono-strand oligomer Q and its G4 form.
Aliquots, 70 ng each, of gel-purified desiccated G4 form of oligomer Q were suspended in 20 p1 of T E buffer and either denatured a t 100 "C for 5 min to generate mono-strand oligomer or left untreated as a 70:30 mixture of G4 and monostranded oligomer Q, respectively. In a procedure derived from Williamson et u1. (1989), the two types of DNA were incubated at room temperature for 10 min with 0.1% dimethyl sulfate and than heated at 90 "C for 15 min with 2 M pyrrolidine. Following desiccation and two washes with water, the intact and hydrolyzed DNA samples were electrophoresed through a 12% polyacrylamide denaturing gel. Measurement of Cerenkov radioactivity in bands of undegraded and hydrolyzed DNA indicated that without dimethyl sulfate treatment, 15% of mono-strand oligomer Q and G4-enriched DNA were broken down under the described conditions. However, following methylation, 16.7% of denatured oligomer Q were hydrolyzed, whereas only 1.5% of the G4-enriched DNA were broken down. The apparent inaccessibility of the W position in guanines of G4 DNA indicated their involvement in Hoogsteen bonding (Hoogsteen, 1959) and affirmed the tetrahelical nature of G4 DNA Gilbert, 1988,1990;Williamson et al., 1989;Sundquist and Klug, 1989).
Purification of the G4 DNA Binding Protein QUAD-Electrophoretic mobility shift analysis revealed the presence of a G4 DNA binding activity in crude extracts of non-histone proteins from rabbit hepatocytes. This activity was purified by sequential steps of column chromatography on Sepharose 6B, DEAE-cellulose, phosphocellulose, and phenyl-Sepharose. As shown in Fig. 4 A , proteins resolved in Sepharose 6B fractions 42 to 55 associated with both mono-stranded oligomer Q and its G4 form. Upon subsequent purification of these fractions by DE52 chromatography, proteins were eluted a t 145 to 245 mM KC1 (fractions 12-20, Fig. 4B) that also associated with both mono-strand and G4 DNA. Chromatography of these fractions on P-11 removed most of the monostrand binding activity and yielded a protein that we termed QUAD and which was eluted from the cation exchanger a t 140 to 215 mM KPi. As seen in Fig. 4C, QUAD associated with both the major and minor G4 DNA forms but bound only negligibly mono-strand oligomer Q. Final purification of P-11-resolved QUAD protein was performed by hydrophobic phenyl-Sepharose column chromatography. QUAD activity was eluted in the final wash with buffer devoid of salt (Fig.  5A). Whereas SDS-PAGE analysis showed that the QUADcontaining fractions of Sepharose 6B, DE52, and P-11 contained a progressively decreasing number of proteins (Fig.  5B), a single major polypeptide band of 58.7 kDa was discerned in fractions of phenyl-Sepharose-resolved QUAD (Fig.  5C). The apparent identity between the elution profiles of the 58.7-kDa protein (Fig. 5C) and the G4 DNA binding activity (Fig. 5A) strongly suggested that the 58.7-kDa polypeptide was QUAD. Multiple electrophoresis analyses indicated that the 58.7-kDa band consisted of at least 80-90% of the total protein in phenyl-Sepharose-resolved fractions of QUAD. This highly purified fraction of QUAD was used throughout this work.
Chemical-Physical Properties of QUAD-Some properties of QUAD protein are summarized in Table I. As demonstrated by its heat lability and inactivation by SDS or trypsin digestion, QUAD is a protein. By contrast, QUAD activity was not diminished by nucleolytic digestion with micrococcal nuclease and thus it did not contain an essential nucleic acid component. In fact, nuclease digestion somewhat increased G4 DNA binding by QUAD (Table I), possibly by removing some competing DNA contamination. Binding of G4 DNA by QUAD was decreased only slightly in the presence of 4-10 mM MalNEt (Table I), and thus reduced sulfhydryl groups are probably not involved directly in G4 DNA binding.
The highly purified QUAD protein migrated on SDS-PAGE as a 58.7 f 2.6-kDa polypeptide (average of determinations in six independent QUAD preparations). Sepharose 6B gel filtration of QUAD yielded a native molecular mass of 57.0 f 2.45 kDa (average of determinations in four independent QUAD preparations). QUAD is, therefore, most probably a monomeric protein.
Specificity of G4 DNA Binding by QUAD-To examine the extent of DNA sequence and structure specificity of the association of QUAD with G4 DNA, molar excesses of a variety of oligonucleotides and polynucleotide were used to compete with 5'-32P-labeled G4 DNA on its binding to QUAD. Results of this experiment are summarized in Table 11. It is  DNA sequence and structure specificity of QUAD binding Phenyl-Sepharose-purified QUAD protein (11.5 units) was incubated a t 4 "C for 20 min in a standard DNA binding reaction mixture that contained 3.0 ng of the 5'-"'P-labeled G4 form of oligomer Q and a 20-fold molar excess of each of the listed unlabeled competitor DNA species. Protein-bound G4 DNA was quantified by mobility shift FIG. 5. Phenyl-Sepharose column purification of QUAD protein. A phosphocellulose-purified fraction of QUAD was chromatographed through a phenyl-Sepharose column as described under "Experimental Procedures." G4 DNA binding activity was eluted from the column in the PS buffer final wash (fractions 4-6). Shown are fractions eluted only by 0.5 to 0.0 M NaCI. A, mobility shift electrophoresis of fractions eluted from the phenyl-Sepharose column. m, mono-strand oligomer G4-Q; G4, quadruplex form of oligomer Q; complex, protein-bound G4 DNA. B, silver-stained proteins in SDS-PAGE-resolved QUAD fractionated by Sepharose 6B, DEAEcellulose, and phosphocellulose chromatographies. C, silver-stained proteins in SDS-PAGE-resolved QUAD fractionated by the phenyl-Sepharose column chromatography shown in A. Arrowheads indicate positions of molecular mass marker proteins. STZ, stabilizing ST1 protein.

TABLE I
Chemical-physical properties of QUAD protein Binding of G4 DNA by 11.5 units of phenyl-Sepharose-purified QUAD was carried out without or with the indicated treatments. QUAD44 DNA complex was resolved by mobility shift electrophoresis, and its amount was quantified as detailed under "Experimental Procedures."

Treatment
% initial activity None 100.0 100 "C, 30 min 2.5 Trypsin digestion" 4.3 0.2% SDS 4.7 Micrococcal nuclease* 145.0 4 mM MalNEt' 86.5 10 mM MalNEt' 87.5 QUAD protein was incubated a t 37 "C for 60 min with 375 pg/ml trypsin, and the proteolytic digestion was terminated by the addition of ST1 to a final concentration of 2.0 mg/ml. Shown is an average result of three independent determinations. * QUAD protein was incubated a t 37 "C for 30 min with 50 pg/ml micrococcal nuclease in the presence of 1.0 mM CaCI2, and digestion was terminated by the addition of EDTA and pTp to final concentrations of 3.3 mM and 0.25 mM, respectively. Shown is an average result of two experiments.
QUAD protein was incubated a t 4 "C for 15 min with MalNEt, and the reaction was terminated by the addition of 20 mM DTT. Average result of two experiments. apparent that only the unlabeled G4 form of oligomer Q competed efficiently with homologous labeled G4 DNA on its binding to QUAD protein. By contrast, neither excess denatured mono-stranded oligomer Q nor oligomer Tet which contained four (dG), clusters chased the G4 form from its complex with QUAD. Further, neither a G4 form of (dG)35 nor denatured mono-stranded (dG)35 competed efficiently with tetrahelical oligomer Q on binding to QUAD (Table 11).
I t is also apparent that most single-and double-stranded DNA species that were tested failed to chase G4 DNA from 85 Unlabeled G4 forms of oligomer Q and of (dG)3s were isolated as described under "Experimental Procedures" except that bands of unlabeled G4 DNA were identified for excision from nondenaturing gels by labeled G4 DNA which was run in parallel lanes. Amounts of DNA that were extracted from the gel slices were determined spectrophotometrically.
its complex with QUAD. An interesting exception is (dT)20 which, at a 20-fold molar excess, partially competed with G4 DNA (Table 11). Hence, QUAD displayed its highest affinity to tetrahelical oligomer Q, it did not bind under the tested conditions guanine run-containing mono-stranded oligomer Q, Tet, and (dG)35 or quadruplex (dG)35. Moreover, that QUAD recognized a specific DNA structure rather than nucleotide sequence was demonstrated by the inefficient competition of G4 DNA binding by denatured or double-stranded DNA that contained guanine clusters or various other base sequences.
In light of the apparent high degree of the specificity of the association of QUAD protein with G4 DNA, we measured the binding affinity of highly purified QUAD to G4 DNA. A constant amount of phenyl-Sepharose-purified QUAD protein was incubated a t 4 "C for 20 min with increasing amounts of 5'-32P-labeled G4 DNA, and DNA-protein complexes were resolved by mobility shift electrophoresis. Radioactivity in free and protein-bound G4 DNA was measured in their respective cut bands to determine their amounts (see "Experimental Procedures"). A typical Scatchard plot of the results of such a measurement is presented in Fig. 6. The range of values of dissociation constants that were derived in different experiments from the negative reciprocal of the slope was between 2.5 X lo-' to 7.0 X lo-' M/liter (four determinations).

DISCUSSION
Single-stranded DNA that contains guanine-rich motifs has been shown to self-associate in the presence of monovalent salt in a concentration-dependent fashion, to form a fourstranded G4 DNA structure in which the strands assume parallel orientation (Sen and Gilbert, 1988, 1990Lu et al., 1992). In agreement with these reports, we found that when stored in solution and in the presence of 0.4-1.0 M NaCl, oligomer Q aggregated in a concentration-dependent manner to generate a form with a lower electrophoretic mobility in a nondenaturing gel than that of the mono-stranded oligomer (Fig. 1). In this work, we showed evidence that a similar slowly migrating and heat-labile G4-like form of oligomer Q Phenyl-Sepharose-purified QUAD protein (11.5 units) was incubated at 4 "C for 20 min with increasing amounts of a 3070 mixture of 5'-"2P-labeled G4:mono-strand oligomer Q. QUAD-G4 DNA complex was resolved by mobility shift electrophoresis in a 6% nondenaturing polyacrylamide gel as described under "Experimental Procedures." The 2-fold excess of labeled mono-strand oligomer Q over its G4 form did not interfere significantly with the binding of G4 DNA to QUAD (see Table 11). A, mobility shift electrophoresis pattern of QUAD with increasing amounts of probe. m, mono-strand oligomer Q; G4, tetrahelical form of oligomer Q; complex, proteinbound G4 DNA. R, Scatchard plot of quantified results presented in A . The gel was dried, bands of mono-strand oligomer Q, its G4 form, and QUAD-bound G4 DNA were cut, their radioactivity was counted, and amounts of DNA in each band were derived from the known specific activity of the DNA.
was generated when it was stored under desiccation (Figs. 1 and 2). The rate of aggregation of the DNA depended on the amount of stored dried oligomer Q (Fig. 3A ), and it dissociated in a dilute solution (Fig. 3B). Methylation-protection analysis indicated that IV' positions of the guanine residues in the G4 form of oligomer Q were involved in its formation (see "Results''). Based on its low electrophoretic mobility, heat lability, concentration-dependent formation and stability, and the involvement of guanine hf' positions in its stabilization, we assumed that the G4 form of desiccated oligomer Q represented a multistranded Hoogsteen-bonded aggregate of the single-stranded DNA. Although a formal proof for the detailed structure of this G4 form was not provided, its properties conform with those of a parallel tetrahelix Gilbert, 1988, 1990).
The principal finding reported in this paper is that a protein termed QUAD, which we purified to near homogeneity from a non-histone protein extract of rabbit hepatocyte chromatin, associated specifically with the G4 form of oligomer Q. A 58.7 & 2.6-kDa polypeptide (Fig. 5C), QUAD behaved as a monomeric protein that was resistant to MalNEt (see "Results" and Table I). Binding competition analyses indicated that QUAD bound specifically to quadruplex oligomer Q. Whereas QUAD associated with G4 oligomer Q a t a high affinity (dissociation constant 2.5 X to 7.0 X lo-' M/liter, Fig. 5), i t failed to bind significantly to denatured, mono-stranded oligomer Q (Table 11). Furthermore, neither mono-strand (dG),, nor its quadruplex form competed efficiently with G4 oligomer Q on binding to QUAD protein (Table 11). Similarly, excess denatured oligomer Tet whose nucleotide sequence 5'-[d(T2G,J4]-3' corresponds to Tetrahymena telomers (Williamson et al., 1989) failed to chase the G4 form of oligomer Q from its complex with QUAD protein (Table 11). It appeared, therefore, that QUAD recognized a specific tetrahelical structure of oligomer Q but failed to associate with either guanine runs in mono-strand oligomer Q, (dG),, or Tct, or with the quadruplex structure assumed by (dG)Ss. Moreover, single-or double-stranded DNA molecules with other nucleotide sequences acted as inefficient competitors of the binding of G4 oligomer Q to QUAD protein (Table 11). Hence, rather than being a sequence-specific DNA binding protein, QUAD behaved as a DNA structure-specific binding protein.
By associating specifically with quadruplex G4 DNA, QUAD differs from previously described DNA binding proteins that bind guanine runs in DNA. Some such proteins are complexed with the guanine-rich telomeric DNA. A heterodimeric protein from Oxytricha binds specifically to the telomeric d(G4T4)2 single strand tail and to its adjacent duplex region (Gottschling and Zakian, 1986;Price and Cech, 1989;Raghuraman et al., 1989). This telomere binding protein has been shown, however, to be unable to bind a folded, presumably guanine quartet-stabilized form of the telomeric sequence (Raghuraman and Cech, 1990). Other guanine cluster binding proteins associate with single-or double-stranded DNA. In yeast, the abundant transcriptional regulator RAP1 binds to the duplex portion of telomeric DNA but not to its singlestranded protrusion (Conrad et al., 1990;Lustig et al., 1990). The intermediate filament subunit .vimentin binds at a low ionic strength single-stranded synthetic oligonucleotides that correspond to a single repeat unit of telomers from Oxytricha (T4G4), Saccharomyces (TGTGTGS), or Tetrahymena (T2G4) (Shoeman et al., 1988). BGP1, a zinc-dependent double strand DNA binding protein from erythroid cells associates specifically with a string of 7 or more guanine residues (Lewis et al., 1988;Clark et al., 1990). None of these proteins, however, was reported to bind tetrahelical guanine-rich DNA.
After this manuscript was submitted for publication, Walsh and Gualberto (1992) reported that recombinant MyoD, a transcription factor that initiates myogenesis, binds guanine tetrads formed within a single strand creatine kinase enhancer probe or within a telomeric DNA probe. In addition to the different physical properties of QUAD and MyoD and their dissimilar tissue localization, they display different DNA binding properties. Although MyoD binds guanine tetrads a t a high affinity, it also binds single-stranded and doublestranded probes (Walsh and Gualberto, 1992). By contrast, QUAD binds to the tetrahelical form of oligomer Q but does not bind significantly to various single-or double-stranded DNA probes that do or do not contain guanine clusters (Table  11). Yet, the fact that these two proteins associate at a high affinity with quadruplex DNA argues for a physiological occurrence and a biological role for this unusual DNA conformer.
It has been proposed that formation of tetrahelical synapses between chromatids aligns homologous chromosomes in preparation for eventual exchange of genetic material (Sen and Gilbert, 1988). It might be that whereas the initial alignment of chromatids is mediated by tetrad formation a t guanine clusters in DNA, proteins are required to stabilize the weakly bonded DNA synapses. It is tempting to speculate that QUAD protein, by virtue of its specific binding to tetrahelical DNA, stabilizes quadruplex DNA synapses to allow recombination to occur.