Identification of the beta-binding domain of the alpha subunit of Escherichia coli polymerase III holoenzyme.

Rapid and processive DNA synthesis by Escherichia coli DNA polymerase III holoenzyme is achieved by the direct interaction between the α subunit of DNA polymerase III core and the β sliding clamp (LaDuca, R. J., Crute, J. J., McHenry, C. S., and Bambara, R. A. (1986) J. Biol. Chem. 261, 7550-7557; Stukenberg, T. P., Studwell-Vaughan, P. S., and O'Donnell, M. (1991) J. Biol. Chem. 266, 11328-11334). In this study, we localized the β-binding domain of α to a carboxyl-terminal region by quantifying the interaction of β with a series of α deletion proteins. Purification and binding analysis was facilitated by insertion of hexahistidine and short biotinylation sequences on the deletion terminus of α. Interaction of β with α deletion proteins was studied by gel filtration and surface plasmon resonance. α lacking 169 COOH-terminal residues still possessed β-binding activity; whereas deletion of 342 amino acids from the COOH terminus abolished β binding. Deletion of 542 amino acids from the NH2 terminus of the 1160 residue α subunit resulted in a protein that bound β 10-20-fold more strongly than native α. Hence, portions of α between residues 542 and 991 are involved in β binding. DNA binding to α apparently triggers an increased affinity for β (Naktinis, V., Turner, J., and O'Donnell, M. (1996) Cell 84, 137-145). Our findings extend this observation by implicating the amino-terminal polymerase domain in inducing a low affinity taut conformation in the carboxyl-terminal β-binding domain. Deletion of the polymerase domain (or, presumably, its occupancy by DNA) relaxes the COOH-terminal domain, permitting it to assume a conformation with high affinity for β.

DNA polymerase III holoenzyme is the replicative polymerase responsible for the synthesis of the majority of the Escherichia coli chromosome. The special replicative role of DNA polymerase III is conferred by its ability to interact with other replication proteins. It apparently functions as an asymmetric dimer, capable of coordinating leading and lagging strand synthesis at the replication fork. Holoenzyme 1 is highly processive (Fay et al., 1982;Wu et al., 1992), perhaps capable of synthesizing the entire E. coli chromosome without dissociation. The ␤ subunit is required for processive DNA synthesis (Fay et al., 1982;LaDuca et al., 1986). A ␤ dimer forms a donut-shaped structure around duplex DNA (Kong et al., 1992) and contacts the polymerase by direct protein-protein interactions (LaDuca et al., 1986;Stukenberg et al.,1991). These properties are consistent with a sliding clamp that can translocate with the polymerase and tether it to the DNA template. Mutations in dnaN, the structural gene for ␤, have been found that suppress mutations in dnaE, the gene encoding ␣, providing genetic support for an interaction between ␣ and ␤ ( Kuwabara and Uchida, 1981).
The catalytic subunit of DNA polymerase III, ␣, in addition to its functional role, must serve as an organization protein to hold the 18-protein holoenzyme complex together . binds the COOH-terminal region of ␣ with high affinity (Kim and McHenry, 1996) to form a dimeric polymerase, pol IIIЈ (McHenry, 1982;Studwell-Vaughan and O'Donnell, 1991). The NH 2 -terminal half of ␣ contains the polymerase active site. 2 Because the high processivity of holoenzyme results from the direct interaction of ␣ and ␤, a specific region of ␣ must be devoted to binding ␤. Here, we identify the ␣ domain responsible for interaction with ␤.
Estimation of the K D for the ␣-␤ Interaction by Gel Filtration-To estimate the affinities of the interaction between ␤ and the ␣ deletion proteins using gel filtration analysis, stained gels of the column fractions were subjected to a laser densitometric scan (Molecular Dynamics). This interaction was modeled as a simple 1:1 interaction between an ␣ monomer and a ␤ dimer (␣ ϩ ␤^␣␤), where the dissociation constant for this reaction is defined as . The molar concentration of bound ␣ monomer (␣␤) was determined from the equation: ␣␤ ϭ (I ␣␤ /I ␣T )␣ T , where ␣ T is the total molar concentration of ␣ loaded onto the column, I ␣T is the total intensity from all the ␣ bands (bound and unbound), and I ␣␤ is the total intensity of bands corresponding to ␣ bound with ␤. The fractions containing ␤ bound to ␣ were determined by comparing the shift in the elution position of ␤ in the presence and absence of ␣. The amounts of free ␤ dimer (␤ f ) and free ␣ monomer (␣ f ) were then determined from mass conservation; ␤ f ϭ ␤ T Ϫ ␣␤ (␤ T is the total initial molar concentration of ␤) and ␣ f ϭ ␣ T Ϫ ␣␤.
Since significant dilution occurs during gel filtration, so that the eluted protein concentration is not equal to the protein concentration loaded onto the column, the K D was also calculated with a correction factor derived from the ratio of the total elution volume of ␣ (generally, 1 ml) to the injection volume (0.2 ml) that yields a 5-fold dilution. Thus, K D was calculated assuming a 5-fold lower ␣ T and ␤ T concentrations than the actual initial molar input concentrations. The corrected and uncorrected K D values represent upper and lower limits, respectively, providing a comparison of the relative affinity between ␤ and the ␣ deletion proteins since all filtration analyses were done under the same conditions.

RESULTS
␣ Containing a Deletion of 542 Amino Acids from the NH 2 Terminus Interacts Strongly with ␤-We previously reported the construction of a series of plasmids that contain progressive deletions from the termini of ␣ (Kim and McHenry, 1996). The terminus of the deletion-containing ␣ subunit was fused to a peptide containing a hexahistidine sequence and a short sequence that is biotinylated in vivo, facilitating purification and immobilization (Kim and McHenry, 1996). We analyzed the ability of these deleted ␣ subunits to interact with ␤ 2 using a gel filtration method initially developed by Stukenberg et al. (1991) to demonstrate an ␣-␤ interaction. As a positive control, the interaction of wild-type ␣ and ␤ was tested. ␤ 2 alone eluted mainly at fractions 61-63 (Fig. 1A), and ␣ eluted at fractions 57-60 (Fig. 1B). When ␤ was incubated with ␣ prior to gel filtration, its elution position shifted to a higher molecular weight (fractions 56 -58, Fig. 1C), indicating an interaction. The densitometric scan of this gel permitted estimation of an ␣-␤ K D between 0.2 and 1 M, consistent with the value 250 nM determined by Stukenberg et al. (1991). ␣N⌬1 (missing the first amino acid and fused to a peptide) showed nearly the same interaction as wild-type ␣ (Fig. 1, D and E, Table I), suggesting that the fusion peptide itself did not perturb ␤ binding. The ␣ deletion protein ␣N⌬542, which has a 542-amino acid deletion from the NH 2 terminus, bound ␤ about 20-fold more strongly than full-length ␣ (Fig. 1, compare C with G) (Table I). Formation of partially unfolded proteins that aggregated precluded examination of deletions greater than 542 residues (Kim and McHenry, 1996). These results showed that the ␤-binding site was located in the COOH-terminal half of the ␣ subunit. The COOH-terminal 169 Amino Acids of ␣ Are Not Required for Interaction with ␤-To determine the COOH-terminal limit of ␣ for ␤ binding, we tested the ␤-binding activities of COOHterminal deletion proteins. The full-length COOH-terminal ␣ fusion protein (␣C⌬0) showed very similar ␤-binding activity to that of wild-type ␣ (Table I). ␤ was shifted to the elution position of ␣C⌬0 due to their interaction (compare Fig. 1A with  2A and B). Both ␣C⌬48 and ␣C⌬169 bound ␤ (Fig. 2, C-F), although binding was 12-and 30-fold weaker than that of wild-type ␣, respectively ( Table I). Deletion of 342 amino acids from the COOH terminus resulted in a protein (␣C⌬342) that did not bind ␤ (Fig. 2, G and H). Thus, residues 542-991 define the approximate terminal limits of an ␣ domain involved in interaction with ␤.
Deletion of the Polymerase Domain of ␣ Increases the Affinity of the ␣ COOH-terminal Domain for ␤-Gel filtration analysis suggested that ␣N⌬542 bound ␤ 2 significantly more strongly than full-length ␣. To verify this observation, we examined the interaction of ␣N⌬542 with ␤ by surface plasmon resonance. Biotinylated ␣ deletion proteins ␣N⌬1 and ␣N⌬542 were immobilized on a sensor chip via streptavidin interaction as described (Kim and McHenry, 1996). When ␣ derivatives were coupled at levels less than 1000 RU, the signal due to ␣-␤ binding was so small that it could not be reliably analyzed. Therefore, ␣ deletion proteins were immobilized at a high level (3500 RU for ␣N⌬542 and 3300 RU for ␣N⌬1). Under these conditions, ␤ bound and dissociated rapidly from ␣N⌬1 and ␣N⌬542 (Fig. 3, A and B). Even at high ␤ concentrations (1 M), binding was substoichiometric (0.08 ␤ 2 /␣N⌬1 and 0.2 ␤ 2 / ␣N⌬542). The on-rate (k on ) and off-rate (k off ) of the interaction between ␣N⌬1 and ␤ were 5.1 ϫ 10 5 s Ϫ1 M Ϫ1 and 7.5 ϫ 10 Ϫ2 s Ϫ1 , respectively, resulting in a K D of ϳ150 nM (K D ϭ k off /k on ). ␣N⌬542 bound ␤ with a K D of ϳ40 nM (k on ϭ 9.5 ϫ 10 5 s Ϫ1 M Ϫ1 ; k off ϭ 3.6 ϫ 10 Ϫ2 s Ϫ1 ). Both K D values were in the range determined by gel filtration analysis. The interaction of ␣N⌬1-␤ was weaker than that of ␣N⌬542-␤, consistent with results from gel filtration analysis, even though there was only a 3.7-fold difference in the K D by BIAcore analysis compared to a 20-fold difference in the K D by gel filtration. and ␤ Can Both Bind to the COOH Terminus of ␣ Simultaneously-Both and ␤ bind to the COOH-terminal region of ␣ (Kim and McHenry, 1996), raising the possibility of steric exclusion by virtue of overlapping binding sites of and ␤. To test this possibility, we performed gel filtration experiments in which ␣N⌬1 and (4-fold molar excess) were incubated followed by addition of ␤. As shown in Fig. 4 (A and B), ␤ was detected in early column fractions indicating an ␣--␤ complex and thus the absence of significant competition between and ␤. Moreover, the binding of to ␣ is much tighter than that of ␤ 2 (70 pM versus ϳ0.2-1 M, respectively), and would have prevented ␤ 2 binding if binding were competitive. binding to ␣N⌬1 appeared to induce a slight decrease in the affinity between ␤ and ␣N⌬1, as indicated by a comparison to the gel filtration results with ␤-␣N⌬1 alone (Fig. 1E); indeed, densito-metric scans of Fig. 1E and 4B showed that approximately 4-fold less ␤ bound to ␣N⌬1 in the presence of . When ␣N⌬542 was used for gel filtration with and ␤, ␤ was also observed to interact with this COOH-terminal domain of ␣ in the presence of excess (Fig. 4C). We were unable to compare the affinity of the ␣N⌬542-␤ interaction in the presence and absence of due to the similar molecular mass of and ␣N⌬542 and their co-migration on polyacrylamide gels, although the level of ␤ bound in the presence of ( Fig. 4C) appears to be less than that observed in the absence of (Fig. 1G).
Stimulation of DNA Synthesis of ␣ Deletion Proteins by ␤-We detected interactions between ␤ and ␣ deletion proteins ␣C⌬0, ␣C⌬48 and ␣C⌬169, but no interaction between ␤ and ␣C⌬342 using gel filtration analysis. DNA synthetic activity of pol III was reportedly stimulated 7-fold by the addition of a stoichiometric excess of ␤ on poly(dA)-oligo(dT) primer-templates (LaDuca et al., 1986). Here we examined ␤ stimulation of ␣ deletion proteins capable of synthesizing DNA using homopolymeric templates as described under "Experimental Procedures." The synthetic activities of three ␣ deletion proteins ␣C⌬0, ␣C⌬48 and ␣C⌬169 and wild-type ␣ were stimulated about 3-fold in the presence of 10,000 -40,000 units (0.4 -1.6 ϫ 10 3 molar excess) of ␤ (Fig. 6). The ␤ subunit did not stimulate ␣C⌬342 activity even in the presence of the same gap-filling units of ␣C⌬342 as ␣C⌬169. Thus, the maintenance of ␤ binding activity in polymerase-proficient ␣ deletion proteins is adequate to permit ␤ stimulation of polymerization.

DISCUSSION
The rapid and processive DNA synthesis by holoenzyme depends on the ␤ sliding clamp (Fay et al., 1982). The DnaX complexes assemble ␤ on DNA in a reaction that is coupled to ATP hydrolysis (Wickner, 1976;Onrust et al., 1991). The ␤ sliding clamp confers high processivity by tethering polymerase to the DNA template (LaDuca et al., 1986;. In this report, we have identified the ␤-binding site of ␣ by measuring the binding of ␤ to modified ␣ subunits that contain a deletion from either their NH 2 or COOH terminus. The results indicate that a region between amino acid residues 542 and 991 of ␣ is involved in ␤ binding. Interestingly, deletion of 542 amino acids from the NH 2 terminus of ␣ resulted in a 4 -20-fold higher affinity for ␤ compared to wild-type ␣. This suggests that the amino-terminal domain of ␣, which contains the polymerase active site, 2 is an inhibitory domain for the binding of ␤. Recently, O'Donnell and colleagues (Naktinis et al., 1996) proposed a model in which DNA occupancy of the polymerase active site converts ␣ to a form that has an increased affinity for ␤. Presumably, the modulation of this interaction is important in permitting rapid cycling of the lagging strand polymerase after completion of FIG. 3. ␣-␤ interactions on BIAcore. A, ␣N⌬1-␤ interaction. ␣N⌬1 was immobilized to the streptavidin chip as described under "Experimental Procedures," and three concentrations (0.2 M, 0.4 M, and 1 M) of ␤ 2 were injected over the ␣N⌬1 chip to examine their interaction. B, ␣N⌬542-␤ interaction. ␣N⌬542 was immobilized to the streptavidin chip as described under "Experimental Procedures," and three concentrations (0.1 M, 0.5 M, and 1 M) of ␤ 2 were injected over the ␣N⌬542 chip.

FIG. 4. Interaction of -␣-␤ by gel filtration analysis. A, -␤ control.
A 200-l mixture of (1.1 nmol) and ␤ (0.51 nmol) in buffer TSG was incubated at room temperature for 30 min, gel-filtered, and analyzed by SDS-polyacrylamide gel electrophoresis as described under "Experimental Procedures," B, -␣N⌬1-␤ interaction. A mixture of (3.32 nmol) and ␣N⌬1 (0.83 nmol) in buffer TSG was incubated at room temperature for 30 min. ␤ 2 (1.53 nmol) was added to the -␣N⌬1 mix, followed by another 30-min incubation at room temperature. The mixture (200 l) of three proteins was gel-filtered and analyzed by SDSpolyacrylamide gel electrophoresis as described under "Experimental Procedures." C, -␣N⌬542-␤ interaction. A mixture of (1.11 nmol) and ␣N⌬542 (0.4 nmol) in buffer TSG was incubated at room temperature for 30 min. ␤ 2 (1 nmol) was added to the -␣N⌬542 mix followed by another 30 min incubation at room temperature. The mixture (200 l) of three proteins was gel-filtered and analyzed by SDS-polyacrylamide gel electrophoresis as described under "Experimental Procedures." The numbers above each gel are column fractions (Fr. No). M, a protein standard for molecular mass: phosphorylase b, 94 kDa; albumin, 67 kDa; ovalbumin, 44 kDa. synthesis of an Okazaki fragment on the lagging strand. Our results are consistent with this proposal and add to our understanding of the mechanism. The unoccupied NH 2 -terminal polymerase domain must constrain the carboxyl-terminal ␤ binding domain to a taut low affinity conformation, since its removal permits the high ␣ affinity COOH-terminal domain to relax to a high ␤ affinity conformer (Fig. 7A). Extending this observation to the model in which DNA modulates the binding of ␤ by ␣, DNA occupancy of the polymerase site would alter its conformation so that the constraint on the COOH-terminal ␣ domain is removed, permitting it to revert to a relaxed conformation with higher affinity for ␤ (Fig. 7B). Upon completion of Okazaki fragment synthesis, and provided the affinity of the polymerase at a nick results in a lower DNA-polymerase affinity (a feature of the model that has not been verified quantitatively), dissociation of the polymerase from DNA would result in a lower affinity for ␤, permitting polymerase cycling. This model may be oversimplified since ␤ presumably functions by tethering the polymerase to DNA, so that if the polymerase transiently dissociates during elongation, it would rapidly rebind. Such a mechanism would not be tenable if DNA dissociation automatically caused a break in the ␣-␤ link. Answers to these questions may rest in the relative rates of polymerase reassociation with a primer terminus versus ␤ dissociation. A mechanism in which the conformational state of the polymerase influences complex stability may have other implications, including destabilization of elongating complexes to permit protein exchange during translesion synthesis in SOS mutagenesis.
Both and ␤ bind to the carboxyl-terminal domain of ␣ (Kim and McHenry, 1996;present study). However, the binding domain appears to be larger and more complex. Like ␤ binding, binding is abolished by deletions beyond residue 542 from the FIG. 5. Effect of on the interaction between ␣N⌬542 and ␤ in BIAcore binding analysis. Streptavidin was coupled to the chip as described (Kim and McHenry, 1996), and ␣N⌬542 was immobilized by streptavidin, resulting in immobilization of 5000 RU of ␣N⌬542 for sensorgram 1 and 4700 RU for sensorgram 2. All binding analyses were carried out in buffer TS at a flow rate of 2 l/min at 20°C. Sensorgram 1 represents the sequential injection of ␣N⌬542 (270 nM, 10 l), ␤ (1) (500 nM, 10 l), (1) (500 nM, 30 l), ␤ (2) (500 nM, 10 l), (2) (500 nM, 20 l), and ␤ (3) (500 nM, 10 l). Sensorgram 2 represent the sequential injection of ␣N⌬542 (270 nM, 10 l), ␤ (500 nM, 10 l), (1) (500 nM, 30 l), and (2) (500 nM, 20 l). The sensorgram at the bottom of the figure shows the subtraction of sensorgram 2 from sensorgram 1 to represent ␤ binding to ␣N⌬542complex without the background of simultaneously dissociating from ␣N⌬542. The thick lines indicate binding of ␤ to ␣N⌬542complex. A, deletion of the polymerase domain (residues 1-542) of ␣ results in a protein with enhanced affinity for ␤ 2 . The ring represents ␤ 2 ; the shaded unlabeled parallelograms, and ovals represent the amino-terminal polymerase domain of ␣; T and R represent the carboxylterminal ␣ domain in the taut and relaxed conformations, respectively. Double line represents DNA. B, DNA occupancy of the polymerase active site may permit it to change conformation, removing its constraints on the ␤-binding carboxyl-terminal domain of ␣, and permitting it to relax from the low affinity taut (T) conformer to the high affinity relaxed (R) conformer. Shaded parallelograms and ovals represent the amino-terminal polymerase domain of ␣. C, map of the domains of ␣. amino terminus. However, binding is also abolished by small deletions (48 residues) from the carboxyl terminus, whereas deletions of up to 169 amino acids from the COOH terminus are tolerated without loss of ␤ binding (Fig. 7C). The mapping in Fig. 7C is approximate and only defines the terminal sequences involved in forming a stable domain that permits protein-protein interaction. The actual number of amino acids within these domains involved in the interactions is probably much smaller.
Present models for holoenzyme function assume simultaneous and ␤ binding during elongation. In fact, the ␥ subunit that is linked to in holoenzyme Onrust et al., 1995b) cross-links to primers in an initiation complex at a site between ␤ and ␣ (Reems et al., 1995). We find that ␤ binds to both free ␣ and ␣with approximately the same affinity, although the affinity for ␤ may be slightly diminished by the presence of .
The close proximity of and ␤ may permit modulation of the ␣-␤ interaction by other mechanisms that are sensitive to the state of the primosome, a component that is linked to the holoenzyme via a -DnaB interaction . DnaB serves a dual role as the cellular helicase that separates two parental strands at the replication fork (LeBowitz and Mc-Macken, 1986) and as a mobile promoter that interacts with the DnaG primase, permitting synthesis of the primer for the next Okazaki fragment at the fork (McMacken et al., 1977;Tongu et al., 1994). These contacts may enable a mechanism whereby synthesis of a new primer is communicated to the lagging strand polymerase, perhaps further destabilizing it for cycling to the new primer. Interestingly, also functions to protect ␤ on elongating polymerases, preventing its premature removal by ␥ complex (Kim et al., 1996b). The conformational state of could influence this property. Proof of these hypotheses awaits precise identification of the interaction domains of the replisome subunits and analysis of the strength and functional consequences of altered conformational states.