Constitution of the Twin Polymerase of DNA Polymerase I 11 Holoenzyme *

It is speculated that DNA polymerases which duplicate chromosomes are dimeric to provide concurrent replication of both leading and lagging strands. DNA polymerase I11 holoenzyme (holoenzyme), is the 10subunit replicase of the Escherichia coli chromosome. A complex of the a (DNA polymerase) and c (3’-5’ exonuclease) subunits of the holoenzyme contains only one of each protein. Presumably, one of the eight other subunit(s) functions to dimerize the a c polymerase within the holoenzyme. Based on dimeric subassemblies of the holoenzyme, two subunits have been elected as possible agents of polymerase dimerization, one of which is the T subunit (McHenry, C. S . (1982) J. Biol. Chem. 257,2657-2663). Here, we have used pure a, c, and T subunits in binding studies to determine whether T can dimerize the polymerase. We find T binds directly to a. Whereas a is monomeric, T is a dimer in its native state and thereby serves as an efficient scaffold to dimerize the polymerase. The c subunit does not associate directly with T but becomes dimerized in the a c ~ complex by virtue of its interaction with a. We have analyzed the dimeric acT complex by different physical methods to increase the confidence that this complex truly contains a dimeric polymerase. The T subunit is comprised of the NHz-terminal two-thirds of T but does not bind to at, identifying the COOH-terminal region of T as essential to its polymerase dimerization function. The significance of these results with respect to the organization of subunits within the holoenzyme is discussed.

existence that each protein has a distinct function.' Study of subassemblies and isolated subunits of the holoenzyme has led to assignment of function to at least half of the subunits so far. Thus, the heterotrimer core subassembly (a,€, e) contains the polymerase and proofreading 3'-5' exonuclease activities (McHenry and Crow, 1979). Genetic overproduction and subsequent purification of individual subunits identified a as the polymerase (Maki and Kornberg, 1985) and c as the 3'3' exonuclease (Scheuermann and Echols, 1985). The core polymerase has only low catalytic efficiency (20 nucleotides/ s, Maki and Kornberg, 1987) and processivity (10-15 nucleotideshinding event, Fay et al., 1981). In contrast, the holoenzyme hydrolyzes ATP to initiate remarkably rapid (500 nucleotides/s) and processive (>8 kilobases) DNA synthesis (Fay et al., 1981;Burgers and Kornberg, 1982). The accessory proteins which confer the ATP-activated speed and processivity onto the holoenzyme were identified by studying which subunits restored these hallmark features onto the core polymerase. Rapid and highly processive synthesis was found to be rooted in formation of a tightly bound preinitiation complex of accessory proteins on primed DNA (Wickner, 1976;Maki and Kornberg, 1988b;O'Donnell, 1987). In assembly of the preinitiation complex, the five protein y complex (y66'x$) hydrolyzes ATP to transfer the p subunit from solution to the primed template (Wickner, 1976;Maki and Kornberg, 1988b;Stukenberg et al., 1991). Once / 3 is transferred to primed DNA by the y complex, the /3 is a mobile "sliding clamp" on duplex DNA and also binds the core polymerase thereby tethering it to the DNA for highly processive synthesis (Stukenberg et al., 1991). Studies aimed at determining the minimal subunit requirements of the y complex for its action in assembling the p clamp showed only the y and 6 subunits were essential to transfer / 3 onto primed DNA (ODonnell and Studwell, 1990). Studies to determine which subunits of core were needed for processive synthesis showed the /3 clamp could not provide a with highly processive synthesis but was able to confer processive synthesis onto a complex of at (Studwell and O'Donnell, 1990).
The studies summarized above have assigned functions to only five subunits of the holoenzyme (a, t, y, 6, p). Due to their limited availability or lack of obtaining them in pure form, little is known about the functions of the 8, b', x , and $ subunits. Several properties have been assigned to the T subunit although its full role in holoenzyme action is still not certain. The studies presented in this report focus on the ability of 7 to dimerize two polymerase subunits. A four-subunit subassembly of holoenzyme, called polIII', * We have recently identified the genes for the remaining subunits of the holoenzyme for which no genes had been identified (0, 6, 6', x, $). They are all distinct genes and map to unique locations on the E. coli chromosome. Z. Dong, R. Onrust, P. S. Studwell-Vaughan, and M. O'Donnell, unpublished observations. is a complex of 7 with core (McHenry, 1982). The a and 7 subunits of polIII' appeared to be equimolar based on their staining intensity in a polyacrylamide gel (McHenry, 1982). Further, hydrodynamic studies indicated a size of polIII' (410 kDa) consistent with two of each subunit (McHenry, 1982). Hydrodynamic studies of the core polymerase indicated that core (160 kDa) contained only one copy of (Y (McHenry, 1982). Hence, it was hypothesized that T dimerized the core. Consistent with a dimeric polymerase within the holoenzyme, hydrodynamic measurements of the holoenzyme indicate a mass of 900 kDa, approximately twice that predicted from summation of the monomeric molecular mass of each of the subunits . Two polymerase subunits within the holoenzyme fits nicely with proposals that replicative polymerases act in pairs for smooth and simultaneous replication of both leading and lagging strands (Sinha et al., 1980;Kornberg, 1982). Subsequent to characterization of polIII', the core polymerase was shown to be capable of dimerizing on its own indicating 7 is not essential for polymerase dimerization . Since the a€ complex showed no tendency to dimerize, the 8 subunit was elected as the probable agent of core dimerization . Furthermore, reconstitution of the processive polymerase on primed DNA indicated a 7 dimer bound only one core suggesting the subunits of polIII' are arranged such that one 7 dimer is bound to only one side of a core dimer . This subunit arrangement of polIII' indicates that 7 does not function to dimerize core by bridging two core molecules? It would appear more likely that the two cores within polIII' become dimerized by interaction between their 8 subunits, and the method used to purify polIII' yielded the dimeric core with a 7 dimer still bound to one of the core halves.
The gene which encodes the T subunit also encodes the y subunit (Kodaira et al., 1983). The y subunit is comprised of the amino-terminal two-thirds of 7 and is formed by an efficient translational frameshift event such that approximately equal amounts of 7 and y are produced from the same dnaZX gene (Tsuchihashi and Kornberg, 1990;Flower and McHenry, 1990;Blinkowa and Walker, 1990). The relationship of y to 7 suggests they may share some functions. In fact, the 7 subunit can act with either 6 or 6' to transfer p to primed DNA suggesting either a 76 or 76' complex may fulfill a similar role to y6 (O'Donnell and Studwell, 1990). 7 differs from y inasmuch as 7 is a DNA-dependent ATPase, whereas y only binds ATP implying the COOH-terminal region of 7 interacts with DNA (Lee and Walker, 1987;Tsuchihashi and Kornberg, 1989). That 7 binds DNA but y does not has been demonstrated (Discussion in Tsuchihashi and Kornberg, 1989). The ability of 7 to bind both core and DNA likely forms a common basis underlying observations of others on the effect of 7 on core polymerase activity. Thus, T increases the processivity of core (Fay et al., 1982), increases the affinity of core for a preinitiation complex on primed DNA , and aids the reconstituted processive polymerase in replication past regions of secondary structure .
The relationship between y and 7 implied these two subunits would have similar molecular interactions within the :' For T to induce two core molecules to dimerize by binding to only one core would require the following two complexities. 1) A conformational change would need to occur in core upon association with T such that the core develops a binding site for another core. 2) Since a T dimer would have two sites for binding core, a conformational change would need to occur in the T dimer upon binding core such that the unoccupied site in the T dimer lost its affinity for another core.
holoenzyme and led to proposals that a 7 dimer was bound to one core, and a y dimer (of the y complex) was bound to the other core within the dimeric holoenzyme (Hawker and McHenry, 1987;McHenry, 1988. Hence, the related 7 and y subunits were hypothesized as the basis of a structural asymmetry within the holoenzyme wherein a different set of accessory proteins were bound to the two halves of a core dimer. The y half of the holoenzyme, proposed to contain the y complex, core, and p, seemed suited to the lagging strand where the y complex is needed to repeatedly initiate replication of lagging strand fragments. The 7 half of the holoenzyme, proposed to contain 7, core, and p, appeared suited to the leading strand where the DNA-binding domain of 7 could anchor the polymerase to DNA for extensive polymerization.
In this report we have used pure a, 6, 7, and y subunits to define the interactions between them. We have utilized hydrodynamic measurements but also other techniques to increase the confidence in conclusions about dimeric complexes. The 8 gene has only recently been identified* and therefore 0 was not available for these studies. Nevertheless, the results obtained with the four subunits studied here demonstrate that 7 dimerizes a without any other subunits. Further, the y subunit showed no affinity for the polymerase. Implications of these results on the organization of subunits within the holoenzyme are discussed.

EXPERIMENTAL PROCEDURES
Materials-Sources were: unlabeled and labeled nucleotides, Pharmacia LKB Biotechnology Inc. and Du Pont-New England Nuclear, respectively. 4x174 ssDNA was prepared by banding the phage down and then up in two successive cesium chloride gradients as described (Ray, 1969). 4x174 ssDNA was uniquely primed with a synthetic DNA 30-mer (map position 2828-2794) as described (Studwell and O'Donnell, 1990). Protein standards for molecular mass studies were from Bio-Rad and Sigma. Proteins were purified to homogeneity as described a (Maki and Kornberg, 1985), t (Scheuermann and Echols, 1985), ac (Studwell and O'Donnell, 1990), T and y , 6 (O' DonnelI and Studwell, 1990), ( 3 (Johanson et et al., 1986), y complex (Maki and Kornberg, 1988b), and SSB (Weiner et al., 1975). The concentration of p was determined by absorbance using an t280 value of 17,900 M" cm" (Johanson et al., 1986). The concentration of y was determined by amino acid analysis. The concentrations of a, c, SSB, y complex, and 6 were determined by the Bradford method using bovine serum albumin as standard (Bradford, 1976). The concentrations of at polymerase and T subunit were determined from their absorbance at 280 nm and their tryptophan fluorescence (explained below). The molar extinction coefficient at 280 nm of a protein in its native state can be calculated from its gene sequence to within k5% (Gill and von Hipple, 1989). Extinction coefficients for a, e , and T (Table 11) were calculated from their gene sequences (respectively, Tomasiewicz and McHenry, 1987;Maki et al., 1983;Flower and McHenry, 1986) using the equation, 6280 = Trp, (5690 M" cm") + Tyr, (1280 M" cm") where m and n are the number of tryptophan and tyrosine residues, respectively, in each subunit. The molar extinction coefficients of tryptophan and tyrosine were from Edelhoch (1967). Proteins were dialyzed to remove DTT prior to absorbance measurements. The concentration of the ae polymerase was taken as the average of measurements by 1) absorbance at 280 nm using ezMI 107,530 M" cm" for a 1:l complex of a and t, 2) by the tryptophan fluorescence intensity of the at polymerase (excitation at 280 nm, emission at 340 nm) in 6 M guanidine hydrochloride in 25 mM Tris-HC1 (final pH 7.5) using N-acetyl-L-tryptophan ethyl ester as standard and the knowledge of 9 tryptophan residues in a 1:1 complex of at. Likewise, concentration of T was the average of measurements by 1) absorbance at 280 nm (using t2w 45,660 M" cm" (as monomer)) calculated from its gene sequence, and 2) measurement of its tryptophan fluorescence intensity (6 tryptophan/T monomer) in 6 M guanidine hydrochloride. The protein concentration determined by absorbance and by fluorescence agreed within 20%.
Specijic Subunit Assays-Activity assays for specific subunits were performed by using them in combination with other subunits to reconstitute the characteristic processive polymerization of the holoenzyme (O'Donnell and Studwell, 1990). Two pl of column fraction was added to 72 ng of primed 4x174 ssDNA, 0.88 pg of SSB, 0.5 mM ATP, 60 p~ dCTP, 60 p~ dGTP, 12 ng of 8, 140 ng of a, 50 ng of e, and 4 ng of y complex in a final volume of 25 pl of Buffer B. Reactions were incubated for 6 min at 37 "C. Replication was initiated upon addition of dATP and [a-"PJd'M'P (specific activity, 3,000-10,000 cpm/pmol) to final concentrations of 60 and 40 pM, respectively, and quenched after 15 s with 25 pl of 1% SDS, 40 mM EDTA. Incorporation of radiolabel was measured by absorption to DE81 paper (Whatman) as described by Rowen and Kornberg (1978). Only the reconstituted processive polymerase which assembles on the primed DNA in the 6-min preincubation gives rise to products in the short 15-s replication time (Maki and Kornberg, 1988b;O'Donnell and Studwell, 1990). Reconstitution of processive activity requires ac, 8, and either the y complex, y b or 16 (or 1 6 ' ) (O'Donnell and Studwell, 1990). Therefore, assays of y used 30 ng of d in place of the y complex, fractions containing T were assayed using 30 ng of d in place of the y complex, and assays of ac, a, and c were performed by omitting the respective subunit(s) from the assay. Assays of acTz and ( a c T ) * in Fig.  5 were performed by omitting a c from the assay.
Gel Filtration-Proteins were incubated together for 1 h at 15°C in 200 pl of Buffer A then injected onto a 30-ml fast protein liquid chromatography Superose 6 column. Columns were developed with Buffer A and fractions of 180 pl were collected. Fractions were assayed for specific subunits as described above and by 12% SDS-polyacrylamide gels (100 pl of fraction) stained with Coomassie Brilliant Blue (Laemmli, 1970). The Stokes radius of each protein complex was determined upon comparison of its elution volume with that of proteins of known Stokes radius analyzed in a parallel column.
Glycerol Gradient Sedimentation-Proteins were incubated together for 1 h at 15 "C in 300 pl of Buffer C containing 10% glycerol then layered onto 11.6 ml of 10-30% linear glycerol gradients in Buffer C. Gradients were centrifuged at 40,000 rpm in a Beckman SW Ti-41 rotor for 20 h at 4 "C. Fractions of 180 pl were collected from the bottom of each tube. Fractions were assayed for specific subunits and were analyzed on 12% SDS-polyacrylamide gels (100 pl of fraction). The sedimentation coefficient of protein complexes was determined by comparison to proteins with known sedimentation coefficients in a parallel glycerol gradient. HPLC-Reversed-phase HPLC analysis was performed on a Beckman HPLC system. Approximately 60 pg of gel-filtered ( a c T ) 2 complex in 1 ml of Buffer A was adjusted to 0.1% trifluoroacetic acid and injected (multiple injections using a 250-pl loop) onto an Aquapore Butyl (C4) HPLC column (330 X 4.6 mm) equilibrated with 40% HPLC grade acetonitrile (Baker) in Buffer D. Subunits were eluted with a 40-70% linear gradient of acetonitrile in Buffer D at a flow rate of 1 ml/min. Elution of subunits was monitored at 220 and 280 nm in separate analyses. Peaks were integrated automatically by the HPLC control system, and their area was confirmed manually by cutting the peaks out and weighing them.
Densitometric Anulysis-Densitometry of Coomassie Blue-stained SDS-polyacrylamide gels was performed using a laser densitometer (Pharmacia LKB Ultroscan XL) and integrated using the gel scan XL system (peak areas were confirmed manually by cutting the peaks out and weighing them).

RESULTS
The T Subunit Binds the ae Polymerase-An interaction between the at polymerase and 7 subunit was examined by gel filtration (Fig. 1) and glycerol gradient analysis (Fig. 2). First, at alone was analyzed. The a and t subunits were mixed in a 1:2 molar ratio and incubated 1 h at 15 "C to constitute the ae polymerase. Fractions from the gel filtration analysis of at were examined by SDS-PAGE (Fig. IA, rightpunel) and by activity in formation of the processive polymerase with the holoenzyme accessory proteins (Fig. LA, left panel). The peak of at polymerase eluted in fractions 83-89 and the excess e  (Lund, 1987). Stokes radii of the other protein standards was calculated from their known diffusion coefficients (Smith, 1970) using the formula, Stokes radius = kT/6m@, where k is Boltzmann's constant, Tis temperature, 7 is viscosity, and D is diffusion coefficient. subunit peaked in fractions 98-102 (not shown). Recovery of polymerase activity from the column was 80% of the activity injected onto it. The ac polymerase had a Stokes radius of 50 A upon comparison of its elution volume with proteins of known Stokes radius (Fig. 1D). This value is close to the 54 A Stokes radius determined for the core polymerase (obtained upon examination of Fig. 5 of McHenry, 1982).
The ac polymerase was also analyzed by glycerol gradient analysis (Fig. 2). The gradient fractions were analyzed by SDS-PAGE ( Fig. 2A, right panel) and activity assays (Fig 2A,  left panel). Recovery of polymerase activity from the gradient fractions was 79% of the total activity loaded onto the gradient. The ac polymerase sedimented with a value of 6.9 S upon comparison with proteins of known S value (Fig. 20). This value is close to the 7 S measurement of the core polymerase (obtained upon inspection of Fig. 6 of McHenry. Analysis of the acr complex by glycerol gradient sedimentation. Glycerol gradients were performed as described under "Experimental Procedures." Subunits in column fractions were identified by reconstitution of processive polymerase activity assays (panels on the left) and by 12% SDS-polyacrylamide gel analysis (panels on the right) as described in the legend to Fig. 1 (Table I).
The 7 subunit ha! previously been analyzed by gel filtration (Stokes radius, 85 A) and glycerol gradient sedimentation (6.8 S) (Tsuchihashi and Kornberg, 1989). It was noted that 7 sedimented as a dimer in the glycerol gradient but behaved larger than a dimer upon gel filtration. The discrepancy was attributed to 7 having an asymmetric shape or possibly a higher aggregation state in the gel filtration analysis. Analysis of 7 by gel filtration and glycerol gradient sedimentation is shown in Fig. 1B and Fig. 2B, respectively. The 7 subunit was visualized in fractions by SDS-PAGE analysis and was assayed for its ability to reconstitute the processive polymerase with at, @, and 6 (O'Donnell and Studwell, 1990). Recovery of 7 activity from gel filtration was 86% and from glycerol gradient sedimentation appeared to be over 95%. The high frictional coefficient of 7 calculated from its Stokes radius and sedimentation coefficient (Table I) indicates 7 has a highly extended structure, consistent with the previous study on the hydrodynamic properties of 7 (Tsuchihashi and Korn-

(FIFO)
a The partial specific volume was calculated by summation of the known partial specific volumes of the individual amino acids (Cohn and Edsall, 1943). The theoretical partial specific volumes of a and c were 0.737 and 0.735 cm'/g, respectively.
berg, 1989). Combination of the Stokes radius and sedimentation coefficient in the Siegel and Monty equation yielded a molecular mass of 206 kDa (Table I), a value more consistent with the predicted mass of a 7 trimer (213 kDa) than a 7 dimer (142 kDa). We will presume here the extended shape of 7 overestimates its size and therefore 7 is a dimer, although a trimeric state for 7 cannot be rigorously excluded.
Mixture of 7 with molar excesses of a and e shows efficient formation of an at7 complex (ie. all the 7 is in complex with a and e) which eluted from the gel filtration column much earlier (Fig. IC) and sedimented in the glycerol gradient faster (Fig. 2C) than either 7 or the at polymerase alone. The SDS-PAGE analysis (right panels) and reconstitution assays (left panels) showed all the 7 comigrated with the at polymerase as an at7 complex. The at7 complex cleanly resolved from the excess at. Out of the total polymerase activity applied to the gel filtration column, 56% was recovered in the at7 peak and 25% was recovered in the at peak for a total recovery of 81%. Eighty-eight % of the 7 activity was recovered from the gel filtration column, and all of it was in the at7 peak. Of the polymerase activity loaded onto the glycerol gradient, 64% was present in the at7 fractions and 20% was localized in the ac fractions for a total recovery of 84%. All of the 7 activity loaded onto the glycerol gradient was recovered (>95%) and was located entirely within the at7 fractions. The fact that all the 7 is found in association with at even at these slight molar excesses of at (over 7 ) demonstrates that the constitution of 7 with at is highly etficient.
The Stokes radius (100 A) and S value (12 S) of the at7 complex, although not identical to the reported valyes for polIII', are similar to the measured Stokes radius (85 A) and S value (11.3 S) of polIII' (McHenry, 1982). Combination of the Stokes radius and sedimentation coefficient of the at7 complex in the Siegel and Monty equation yielded a mass of 520 kDa, closer to 457 kDa calculated for a subunit stoichiometry of ( a t r ) , than 300 kDa calculated for an at72 complex (Table I)  The Molar Ratio of Subunits in the at7 Complex-We next determined the molar ratio of subunits in the at7 complex to further distinguish between the ~C T~ or at^), assignment. Consistent with an (at7)* complex, the a, t, and 7 subunits were found to be equimolar in the complex by two methods. One method was to purify the at7 complex, resolve the subunits in a SDS-polyacrylamide gel, then stain the gel with Coomassie Blue and determine the molar ratio of the subunits by densitometry (Fig. 3). The different subunits may not take up equivalent amounts of Coomassie stain. Therefore molar ratio "standards" were prepared by mixing a, e, and T of known concentration in the ratios 1:l:l (Fig. 3A) and 1:1:2 (Fig. 3B). Analysis of the at7 complex isolated by gel filtration (Fig. 3C) and by glycerol gradient sedimentation (Fig. 30) showed a strong resemblance to the 1:l:l molar ratio standard (Fig. 3A). The ratio of subunits in the at7 complex isolated by gel filtration was calculated to be 1:0.81.2 (n/t/7, respectively) by normalizing the area under their respective peaks to those of the standards ( Table 11). The ratio of subunits in the at7 complex isolated by sedimentation in a glycerol gradient was 1.0:0.5:1.2 (a/t/7, respectively).
The second method used to analyze the molar ratio of subunits in the at7 complex was measurement of the relative ultraviolet absorbence of each subunit after dissociating the at7 complex in trifluoroacetic acid and resolving the subunits by reversed-phase HPLC (Fig. 4). Recoveries of the a, 7 , and t subunits from the HPLC column were 77, 80, and 71%, respectively: Elution of the a, 6 ) and 7 subunits from the Since HPLC was performed under denaturing conditions, the by evaporating the HPLC column fractions, followed by SDS-poly-recoveries of a, 7 , and c were not determined by activity assays, but acrylamide gel electrophoresis, and then recoveries were quantitated by determining the amount of each subunit in the column fractions by laser densitometry of the Coomassie Blue-stained gel. The total amount of each subunit was compared to the total amount of each respective subunit loaded onto the HPLC by analysis of the same at7 complex prior to HPLC, analyzed in the same polyacrylamide gel. The observed recoveries are a minimum estimate of the true recovery as approximately 10% of the sample is lost during the multiple injections required to load the complex onto the column. Molar ratios of subunits in the at7 complex isolated by gel filtration were calculated from the densitometer tracing (Fig. 3C) upon dividing the observed peak area of a subunit by the peak area of that respective subunit in the 1:1:1 molar ratio standard (Fig. 3A). The values obtained for c and T were normalized to the value of a. Molar ratios of subunits in the glycerol gradient purified ~C T complex (Fig. 6D) were determined likewise. In the HPLC analysis, the molar ratio of subunits in the at7 complex was determined by monitoring their elution from the HPLC column at 220 nm and 280 nm (see "Experimental Procedures"). The peptide bond is the major chromophore at 220 nm, and thus the molar ratio of subunits was calculated by dividing the area under each peak (Fig. 4A) by the number of amino acids in the respective subunit and normalized relative to the value for a. The areas under the peaks of 280-nm absorbance were divided by their respective tm (see "Experimental Procedures") and normalized relative to the value for cy. The number of amino acids and the cZm values for cy, t, and T were determined from their respective gene sequences. "Experimental Procedures." HPLC column was monitored at 220 nm (Fig. 4A). The peptide bond absorbs strongly at 220 nm, and therefore the area under the peak of absorbance is directly proportional to the number of amino acids in the respective protein. The molar ratio, calculated by dividing the peak area by the number of amino acids, was 1:0.8:1.3 for a, e, and 7 , respectively (Table 11). In a parallel analysis, elution of a, t, and 7 from the HPLC column was monitored at 280 nm where the amino acids tryptophan and tyrosine absorb (Fig. 4B). The molar extinction coefficients of a, e, and 7 were calculated from the number of tryptophan and tyrosine residues in each subunit predicted from their respective genes (Table 11). A molar ratio of 1:0.9:1.2 for a, t, and 7 , respectively, was calculated by dividing the area under the peak of 280 nm absorbance by the extinction coefficient of the respective subunit (Table 11). Whereas the molar ratio of a to t in the at polymerase is one-to-one Studwell and O'Donnell, 1990), the at7 complex appears on average 25% deficient in t (Table 11). Perhaps the interaction of 7 with a destabilizes the binding of E to a.
The nearly equimolar ratio of subunits in the at7 complex combined with its large molecular weight indicates the (at7), stoichiometry rather than the (Ye72 stoichiometry. An altl~l complex is unlikely in light of the dimeric structure of 7 and large size of the ( a € 7 ) 2 complex. Further evidence for the at^)^ assignment is given below. Test of the Dimeric Polymerase-Assembly of two polymerases onto a 7 dimer scaffold allows a testable prediction. Namely, under limiting at polymerase (molar excess of 7 ) , each 7 dimer should obtain only one at molecule thus forming an at72 complex as diagrammed in Fig. 5A. This prediction assumes the two molecules of at bind independently (without high cooperativity) to the 7 dimer. Indeed, constitution of at with a &fold molar excess of 7 dimer yielded the (Ye72 complex (Fig. 5B, tricmgles). The (Ye72 complex eluted @om the gel filtration column later (Stokes radius of 91 A) than the dimeric ( 0 € 7 ) 2 complex (Fig. 5B, circles) but earlier than either at or r alone ( Fig. 1 A and B ) , consistent with the predicted size of an (YE72 complex (300 kDa). Visualization of the subunits in the SDS-polyacrylamide gel showed the at^^ complex did not fully resolve from the excess free 7 subunit (data not shown) which precluded subunit molar ratio studies on the at^^ complex.
Subunit Contacts in the (01€7)2 Complex-We used gel filtration to show the 7 subunit binds the at polymerase through contact with a (Fig. 6, C and D). Mixture of T with a molar excess of a followed by gel filtration resulted in an ( (~7 )~ complex which consumed all the T subunit (Fig. 6C). The left panel in Fig. 6C shows all the T activity coeluted with the a activity. The SDS-polyacrylamide gel analysis confirmed the coelution of a and 7 (rightpanel, Fig. 6C). In this experiment, 83% of the 7 activity was recovered from the column, and all of it was in the (a7)2 peak. Of the a activity loaded onto the column, 49% eluted with the ( (~7 )~ complex and 32% was present as free a for a total recovery of 81%. The ( a~)~ assignment is based on 1) the molar ratio 1:O.g (a/.) obtained by densitometry analysis (not shown) of the gel-filtered ( (~7 )~ complex in the Coomassie Blue-stained polyacrylamide gel ( Fig. 6C), and 2 ) the large Stokes radius (97 A) of the (LIT), complex determined from its elution volume (nearly the same as the ( m 7 ) 2 complex) from the gel filtration column (Fig.  6C).
Incubation of a mixture of 7 and t did not yield a complex in the gel filtration analysis (Fig. 6 D ) . The SDS-PAGE analysis showed t did not comigrate with T in the gel filtration fractions (Fig. 6D, right panel). Furthermore, t activity did not comigrate with 7 activity (Fig. 6D, left panel).
The y subunit of the holoenzyme is derived from the same gene as T by a translational frameshift event which occurs after synthesis of approximately two-thirds of T (Tsuchihashi and Kornberg, 1990;Flower and McHenry, 1990;Blinkowa and Walker, 1990). The frameshift is followed within a few codons by a stop codon. Therefore, except for one amino acid near the COOH terminus, y is comprised of the NHz-terminal 431 amino acids of 7 . We examined the y subunit for ability to bind a€ (Fig. 6, A and B ) . Previous hydrodynamic measurements of y showed it had a Stokes radius of 62 A and S value of 6.6 (Tsuchiahashi and Kornberg, 1989). Gel filtration of y alone showed it eluted (Stokes radius, 63 b;) near the position of 7 (Fig. 6 A ) indicating the y subunit, like 7 , has an extended structure. The y subunit sediments (6.8 S) slightly slower than 7 in a glycerol gradient consistent with the smaller size of a y dimer (data not shown). The y, a, and subunits were incubated together and then gel-filtered, but there was no evidence of complex formation between at and y (Fig. 6 B ) . The y subunit eluted in the same position whether at was present or not (compare Fig. 6, A with B ) . Likewise, the elution volume of at was unchanged by y (compare Fig. L4 with Fig. 6 B ) . Recoveries of y and at activity from the column were 80 and 77%, respectively. We have repeated the gel filtration analysis in lower salt (100 mM NaC1) both in the presence and absence of 0.5 mM ATP and 8 mM MgC12 with no different results (no interaction between y and ae). Further, we observed no interaction between y and 7 , or between y and ae7 in gel filtration experiments (data not shown). The observation that 7 binds to a but y does not, combined with the fact that y and T share their NH2-terminal sequences, implies the COOH-terminal region of 7 is essential to the interaction with a. Whether only the COOH-terminal region of 7 is sufficient for the interaction with a will require further studies.

DISCUSSION
We have reconstituted the twin polymerase of DNA polymerase I11 holoenzyme. The polymerase subunit, a, is a monomer and remains so when complexed with e, the 3'-5' exonuclease subunit. The dimeric 7 subunit binds two ae polymerase molecules to form an ((Yt7)Z complex with an observed mass of 520,000 daltons (predicted mass of 456,000 Da). The three subunits were equimolar as expected for a (a€7)2 complex. These results are consistent with the initial proposal by McHenry (1982) that T dimerizes core within the polIII' complex ( a e % 7 ) 2 . The polIII' complex is naturally purified from E. coli, evidence for the existence of the dimeric polymerase inside the cell. The intact holoenzyme purified from E. coli contains the a, e, and 7 subunits in approximately equimolar ratios and is large enough to accommodate a dimer of all the subunits (Maki et aZ., 1988) suggesting the presence of the (aer), dimer within the holoenzyme.
The pol111 core (a, c, and % subunits) is reported to dimerize m FIG. 7. Proposed arrangement of subunits in the (acT)s complex. The native state of T appears dimeric. Two a monomers bind the halves of the T dimer. The e subunit directly binds to a but shows no evidence of direct interaction with 7 . The COOH attached to T represents the essentiality of the COOH-terminal region of 7 for interaction with a.
when sufficiently concentrated (18 p~) (Maki et aZ., 1988). Since the at polymerase showed no tendency to dimerize, it was suggested that the % subunit was the agent that mediated dimerization of core (Maki et aZ., 1988). The % subunit, and possibly even others, may contribute to polymerase dimerization, but the present work shows the 7 dimer can achieve this function at a relatively low concentration (Kd 5 17 nMl5 in the absence of 8. Based on the estimate of 20 molecules of the holoenzyme/cell (McHenry and , the intracellular concentration of its subunits would be approximately 40 nM. Hence, the binding strength of 7 to ac is strong enough to ensure formation of the ( L Y C T )~ complex at the level of these subunits found in vivo. We would like to study the relative contribution of % to polymerase dimerization by the reconstitution approach, but these studies must await overproduction of % as it has never been purified free of a and t, and its gene has only recently been discovered.2 The simplest interpretation of results in this report suggests the 7 subunits are symmetrically disposed about the two ae polymerases (Fig. 7).3 The 7 dimer is presented in Fig. 7 as long and thin to reflect its extended structure predicted from its Stokes radius and sedimentation coefficient. The y subunit contains the NH2-terminal two-thirds of the 7 protein but does not bind ae indicating that at interacts with the COOHterminal portion of 7 . The two a subunits are dimerized by virtue of one a binding each half of the 7 dimer. The 7 subunit interacts with a e mainly through contact with a, since 7 forms The affinity of ac for T was too great to observe an equilibrium of ac with a c~ at amounts of these proteins suitable for visualization in SDS-polyacrylamide gels or by reconstitution of processive activity assays. Thus, we prepared [3H]a to increase the sensitivity of the analysis. The [3H]a (95 Ci/mmol) was prepared using NaB[3H]4 (74 Ci/ mmol) and formaldehyde (Rice and Means, 1971)  This value may be an upper limit if equilibrium was not reached in the 24-h incubation. a complex with a but not with t. The t subunit is shown bound only to a since a complex between 7 and t was not observed.
It has been hypothesized that the accessory proteins of the E. coli DNA polymerase I11 holoenzyme are disposed asymmetrically about the two polymerase subunits to confer the separate properties anticipated in replication of the leading and lagging strands McHenry, 1988). The close relationship between the 7 and y subunits was proposed as the basis for an asymmetric orientation of accessory subunits about the two core polymerases (i.e. 7 on one core and y on the other) McHenry, 1988). Ability to reconstitute a complex containing a T dimer bound to a single core supported the notion that a dimer of y may bind the other core molecule within the holoenzyme (Maki et aL, 1988). However, in light of the studies presented here, the previous observation of a 7 dimer bound to one core is likely explained by the 10-fold molar excess of T dimer relative to core monomer used in the previous study. This would lead to formation of a core -7 2 complex, much like the ( Y C T~ complex formed here using a molar excess of T over at (Fig. 5). Another previous observation which indicated y bound to core was that addition of T to primed DNA "coated" with SSB in the presence of core, B, and y complex decreased the amount of y complex bound to the DNA by %fold, possibly explained by competition of 7 and y for core . We did not detect an interaction of y with at in these studies, although it is possible that the interaction of y with at was too weak to detect it (Kd >> 5 PM, the concentrations of at and y used in the Fig. 6 experiment). With such a low affinity of y for at relative to the strength of the 7 -a~ interaction, it seems y would have a difficult time ever appropriating a polymerase from 7 in assembly of an asymmetric dimer. Despite the results presented here it is still possible that 7 and y are positioned on opposite halves of a polymerase dimer. For example, perhaps y interacts with 8 or develops a strong interaction with core when y is within the y complex or in the presence of @ and primed DNA (y does not bind at in the presence of @ in solution, data not shown). It is also possible that T competes with another subunit of the y complex for core, although the asymmetry achieved by such an interaction would not be based in the relationship of y to T .
The COOH-terminal one-third of 7 (missing in y), besides a role in binding a, is also essential for the interaction between 7 and DNA (evidence outlined in the Introduction). It has been proposed that since 7 strengthens the grip of the polymerase to DNA, the 7 subunit would not be present on the lagging strand where the polymerase must loosen its grip on DNA and dissociate from a completed Okazaki fragment in order to cycle to a new RNA primer . A mechanism by which the lagging strand polymerase can have a tight grip on DNA yet still rapidly dissociate from the DNA after it has completely replicated the available template has been addressed in our previous studies. Thus, core rapidly dissociates from the @ clamp only after completing processive replication of the DNA template, followed by (or concerted with) rapid reassociation of the core with a new preinitiation complex (p clamp) on a new primed template (O'Donnell, 1987;O'Donnell and Studwell, 1990;Studwell and O'Donnell, 1990;Studwell et al., 1990). Further, 7 did not prevent the polymerase from cycling to new primed templates (endowed with a / 3 clamp) in this experimental system (O'Donnell and Studwell, 1990).
The simplest arrangement of subunits to explain how a T dimer binds two at polymerases is that a 7 dimer bridges two at polymerase molecules (e.g. Fig. 7). It is reasonable to presume that 8 is also present on each half of the polymerase dimer since 8 is isolated in a tight complex with a and t (Le. core). Additionally, each core must interact with a sliding clamp / 3 dimer in order to become tethered to DNA for processive synthesis (Stukenberg et al., 1991). There are approximately two 0 dimers within the holoenzyme consistent with a dimer for each half of the polymerase dimer . This leaves the y complex subassembly to impose a structural asymmetry onto the holoenzyme. The y complex may be needed only once to initiate processive replication of the polymerase on the leading strand, but the y complex is needed repeatedly on the lagging strand to initiate processive replication by the lagging strand polymerase on each new primer. To determine whether only one y complex is present in the holoenzyme asymmetrically disposed relative to the two polymerases will require further studies.