Characterization of a cooperativity domain mutant Lys3 --> Ala (K3A) T4 gene 32 protein.

The N-terminal “B” domain of T4 gene 32 protein contains a Lys3-Arg4-Lys5 sequence that has been postulated to provide a major determinant for cooperative binding. In this report, the equilibrium binding properties of a Lys3 → Ala substitution mutant of gp32 (K3A gp32) and described and compared to a set of substitution mutants of Arg4 previously described (Villemain, J. L., and Giedroc, D. P. (1993) Biochemistry 32, 11235-11246) and further characterized here. K3A gp32 exhibits binding behavior which mirrors that of R4Q gp32. Despite an 6-8-fold decrease in overall binding affinity (Kapp = Kint × ω) at pH 8.1, 0.20 M NaCl, 20°C, the magnitude of the cooperativity parameter is at most 2-3-fold smaller than that of the wild-type protein. The magnitude of ω is independent of salt concentration and type over the range in [NaCl] from 0.125 to 0.225 M and [NaF] between 0.20 and 0.32 M (log ω = 2.86 ± 0.19). For comparison, log ω for wild-type gp32 is 2.91 (± 0.21) resolved at 0.275 M NaCl and 3.39 ± 0.11 in [NaF] between 0.40 and 0.45 M. In contrast to ω, the [NaCl] dependence of Kapp is large and markedly nonlinear for both wild-type and K3A gp32s over a [NaCl] range extending from 0.05 M to 0.40 M NaCl. Modeling of the complete salt dependence of Kapp for wild-type, K3A, and R4T gp32s in NaCl and NaF with a simple ion-exchange model suggests that substitutions within the basic Lys3-Arg4-Lys5 sequence do not strongly modulate the net displacement of cations and anions upon poly(A) complex formation by gp32.

Bacteriophage T4 gene 32 protein (gp32) 1 is the paradigm for helix-destabilizing or single-stranded (ss) DNA-binding proteins, which play accessory roles in DNA replication, recombination, and repair (for a review see Ref. 1). gp32 binds ss nucleic acids with an "unlimited" type of cooperativity to ssDNA segments transiently formed during these processes (2). Formation of clusters of gp32 monomers directly derives from the high cooperativity of binding and probably imparts on the ssDNA a particular conformation that is readily utilized by the DNA polymerase or recombinase as well as serving a protective role against the action of intracellular nucleases. In both DNA replication and recombination, it is also clear that apart from interactions with ssDNA, one or more gp32 monomers also engage in functional heterologous contacts with other proteins of the multienzyme complexes that carry out these processes (3,4).
Limited trypsinolysis experiments generally reveal that gene 32 protein (301 amino acids) contains three functional domains, which map to distinct regions of the primary structure (5). Two terminal domains encompassing residues 1-21 (the N-terminal "B" or basic domain) and the C-terminal "A" or acidic domain (residues 254 -301) are probably exposed to solvent and are structural appendages on the ssDNA-binding core domain (residues 22-253) (6). The core domain binds an intrinsic Zn(II) ion (7) and contains major determinants for ss nucleic acid binding. The crystal structure of the core fragment complexed to oligo(dT) 6 reveals that the ssDNA fits into a large cleft of a C-shaped molecule (8). The cleft is lined with basic amino acids, in particular those derived from a Lys 110 -Arg 111 -Lys 112 sequence (9), as well as aromatic amino acids (10). The Zn(II) domain organizes one of the two globular domains of the molecule, which forms part of the ssDNA binding groove, as anticipated from spectroscopic and biochemical experiments (11,12). It is well established that the A domain plays a functionally important role in heterologous protein-protein interactions, which are required for the formation of multiprotein complexes that carry out replication (4,13) and recombination (14,15).
Cooperative binding contributes up to 40% of the binding energy of the gp32 monomer at equilibrium under physiological salt concentrations; however, there exists no information available that addresses the extent to which cooperative binding is required for gp32 function in DNA metabolism. We have initiated a program to determine how the N-terminal domain modulates the mode and mechanism of gp32 binding on model ss nucleic acid substrates. Our approach is to generate a series of N-terminal domain point mutants of gp32 with well defined thermodynamic and kinetic defects, which could then be used as tools to probe the function of the N-terminal domain in vivo and in in vitro replication and recombination assays.
The simplified view of the three-domain model of gp32 has the "B"-domain participating in contiguous monomer-monomer interactions on the ssDNA, while the gp32 monomer-ssDNA interface is functionally and physically separate (16). In a previous report, we determined that Arg 4 in the Lys 3 -Arg 4 -Lys 5 sequence plays a critical role in maintaining high affinity and highly cooperative binding to ss nucleic acids (17) in a manner strongly dependent on the nature of the substitution of Arg 4 . As a result, point mutants with K app reduced as little as 2-1000fold were characterized. In particular, quantitative investigation of moderately defective gp32s allowed us to conclude that this model is an oversimplification, since these mutants retain a high cooperativity of binding () even in the presence of large 10 -100-fold decrements in the overall binding constant (K app ϭ K int ϫ ) (17). This suggested that the N-terminal domain controls the magnitude and energetics of both and K int terms. Although the recent crystallographic structure of the core fragment complexed with a ssDNA oligonucleotide (8) does not include the B-domain, the proximity of the N-terminal residue (Gly 22 ) of the core fragment to the ssDNA cleft provides structural support for the suggestion that the N-terminal domain and ssDNA cleft cooperate to enable gp32 to form tightly and cooperatively bound clusters of monomers at equilibrium.
In this report, we extend these findings to Lys 3 with our characterization of the binding of K3A gene 32 protein on the model ribohomopolymer poly(A) under solution conditions previously used to characterize wild-type gp32 and a number of Arg 4 substitution mutants of gp32 (17). We show that this mutant behaves quantitatively like R4Q gp32 (17) and retains a high cooperativity of binding at equilibrium largely indistinguishable from the wild-type protein. Given the widely disparate nature of the K3A and R4Q substitutions, the data presented here provide crucial support for the contention that the N-terminal B-domain substitutions of gp32 truly reflect a perturbation of the wild-type protein rather than reporting on new or novel properties acquired by the mutant proteins. More importantly, these data provide the equilibrium basis for an interpretation of the stopped-flow fluorescence dissociation kinetic behavior and biochemical function of gp32s described elsewhere. 2,3

Materials
All buffers were prepared with doubly distilled Milli-Q water. Buffer salts were obtained from Sigma. Poly(A) obtained from Midland Certified Reagent Co. (Midland, TX) was fractionated on Sepharose S-200 (Pharmacia Biotech Inc.), and the first 30% of the fractions containing absorbance were collected and dialyzed exhaustively against T/0.10 M NaCl. The concentration of poly(A) nucleotides was determined from duplicate dilutions in the samebuffer using ⑀ ϭ 10,300 M (nucleotide) Ϫ1 ⅐ cm Ϫ1 at 260 nm (18). The fractionated poly(A) was polydisperse and migrated roughly along side a Ϸ500-base pair duplex DNA fragment on a 1% agarose gel. Other fractionations are described in the text.
K3A gene 32 protein was prepared from an overproducing plasmid harboring the K3A gene 32 coding sequences under the transcriptional control of the inducible P L promoter exactly as described previously (17). The details of the degenerate oligonucleotide cassette-based mutagenesis approach used to construct the expression plasmid are essentially as outlined (17). Briefly, an NdeI-SacI double-stranded cassette, which would direct a random set of mutations at Lys 3 , was synthesized with 25% A, G, T, and C in the first two positions and 50% GϩC in the third position of the Lys 3 codon in the sense strand. Three inosine residues were inserted in the nonsense strand (19). Following ligation into NdeI/SacI-digested pT7gp32.wt and bacterial (Escherichia coli TB1) transformation, transformants were selected, double-stranded miniprep DNA prepared and subjected to DNA sequencing using the dideoxy method in order to identify the mutation at codon 3 (Ala ϭ CGC). The mutant gene containing upstream translational regulatory signals was subcloned into the pP L expression plasmid pTL9W to create pP L g32.K3A for protein expression (17). Wild-type and R4T gp32s were purified as described (17).

Methods
Buffers-T buffer is 0.01 M Tris-HCl, 0.1 mM Na 3 EDTA, pH 8.1, at 20°C. To make the buffers for binding measurements, a 10-fold stock solution of T buffer was mixed with the appropriate amount of 3.94 M NaCl and volume made up with doubly distilled water. The stock NaCl concentration was determined by refractive index. The conductivity of all buffer solutions was also determined at ambient temperature to ensure their relative consistency of [NaCl].
Fluorescence Binding Experiments-The fluorescence of K3A gp32 was monitored with an SLM 8000 spectrofluorimeter by excitation at 292 nm or 296 nm (excitation bandpass ϭ 2 nm) and emission at 347 nm (bandpass ϭ 4 nm). The temperature was maintained with a thermostatted sample compartment at 20 Ϯ 0.1°C. Protein samples, prepared at the indicated concentrations in 2.0 ml of buffer, were continuously stirred gently throughout the course of titration with small (2-5 l) aliquots of a concentrated stock solution of polynucleotide added. The measured fluorescence values (F i ) were converted to corrected fluorescence values (F corr,i ) after accounting for dilution, inner filter correction (20) and photobleaching as described (17). The extent of photobleaching was always less than 5% under the excitation conditions used. Q obs is given by Q obs ϭ (F corr,max Ϫ F corr,i )/F corr,max . We determined that the binding density obtained from the degree of quenching, Q ϭ (L B /D T ) where D T ϭ [poly(A)] T , directly reflects the true binding density, , using a general method of analysis (21) (data not shown) (17). Individual titrations were subjected to a nonlinear least-squares minimization using the algorithm JANA (17) linked to NONLIN (24) running on a PC486 platform to extract values of K int and with n fixed at 7.5 from the McGhee-von Hippel closed-form expression for the cooperative large-ligand overlap binding model (22,23) L B determined from Equation 1, L T , and D T were used as input into JANA to solve for L B numerically at iteratively adjusted K int and such that the sum of the residuals between calculated and measured L B is minimized (24). The results of fitting individual titrations are reported with only those data points where Յ 0.011 used in any analysis, unless otherwise noted (17). Values of K int and are returned by NONLIN with 67% confidence intervals, which incorporate the high negative cross-correlation coefficient (Ϫ0.99) between these two parameters calculated by NONLIN. The validity of the returned parameters and associated confidence intervals was routinely assessed visually by superimposing the experimental data with three theoretical isotherms: one described by the minimized parameters, one calculated with the upper confidence limit of K int and the lower limit of , and a third calculated with the lower limit of K int and the upper limit of . Saltback-induced dissociation titrations of gp32-poly(A) complexes were carried out and analyzed with fixed at 1000 as described (17,20,25). Circular Dichroism Binding Experiments-The decrease in the positive Cotton band of poly(A) at 264 nm was monitored upon titration with small aliquots of gp32 protein on a Jasco C-600 spectropolarimeter. About 15-25 M poly(A) in 1700 l of T buffer with an indicated concentration of NaCl was incubated at 20°C for 5 min before data acquisition. Typical acquisition parameters were a 1-2-s time constant, a 1-nm bandwidth, a 0.2-nm step size, and a sensitivity of 20 mdeg. Six scans in a window of 262-266 nm were averaged to increase the signalto-noise ratio. After each addition of gp32, the cuvette was stirred for 1 min at which point stirring was ceased and the CD spectrum recorded. The aromatic CD contribution from gp32 was subtracted from the spectrum, which was then corrected for dilution. The resulting CD intensity at 264 nm was used for the calculation of fractional saturation, ⍜. Fractional saturation for a gp32 addition point was calculated from ⍜ ϭ (mdeg i Ϫ mdeg init )/(mdeg final Ϫ mdeg init ), which assumes that a change in the CD intensity at 264 nm is proportional to the protein binding density. The data were then subjected to the same nonlinear least squares analysis as above, except that n ϭ 10 nucleotides (determined independently at low salt) and data points at ⍜ Յ 0.40 were used in the analysis.

K3A gp32 Binds Poly(A) More Weakly than Wild-type gp32 but with a Largely Indistinguishable Cooperativity of Binding-
The interaction of gp32 with poly(A) was monitored by measuring the decrease in the intrinsic protein (Trp) fluorescence that occurs upon RNA binding. Titrations were carried out in the "reverse" mode in which a fixed input concentration of ligand gp32 (L T ) is titrated with lattice poly(A). All measurements unless otherwise noted were carried out at pH 8.1 in 10 mM Tris-HCl, 0.1 mM Na 3 EDTA (T buffer) with the indicated concentration of NaCl at 20°C. Independent determination of the apparent site size (n), the number of occluded nucleotides per gp32 monomer, was carried out under low salt conditions (T/0.05 M NaCl). Like found for wild-type gp32 and other Bdomain mutants (17), K3A gp32 was shown to bind stoichiometrically with n ϭ 7.5 (Ϯ 0.5) nucleotides (1,5). The maximal extent of quenching of the Trp fluorescence (Q max ) of K3A gp32 was determined to be 0.30 (Ϯ 0.01) comparable to the value (0.29) determined for wild-type gp32.
In order to determine how substitution of Lys 3 with Ala affects the binding properties of gp32, resolution of K int and according to McGhee-von Hippel linear lattice theory (22) is required. This can only be done under nonstoichiometric binding conditions. Solution conditions were therefore identified for K3A gp32 under which nonstoichiometric binding could be readily detected in reverse titrations using methods exactly as described previously (17). In initial experiments carried out in T buffer/0.275 M NaCl, pH 8.1, and 20°C under which wild-type gp32 binding could be readily measured, K3A gp32 was found not to bind to an appreciable extent, signaling a significantly lower binding affinity. T/0.125 M NaCl, pH 8.1, was subsequently identified as solution conditions that could be used to resolve K int from . Fig. 1 shows a set of representative binding isotherms obtained for K3A gp32 with the initial input ligand concentration, [L T ], ranging from 1.1 to 3.0 ϫ 10 Ϫ7 M, conditions under which the wild-type protein shows little or no ligand aggregation (20,26). The multiple titrations spanned overlapping ranges of binding density () or polynucleotide lattice saturation (⍜) (indicated in the legend to Fig. 1). To minimize any lattice end effects, only those data points where Յ 0.011 (or ⍜ Յ 0.083) were considered in an analysis, thus removing points at the beginning of some reverse titrations at higher [L T ] (open symbols in Fig. 1). The range in values of K int and obtained from a nonlinear least squares parameter optimization of individual titrations, some of which are shown in Fig. 1 as best-fit theoretical isotherms superimposed on the data. As can be seen, the apparent maximal extent of quenching in each isotherm appears to increase with increasing imput ligand concentration but consistently remains far smaller than Q max (Q max ϭ 0.30). This is behavior diagnostic of highly cooperative ligand binding (27). This is borne out by the quantitative analysis. The continuous curve through each set of the data points represents the function described by the optimized parameters (K int and ) with n fixed at 7.5 nucleotides. K app was found to be 6.7 (Ϯ 2.0) ϫ 10 6 M Ϫ1 with the resolved values of ranging from 430 to 1850 and K int ranging from 0.5 to 1.1 ϫ 10 4 M Ϫ1 . Fig. 2 presents the results of multiple independent determinations of made from additional reverse titrations like those shown in Fig. 1 (17), while between 0.40 and 0.45 M NaF, log ϭ 3.39 (Ϯ 0.11) or Ϸ 2400, or no more than 3-fold larger than for K3A gp32.
These data suggest that the magnitude of is more similar than different from wild-type gp32 (17), although the finite and polydisperse length of the poly(A) used to make these measurements places limitations on the maximum magnitude of accessible in these experiments. As an illustration of this, was determined for K3A gp32 using three different pools of poly(A) obtained from a fractionation of commercially available poly(A) on a Sephacryl S-200 column. A broad peak of absorptivity was obtained. Pool I was obtained from the very tip of the peak (first 5% of the fractions containing absorbance), pool II represented the next Ϸ15% of the peak, while pool III represented approximately the next 25% of the peak. values resolved from reverse titrations in 0.20 M NaF using each of these pools are shown in Fig. 2 as the data points labeled I, II, and III. As can be seen, log for the longest poly(A) fraction (pool I) is at the high end of the range obtained for K3A gp32 ( Ϸ 1500), while pool III underestimates the true cooperativity of binding ( Ϸ 350). To minimize the impact of polydisperse and finitelength poly(A), the quantitative analysis of all reverse titrations was limited to data points obtained at a fractional satu- ration of the nucleic acid by gp32 to 10% or less ( Յ 0.10 or Յ 0.013). gp32 cluster size simulations (27) reveal that, at this extent of cooperativity and fractional saturation, it is unlikely that greater than 5-10% of the gp32 clusters would exceed the length of the lattice (simulations not shown). All of the experiments reported here were carried out with a poly(A) pool identical or comparable to pool II.
With a highly cooperatively binding ligand, reverse fluorescence quenching titrations can only be used to estimate K app and accurately resolve K int from over a fairly narrow range of K app (Ϸ10 6 to 10 7 M Ϫ1 ) (27). Depending on the salt dependence of K app , this can also severely limit to what degree the solution conditions can be varied to probe the effect of solution variables on the interaction. To extend the usable range in solution conditions, in particular, salt concentration, which can be used to measure the binding properties of K3A gp32, independent determinations of K int and were made for wild-type and K3A gp32s from "forward" titrations of poly(A) with gp32 using circular dichroism (CD) spectroscopy (28). Fig. 3 shows a set of representative forward titrations carried out at various [NaCl] from 0.30 to 0.375 M. Due again to the fact that the McGhee-von Hippel model assumes an infinite-length nucleic acid, we confined the quantitative analysis to data points where Յ 0.40 as a compromise value. Binding parameters found for the wildtype and K3A gp32 compared to other Arg 4 substitution mutants (17) as well as the [NaCl] used to obtain parameters are shown in Table I. The determination of for the wild-type gp32 is consistent with previous studies carried out over a wide range of temperature (20 -42°C) and [NaCl] in 10 mM HEPES, pH 7.7 (29). In comparison with values found from the reverse titrations (Figs. 1 and 2), these values are on the low side of the range reported, presumably again due to lattice end effects. However, these findings provide additional evidence that values for the wild-type gp32, K3A gp32 and other B-domain mutants are more similar than different.
The Effect of Salt Concentration and Type on the Binding of K3A and N-terminal Domain gp32 Mutants to Poly(A)-Previ-ous studies have shown that the overall binding affinity (K app ϭ K int ϫ ) of wild-type gp32 for ss nucleic acids is strongly dependent on the salt concentration and type (18). Since the magnitude of is independent of salt concentration (Fig. 2), this requires that the K int term contain most or all of the salt dependence of K app (18). According to Record's polyelectrolyte theory (30,31), the change in K app as a function of salt concentration, Ѩ log K app /Ѩ log [M ϩ X Ϫ ], reflects (neglecting the stoichiometry of water release) the net uptake or release of cations (a) and anions (c) from the free protein and nucleic acid as a result of complex formation according to the following relationship, where K X is the average anion association constant and [X Ϫ ] is the total anion concentration.  (18). According to this model, these data suggest that as many as three net anions may be released from gp32 as a result of cooperative binding to poly(A). In addition, the absolute magnitude of K app at 0.40 M NaX is Ϸ20-fold larger in NaF versus NaCl (cf. Fig. 4 below), as previously observed (18). Reverse titrations with the weaker binding mutant K3A gp32 were then used to obtain K app at [NaX] Յ 0.25 M. Using only data obtained from such titrations, we showed previously that the [NaCl] dependence of K app for R4T and R4Q gp32s was each markedly more shallow than that obtained for wild-type  . This has in fact been shown to be the case for basic oligopeptides (34). However, in some instances, including this one (Fig. 4) We have used Equation 11 in an attempt to simulate the dependence of K app on [MX] using both the NaCl (Fig. 4A) and NaF (Fig. 4B) (32,33) and starting with n tot ϭ 4 (18) a nonlinear least squares fit to the wild-type gp32 data gives m tot Ϸ 5, ⌬q Ϸ 6, and log K 1M ϭ 2.4. Since the values of n tot and ⌬q are correlated, their absolute magnitudes cannot  (17). The data at 0.05 M NaCl (log [NaCl] ϭ Ϫ1.3) were obtained from duplicate reverse titrations for wild-type, K3A, and R4T gp32s. B, [NaF] dependence of K app for wild-type (E), K3A (q), and R4T (J) gp32s. be precisely determined. However, knowing that n tot ϭ 3-4 (Ref. 18 and this study), an (n tot and ⌬q) Ϸ 10 and an m tot value in the range of 4 -5 is required to adequately fit all of the data, particularly at [NaCl]Ͻ0.20 M. We then asked if the complete [NaCl] dependence of K app for K3A and R4T gp32s could be modeled with the same coefficients of participation of ions for the wild-type protein (m tot , n tot , and ⌬q equal to 5, 4, and 6, respectively) with the magnitude of log K 1M simply reduced. An acceptable fit to the data is obtained requiring, only in the case of K3A gp32, that the magnitude of K d Mϩ or K d XϪ be allowed to vary by a small amount (less than 50%). For K3A and R4T gp32s, log K 1M was found to be 1.6 and 0.8, respectively. The fits are shown superimposed on the experimental data in Fig. 4A. 4 Next, the [NaF] dependence of K app was analyzed with ion stoichiometries fixed at their values obtained above with only the magnitude of K d XϪ made much larger (K d XϪ ϭ 1 or 10 M) (32), consistent with the fact that F Ϫ binds weakly to proteins. An excellent fit to the data is obtained with the observed degree of curvature in the log-log plot, with the magnitude of log K 1M again significantly different among the three proteins. Log K 1M values of 4.8, 3.9, and 3.1 for wild-type, K3A, and R4T gp32s were obtained. As required by the model, this systematic decrease in log K 1M in NaF as a result of substitution of the B-domain residues exactly parallels that which is found in NaCl solutions. This analysis suggests that it is unnecessary to implicate differential ion uptake and release (at 0.20 M NaCl, 20°C) as the underlying origin of the Ϸ6 -8-fold and 30 -50-fold decrease in K app determined for K3A and R4T gp32s, respectively.

DISCUSSION
In this paper, we present a physicochemical characterization of an N-terminal domain mutant of gp32, K3A gp32, and place it in the context of a set of Arg 4 substitution mutants described previously (17). We show that like Arg 4 mutants, K3A gp32 exhibits a high cooperativity of binding to the homopolymer poly(A) at equilibrium despite exhibiting an approximately order of magnitude decrease in overall binding affinity, K app . This binding behavior is fully consistent with the behavior of other B-domain mutants (17). It is, however, contrary to expectations from the simple models of cooperativity of binding by gp32, where substitutions introduced into the B-domain would be expected to affect only the magnitude of the cooperativity parameter . Interestingly, the quantitative binding properties of all N-terminal gp32 mutants bear a qualitative resemblance to substitution mutants of bacteriophage fd (M13) SSB, gene V protein (gVp), which strongly disrupt the dimer-dimer interface (e.g. residues Glu 40 , Tyr 41 , and Arg 82 ) (35)(36)(37)(38). These residues in gVp have been modeled as directly participating in the cooperativity of gVp binding; however, substitution of these residues gives rise to a significant reduction in the magnitudes of both the K int and terms resolved on poly(dA) (37). Clearly, the origin of the cooperative binding energy must derive from multiple regions of the gVp polypeptide chain. This may also be the case for gp32. Inspection of the crystal structure of a gp32 core protein-oligonucleotide complex (8) reveals that N-terminal domain may be ideally positioned to stabilize both proteinnucleic acid and protein-protein interfaces, thereby contributing to the magnitude of both K int and terms. Since a simple reduction in the magnitude of the cooperativity parameter is not observed, this suggests residues Lys 3 and Arg 4 do not simply independently engage in a specific set of protein-protein interactions in the cooperatively bound gp32 cluster like that envisioned for the C-terminal tail of adenovirus single-strand binding protein (39).
The detailed binding properties of K3A gp32 strongly mirror those of R4Q gp32 in every respect (17). Both proteins exhibit a modest 6 -8-fold decrement in overall apparent binding affinity, yet retain a cooperativity of binding () that is not readily distinguished from that of the wild-type protein. The same nonlinear [NaCl] dependence of K app is observed for both proteins. Given the widely disparate nature of the K3A and R4Q substitutions, these findings strongly suggest that the N-terminal B-domain substitutions of gp32 reflect a perturbation of the wild-type protein rather than reporting on new or novel properties acquired by the mutant proteins. For this reason, although it is unknown to what extent these findings can be extrapolated to Lys 5 or other residues within the Nterminal domain, we anticipate this to be case. This has prompted us to use K3A and R4Q gp32s as tools to probe the role of the N-terminal domain in gp32 function using stopped flow fluorescence techniques to monitor the kinetics of association and dissociation of gp32 from poly(A) under solution conditions described here 2 as well as biochemical reconstitution of multiprotein recombination and replication complexes. 3 The apparent binding affinity, K app , of wild-type gp32 (18) as well as many sequence-specific and nonspecific gene regulatory proteins is strongly dependent on the salt concentration, [MX], and type. In Cl Ϫ -containing buffers, K app has been shown to contain significant contributions from both cation and anion release upon complex formation (18) (Fig. 4). It was therefore of interest to examine the salt dependence of K app of K3A gp32 and other B-domain mutant proteins to determine if the decrease in K app (at 0.20 M NaCl, 20°C) originates with differential ion uptake and release in any case. Analysis of the complete [NaX] dependence of K app of wild-type and B-domain mutant gp32s with a simple ion-exchange model (30,33) qualitatively reveals that large changes in the stoichiometry of the participation of anions and cations in the binding equilibria cannot account for the significantly decreased binding affinity and [NaCl] dependence of K app reported previously for B-domain mutant gp32s (17). Stated another way, approximately the same extent of diminution in K app determined for N-terminal mutants appears to persist over a relatively wide range of salt concentration. This suggests that substitution of one of the positively charged amino acids within the N-terminal B-domain does not result in a large perturbation in the participation of ions in the binding reaction. The significant decrease in the slope of the Ѩ log K app /Ѩ log [NaCl] plot at [NaCl]Ͻ0.20 -0.25 M implicates the significant cation uptake as a result of gp32 binding to poly(A) (33). Additional studies to far lower [NaCl] (e.g. Յ0.01 M) are required to quantitatively assess the extent of cation uptake by gp32 and mutants (33). A significant cation uptake term associated with gp32-poly(A) binding would not be unexpected, since E. coli SSB has been shown to bind to poly(A) with significant uptake of cations (32). The significant effect of anions on the wild-type gp32-poly(A) equilibria at [Na ϩ ]Ͼ0.20 M is closely similar to that observed for the E. coli SSB-poly(U) interactions (25,32), E. coli lac repressor (40), and E. coli RNA polymerase (41), although in the latter two cases, the effect of the anion is largely confined to the magnitude of K app rather than on the magnitudes of both K app and Ѩ log K app /Ѩ log [NaX].
The nonlinearity in the Ѩ log K app /Ѩ log [NaCl] plots for wild-type gp32 that we observe here is consistent with previous kinetic estimates of K int derived from stopped flow association 4 These fits are simulations only, which depend strongly upon the input magnitudes of the fixed parameters. This is the case because unique values of ⌬n and ⌬q cannot be resolved with this analysis (33) and the data sets are somewhat underdetermined. With these considerations, the limits on m tot , ⌬q, and log K 1M are at Ϯ 1.0, Ϯ 0.5 and Ϯ 0.3, respectively. kinetic experiments carried out on poly(A) at 0.10 and 0.20 M NaCl under solution conditions similar to those used here (42). These studies show that the value of K int at 0.10 M NaCl falls significantly below the linearly extrapolated estimate determined at [NaCl] Ն 0.20 M. 5 In addition, kinetic K int values determined for poly(U) (to which wild-type gp32 binds with a similar affinity as poly(A)) at 0.10 M NaF versus 0.10 M NaCl were shown to be identical within experimental error, in contrast to determinations made at [NaX] Ͼ Ͼ 0.20 M (42). This predicts that K app for the wild-type protein would strongly deviate from linearity in this salt concentration region and become relatively insensitive to the nature of the anion, as is directly observed or extrapolated (Fig. 4).
Other more complex explanations beyond that of a simple protein-ion exchange model applied here can give rise to significant curvature or nonlinearity in log-log plots, e.g. a protein conformational change or a change in the mechanism of binding at low salt. With regard to the latter, the entire [NaCl] dependence of K app has been shown to reside in the rate of dissociation induced by [NaCl] jumps, k d(app) , at [NaCl] Ն 0.20 M, i.e. Ѩ log K app /Ѩ log [NaCl] ϭ Ѩ log k d(app) /Ѩ log [NaCl] ϭ Ϫ7.0 (Ϯ 1.0) (43,44). Since the salt dependence of the bimolecular association rate constant for the formation of a non-cooperatively bound gp32 monomer is essentially zero (42), this is consistent with a gp32 monomer dissociating directly from the ends of cooperatively bound clusters as a result of a jump in [NaCl]. We have recently extended these observations to K3A gp32 and other B-domain mutants studied here. 2 Strikingly, at [NaCl]Յ0.20 M NaCl, Ѩ log k d(app) /Ѩ log [NaCl] for wild-type gp32 on poly(A) was shown to fall to Ϫ2.3 (Ϯ 0.3); in NaFcontaining buffers, the magnitude becomes only slightly smaller and the [NaF] dependence of k d(app) , Ѩ log k d(app) /Ѩ log [NaF] drops slightly to Ϫ1.9 (Ϯ 0.2) (42). This trend in the relatively insensitivity of the magnitude and [NaX] dependence of k d(app) on the nature of anion observed for wild-type protein gp32 at low salt concentration mirrors exactly what we observe in the magnitude and [NaX] dependence of K app for B-domain mutant proteins. Lohman previously postulated that gp32 undergoes a change in mechanism of dissociation at [NaX] Յ 0.20 M, where instead of dissociating directly from the ends of clusters, significant populations of gp32 monomers first slide or translocate to a non-contiguous site lattice position and then dissociate into solution (44).
In summary, examination of the salt dependence of the binding of K3A gp32 and other B-domain mutants to poly(A) over a much wider range of [NaX] accessible in the previous study (17) suggests that substitutions within the Lys 3 -Arg 4 -Lys 5 sequence significantly reduce the magnitude of K int as well as , but in the absence of a strong perturbation in the stoichiometry of cation and anion release associated with complex formation. This leads to approximately the same relative degree of diminution in K app for poly(A) over a relatively wide range of salt concentration. If, as seems likely (16), these findings can be extrapolated to other nucleic acid ligands, studies which seek to examine to what degree modestly defective B-domain mutants of gp32 support replication and recombination processes in vitro and in vivo will be greatly facilitated since these will necessarily be done under different solution conditions. These studies are currently in progress. 3