DNA “Melting” Proteins

IV. FLUORESCENCE MEASUREMENTS OF BINDING PARAMETERS FOR BACTERIOPHAGE T4 GENE 32-PROTEIN TO MONO-, OLIGO-, AND POLYNUCLEOTIDES*

(Received for publication, March 11, 1976)

RAYMOND C. KELLY,+ DAVID E. JENSEN,~ AND PETER H. VON HIPPEL

From the Institute of Molecular Biology and Department of Chemistry, University of Oregon, Eugene, Oregon 97403

The quenching of the intrinsic tryptophan fluorescence of T4-coded gene 32protein on binding to nucleotide ligands, which was described in the preceding paper, is here exploited to measure thermodynamic parameters of the single-stranded nucleic acid-gene 32-protein interaction. It is shown that binding of small ligands follows a single site binding isotherm, with association constants increasing from -20 Mm’ for phosphate, to -lo3 M for ribose or deoxyribose 5’-phosphate, to -10’ M- ’ for mononucleotides, and to -lo5 Mm’ for dinucleoside monophosphates (all in 0.1 M Na+). The measured binding constants appear to be about the same for homologous ribose- and deoxyribose-containing ligands and to be independent of oligonucleotide base sequence and composition. Furthermore, beyond the dinucleotide level and up to octanucleotides, the increase in binding constant with increasing chain length is only about that expected from the statistical factor resulting from the increased number of ways a longer oligonucleotide can form a protein complex. This suggests that the basic binding unit involved in gene 32-protein associations with single-stranded nucleic acids can be approximated by a dinucleoside monophosphate.

Oligonucleotides long enough to accomodate two or more protein monomers are characterized by much

larger association constants, indicating that binding is cooperative in protein concentration. A cooperativity parameter (w,) of -lo3 is estimated from these data, in good agreement with that deduced from the application of ligand-perturbed helix ..i coil transition calculations. Values of association constants (K,w,) of -lOa Mm’ (in 0.1 M Na+) and site size (n,) of -5 (il) nucleotide residues/protein

monomer are determined by the fluorescence titration technique for the cooperative binding of gene 32-protein to both poly(dA) and poly(rA); these values are also in agreement with those measured by Jensen et al. (Jensen, D. E., Kelly, R. C., and von Hippel, P. H. (19’76) J. Biol. &em. 251, 7215-7228). Possible in uiuo consequences and correlations of these findings with proposed roles for gene 32-protein in

replication and recombination are discussed.

In the preceding paper (l), we examined the intrinsic protein fluorescence of gene 32-protein in various conformational and interactional states and studied the quenching of the fmores-

cence brought about by solute perturbants and by the binding of various nucleotide ligands. In this paper, we extend these

*This research was supported in part by United States Public Health Service Research Grants GM-15792 and GM-15423, as well as by Predoctoral Traineeships (to R.C.K. and D.E.J.) from United States Public Health Service Training Grant GM-00444. The fluores- cence apparatus primarily used in these studies was obtained through National Science Foundation (Matching) Equipment Grant MPS 75.06606. This work was submitted (by R.C.K.) to the Graduate School of the University of Oregon in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemistry.

$ Present address, Institute of Pathology, Case Western Reserve University School of Medicine, Cleveland, Ohio 44106.

8 Present address, Department of Biochemistry and Biophysics, Oregon State University, Corvallis, Oregon 97331.

approaches to obtain thermodynamic parameters characteriz- ing the cooperative and non-cooperative binding of gene 32-protein to a variety of such ligands. In conjunction with the parameters of this system measured by examining the effect of gene 32-protein binding on nucleic acid melting curves and polynucleotide spectral properties (2), these results permit a quantitative definition of the molecular nature and specificity of some of the interactions presumably involved in the physio- logical functioning of this protein.

MATERIALS AND METHODS

Details of most of the materials and techniques used in this studv are summarized in Kelly and van Hippel il); minor procedural modifications are described under “Results” as they arise. We re- emphasize that, unless otherwise indicated, all experiments were conducted in “standard” buffer containing: 50 mM Na,HPO,, 1 rn~ Na,EDTA, and 1 mM P-mercaptoethanol, pH 7.7.

7240

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

Gene 32.Protein-Nucleic Acid Binding Parameters 7241

RESULTS

As described in the preceding paper (l), the binding of a variety of nucleotide ligands to gene 32.protein results in a partial quenching of the intrinsic protein fluorescence. This quenching is saturable with increasing ligand concentration, and at least for the shorter ligands, the degree of quenching shows a hyperbolic relationship to ligand concentration. Bind- ing isotherms, generated by titration of the protein fluores- cence, have been measured to obtain a variety of quantitative parameters characterizing the interactions.

Binding of Short Oligonucleotides

Single Site Binding Constants Are Calculated from Fluores- cence-Quenching Titration &rues-The peak fluorescence intensity of gene 32-protein is quenched 3 to 4% by the addition of saturating concentrations of d(ApA);’ see Fig. 2 and Table II in Ref. 1. A typical fluorescence titration curve for the binding of this compound is shown in Fig. 1 (this paper). The points of Fig. 1 represent the actual data, and the solid curve the best fit theoretical single site binding isotherm, plotted according to Equation 1:

LEE F0

-Au--.+ 1 + K[N]

(1)

where AF is the decrease of fluorescence intensity at the emission maximum in the presence of a concentration [N] of free nucleotide ligand, AF, is the decrease in fluorescence intensity at infinite ligand concentration, and K is the associa- tion constant of the ligand.protein complex. F’, the initial fluorescence intensity before the addition of ligand, is used as a normalization factor. Thus, in Fig. 1, the quantity actually plotted (lOOAF/F”) is the percentage of the original fluores- cence quenched.

The following procedure has been used to determine the necessary parameters. First, an estimate of AF, is made from the raw titration data. Using this value we establish AFIF” and calculate the free ligand concentration [N] at each point by subtracting the amount bound to the protein (assuming only one ligand bound per monomer at saturation) from the total ligand added. Equation 1 may be recast as:

r = 1 1 -+- (2) AF K[N]AF, AF_

and the parameters obtained as above used to construct a plot, of 1lAF against l/[N]. Such a double reciprocal plot of the data for Fig. 1 is presented in Fig. 2. The linearity of the plot confirms the one binding site assumption and the intercept with the l/AF axis provides a refined measure of AF, , now based on all the experimental points. The slope of plots of the type shown in Fig. 2 yields l/K, again based on all the experimental data. Due to the distribution of the data, the intercepts of these double reciprocal plots are often more accurately obtainable than the slopes. Thus, given a refined value of AF,, we generally have also determined K directly from the midpoint of the binding isotherms; i.e. K = l/[N], (see Fig. 1); where [NIV, = the free ligand concentration at half-saturation of the protein. Association constants measured in this way on the actual binding isotherms (based only on midpoint data) were generally in good agreement with those

‘The definitions for all compounds are given in Footnote 1 of Ref. 1.

b(ApAfl,: I06M

FIG. 1. Fluorescence titration curve for the binding of d(ApA) to gene 32.protein in standard buffer (50 mM Na,HPO,, 1 mM Na,EDTA, 1 mM P-mercaptoethanol, pH 7.7), 25.1 i 0.1”. Protein concentration = 3.6 x 10.'M.

o.2, 0 4 6 12 16 20

A x 104 [d(APA)l free

FIG. 2. Double reciprocal plot for the binding of d(ApA) to gene 32.protein in standard buffer (50 mM Na,HPO,, 1 mM Na,EDTA, 1 mM P-mercaptoethanol, pH 7.7), 25.1 + 0.1”. Protein concentration = 4.7 X lo-‘M.

determined directly from the double reciprocal plots. Associa- tion constants for deoxyadenosyl and deoxythymidyl oligomers ranging up to 8 nucleotides in length, as well as for P’-deox- yribose 5’.phosphate (dRib-P)* and ribose 5’-phosphate (Rib-P), are collected in Table I. We note that the binding constants for the shorter ligands show a substantial standard error, due to the very low affinity of the protein for Rib-P, dRib-P, dAMP, and dTMP.

Short Oligonucleotides Bind to Protein Monomers-When a protein monomer binds to a long single-stranded nucleic acid lattice, it “covers” (occludes) n bases, which then, by defini- tion, are not simultaneously accessible to another protein molecule. For long lattices n is thus termed the protein site size (e.g. see Ref. 3). I f binding is polar (i.e. the protein binds to the lattice in a preferred direction), as appears to be the case for gene 32-protein binding to oligonucleotide ligands (Ref. 1 and see below), oligonucleotides of length 1 less than or equal to n can bind only to a single protein monomer.

In order to permit unambiguous interpretations of short oligonucleotide ligand binding isotherms, it is necessary to demonstrate, under the conditions of the experiment, that the protein is present primarily in the monomer form. The results of Carroll et al. (Ref. 4, see Fig. lb), as well as preliminary suberimidate cross-linking experiments (l), suggest that at the protein and salt concentrations used in this study the

’ The abbreviations used are: dRib-P, 2’-deoxyribose 5’.phosphate; Rib-P, ribose 5’-phosphate.

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

7242 Gene 32-Protein-Nucleic Acid Binding Parameters

TABLE I

Binding constants for short nucleotide-containing and model ligands to gene 32.protein”

All binding constants measured in standard buffer, containing 50 mu Na*HPO,, 1 mu Na, EDTA, and 1 mu 8-mercaptoethanol, pH 7.7.

dAMP d(ApN d(pA), d(pA), TMP dUbT) d(pT), d(pW, d(ApT) d(TpA) dRib-P Rib-P d(pA), d(b), r&N

.I, L

1.3 i 0.3 x 10’ 1.5 F 0.2 x 105 4.3 i 0.6 x 10’ 4.6 i 0.3 x lo5 8.6 F 4.0 x 10’ 1.5 i 0.1 x 105 7.6 A 2.0 x lo5 3.3 * 0.9 x 105 2.0 i 0.2 x 105 2.1 f 0.1 x 105 2.6 i 0.7 x lo3 2.2 + 1.1 x 103 1.8 t 0.6 x lo5 7.2 i 5.0 x 10’ 3.9 i 1.7 x 105

*Bimolecular association constants in standard buffer at 25.1 * 0.1” obtained by quenching of protein fluorescence. Error limits cal- culated from standard deviations of intercepts and slopes of double reciprocal plots; see text.

protein should indeed be present primarily in the monomer form. We thus assume that the binding constants of Table I, for oligonucleotides 8 residues or less in length, represent the binding of single protein monomers to the oligonucleotide ligands.

Two Nucleotide Residues Are Bound to Gene 32-Protein Binding Site.-The number of bases “covered” in binding a protein monomer to a polynucleotide chain (n) is not necessar- ily equal to the number of nucleotide units (m) which actually participate in binding the ligand to the protein. The relation- ship between these parameters is illustrated in Fig. 3. In this section we attempt to determine m for gene 32-protein.3

The data of Table I show that the apparent affinity constant (K) for the binding of gene 32.protein to mononucleotides is -10’ Mm’; for dinucleoside mono- or diphosphates, K = lo5 Mm’; and for tetranucleotides K increased by an additional factor of 3 to 5. For oligonucleotides in which 1 > m, the addition of more nucleotides can increase the apparent binding constant due to entropic (statistical) factors. For example, if 1 = 3 and m = 2, there are two ways the protein can bind to the ligand (with constant polarity), and thus K, = 2 K, (where K, is the apparent association constant for a protein to a ligand length 1, while K, is the intrinsic binding constant to the minimal (length m) ligand). In general, then, for ligands where L > m:

K, = (1 m + l)K,, (3)

‘The parameter m should be differentiated from m’, which is the number of electrostatic (ion pair; e.g. DNA phosphate with lysine or arginine side chain) interactions in which the protein ligand and the DNA lattice are actually engaged (see Ref. 5). Record et al. (5) have used the dependence of the binding constant on Na’ concentration measured by Jensen et al. (2) to estimate m’ for the native DNA-gene 32.protein interaction at one to two charge-charge interactions, de- pending upon whether or not the free ligand binds negative counter- ions. This estimate is in reasonable accord with the value of m found here for the single-stranded DNA-gene 32.protein interaction, suggest- ing again that the native and denatured DNA interactions are topologically rather comparable (see Ref. 2).

- tm+

-f.

FIG. 3. Schematic representation of binding site sizes and ligand lengths. I represents the length of the oligonucleotide lattice (here 10 bases), n represents the lattice length covered by the protein ligand (here 5 bases), and m represents the length of the base sequence actually bound to the protein ligand (here 2 bases). (Note that all the bases of the sequence m need not be bound to the ligand (i.e. in general m is the length of the nucleotide sequence whose outer members are bound to the protein), but for gene 32.protein (see text) all the bases of the m sequence do appear to be involved in binding interactions with the protein.)

Thus, Table I shows that the increase in the apparent association constant in going from mono- to dinucleotides is considerably larger than the statistical factor of 2 expected when 1 is taken as 2 and m as 1. However, the increase in K in

going from di- to tetranucleotides is approximately that expected from statistical considerations (m = 2; I = 4). These results tentatively suggest that the “active site” of gene 32-protein involves actual binding interactions with the func- tional groups of approximately 2 adjacent nucleotide units.’

Further insight is gained by considering the binding con- stants measured for dRib-P and Rib-P. As Table I shows, the affinity of these molecules for gene 32.protein is only 3- to 5-fold smaller than that of the corresponding mononucleotides, indicating that the primary interaction may be with a central sugar-5’-phosphate moiety. The dinucleoside monophosphate binding data (Table I) suggest that this interaction is strength- ened by the placement of a base on either side of the central sugar-phosphate moiety.’ Additional “flanking” phosphates change the apparent binding constant relatively little and, as pointed out above, increasing the ligand chain length to the tetranucleotide level increases the protein affinity further only to the extent expected from the statistical factor (when m is assumed to be 2). It is noted, however, that on this basis octanucleotides should bind about 3-fold tighter than tetranu- cleotides; instead, they show an apparent affinity which is not very different from that of the tetranucleotide ligands. It is possible that for these longer ligands other factors (e.g. decreased conformational entropy of the entire oligonucleotide chain in the presence of the protein) may offset the expected statistical factor in determining the apparent binding con- stant.

Binding Is Not Base-specific-Gene 32.protein might be

I This conclusion is qualitatively supported by the apparent shape of some of the mononucleotide binding isotherms. While the error in these measurements is large (because of the low binding constants involved), binding isotherms obtained, in particular for dAMP, indicate some deviation from single site binding, suggesting perhaps some positive cooperativity (presumably involving two sites) for the binding of these mononucleotides. No measurable deviations are observed for ligands of 1 2 2 nucleotide residues.

‘The binding constant for d(ApA) to the protein in Tris buffer is approximately 2-fold increased over that in 0.05 M phosphate buffer (at the same total ionic strength and pH). If we assume on this basis that in 0.05 M PO,*- the protein is half-saturated with phosphate anion, we can estimate the phosphate binding constant of gene 32-protein as -20 Me’ under these conditions at pH 7.7. This is very close to the binding constant estimated by others for the active site interaction of ribonu- clease with phosphate at pH 7.7 (-25 Me’; see Ref. 9).

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

Gene 32-Protein-Nucleic Acid Binding Parameters 7243

expected not to show a large degree of base specificity in its binding behavior on teleological grounds; i.e. we presume it must bind to all sequences in order to function effectively in DNA replication and recombination. A comparison of the binding affinity of oligonucleotides of varying sequence is a sensitive way to examine quantitative aspects of the base sequence specificity of the binding process. The observation that the sugar-phosphate group binds nearly as tightly as the mononucleotides to gene 32-protein was interpreted to indicate that a major fraction of the interaction between the protein and nucleotides involves the sugar-phosphate moiety. Here, the role of the base in the binding interaction is further defined by comparing the affinity of the protein for the four dinucleoside monophosphates: d(ApA), d(ApT), d(TpA), and d(TpT).

The results of these experiments are summarized in Table I and show, within the limits of error of the measurements, that all these dinucleoside monophosphates bind with the same affinity. The data of Table I also indicate that while there are some (small) apparent differences between the binding of the various deoxyadenylic and deoxythymidylic oligomers, these differences fit no pattern indicative of base specificity. Thus, we conclude that gene 32.protein binding to oligonucleotide ligands is not markedly base-specific, at least with respect to discriminating between dA and dT bases located in the “interaction positions” of the binding site. (Oligonucleotides containing dG and dC were not tested.)

Since the natural “substrate” for gene 32-protein is T4 DNA, which contains glucosyl (and hydroxymethyl) groups at posi- tion-5 of cytosine, we assume these groups do not inhibit gene 32-protein binding to single strand sequences. This is crudely confirmed by the fact that the isolation method of Alberts and co-workers (6), employing denatured DNA cellulose chroma- tography, is unperturbed by the replacement of calf thymus DNA by T4 DNA on the column (6). To further probe the sensitivity of binding to substitutions on the nucleotide bases, we have examined the binding of the N’-methyl derivative of dAMP to the gene 32-protein site. The measured binding constant of N’-methyl-dAMP was 1.1 x lo4 Mu.‘, which is not significantly different from the dAMP value of 1.3 x lo4 Mm’

(Table I). Since, in particular, the N’-methyl substitution both blocks the Watson-Crick base-pairing position of the dA residue, and places a positive charge on the ring, this result reinforces the suggestion that the chemical details of the base portions of the nucleotides are not discriminated in the interaction of gene 32.protein with nucleotide ligands.

Binding Is Not Sugar-specific-In the sucrose gradient sedimentation experiments of Alberts and Frey (7), it was reported that gene 32-protein did not form a complex with single-stranded R17 RNA. The data for this experiment were not shown and it is not possible to say whether a low level of binding might have gone undetected. Nevertheless, this result appeared reasonable from a teleological standpoint since a protein that binds to and denatures RNA secondary structure in the cell might be troublesome. However, the apparent preference of gene 32-protein for deoxyribopolynucleotides suggested by this experiment could have had a structural origin, since it is well known that R17 RNA, in particular (8), forms quite stable double helical hairpin structures which could appreciably diminish the apparent affinity of gene 32.protein for this ligand.

To approach this problem more directly, we have measured and compared the affinity of the protein for ribo- and deox- yribo-containing model compounds which do not form double

helical structures, ranging from the sugar-phosphate com- pounds dRib-P and Rib-P, through the dinucleoside mono- phosphates d(ApA) and r(ApA), to the polynucleotides poly- (dA) and poly(rA). The binding data obtained with the short ligands are summarized in Table I; those with the polynucleo- tides are in Table II and are discussed below. Both sets of data indicate, again within the limits of error of the measurements, that the affinities of gene 32-protein for ribo- and deoxyribo- containing ligands are very similar; in addition Jensen et al. (2) have shown that gene 32-protein binds to both poly(dA) and poly(rA), inducing significant changes in the nucleic acid backbone geometry.

The Bincling Site Is Polar-In principle, an oligonucleotide could bind to the gene 32-protein monomer binding site (relative to a particular axis on the protein) either in a polar (Le. with preferred 5’ + 3’ or 3’ + 5’ direction) or a nonpolar (random) orientation. The binding data of Table I, together with the quenching data of Table II of Ref. 1, can be used to show that the binding site does have a polarity relative to the oligonucleotide chain. The binding constants of d(pA), and d(Ap)*, as well as d(ApA), are shown in Table I. The binding constant of d(pA), is identical to that of d(ApA) within experimental error and the amount of the protein fluorescence quenched by both at saturation is very comparable. The binding constant of d(Ap), is somewhat lower than that of the other compounds and the ligand quenches about 5 times as much of the intrinsic protein fluorescence. The latter result, in particular, shows that the chain is bound to the protein binding site only in one direction, since otherwise the divalent phos- phate 3’- and 5’-end groups should be equally effective in quenching the tryptophan fluorescence.

Binding of Long Oligonucleotides

The binding behavior of large oligonucleotides and polynu- cleotides provides information about the cooperative mode of gene 32.protein binding to nucleotides. This type of binding, of course, cannot occur with nucleotide ligands shorter than some minimum length. The expected minimum length depends on whether adjacent protein molecules can bind in opposite directions. If they can, the protein could bind cooperatively to ligands shorter than the length physically covered by one protein molecule. For example, if the gene 32-protein “active site” which interacts with approximately two contiguous bases is located at one end of the molecule, then two proteins might bind in opposite directions to an oligonucleotide containing as few as four bases. On the other hand, if protein molecules bind to the lattice in the same direction, a nucleotide at least two bases longer than the number physically covered by a protein molecule would be necessary to achieve contiguous binding.

There are two types of evidence which suggest that the cooperative binding of gene 32-protein to nucleotide lattices is unidirectional. (a) If the cooperative mode of protein binding on polynucleotides involves protein-protein interactions then, based on symmetry considerations, there must be heterologous interactions (1); i.e. the protein domains involved in contacts between adjacent molecules must be different. This cannot occur on a lattice unless the molecules are all “pointed” in the same direction. (b) As cited above, the nucleotide binding site itself displays polarity as evidenced by the observed differences in the fluorescence quenching and binding properties of d(pA), and d(Ap),

Since m = 2, and n = 5 to 7.5 nucleotide residues (see

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

7244 Gene 32-Protein-Nucleic Acid Binding Parameters

below), the shortest chain length of oligonucleotide which will allow the contiguous binding of two protein molecules should be 7 to 10 nucleotide residues. Octanucleotides bind to gene 32-protein with about the same affinity as tetranucleotides (Table I) and show no cooperative binding behavior. However, for some oligonucleotide length slightly larger than 8, the comparative mode of binding should assert itself.

Cooperative Binding Site Size for Gene 32-Protein Binding Is About 5 Nucleotide Units per Protein Monomer-The binding of gene 32.protein to a mixture of deoxyadenylate oligomers (cUpA),,,,)> to d(pT),, , and to the polynucleotides poly(dA) and poly(rA), was examined by the fluorescence quenching technique. Fig. 4 shows a fluorescence titration curve for d(pT) 16 ; the d(pA),,,, mixture produced a similar curve. Fig. 5 presents the analogous data for poly(dA) and poly(rA). The binding of the protein to these polynucleotides in standard buffer (0.1 M Na+) is very tight; see Table II. In Figs. 4 and 5, the change in fluorescence is plotted against the ratio of the added nucleotide concentration to the protein concentration to permit determination of the stoichiometry of binding (n, the number of nucleotides covered by each protein molecule in the complex). The estimated binding constants (see below) for these compounds relative to those for octanucleotides (see Tables I and II) implies that cooperative interaction of gene 32.protein molecules occurs at a chain length greater than 8 but less than about 12 nucleotides.

If the binding constant of a ligand to the protein were infinitely large, the change in protein fluorescence with each addition of ligand would be a linear function of the total ligand concentration until a stoichiometric quantity of the ligand had been added; beyond this point the fluorescence should show no further decrease with the addition of more ligand. A sharp break in the isotherm would thus be observed at the saturating ratio of ligand to protein. For finite, but relatively large, values of the binding constant, extrapolation to the “infinite” case is possible to determine stoichiometry. The straight lines in Figs. 4 and 5 represent the best linear approximation to the initial and final slopes of the plots. The value of the abscissa at the intersection of these lines measures the site size (n) of the gene 32-protein monomer. The data of Figs. 4 and 5 indicate that n = 5 nucleotide residues; in reasonable accord with the value of

n = 7.5 nucleotides estimated from melting profile data and n = 6.7 estimated from the gene 32.protein-induced ultraviolet hyperchroism of poly(dA) (2).6

The measured values of binding stoichiometry for d(pN 12.18, d(pT) 16, and poly(dA) and poly(rA) are summa- rized in Table II. These values for the stoichiometry (n) differ from that of about 10 nucleotides per protein reported by Alberts and co-workers (7, 11). A simple explanation for this discrepancy might be that approximately one-half of the protein used in our experiments was inactive with respect to ligand binding. A DNA-cellulose control chromatography ex-

‘It should be pointed out that this site size (n) for gene 32-protein may also modify some current concepts of the shape of the gene 32-protein monomer. Electron microscopy of single-stranded fd phage circular DNA in the presence of saturatipg gene 32.protein shows that the DNA length in the complex is -4.6 A/nucleotide (10). Thus, if the site size for gene 32protein is -5 (this wgrk) to 7.5 (Ref. 2) nucleotide residues, the molecule must be 25 to 40 A in (nonoverlapping) length along the dimension aligned with the polynucleotide lattice. The diameter of a spherical protein cf average partial specific volume and molecular weight 35,000 is -40 A. Thus, based on the present site size, it is not necessary that gene 32.protein monomer be particularly asymmetric in shape.

TABLE II

Apparent binding constants and stoichiometry for association of large oligonucleotides and polynucleotides with gene 32.protein”

All binding constants measured in “standard: buffer containing: 50 mM Na,HPO,, 1 mM Na,EDTA, and 1 mM P-mercaptoethanol, pH 7.7.

K “,l,lcr

M I

dWJm,s 7.8 x 10’ 5.1 d(pV,, 7.0 x 10’ 6.2 poly(dA) 2.4 x 10’ 4.9 poly(rA) 4.4 x 108 4.7

“Determined using Equation 7, assuming the values for binding stoichiometry tabulated here.

*Determined from the intersection of the extrapolated initial and final slopes of binding plots.

Nucleotides

32- Protein

FIG. 4. Fluorescence titration curve for the binding of d(pT) ,C to gene 32.protein in standard buffer (50 mM Na,HPO,, 1 mM Na,EDTA, 1 mM &mercaptoethanol, pH 7.7), 25.1 i 0.1”. Protein concentration = 1.4 x 10-V M.

I I I I

I I / I I 0 IO 20 30 40

Nuckotides 32- Protein

FIG. 5. Fluorescence titration curves for the binding of poly(dA) and poly(rA) to gene 32.protein in standard buffer (50 mM Na,HPO,, 1 mM

Na,EDTA, 1 mM P-mercaptoethanol, pH 7.7), 25.1 i 0.1’. Protein concentration = 1.4 x 10 ’ M.

periment was performed to check this possibility. Fifty micro- liters of the same solution of purified gene 32-protein solution (approximately 50 rg) used in determining the isotherms of Figs. 4 and 5 was loaded on a small denatured calf thymus DNA-cellulose column (-1 ml in total column volume contain- ing -1 mg of denatured DNA). The column was washed with about 5 column volumes of standard buffer and fractions were

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

Gene 32.Protein-Nucleic Acid Binding Parameters 7245

collected. It was then eluted with standard buffer containing 2.0 M NaCl and fractions were again collected. The fractions were assayed by fluorescence at 342 nm; over 92% of the gene 32-protein added to the column was eluted by the 2 M salt wash. This indicates that almost all of the protein used in the titration experiments was active with respect to denatured

DNA binding, and thus, that the stoichiometry reported here should not be in error by more than about 10%.

The value of 10 nucleotides per protein for the stoichiometry of gene 32-protein binding to polynucleotides was originally

determined by sedimenting increasing amounts of single- stranded fd DNA with a constant amount of gene 32.protein in sucrose gradients (7). The data which produced this estimate of n nucleotides were not presented in the original publication,

but did appear in a later paper (11). Applying an extrapolation procedure similar to that used in Figs. 4 and 5 to these data yields a site size estimate of 5 to 6 nucleotides per protein monomer, in good agreement with those reported here.’

The similarity of the site sizes for single-stranded DNA and RNA suggests also. that binding to both ligands probably involves similar interaction geometries, etc.

Estimates of Affinity Constants for Cooperative Binding- While the large magnitudes of the cooperative binding constants for gene 32-protein to long oligonucleotide ligands permits a straightforward determination of binding stoichiom- etry, this tight binding makes it more difficult to use these data to measure binding constants since only the data points indicating some curvature at the “break” in Figs. 4 and 5 contain relevant information. However, these data do permit us to estimate apparent (cooperative) binding constants (per protein monomer) and to establish lower limits to these binding constants with some certainty. Such an estimate of an anparent binding constant (K,,,,, ; see below for a molecular interpretation of this parameter) can be achieved by the following procedure.

The experimentally measured fractional fluorescence change of the protein (AFIAF,) may be represented by %, where AF and AF, are, as previously, the protein fluorescence change at a particular ligand concentration and the fluorescence change at a saturating ligand concentration, respectively. The binding

constant for the cooperative interaction of gene 32.protein with a single oligonucleotide site may be written:

K a,,” = V’~llI~‘l[Nl (4)

where [N] is the concentration of free nucleotide binding sites, [P] the concentration of free protein, and [PN] the concentra- tion of the nucleotide ‘protein complex. The fractional satura- tion of the protein, 8, is then just the ratio of the bound protein to total protein:

LPN1 e =

[P] + [PN] =

f&,&N]

1 + $,,,[Nl

‘Very recently, Anderson and Coleman (12) have published a site size of -11 nucleotide residues for gene :i2-protein binding to single- stranded DNA, based on a CD titration technique. Inspection of their Fig. 2B suggests that perhaps under the ionic conditions of their experiment binding was not sufficiently tight to permit observation of an unambiguous end-point. Otherwise, we cannot account for this discrepancy.

This expression is rearranged and solved for [N] to yield:

e [Nl =

$,&l-e) (5)

The free concentration of nucleotide sites, [N] is assumed equal to the total concentration minus the concentration bound

to protein:

VI = [NOI BP01 (6)

where [N,] and [P,] are the total concentrations of nucleotide

sites and of protein, respectively. Equating the right sides of Equations 5 and 6 gives:

t&l - e[pJ = e

Kapp(l-e)

which when rearranged and solved for K,,,,, yields:

K arv =

e

(1-e) ([No1 - e[PJ)

The magnitude of this expression is of particular interest at the stoichiometric concentration of nucleotide sites; i.e. where N, = P, . At this nucleotide concentration it assumes the form:

K w =

e

(i-e)* [p,l (7)

If the binding is infinitely tight, then % = 1 at [N,] = [P,] and the equation is indeterminant, as expected. A finite apparent

binding constant, however, can in principle be determined if the protein concentration and the value of the fractional saturation at the stoichiometric ratio of nucleotides to protein are both known.’

The apparent binding constants for d(pA),,,,, d(pT),,, poly(dA), and poly(rA) are summarized in Table II. These values are based, in each case, on the listed stoichiometries which are needed to determine the point where [N,] and [P,] are equal. These values are also in reasonable accord, consider- ing the differences in ionic strength involved, with those determined from nucleic acid melting profiles (Ref. 2 and see below). The data of Tables I and II also suggest, within the large limits of error of the apparent binding constant analysis for the long ligands, that oligo- and polyribonucleotides and the corresponding deoxyribonucleotides bind to the protein with essentially equivalent affinities.

A few comments should be made about the accuracy of these

‘Obviously, this formulation ignores the effect of “overlap” binding; i.e. that “gaps” (g) of free nucleotide bases between bound proteins which are smaller than n (the occlusion size of the protein) cannot bind protein and are thus over-represented in [N], while gaps larger than n can bind (g - n) protein monomers (not just g/n) and thus are under-represented in [IV]. In other words, this approach assumes Scatchard binding to independent sites. The corrections required to obtain actual binding constants to long DNA chains where considera- tions of overlap apply have been treated in McGhee and van Hippel (3); the situation for finite length chains is considred under “Discus- sion.” We anticipate the discussion to follow by pointing out that for highly cooperative binding situations the effect of “overlap” becomes insignificant and with relatively little error we may represent the free nucleotide site concentration as in Equation 5, and K,,,, by Ku.

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

7246 Gene 32-Protein-Nucleic Acid Binding Parameters

estimates of the apparent binding constants of gene 32.protein to long nucleotide chains from data such as those of Figs. 4 and 5. Obviously, these constants represent lower limits, since contamination with shorter chain lengths capable of binding only one protein will appreciably “round off” the breaks and lead to lower values of K,,,, However, this should also lead to nonsaturation at higher ligand concentrations, and this is not apparent, even in Fig. 4. In summary, we can state with some confidence that K,,, , for binding to single-stranded polymers under the conditions used here, is no smaller than 10’ Mm’.

Magnitude and Nature of Cooperative Interaction-The marked increase in apparent binding affinity for gene 32.pro- tein of ligands greater than -10 to 12 nucleotides in length indicates that more than one protein monomer binds to these ligands and that the binding is cooperative; i.e. the binding of a second protein monomer adjacent to the first involves a favorable free energy of interaction greater than that binding an isolated protein to the polynucleotide chain. Furthermore, this favorable interaction free energy must overcome the unfavorable contribution resulting from the loss of “shuffling” free energy experienced by the first protein when the second is bound.

Using the data assembled here, we may attempt a quantita- tive formulation of this model. To this end, the free energy of interaction of gene 32.protein monomers with a polynucleotide chain long enough to accommodate two or more proteins may be written:

AGbind = AGint ' AGoverlap ' AGcoop (8)

where AG blrrd is the intrinsic free energy change for binding to the polynucleotide, AGoVeTlall is the contribution from the statistical factor which depends on the number of ways the protein can bind to the chain, and AGcoO,, is the free energy of cooperative interaction between contiguously bound protein monomers on the chain. In addition, as above, we define n as the number of bases covered by a protein monomer, m as the length of the nucleotide sequence actually bound to the protein “active site,” and 1 as the number of residues in the oligonu- cleotide chain under consideration.

For purposes of the discussion to follow, m will be taken as equal to 2 and n is 5. The protein binding site will be assumed to lie at the end of the molecule as illustrated in Fig. 3. Thus, cooperative (and polar) binding could, in principle, occur with oligonucleotides containing 7 residues. (This assumption may be modified if necessary.) In terms of the above, we can write down the predicted free energy change on binding a single gene 32.protein molecule to an oligonucleotide 1 residues long, where 1 is assumed to be greater than or equal to m (and less than 7) as:

AGbind = AGm - RT In ((l-m+l)

For the binding of two protein molecules to an oligonucleo- tide of length 1 there are two association equilibria:

P + NQ Z PNk + P Z P2NQ

characterized by association constants:

K, = LPN&l

and K2 = [P2NLl

[PI INI [PI [PN&

where K, and K, refer to the binding of the first and second

protein molecule to the oligonucleotide chain and [PI, [N,], and [PN,] represent the concentrations of the relevant species. These equations can be rearranged to give: [PN, ] = K, [P] [NL ] and [P2Nl] = K, [P] [PN!] = K,K, [PI’ [N,]. The fractional saturation of the protein (e), is the ratio of bound protein to total protein (P,):

e [PNRl + 2 [P2Ni.3

=

pcl

K, [PI [N 1 + 2 K,K2 [PI2 [N I = (9)

po

For the binding of a single protein molecule, the nucleotide concentration at which the protein is half-saturated is the inverse of the association constant. This is clearly not the case if more than one molecule is bound per chain, but if K, and K, are known (or can be estimated), an apparent binding constant can be generated at half-saturation of the protein for compari- son with a measured value. If 0 = 0.5, then [P] = [P,]/Z. Applying these conditions to Equation 9 and dropping the brackets:

K1PoN9. + 2

0.5 = KlK2Po Nf,

2po

which, given that K,,,,, = l/N,, rearranges to:

K assoc = K, + K&J’, (10)

A similar development for the binding of three proteins to an oligonucleotide gives:

K assoc = K, + K,K2Po + K,K2K3P02 (11)

and, in general:

K assoc = i (PO)"-' . (;K,)

The contribution of each higher K (n > 1) is scaled by the protein concentration and therefore is of significance only if the K is at least comparable to the inverse of the protein concentration.

The values of K can be chosen in accord with Equation 8, using as K,,,t the binding constant for gene 32..protein to d(ApA) (Kd( APA) z lo5 M-l) and the statistical factor (1 - m + l), as above. The binding of a second protein molecule to a chain long enough to permit this will restrict the number of ways the first can bind. The reasonable convention is adopted that this loss in shuffling entropy is suffered by the second molecule. Given the assumed values for m and n, and neglect- ing cooperativity, one predicts (e.g.) that a second protein binds to an oligonucleotide 7 residues long with an affinity constant (K,) = 1/6(K,).

Cooperativity may be incorporated by defining the parame- ter w as the increase in binding constant, for a second contiguously bound protein monomer, above the value ex- pected in the absence of cooperative interactions between contiguously bound monomers. (Thus w = 1 for non-coopera- tive binding; for a more complete definition of w for infinite chains, see Ref. 3.) Here w is taken to be independent of

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

Gene 32-Protein-Nucleic Acid Binding Parameters

nucleotide chain length. Then, for the binding of the second and subsequent protein molecules to the oligonucleotide we may write:

IO” -

IO’O -

5 f 6 10s 0

H

; lo*-

IO7

IO6 -

K2 = Kint*w. (!I.-m+l-n) (Il-m+l ) ; for 9. > 6

K3 = Kint.u for L > 11

and in general, for the binding of the ith protein molecule:

(t-mtl-[i-l]n) Ki = Kint*u*

(a-m+l) ; for 1 > ([i-l]n+l)

The values of K,, K, . generated by these equations can then be used to produce predicted binding constants (recipro- cal nucleotide concentrations at half-saturation) using Equa- tions 10 and 11. The predicted dependence of this binding constant on oligonucleotide chain length is shown in Fig. 6 for four values of w. The symbols are the experimentally measured values; the circles representing the deoxyadenosine oligomers and the squares the deoxythymidine oligomers. (The points on the right-hand side are for poly(dA) and poly(rA).) Given the assumptions involved in formulating the model, it appears that the best fit to the data for the oligonucleotide chains comes at 0 = 250.

For cooperative binding of the protein to a polynucleotide, the binding of each protein to the chain yields an apparent binding constant Kw, neglecting end effects and “overlap” (3). We can estiamte w for poly(dA) and poly(rA) as follows, using the apparent polymer binding constants (K,,,) listed in Table

I:

wpoly dA = KPoly dA ze 1 6 x lo3

Kd(ApA) '

K poly rA

Woolf rA 1 Kr(ApA) = 1.1x103

Thus, the apparent cooperativity parameter may be somewhat greater for protein binding to very long lattices (by factors of 4

to 6), but given the approximate value of the binding constant estimation (see above) these differences may not be significant. For present purposes an w of approximately lo3 may be taken as a working estimate for this parameter; this value is in good agreement with that estimated from melting profiles (2). Since the melting profile measurements were made at considerably lower ionic strengths than the fluorescence experiments re- ported in this paper, we may surmise that the cooperativity parameter (unlike the intrinsic affinity constant of gene 32.protein for the polynucleotide lattice, see Ref. 2) is rela- tively insensitive to ionic strength changes.

DISCUSSION

Molecular Findings-In this and the preceding paper (l), we have utilized the quenching of the intrinsic tryptophan fluores- cence of gene 32-protein by nucleotide ligands to partially map the nucleic acid binding site of this protein and to study the conformational alterations associated with cooperative and non-cooperative binding to DNA. This fluorescence quenching has also been exploited to measure the binding constants of a variety of mono-, oligo-, and polynucleotide ligands to gene 32.protein, and by determining apparent binding constants as

jwio5

I

i

lo5 1 I I I b

4 8 12 I6 20 ” OD

7247

Chain Length

Fit. 6. Binding constant (K,,,,,,) for gene 3%protein to oligonucleo- tide lattices as a function of lattice length for different values of the cooperativity parameter, w. (See text for definitions of K,,,,,, , w, and the assumptions involved in generating these plots.) The points correspond to the actual data (K,,,,, ) for adenosine- (0) and thymidine- (Cl) containing oligomers, respectively.

a function of oligonucleotide chain length, to measure m (the number of nucleotide residues which actually interact with the protein), n (the number of nucleotide residues covered by the protein), and w (the cooperativity parameter).

The results reported here, together with those of Jensen et al. (2) and Carroll et al. (4, 13), permit a further analysis of the molecular origins of the cooperativity of gene 32.protein binding. By definition (as used here), cooperative protein binding means that binding of a protein monomer to a polynucleotide lattice in a “contiguous” position (adjacent to a previously bound protein monomer) is favored, in free energy terms, over binding the same monomer to an “isolated” lattice site. Of course, this favored contiguous binding is always opposed by “shuffling” (mixing) entropy favoring isolated binding, because there are many more ways (at low lattice saturation) to bind an “isolated” monomer. However, as the ratio of protein monomers to lattice sites is increased, positive cooperativity of binding increases the extent of lattice satura- tion with increasing protein concentration much more rapidly than non-cooperative binding (3).

Positive cooperativity can arise either because adjacent, protein monomers engage in direct and favorable protein- protein interactions in the relative orientation in which they align on the polynucleotide lattice, or because binding involves a free energy-requiring distortion of the polynucleotide lattice. If such a distortion is easier for a contiguously bound protein monomer to propagate than for a singly bound monomer to initiate, binding is cooperative. Of course, both effects may be operating in a given system and the data available for the gene

32-protein-polynucleotide interaction suggest both may apply here.

Thus, Carroll et al. (4, 13) have shown that gene 32-protein tends to form indefinite aggregates at moderate temperatures in the absence of polynucleotides and have (implicitly) as- sumed that this aggregation reflects the protein-protein

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

7248 Gene 32-Protein-Nucleic Acid Binding Parameters

interactions responsible for the observed cooperativity of the binding to polynucleotides. This assumption is certainly quali- tatively consistent with our findings, although we have sug- gested that the self-limited “high salt” dimers of gene 32-pro- tein involve different interactions. Based on a mathematical analysis of their data, Carroll et al. (4) have estimated the intrinsic monomer-monomer indefinite association constant for aggregation in the absence of nucleotide ligands to be approximately lo5 Mm’. This corresponds to a unitary free energy of interaction (at 25”) of - -9 kcal/mol (molar standard states) and thus to a cooperativity parameter (w) of -4 x 106. The value of w measured in the present studies (10’) corre- sponds to a (unitary) monomer-monomer interaction free energy (on the lattice) of - -4 kcal/mol. Thus, it appears that the intermonomer protein-protein interactions may be appre- ciably less favored on the lattice than in free solution. This difference might reflect either the involvement of different binding sites in the interaction in the presence and absence of polynucleotides, or it might arise because the putative protein- protein interaction sites are somewhat unfavorably constrained when the protein is attached to the lattice. Thus, lattice attachment per se could orient the functional groups responsi- ble for the interprotein interactions into a less than optimal geometry, or binding to the lattice could induce a conforma- tional change in the protein monomers which decreases the interprotein affinity. In keeping with the latter possibility, the results of the preceding paper (1) have indeed demonstrated that binding to nucleotide ligands alters the conformation of the gene 32-protein and Carroll et al. (4) have shown that binding of gene 32.protein to oligonucleotides of sufficient length to permit cooperative binding partially inhibits the protein self-association seen in the absence of oligonucleotides.

!Iowever, Jensen et al. (2) have also demonstrated that gene 32-protein binding appreciably distorts the polynucleotide lattice from the “normal” single chain conformation present free in solution, a finding consistent with the notion that some of the favorable free energy of contiguous binding might also be attributed to the fact that contiguous monomers “collaborate” in distorting the lattice, and thus the unfavorable free energy required for lattice distortion (per protein monomer) is less for contiguously bound protein molecules. Additional studies will be required to further define the molecular bases of the observed cooperativity.

Possible in Vivo Functional Correlations-Gene 32-protein, together with the other DNA “melting” (“unwinding”) proteins for which function has been at least partially established, appears to play an important role in the events of recombina- tion and replication (for a recent review see Kornberg, Ref. 14). In both of these processes, as presently conceived, single- stranded DNA regions serve as crucial intermediates and melting proteins are generally assigned the role of facilitating the molecular interactions in which these single-stranded intermediates are involved.

Gene 32.protein could serve at least two purposes in this connection: (a) due to its T,-lowering properties, gene 32-pro- tein could (at equilibrium) facilitate either the denaturation or (by selectively destabilizing local double-stranded hairpin regions formed at random in single-stranded DNA) the renatu- ration of double-stranded DNA; and (b) by binding tightly and cooperatively to single-stranded intermediates it could protect these regions against attack by single strand-specific exo- and endonucleases.

As discussed in detail in a preceding paper (a), it is very

likely that under cellular conditions gene 32-protein facilitates double helix renaturation rather than denaturation; i.e. that the total cellular environment (including physiological concen- trations of gene 32-protein) is constituted so that short random double-stranded hairpins are unstable at equilibrium, while longer complementary double helical regions are stable. As a consequence (15), the decrease in the equilibrium T, resulting from the presence of gene 32.protein should increase the renaturation rate of the “correct” double helix by destabilizing many of the metastable (“hairpin”) double helical states which would otherwise slow the process. We note that cooperative binding of the protein to the nucleotide lattice is crucial to this function, since non-cooperative binding would tend to “skip- over” the hairpin regions. This surmise is consistent with the finding that melting proteins which bind non-cooperatively also do not appear to facilitate double helix renaturation (16).

It is tempting to speculate that a major function of gene 32-protein in uivo may be to protect the single-stranded regions of DNA which are generated during genetic processes from degradation by nucleases. A variety of evidence supports this idea. (a) Gene 32-protein protects single-stranded DNA from attack by several nucleases (17-19). (b) The DNA extracted from 32-phage-infected cells consists primarily of pieces less than one phage equivalent in length (20). (c) Putative “recom- binational” intermediates isolated from T4 amber mutants in gene 32 do not demonstrate single-stranded regions, termini, or the kind of branched structures seen in wild type T4 and several other DNA-deficient mutants (20, 21). (d) A T4-in- fected Escherichia coli contains nucleases of almost every conceivable specificity (14).

I f this speculation on the function of gene 32-protein is correct, it again becomes obvious why the binding of the protein to polynucleotides is cooperative; this is the only prac- tical way to approach saturation of the lattice at reasonable protein concentrations (3). The results of Gold and co-workers (22), who have shown that the amount of gene 32.protein made in a T4-infected cell appears to be approximately proportional to the amount of single-stranded DNA present, are consistent with this view.

Gene 32.protein is one of a number of T4-coded proteins which are considered to be part of the replication complex (11). However, the presence or absence of this component appears to have little effect on the affinity for DNA of 18 other T4 DNA-binding proteins (23). Gene 32-protein specifically stim- ulat& T4 DNA polymerase activity in vitro and also binds to the enzyme (24). It is not known if the formation of a polymerase.gene 32-protein complex is necessary for stimula- tion of the polymerase; gene 32-protein could also stimulate indirectly by aligning and distorting the DNA template into a specific form optimal for T4 polymerase-mediated DNA syn- thesis.

The other properties of gene 32.protein are clearly of considerable significance to its presumed in vitro functions. Its apparently quite uniform binding constant to single-stranded polynucleotides of differing base composition and base sequence is a necessary feature if it is to interact with all single-stranded DNA regions, while its much lower binding constant to double helical DNA prevents its loss from solution onto the vast majority of the DNA which is always present in this form (see discussion of such coupled equilibria in von Hippel et al. (25). However, the approximately equal affinity of this protein for single-stranded DNA and RNA is surprising and introduces new complications and opportunities into

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

Gene 32-Protein-Nucleic Acid Binding Parameters 7249

consideration of its in oiuo function. The accompanying paper by Gold and co-workers (19) further develops this theme.

Cooperativity and Origins of Binding Specificity-In this and a preceding paper (2), we have demonstrated that binding of gene 32-protein monomers to oligo- and polynucleotides appears to be quite nonspecific with respect to both sugar-type and nucleotide composition. However, we must emphasize that the measurements presented here are direct binding measure- ments and that the binding constants and cooperativity factors are probably not accurate to better than factors of 2 to 5. Thus, the possibility remains that there will be small differences between association constants (and/or cooperativity factors) for the binding of gene 32-protein to polynucleotides of varying composition and sequence. Furthermore, these small differ- ences can be amplified by binding cooperativity and thus appear as substantial differences in certain types of relative affinity measurements.

Recently Bobst and Pan (26) have reported apparent differ- ences in the relative affinity of various types of polynucleotides for gene 32-protein, based on monitoring the competitive displacement of limiting quantities of gene 32-protein from a spin-labeled polyriboadenylic acid probe to a variety of other polynucleotides. This apparent discrepancy between Bobst and Pan’s study and our own could easily be accounted for by binding constant and cooperativity parameter differences (for different polynucleotides) well within the standard error of our measurements. As indicated above, such differences are ampli- fied in displacement experiments by binding cooperativity under conditions of limiting protein, since the relative proba- bilities for occupancy of a given pair of polynucleotide se- quences will be the ratio of the products of the relevant monomer binding constants and cooperativity parameters.

Similar small differences in binding affinity for various nucleotide sequences, coupled with the amplifying effect of cooperativity, can be used to explain the apparent specific binding effects involved in the in uiuo function of gene 32-protein, as discussed by Gold and co-workers in the accom- panying papers (19, 22). These workers have shown that gene 32protein biosynthesis is regulated by titration onto single- stranded DNA sequences present as intermediates in replica- tion, recombination, and repair; and that when these regions (and others present as gaps in certain mutant T4 production systems) are saturated, gene 32-protein appears to bind to its own messenger, resulting in specific translational repression of synthesis of this protein.

Russel et al. (19) have presented a specific model for such a sequence of events, based on gene 32-protein binding to a particularly long, non-base-paired sequence on its own messen- ger RNA, and our purpose here is simply to provide some physiochemical support and quantitative elaboration of struc- turally based specificity models of this type. Thus, we note that in the absence of any nucleotide residue composition or sequence specificity of binding whatsoever, gene 32.protein will preferentially bind to, and saturate, all single-stranded se- quences which are unencumbered by regions of secondary structure with net stability exceeding approximately - 4 kcal; this is the free energy associated with one cooperative protein- protein interaction (w 1 103). I f the conformational free energy of a region of secondary structure (e.g. a hairpin loop) lying in the path of an advancing wave of contiguously binding gene 32-protein exceeds w-4 kcal, then the extension of that advancing sequence of contiguous gene 32-protein monomers will cease if other single-stranded regions are available in

which a new sequence of contiguous gene 32.protein binding can be initiated, and the next longest polynucleotide region unencumbered by relatively stable secondary structure will be coated. Thus, in the absence of any nucleotide-based binding specificity, single-stranded polynucleotide regions will be covered in order of length devoid of stable (by the above criterion) secondary structure and we would predict that only DNA gaps exceeding the length of that of the postulated structure-less single strand sequence for gene 32 mRNA pro- posed by Russel et al. (19) will be covered prior to translational repression. Small sequence-based differences in binding affin- ity (and/or binding cooperativity) can be used to further sharpen or modulate the above structure-based specificity. However, in light of the data presented in these papers and the physiochemical rationale outlined above, it appears both unlikely and unnecessary that gene 32-protein will be found to show a strong and specific sequence-dependent binding to its homologous messenger RNA.

As is apparent in the repression model developed by Russel et al. (19), it is only required that other T4 messages not contain sequences devoid of secondary structure in translation initiation regions which .are as long as that of the gene 32 mRNA, in order for repression to be specific. Binding to other mRNA molecules, in noninitiation regions, may be possible (and perhaps desirable, to provide a further available pool of presynthesized gene 32.protein, see Ref. 19) as long as this protein can be displaced by the progression of translation per se.

Conclusions-The detailed physiochemical analyses of the non-cooperative (ribonuclease) and cooperative (T4 gene 32.protein) binding of these melting proteins to nucleic acids and nucleic acid models, presented in this and the preceding papers (1, 2, 27), may be useful to other workers in probing the physical chemistry of many other melting protein-nucleic acid systems currently under study. Such probing, in turn, should eventually help to provide a molecular understanding of the extremely complex processes of genome manipulation involved in DNA replication, recombination, transcription, and repair.

Achowledgments-We are pleased to acknowledge many very helpful discussions of aspects of this work with our colleagues in the Institute, including particularly Drs. Frede- rick Dahlquist and James McGhse.

REFERENCES

1. Kelly, R. C., and van Hippel, P. H. (1976) J. Rio/. Chem. 251, 7229-7239

2. Jensen. D. E., Kelly, R. C., and van Hippel, P. H. (1976) J. Hiol. Chem. 251, 7215-7228

3. McGhee, J. D., and van Hippel, P. H. (1974) J. Mol. Rio/. 86, 469-489

4. Carroll, R. B., Neet, K. E., and Goldthwait, D. A. (1975) J. Mol. Biol. 91, 275-291

5. Record, M. T., Jr., de Haseth, P., and Lehman, T. M. (1976) J. Mol. Biol., in press

6. Alberts, B. M., Amodio, F. J., Jenkins, M., Gutmann, E. D., AND Ferris, F. L. (1968) Cold Spring Harbor Symp. Quant. Rio/. 33, 289-305

7. Alberts, B. M., and Frey, L. (1970) Nature 227, 1313-1318 8. Steitz, J. A. (1969) Nature 224, 957-964 9. Richards, F. M., and Wyckoff, H. W. (1971) in The Enzymes

(Boyer, P., ed) Vol. 4, pp. 647-806, Academic Press, New York 10. Delius, H., Mantell, N. J., and Alberts, B. (1972) J. Mol. Biol.

67, 341-350 11. Alberts, B. M. (1971) in Nucleic Acid-Protein Interactions:

Nucleic Acid Synthesis in Viral Infection (Ribbons, D. W., Woessner, J. F., and Schultz, J., ed) pp. 128-143, North Hol-

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

7250 Gene 32.Protein-Nucleic Acid Binding Parameters

land, Amsterdam 12. Anderson, R. A., and Coleman, J. E. (1975) Biochemists 14,

5485-5491 13. Carroll, R. B., Neet, K. E., and Goldthwait, D. A. (1972) Proc.

Natl. Acad. Sci. U. S. A. 69, 2741-2744 14. Kornberg, A. (1974) DNA Synthesis, W. H. Freeman and Co., San

Francisco 15. Marmur, J., and Doty, P. (1961) J. Mol. Biol. 3, 585-593 16. Herrick, G., and Alberts, B. M. (1976) J. Biol. Chem. 251,

2133-2141 17. Huang, W. M., and Lehman, I. R. (1972) J. Biol. Chem. 247,

3139-3146 18. Wu, J.-R., and Yeh, Y.-C. (1973) J. Viral. 12, 758-765 19. Russel, M., Gold, L., Morrissett, H., and O’Farrell, P. 2. (1976) J.

Biol. Chem. 251, 7263-7270

20. Kozinski, A. W. and Felgenhauer, Z. Z. (1967) J. Viral. 1, 1193- 1204

21. Broker, R. R. (1973) J. Mol. Riol. 81, 1-16 22. Gold, L.. O’Farrell, P. Z., and Russel, M. (1976) J. Biol. Chem.

251, 7251-7262 23. Huang, W. M., and Buchanan, J. M. (1974) Proc. Natl. Acad. Sci.

U. s. A. 71, 2226-2230 24. Huberman, J. A., Kornberg, A., and Alberts, B. M. (1971) J. Mol.

Riol. 62, 39-52 25. van Hippel, P. H., Revzin, A., Gross, C. G., and Wang, A. C. (1974)

hoc. Natl. Acad. 5%. U. S. A. 71, 4808-4812 26. Bobst, A. M., and Pan, Y.-C. E. (1975) Biochem. Biophys. Res.

Commun. 67. 562-570 27. Jensen, D. E., and van Hippel, P. H. (1976) J. Biol. Chem. 251,

7198-7214

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from

R C Kelly, D E Jensen and P H von Hippelfor bacteriophage T4 gene 32-protein to mono-, oligo-, and polynucleotides.

DNA "melting" proteins. IV. Fluorescence measurements of binding parameters

1976, 251:7240-7250.J. Biol. Chem. 

  http://www.jbc.org/content/251/22/7240Access the most updated version of this article at

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/251/22/7240.full.html#ref-list-1

This article cites 0 references, 0 of which can be accessed free at

by guest on July 22, 2020http://w

ww

.jbc.org/D

ownloaded from