Metal binding to DNA polymerase I, its large fragment, and two 3',5'-exonuclease mutants of the large fragment.

DNA polymerase I (Pol I) is an enzyme of DNA replication and repair containing three active sites, each requiring divalent metal ions such as Mg2+ or Mn2+ for activity. As determined by EPR and by 1/T1 measurements of water protons, whole Pol I binds Mn2+ at one tight site (KD = 2.5 microM) and approximately 20 weak sites (KD = 600 microM). All bound metal ions retain one or more water ligands as reflected in enhanced paramagnetic effects of Mn2+ on 1/T1 of water protons. The cloned large fragment of Pol I, which lacks the 5',3'-exonuclease domain, retains the tight metal binding site with little or no change in its affinity for Mn2+, but has lost approximately 12 weak sites (n = 8, KD = 1000 microM). The presence of stoichiometric TMP creates a second tight Mn2+ binding site or tightens a weak site 100-fold. dGTP together with TMP creates a third tight Mn2+ binding site or tightens a weak site 166-fold. The D424A (the Asp424 to Ala) 3',5'-exonuclease deficient mutant of the large fragment retains a weakened tight site (KD = 68 microM) and has lost one weak site (n = 7, KD = 3500 microM) in comparison with the wild-type large fragment, and no effect of TMP on metal binding is detected. The D355A, E357A (the Asp355 to Ala, Glu357 to Ala double mutant of the large fragment of Pol I) 3',5'-exonuclease-deficient double mutant has lost the tight metal binding site and four weak metal binding sites. The binding of dGTP to the polymerase active site of the D355A,E357A double mutant creates one tight Mn2+ binding site with a dissociation constant (KD = 3.6 microM), comparable with that found on the wild-type enzyme, which retains one fast exchanging water ligand. Mg2+ competes at this site with a KD of 100 microM. It is concluded that the single tightly bound Mn2+ on Pol I and a weakly bound Mn2+ which is tightened 100-fold by TMP are at the 3',5'-exonuclease active site and are essential for 3',5'-exonuclease activity, but not for polymerase activity. Additional weak Mn2+ binding sites are detected on the 3',5'-exonuclease domain, which may be activating, and on the polymerase domain, which may be inhibitory. The essential divalent metal activator of the polymerase reaction requires the presence of the dNTP substrate for tight metal binding indicating that the bound substrate coordinates the metal.(ABSTRACT TRUNCATED AT 400 WORDS)

DNA polymerase I (Pol I) is an enzyme of DNA replication and repair containing three active sites, each requiring divalent metal ions such as Mg2+ or Mn2+ for activity.
As determined by EPR and by l/T1 measurements of water protons, whole Pol I binds Mn2+ at one tight site (Ko = 2.5 PM) and -20 weak sites (& = 600 MM). All bound metal ions retain one or more water ligands as reflected in enhanced paramagnetic effects of Mn2+ on 1/T, of water protons.
The cloned large fragment of Pol I, which lacks the 5',3'-exonuclease domain, retains the tight metal binding site with Iittle or no change in its affinity for Mn'*, but has lost -12 weak sites (n = 8, KB = 1000 PM). The presence of stoichiometric TMP creates a second tight Mn2+ binding site or tightens a weak site loo-fold. dGTP together with TMP creates a third tight Mn2+ binding site or tightens a weak site 166-fold. The D424A (the Asp'= to Ala) 3',5'-exonuclease deficient mutant of the large fragment retains a weakened tight site (& = 68 pM) and has lost one weak site (n = 7, & = 3500 j.tM) in comparison with the wild-type large fragment, and no effect of TMP on metal binding is detected. The D355A, E357A (the Asp356 to Ala, Glu"' to Ala double mutant of the large fragment of Pol I) 3',5'-exonuclease-deficient double mutant has lost the tight metal binding site and four weak metal binding sites. The binding of dGTP to the polymerase active site of the D355A,E357A double mutant creates one tight Mn2+ binding site with a dissociation constant (KD = 3.6 PM), comparable with that found on the wild-type enzyme, which retains one fast exchanging water ligand.
Mg" competes at this site with a KD of 100 pM. It is concluded that the single tightly bound Mn2+ on Pol I and a weakly bound Mn2+ which is tightened loo-fold by TMP are at the 3',5'-exonuclease active site and are essential for 3',5'-exonuclease activity, but not for polymerase activity.
Additional weak Mn2+ binding sites are detected on the 3',5'-exonuclease domain, which may be activating, and on the polymerase domain, which may be inhibitory.  (Brutlag et al., 1969;Klenow and Overgaard-Hansen, 1970;Ollis et al., 1985;Freemont et al., 1986;Derbyshire et al., 1988). NMR studies of the conformations of substrates and templates bound at the polymerase site Mildvan, 1985,1986;Mullen and Mildvan, 1988) and kinetic studies of DNA polymerization (Kuchta et al., 1987(Kuchta et al., ,1988 have provided insight into the reaction mechanism of high fidelity DNA synthesis by this enzyme. X-ray crystallographic studies of complexes containing DNA bound at the 3',5'-exonuclease site of the large (Klenow) fragment of Pol I have provided structural insight into the mechanism of the order of magnitude increase in fidelity produced by the 3',5'-exonuclease . Each enzymatic activity of Pol I requires divalent metal ions such as Mg2+ or Mn'+. Slater et al., (1972) established that Pol I has multiple classes of divalent metal ion binding sites. Employing x-ray crystallography and genetic modification, Ollis et al., (1985) and Derbyshire et al., (1988) detected two adjacent divalent cations at the 3',5'-exonuclease active site in an enzyme-dNMP complex. The 3',5'-exonuclease activity was profoundly decreased by the D424A single mutation of a ligand for one of these metals. This activity was also greatly decreased by the D355A,E357A double mutation of two ligands for the other metal, one of which, D355, is shared by both metals. The single mutation resulted in undetectable metal binding at one of the two sites, and the double mutation abolished metal binding at both sites. These mutations had no effect on polymerase activity (Derbyshire et al., 1988). The cloning and overproduction of these mutants (Derbyshire et al., 1988) and of the wild-type large fragment (Joyce and Grindley, 1983) Jovin et al. (1969) as modified previously by Slater et al. (1972). Additionally, gel filtration was carried out on Sephadex G-100 to remove degraded fragments. In this gel filtration step, 10 mg of purified Pol I was loaded onto a 1 x 100~cm column and eluted using upward flow with 100 mM Tris-HCl, pH 7.5, containing 1 mM dithiothreitol.
Purification of the large fragment of Pol I from the E. coli strain CJ155 was performed as described previously (Joyce and Grindley, 1983). The D424A and D355A,E357A enzymes were similarly purified from the appropriate genetically engineered mutants of E. coli (Derbyshire et al., 1988). The enzymes were stored in either 50 mM K+ phosphate containing 0.5 mM dithiothreitol or 10 mM K+Pipes containing 1 mM dithiothreitol, pH 7.0, with 50% v/v glycerol at -20 "C or as an 85% ammonium sulfate suspension at 0 "C.
All enzymes were further purified free from divalent metal ions by elution from a G-25 Sephadex column at 4 "C, using 10 mM Tris-HCl, pH 7.5, and 32 mM KC1 as elution buffer. Before loading the enzymes the column was thoroughly washed with 20 mM EDTA, 10 mM Tris-HCl, pH 7.5, and 32 mM KC1 and equilibrated with this buffer in the absence of EDTA. Buffers were passed through K+-Chelex-100 before use.
Protein concentrations were determined spectrophotometrically using A% = 8.5 for whole Pol I (Jovin et al., 1969) and A&% = 9.3 for the wild-type and mutant forms of the large fragment (Setlow et al., 1972) assuming molecular weights of 103,000 for whole Pol I and 68,000 for the large fragment (Joyce et al., 1982).

Methods
Enzyme Assay-The enzymes were assayed by the method of Setlow (1974) using poly(dA-dT) as template-primer. The polymerase activity of whole Pol I was 5300 units/mg. The polymerase activities of the large fragment, the D424A mutant, and the D355A,E357A mutant were 12,000 f 2000 units/mg. One unit of activity is defined as 10 nmol of total nucleotides polymerized in 30 min (Setlow, 1974;Richardson et al., 1964). The specific activities of these enzymes were unaffected by the subsequent metal binding experiments.
Water Proton Relaxation Rate Measurements-The longitudinal relaxation rate of water protons was measured at 16 "C with a Seimco pulsed NMR spectrometer at 24.3 MHz by using a 180 "-s-90 ' pulse sequence as described previously Cohn, 1970, Mildvan andEngle, 1972). The observed enhancement of the relaxation rate is defined as t* = (l/T,r*)/(l/Tip), where l/Tip is the paramagnetic contribution to the relaxation rate in the presence (*) and absence of the enzyme (Mildvan and Cohn, 1970).
The number of fast exchanging waters coordinated to Mn2+ at 24 "C were determined by measuring the frequency dependence of the water proton relaxation rate as described above at 15.0, 24.3, 30.0, 40.0, and 59.8 MHz on a Seimco pulsed NMR spectrometer equipped with a variable frequency probe. The coordination number for fast exchanging water ligands was determined as previously described (Mildvan and Gupta, 1978).
Binding Studies-The concentration of free Mn'+, Mm, in each sample containing enzyme was determined by measuring the electron paramagnetic resonance signal intensity due to free Mn*+ in the absence and presence of enzyme, using a Varian E-4 EPR spectrometer (Cohn and Townsend, 1954;Mildvan and Engle, 1972). For the proton relaxation rate (PRR) experiments, determination of concentrations of free and bound Mn*+ required the evaluation of tt, the enhancement factor of enzyme-bound Mn2+ for the tight and weak metal binding sites. The eb at each concentration was calculated using the equation tb = [c* -MnJMn,]/[l-MnJMn,] (Mildvan and Engle, 1972). The ratio of free Mr?+ to total Mn*+, Mm/Mm, was determined from the EPR data at each Mn*+ concentration, The tb values were averaged for each type of site, and the Mnr and Mnb concentrations were calculated from c* as described previously (Mildvan and Engle, 1972). Whenever possible data were collected between 20 and 80% occupancy of sites since the statistical variation minimizes in this region, and Scatchard analysis was carried out since this represents the best linear method for analyzing binding data (Deranleau, 1969a The binding study was performed at 16 "C in 10 mM Tris-HCl, pH 7.3 + 0.2, 32 mM KC1 at an enzyme concentration of 10 @M. The dissociation constants (Ko) and stoichiometries (n), which were used to calculate the theoretical curve, are given in Table I . 1969b). In the Scatchard analyses of data in which there were multiple noninteracting sites, Mnr was calculated from the cubic Equation 1 derived by assuming the simultaneous operation of both equilibria (Miziorko and Mildvan, 1974).
In Equation 1 K, and K2 represent the dissociation constants, and Ci and C2 represent the concentrations of tight and weak sites respectively. Alternatively the data were fit using a graphical method which yielded identical results (Rosenthal, 1967). The number of binding sites (n) and dissociation constants (Kn) were chosen to minimize the deviation of the calculated curve from the data points. This analysis also yielded upper limit estimates of the errors in n and KD values. An independent statistical analysis of the random errors in n and Ka expressed as f2 standard errors of the mean was made, based on the deviations of the individual data points from the best-fit calculated curve. The two S.E. uncertainty of the 1: intercept yielded the 95% confidence limits in n, and the two SE. uncertainties in both the x and y intercepts yielded the 95% confidence limits in KD. The statistical errors were generally smaller than those estimated by the curve fitting procedure. In all cases, the larger of the two errors is reported.

AND DISCUSSION
Mn2' Binding to Whole Pol I and the Large Fragment of Pol I-Mn" binding to whole Pol I and to its large fragment was studied by the EPR method which measures free Mn2+ in a mixture of free and bound Mn'+. A Scatchard plot of Mn2+ binding to whole Pol I based on EPR data ( Fig. 1) may be tit most simply by assuming that the complete enzyme binds a single Mn2+ ion tightly (Ko = 2.5 f 1.1 pM) and 20 f 5 Mn2+ ions more weakly (Ko = 600 -+ 300 PM) (Table l).' The presence of Mn'+-binding sites of intermediate affinity as described previously @later et aZ., 1972), while not excluded, are not necessary to improve the fit to the EPR data.
Genetic deletion of the 5',3'-exonuclease domain and overexpression of the gene (Joyce and Grindley, 1983) have yielded the large fragment of Pol I (M, = 66,000), which corresponds to the large proteolytic fragment (Klenow fragment) of the enzyme (Brutlag et al., 1969, Klenow andOvergaard-Hansen, 1970). The large fragment of Pol I retains the tight Mn*+ binding site, as found in whole Pol I, with little or no change * The slightly reduced stoichiometry for Mn2+ binding at the tight site may be due to partial occupancy of this site by trace amounts of Zn*+ which is known to bind tightly at this site (Ferrin et al., 1983;Ollis et al., 1985). in the affinity of this site for Mn2+ (Table I) as determined by Scatchard analysis of both EPR and PRR data (Fig. 2, A  and B). Therefore, the tight Mn" binding site is not located on the 5',3'-exonuclease domain in Pol I. Differences between whole Pol I and the large fragment are found in a significantly reduced number of weak divalent cation binding sites on the large fragment (n = 8.0 f 1.0) ( Table I), suggesting that they exist on the 5',3'-exonuclease domain.
While the EPR method measures free Mn*+, the PRR method detects enzyme-bound Mn2+ by its enhanced paramagnetic effects on l/T, of water protons. The large enhancement factors (tb >> 1) for Mn2+ bound at both the tight (e, = 13.2) and weak sites (tb = 5.7) indicate that one or more fastexchanging water ligands remain coordinated to Mn", i.e. the bound metals are not buried in the protein.
Effect of TMP and dGTP on the Mn2+ Binding Properties of the Large Fragment of Pol I-In the presence of a stoichiometric amount of TMP, which binds at the exonuclease site (Derbyshire et al., 1988), the large fragment of Pol I binds two Mn2+ ions tightly (ED = 9.7 f 2 pM) and 7.0 f 1.0 weakly (En = 1200 f 600 FM) (Fig. 2, C and D, Table I). Hence the presence of TMP has created an additional tight Mn"-binding site on the enzyme probably by tightening one of the weak sites. The high affinity for Mn" in the presence of TMP and a systematic decrease in the enhancement factor at the tight sites with increasing Mn2+ occupancy (tb = 11.0-7.8) (Table  I) establish the formation of a quaternary E-Mnz' TMP complex, presumably at the 3',5'-exonuclease site. The systematic decrease in tb with increasing occupancy of Mn'+binding sites may be due to magnetic dipolar interaction between two Mn2+ ions bound near each other at the 3',5'exonuclease site (Leigh, 1970;Steitz et al., 1987). Within experimental error, no positive cooperativity in metal binding is detected in the I& values. However the weak binding of Mn*+ at one of these sites in the absence of TMP may be due to negative cooperativity between them.
In the presence of stoichiometric amounts of both TMP and dGTP, a deoxynucleoside triphosphate substrate which binds tightly at the polymerase active site in the absence of metal ions (En = 2.9 MM) (Mullen et al., 1989), three tight Mn2+-binding sites on the enzyme are detected (I& = 7.2 + 2 pM) (Fig. 2, E and F, Table I). The third Mn"-binding site created by the presence of dGTP is clearly on enzyme-bound dGTP rather than on free dGTP since its enhancement factor, tT = 10.5 f 2.7, greatly exceeds the value of 2.2 found for binary Mn*+ .dNTP complexes (Slater et al., 1972). The eb value for E-dGTP-Mn2+ was determined from the average enhancement factor for the three tight sites (9.7 + 0.9) by factoring out the average enhancement factors of the two other tightly bound Mn2+ ions independently measured in the E-Mn$+ TMP complex. This value for tb will be confirmed in a later section by genetic deletion of the other two tight sites.

Mn"
Binding to the D424A 3',5'-Exonuclease-deficient Mutant of the Large Fragment of Pol I-With TMP bound at the 3',5'-exonuclease site, two metals were detected at this active site by a combination of genetic and x-ray crystallographic studies (Ollis et al., 1985;Steitz et al., 1987;Derbyshire et al., 1988) (Fig. 3). One of these metals (metal A) was coordinated by the 5'-phosphate of TMP and the carboxylates of ASPIRE, G1u357, and Asp501, whereas the other metal (metal B) was also coordinated by the carboxylate of Asp355 and formed a second-sphere complex with the carboxylate of Asp424 and the phosphate of TMP (Fig. 3). In the absence of TMP, metal A remained bound to the wild-type crystalline enzyme, but metal B was not detected in the x-ray structure, suggesting that metal B was bound more weakly than metal A. The exonuclease-deficient mutant D424A also retained metal A but not metal B under the conditions of the crystallographic experiment, i.e. in the presence of TMP and 3 M M-IdzS0~.
We have studied Mn2+ binding to the D424A mutant in the absence of nucleotides or metal-liganding salts (Fig. 4, A and  B). Scatchard analysis of the binding data and PRR enhancements reveal the loss of approximately one weak site in comparison with the native large fragment and an order of magnitude weakening of the tight metal binding site (Table  I   I   Hence the weaker binding of Mn2+ at this site in the D424A mutant could have resulted from loss of the electrostatic effect of the nearby Asp4'". Unlike the wild-type large fragment, the D424A mutant showed no tightening of a weak metal-binding site in the presence of stoichiometric amounts of TMP (Fig. 4, C and D). Stated another way, in the presence of TMP, the D424A mutation caused the loss of one of the two tight Mn*+-binding sites, and a weakening of the other one on the large fragment of Pol I ( Table I).
The average enhancement factor of the residual weakly bound Mn2+ in the binary complex of the D424A mutant (6 = 9.7 f 2.1) was greater than that obtained with the wildtype enzyme (tb = 5.7 f 0.6). This increase in the average tb value cannot be explained solely by the loss of a site with low enhancement since the tb value of the lost site would have to be negative, a physical impossibility.
Hence the observed increase in the average th for weakly bound Mnz+ remaining in the D424A mutant must result at least in part from the loss of dipolar interaction with Mn2' at site B which no longer exists (Fig. 3). Such dipolar interaction, which decreases cb due to a shorter electron spin relaxation time of Mn2+ (Leigh, 1970;Gupta, 1977;Mildvan and Gupta, 1978) provides independent evidence for the loss of a weak binding site and requires that one or more of the weakly bound Mn2' ions in 14331 wild-type enzyme for the weak Mn2'-binding site. The D424A mutation abolishes Mn*' binding at the weak site and weakens Mn2+ binding at the tight site, likely due to proximity of the two sites. The tight Mn2+-binding site is probably site A, and the weak Mn*+-binding site is probably site B detected in the crystal structure (Fig. 3). Our ability to detect metal occupancy at site B even in the absence of dNMP may well be due to the lower ionic strength of the present Mn2+-binding experiments.
Metal Binding to the D355A,E357A 3',5'-Exonuclease-deficient Mutant of the Large Fragment-Further evidence that the tight Mn2+-binding site, which is weakened by the D424A mutation, is indeed at the 3',5'-exonuclease site was obtained by solution studies of Mn2+ binding to the D355A,E357A double mutant. EPR studies of solutions containing the double mutant (58-190 pM) and MnC12 (11 pM to 4.6 mM) did not detect tight binding of Mn2+ to the enzyme, but only weak binding sites, Since EPR detects free Mn2+, and since most of the Mn2+ was free in these studies, the EPR data were not suitable for analysis to determine the n and & values of the residual weak sites. The PRR data, which detects the enhanced effect of bound Mn2+, was more sensitive and could be analyzed by a Scatchard plot (Fig. 5) to yield both n and K. values (Table I).
the wild-type enzyme is near site B, the weak site which was lost in the D424A mutant.
These results, obtained in solution, are consistent with the x-ray studies of the crystalline enzyme and provide additional quantitative information.
Thus, we conclude that two Mn" ions bind at the 3',5'-exonuclease site on the native enzyme in the absence of dNMP. One Mn2' ion binds tightly and one binds weakly. The presence of TMP raises the affinity of the FIG. 3. The 3',5'-exonuclease active site as determined by x-ray crystallography (Ollis et al., 1985;Steitz et al., 198'7;Joyce and Steitz, 1987).
The two divalent metal-binding sites are labeled A at the tight site and B at the weak site. The figure is modified from Steitz et al. (1987)  the analysis of the PRR data revealed that the tight Mn*+-binding site was lost in the double mutant. In addition, 4 +: 1 weak sites were also lost in the double mutant, in comparison with the wildtype large fragment (Table I). Hence the mutated residues Asp3= and G1u357 contribute significantly to the binding of Mn2' at the tight site, presumably by direct coordination, and at 4 f 1 weak sites either directly or indirectly.
One of the weak sites which is lost in the D355A,E357A double mutant is probably site B detected in the crystal structure, since one Conditions are described in Fig. 1. Enzyme concentrations of 165, 217, and 239 pM (A and B) and 42 pM (C and D) were used. The concentration of TMP was 42 @M (C and D). Parameters used to calculate the theoretical curves are given in Table I. Conditions are described in Fig. 1. Enzyme concentrations of 58-190 pM were used. Parameters used to calculate the theoretical curve are given in Table I. of the mutated residues, Asp355, is also a ligand for Mn2+ bound at site B (Fig. 3). These studies in solution thus confirm and extend the x-ray data which show the loss of two metal binding sites in the double mutant (Derbyshire et al., 1966).
In the double mutant, the average enhancement factor tb for the residual weakly bound Mn*+ (tb = 15.7 f 1.6) was significantly greater than the corresponding & value of the wild-type enzyme (5.7 * 0.6) ( Table I). As in the case of the D424A mutant, this increase in cb cannot be due solely to the selective loss of Mn*+ sites with low es, since these eb values would have to be negative, but requires an increase in tb of one or more of the weakly bound Mn2+ ions remaining in the D355A,E35'7A mutant. As discussed for the D424A mutant where a similar effect was observed (Table I), such an increase in tb probably results from the loss of dipolar interactions, which were present in the wild-type enzyme, between weakly bound Mn2+ ions remaining in the double mutant and Mn2+ bound at sites B, A, or at one or more of the other lost sites.

Metal
Binding at the Polymerase Active Site-Using the D355A,E357A double mutant which lacks the tight metalbinding site, we have investigated Mn2+ binding at the polymerase active site in the presence of the substrate, dGTP. A Scatchard analysis of the EPR data for Mn2+ binding to the enzyme-dGTP complex indicates the appearance of one tight metal binding site (& = 3.6 pM) in addition to the 4 ? 1 weak metal binding sites which were detected in the absence of the nucleotide (Fig. 6A, Table I). Because of the high concentrations of enzyme and dGTP required for >90% complexation, the levels of free Mn2+ were too low for detection until saturating levels of Mn2+ were approached, resulting in a larger error (& 50%) in the KD of Mn2+ at the tight site. In a competition experiment monitored by PRR, Me displaced Mn2+ from the tight polymerase site created by dGTP with a Ko of 100 f 20 pM (Fig. 6B).
The eb value for the tightly bound Mn2+ (10.8 f 0.9) is much greater than that found for a binary Mn2+. dNTP complex (tt, = 2.2 f 0.1) (Slater et al., 1972) establishing the existence of a ternary enzyme-dGTP-Mn2+ complex in which the binding of dGTP to the enzyme has created a tight Mn2+-binding site. This increase in the enhancement factor is a result of the increased correlation time of Mn2+ due to binding to the enzyme-dGTP complex and is in quantitative agreement with the average value (10.5 2 2.7) calculated above for the wildtype enzyme. The dissociation constant of Mn2+ from the ternary complex (KB = 3.6 pM) is 2.9-fold tighter than the KLI value of a binary Mn2+.dNTP complex under similar conditions @later et al., 1972), indicating that the enzyme has raised the affinity of bound dGTP for Mn2+, possibly by contributing a ligand to the metal. Since dGTP would be I  I  I  I  I  I  I  I  I  I were 86 pM. Other conditions are described in Fig. 2 (Mildvan and Gupta, 1978;Mildvan et al., 1980): l/fTIp = q(C/#f (Tc), f(TJ = 37,/(1 + WIT,*) + 7r,/(l + I&~~) and l/re -l/rs = B[r,/(l + wsW) + 47,/(1 + 4~s*r~*)],where q is the number of fast-exchanging water ligands, C is a product of physical constants, equal to 812 A/s'~ for Mn'+-proton interactions, r is the metal nucleus distance, 7, is the dipolar correlation time, 7s is the longitudinal electron spin-relaxation time, B is the zero-field splitting parameter, and 7, is a time constant for motion of the water ligands which modulates B. The dipolar correlation times re are given for 24.3 MHz, and the q value is averaged over all frequencies.
expected to donate probably two (Burgers and Eckstein, 1979) and, at most, three ligands to octahedral Mn*+ (Cohn and Hughes, 1962;Sternlicht et al., 1965;Sloan et al., 1975;Sloan and Mildvan, 1976) Overhauser effect studies Mildvan, 1985, 1986). The location of the dNTP substrate is based on intermolecular nuclear Overhauser effect studies Mildvan, 1985, 1986), photoaffinity labeling (Joyce et al., 1985), and on the binding of substrates to a peptide fragment of Pol I consisting of residues 728-777 (Mullen et al., 1989). The location of the triphosphate moiety is speculative, and the P,r coordination of the metal is based on kinetic studies (Burgers and Eckstein, 1979) and NMR measurements (Sloan et al., 1975).
Mn2+ was determined by analysis of the frequency dependence of the PRR of water (Table II). The results indicate that in the ternary enzyme-dGTP-Mn" complex, only one rapidly exchanging water ligand remains coordinated to Mn". Hence two or three water ligands of Mn*+ have either been replaced by ligands from the enzyme or alternatively have been occluded by the enzyme such that they exchange with solvent at a rate slower than 10%'.

CONCLUSIONS
From the Mn2+ binding data for Pol I, its large fragment, and the 3',5'-exonuclease-deficient mutants of the large fragment, we can infer the distribution of Mn*+-binding sites among the three domains of Pol I and some of the properties of these sites. Based on affinity for the enzyme, two types of sites are detected, tight sites with dissociation constants in the micromolar range and weak sites with dissociation constants in the millimolar range. The locations of these sites are summarized in Fig. 7. Approximately 12 weak Mn'+binding sites reside on the 5',3'-exonuclease domain, since -12 fewer weak sites are detected on the large fragment than on the whole Pol I. One or more of these sites may be essential for the 5',3'-exonuclease reaction, since this reaction is known to require a divalent cation (Klett et ai., 1968) and this requirement is typically satisfied by 7 mM Mg2+ (Lehman and Richardson, 1964;Deutscher and Kornberg, 1969). The 3',5'-exonuclease domain binds one Mn2+ tightly using the carboxylate ligands Asp355 and G~IY'~~, since mutation of these residues abolishes the tight binding of Mn2+ (Table I) as well as 3',5'-exonuclease activity (Derbyshire et al., 1988).
This site corresponds to site A in the crystallographic model of the exonuclease active site (Fig. 3). The previous view that the tight Mn'+-binding site on Pol I detected in the absence of substrate is at the polymerase site @later et al., 1972) is ruled out. This tight site, now known to be site A, binds Mg2+ with a dissociation constant of 38 f 9 pM (Slater et al., 1972) and has a high affinity for Zn2+ (Ferrin et al., 1983;Ollis et al., 1985). The occupancy of this site alone by Mg2' is insuf-ficient to activate the 3',5'-exonuclease reaction since the apparent KM for Mg2+ is approximately 1.7 mM, and the optimum Mg2+ is between 7 and 20 mM with whole Pol I (Lehman and Richardson, 1964). A more recent estimate of the KM for Me with the cloned large fragment is approximately 3 mM.' An adjacent Mn" on the 3',5'-exonuclease domain is weakly bound at or near AS~~'~ in the absence of TMP and tightly in the presence of TMP, since mutation of this residue removes the tight site created by TMP, induces an order of magnitude lowering of the affinity for the tightly bound Mn2+ (Table I), and abolishes 3',5'-exonuclease activity (Derbyshire et al., 1988). This site, which binds Mn2+ lOOfold more tightly in the presence of TMP (Table I), is most likely site B in the crystallographic model ( Fig. 3). At least four additional weak Mn'+-binding sites are located on the exonuclease domain, since three in addition to site B are lost in the D355A,E357A double mutant, and at least one remaining Mn2+ is near enough to Mn2+ that was bound at site B or A to undergo dipolar relaxation.
Occupancy of site B and possibly one or more of the other weak metal binding sites is necessary to explain the high KM for Mg2f in the 3'-5'exonuclease reaction.
The remaining 3 f 1 weak Mn'+-binding sites are probably on the polymerase domain or, less likely, on the 3',5'-exonuclease domain, remote from the exonuclease active site. Although the polymerase domain does not bind Mn2+ tightly in the absence of a substrate, the binding of dNTP induces the tight binding of Mn2+ at the polymerase active site, indicating that the Mn2+ is coordinated by the enzyme-bound substrate. A ligand to Mn2+ may also be donated by the enzyme which would explain the higher affinity for Mn*+ and the single fast exchanging water ligand remaining on the metal. A comparison of the dissociation constant for Mn2' at the site induced by dGTP (KD = 3.6-7.8 pM, Table I) with the kinetically  determined Michaelis constant for free Mn*+ in the polymerization reaction (Ka = 9.8 f 2.8 pM, Slater et al., 1972) 'V. Derbyshire, N. D. F. Grindley, and C. M. Joyce, personal communication. indicates this to be the active site for DNA polymerization. At higher levels, Mn2+ inhibits polymerization with a Kr of 600 f 300 PM @later et al., 1972) consistent with the occupancy of one or more of the weak binding sites (Table I). In uiuo, the relevant activator is Mg2+ which binds ligands more weakly than Mn'+ does and is spectroscopically inert. By competition with Mn'+, the Ko of M< from the polymerase-dGTP complex found here (100 + 2 FM) is comparable with the dissociation constants of binary Mp'+.dNTP complexes (Mildvan and Cohn, 1966), indicating little additional contribution to affinity by the enzyme at the polymerase active site. Like Mn'+, at higher levels, Mg2' also inhibits polymerization (KI = 3 mM, Travaglini et al., 1975) probably due to the occupancy of one or more of the weak metal binding sites. The modest (5 3-fold) increase in the affinity of dGTP for divalent cations induced by binding of the nucleotide to the enzyme may be due to a change in metal-nucleotide coordination from tridentate q&-y in the binary metal. dGTP complex, to bidentate @,r in the ternary complex, consistent with activation of the leaving pyrophosphate group by the activating metal (Fig. 7).