Biotin tagging deletion analysis of domain limits involved in protein-macromolecular interactions. Mapping the tau binding domain of the DNA polymerase III alpha subunit.

The τ subunit dimerizes DNA polymerase III via interaction with the α subunit, allowing DNA polymerase III holoenzyme to synthesize both leading and lagging strands simultaneously at the DNA replication fork. Here, we report a general method to map the limits of domains required for heterologous protein-protein interactions using surface plasmon resonance. The method employs fusion of a short biotinylation sequence at either the NH2 or COOH terminus of the protein to be immobilized on streptavidin-derivatized biosensor chips. Inclusion of a hexahistidine sequence permits rapid purification and separation of the fusion protein from the endogenous Escherichia coli biotin carboxyl carrier protein. Ten deletions of the α subunit were constructed and purified by Ni2+-nitrilotriacetic acid chromatography and, when required, monomeric avidin chromatography. Each α deletion protein was captured by streptavidin immobilized on a Pharmacia Biosensor BIAcore chip, and the τ binding activity of each α deletion was analyzed using surface plasmon resonance. The τ subunit bound very tightly to a full-length amino-terminal fusion of the biotinylation sequence with α (KD ∼ 70 pM). Four additional NH2-terminal α deletion proteins (60, 240, 360, and 542 residues deleted) retained strong binding activity to the τ subunit (KD = 0.19-0.39 nM), whereas deletion of 705 residues or more from the NH2 terminus of the α subunit abolished τ binding activity. Full-length α that contained a carboxyl-terminal fusion with the biotinylation sequence bound τ strongly (KD = 0.37 nM). However, deletion of 48 amino acids from the COOH terminus totally eliminated τ binding. These results indicate that the COOH-terminal half of the α subunit is involved in τ interaction.

The subunit dimerizes DNA polymerase III via interaction with the ␣ subunit, allowing DNA polymerase III holoenzyme to synthesize both leading and lagging strands simultaneously at the DNA replication fork. Here, we report a general method to map the limits of domains required for heterologous protein-protein interactions using surface plasmon resonance. The method employs fusion of a short biotinylation sequence at either the NH 2 or COOH terminus of the protein to be immobilized on streptavidin-derivatized biosensor chips. Inclusion of a hexahistidine sequence permits rapid purification and separation of the fusion protein from the endogenous Escherichia coli biotin carboxyl carrier protein. Ten deletions of the ␣ subunit were constructed and purified by Ni 2؉ -nitrilotriacetic acid chromatography and, when required, monomeric avidin chromatography. Each ␣ deletion protein was captured by streptavidin immobilized on a Pharmacia Biosensor BIAcore chip, and the binding activity of each ␣ deletion was analyzed using surface plasmon resonance. The subunit bound very tightly to a full-length amino-terminal fusion of the biotinylation sequence with ␣ (K D ϳ 70 pM). Four additional NH 2 -terminal ␣ deletion proteins (60, 240, 360, and 542 residues deleted) retained strong binding activity to the subunit (K D ‫؍‬ 0.19 -0.39 nM), whereas deletion of 705 residues or more from the NH 2 terminus of the ␣ subunit abolished binding activity. Full-length ␣ that contained a carboxyl-terminal fusion with the biotinylation sequence bound strongly (K D ‫؍‬ 0.37 nM). However, deletion of 48 amino acids from the COOH terminus totally eliminated binding. These results indicate that the COOH-terminal half of the ␣ subunit is involved in interaction.
DNA polymerase III holoenzyme (holoenzyme) 1 is the major DNA polymerase responsible for Escherichia coli chromosomal DNA synthesis. Holoenzyme consists of 10 individual subunits: ␣, ⑀, , , ␥, ␦, ␦Ј, , , and ␤ (McHenry, 1988;Kornberg, 1988), all of which act cooperatively in the coordinate and processive synthesis of leading and lagging strands (Wu et al., 1992a(Wu et al., , 1992bFay et al., 1982). Each subunit is encoded by a separate gene on the chromosome, except for and ␥, which are both expressed from dnaX. The subunit (71 kDa) is the full-length product of the dnaX gene; the ␥ subunit (47 kDa) is synthesized by a Ϫ1 translational frameshift and comprises the NH 2 -terminal two-thirds of the subunit (McHenry et al., 1989;Tsuchihashi and Kornberg, 1990;Blinkowa and Walker, 1990;Flower and McHenry, 1990).
Polymerase III core (pol III) is composed of three different subunits: ␣, ⑀, and (McHenry and Crow, 1979). The ␣ subunit (dnaE product), a 130-kDa polypeptide, has intrinsic DNA polymerase activity (Welch and McHenry, 1982;Maki et al., 1985), the ⑀ subunit (dnaQ product) contains 3Ј 3 5Ј exonuclease activity for proofreading (Scheuermann and Echols, 1984), and the function of the subunit (holE product) is unknown (Carter et al., 1993;Studwell-Vaughan et al., 1993;Slater et al., 1994). Pol III can be dimerized through the interaction of the ␣ subunit of pol III and the subunit (McHenry, 1982;Studwell-Vaughan and O'Donnell, 1991), presumably allowing two polymerases to synthesize lagging and leading strands simultaneously without dissociation. Pol III itself has a low processivity (10 nucleotides/binding event) in low ionic strength and is completely distributive under physiological conditions (Fay et al., 1981). The processivity of pol III can be increased by the addition of the subunit to form pol IIIЈ in the presence of spermidine (60 nucleotides/ binding event). Pol III becomes highly processive on a primed template with two other components, the ␤ subunit and dnaX complex (Fay et al., 1981).
The subunit binds tightly to the ␣ subunit, but the ␥ subunit, a product missing the COOH-terminal third of the subunit, does not bind to ␣ , suggesting that the COOH-terminal region of the subunit is absolutely required for ␣ binding. Nevertheless, the ␥ subunit can replace the subunit in reconstituted holoenzyme assays on long single-stranded templates . The ( 4 ␦␦Ј) or ␥ (␥ 4 ␦␦Ј) complex loads the ␤ sliding clamp onto a primed template in an ATP-dependent reaction . The ␦ and ␦Ј subunits are required for the DnaX proteins to be a functional clamp loader (Onrust and O'Donnell, 1993). Two other subunits, or , help the DnaX subunits to assemble a functional complex with ␦␦Ј at physiological levels (Olson et al., 1995). The ␤ sliding clamp confers high processivity to the holoenzyme through the interaction with ␣ (LaDuca et al., 1986;Stukenberg et al., 1991).
Clearly, the special replicative role of DNA polymerase III is facilitated by a number of protein-protein interactions with auxiliary proteins. We would expect the relatively large polymerase catalytic subunit ␣ (130 kDa; 1160 amino acids) to contain a polymerase domain plus additional domains that enable these special interactions within the DNA polymerase III ho-loenzyme replicative complex. Here, we report the development of a technique to permit rapid assessment of domain boundaries required for ␣interaction. This was made possible by development of vectors that permit expression of deletion-containing ␣ subunits with a biotinylated peptide fused to the terminus containing the deletion. This biotin tag together with an adjacent hexahistidine sequence can be used to detect, purify, and immobilize the fusion proteins on a streptavidinderivatized BIAcore biosensor chip. Immobilized ␣ derivatives were analyzed for binding using surface plasmon resonance. An important development in this technique was the discovery of a short 13-nucleotide consensus sequence that can be biotinylated in vivo and in vitro (Schatz, 1993). This permits the use of a relatively short unstructured biotinylated peptide rather than the minimal 75-amino acid sequence from the biotin carboxyl carrier protein of acetyl-CoA carboxylase in terminating the fusion proteins. The technique reported here should be generally applicable to a variety of interacting macromolecular systems.

Chemicals and Reagents
d-Biotin was purchased from Sigma. Nitro blue tetrazolium chloride and 5-bromo-4-chloro-3Ј-indolyl phosphate p-toluidine salt were obtained from Life Technologies, Inc. Ni-NTA (nitrilotriacetic acid) resin, the QIAquick Gel extraction kit, and the plasmid preparation kit were from QIAGEN. Immobilized avidin AffinityPak column and SoftLink SoftRelease avidin resin were purchased from Pierce and Promega, respectively. CM5 sensor chips (research grade), Nhydroxysuccinimide,carbodiimide and ethanolamine hydrochloride were obtained from Pharmacia Biosensor.

Enzymes
Alkaline phosphatase-conjugated streptavidin and immunopure streptavidin were obtained from Pierce. Pfu DNA polymerase was purchased from Stratagene. The ␣ subunit was purified to homogeneity from an overproducing E. coli strain and had a specific activity of 5.9 ϫ 10 6 units/mg on activated calf thymus DNA (Kim and McHenry, 1996). The subunit was purified as described .

Promoters
The P A1/O4 / O3 promoter/operator (referred to as P A1 ) was constructed by the combination of an early E. coli RNA polymerase-dependent T7 promoter A1 and two lac operators (Lanzer, 1988;Lanzer and Bujard, 1988); O4 sequences carry a 17-bp core region (TTG TGA GCG GAT AAC AA) of the E. coli lac operator between the Ϫ10 and Ϫ33 hexamers of the promoter, and O3 sequences contain the 29-bp wild-type lac operator downstream of the Ϫ10 hexamer of the promoter. P A1 is 136-fold more tightly repressible than the tac promoter (P tac ), but upon induction with isopropyl-␤-D-thio-galactoside, is 2-fold more active than GATCCGGTACCACCCAGTGCGGCCGCACTAGTG P tac . 2 The T7 promoter (Studier, et al., 1990) was derived from pET-11c vector (Novagen).

Construction of Fusion Vectors with the P A1 Promoter
Vectors were constructed to enable NH 2 -or COOH-terminal fusions with a peptide containing a biotinylation site, a hexahistidine tag, and a thrombin cleavage site. Four oligonucleotides were used to generate a double-stranded DNA fragment encoding the fusion peptide. Fig. 1 shows the overall cloning scheme. For the NH 2 -terminal fusion vector, oligonucleotides 2254 and 2255 were phosphorylated by T4 polynucleotide kinase and annealed to oligonucleotides 2253 and 2256 to generate a DNA duplex. These four annealed oligonucelotides were ligated to each other in the presence of T4 DNA ligase. This DNA fragment was directly ligated to the XbaI-SphI fragment of plasmid pDЈDSP.4 (Pritchard et al., 1996) containing a P A1 to generate pDRK-N (Fig. 1A). DNA sequencing of this vector revealed that two nucleotides were missing from the oligonucleotide insert. The pDRK-N vector was repaired by replacing a BstBI-PstI DNA fragment with an annealed product of oligonucleotides 2406 and 2407, resulting in plasmid pDRK-N(M). This replaced fragment encoded the same amino acid sequence as the original sequence, although some bases were changed to avoid a potential hairpin structure. The DNA sequence at the oligonucleotide-generated region of pDRK-N(M) was verified.
For the COOH-terminal fusion vector, oligonucleotides 2258 and 2259 were phosphorylated and ligated to oligonucleotides 2257 and 2260 to generate a DNA duplex. This DNA fragment was ligated to the NcoI-KpnI fragment of pDЈDSP.4 to generate pDRK-C (Fig. 1B).

Construction of Fusion Vectors with the T7 Promoter
To make NH 2 -and COOH-terminal fusion vectors containing a T7 promoter, the NdeI-BamHI fragment of pET-11c was replaced by a DNA fragment with annealed oligonucleotides 100 and 200, containing restriction enzyme sites for NdeI, NheI, NotI, DraIII, KpnI, and BamHI in order to provide restriction enzyme sites after the T7 promoter to facilitate subsequent cloning steps (Fig. 1A). The resulting plasmid was called pET11c(M). The NdeI-DraIII fragment, containing a fusion peptide region of pDRK-N(M), was inserted into the NdeI-DraIII sites of pET11c(M) to generate pET11c(M)-N. The PstI site at the ␤-lactamase gene of pET11c(M)-N was removed by replacing the PvuI-AlwNI fragment with one derived from pHN1 (Kim and McHenry, 1996) because the other PstI was required to make the fusion protein of the desired gene. This plasmid was derived from pJF118EH (Fü rste et al., 1986) containing a ␤-lactamase gene in which the PstI site was modified but still displayed an Amp R phenotype. The resulting NH 2 -terminal fusion vector, containing a T7 promoter, was pET11-N. The COOH-terminal fusion vector, pET11-KC, was created by ligation of the XbaI-KpnI fragments of pDRK-C and pET11c(M). Fig. 2 depicts the final fusion vectors and their fusion peptide regions. The biotin tag in the fusion protein was used for detection, purification, and immobilization through a specific interaction with streptavidin or avidin. A hexahistidine sequence was introduced for protein purification using Ni 2ϩ -NTA chelating chromatography. Four glycines provided a hinge region between the fusion peptide and the protein of interest.

Growth and Induction of Overexpressing E. coli Strains
E. coli strains containing overexpressing plasmids were grown in 1 liter of F-medium (1.5% yeast extract, 1% peptone, 1.2% K 2 HPO 4 , 0.02% KH 2 PO 4 and 1% glucose) plus 50 g/ml ampicillin at 37°C unless otherwise stated. Cells were induced with isopropyl-␤-D-thio-galactoside (1 mM final concentration), followed by addition of d-biotin and ampicillin to 10 M and 50 g/ml, respectively. After 3 h of induction, cells were harvested by centrifugation at 5,860 ϫ g for 10 min at 4°C and resuspended in 1 ml of Tris-sucrose buffer (50 mM Tris-HCl (pH 7.5) and 10% sucrose)/g of cells. Cells were quickly frozen in liquid N 2 and stored at Ϫ80°C. Overproduction of ␣N⌬812 was performed at 15°C in a 200 L fermentor using similar conditions, except the pH was maintained at 7.5 and cells were rapidly harvested with chilling as described (Cull and McHenry, 1995).

Purification of the ␣ Deletion Proteins
Ni 2ϩ -NTA Ion Chelating Chromatography-Cells were lysed as described (Cull and McHenry, 1995) in the presence of 2 mg of lysozyme/g of cells, 5 mM benzamidine, and 1 mM PMSF. Cells containing ␣N⌬812 deletion mutant were lysed in 1 mg/ml lysozyme and 0.1% Triton X-100 2 H. Bujard, personal communication. non-ionic detergent without heat shock to minimize protein precipitation. Lysates were centrifuged at 23,300 ϫ g at 4°C for 1 h. Supernatants were precipitated by the addition of an equal volume of saturated ammonium sulfate solution and centrifuged at 23,300 ϫ g at 0°C for 1 h. The protein pellet was dissolved in buffer N and applied to a pre-equilibrated Ni 2ϩ -NTA column and recirculated three times at a flow rate of 2 column volumes/h. The column load was 100 mg of protein/ml for proteins that are expressed at ϳ5% of total cell protein; for those that were not expressed at that level, column loads were adjusted up to 300 mg/ml. The column was washed with 10 column volumes of buffer S plus 10 mM imidazole (pH 7.8). Proteins were eluted with 10 column volumes of a 10 -100 mM imidazole gradient in buffer S. Activity eluted approximately halfway through the gradient. All protein preparations were performed at 0 -4°C unless stated otherwise.
Monomeric Avidin Affinity Chromatography-When additional purification was required, the pooled material from the Ni 2ϩ -NTA column was loaded directly (ϳ20 nmol of biotinylated protein/ml of resin) onto a pre-equilibrated SoftLink-SoftRelease avidin column 3 and recirculated five times at a flow rate of 5 column volumes/h. The column was washed with 20 column volumes of buffer H. Buffer H containing 5 mM biotin was added until one column volume of buffer passed through the column, and the flow was stopped. The column was incubated at 4°C for 30 min to allow biotinylated proteins to dissociate from avidin, and biotinylated proteins were eluted with 10 column volumes of buffer H containing 5 mM biotin, typically between 3 and 6 column volumes. The pool of the monomeric avidin fractions was exhaustively dialyzed against buffer H to remove free biotin.
Biotin Blot-Protein samples (ϳ0.1-1 g of biotinylated protein) were run on 10% SDS-polyacrylamide gels, and transferred to a nitrocellulose membrane at 500 A for 3 h in 25 mM Tris-HCl, 192 mM glycine (pH 8.3), 20% methanol and blocked in TBS containing 3% nonfat milk for 1 h. The membrane was incubated in TBS containing 0.3% nonfat milk and alkaline phosphatase-conjugated streptavidin (2 g/ml) for 1 h, washed in TBS plus 0.3% nonfat milk three times, developed in 10 ml of a solution containing nitro blue tetrazolium chloride (1.65 mg) and 5-bromo-4-chloro-3Ј-indolyl phosphate p-toluidine salt (2.2 mg) in 10 mM diethanolamine (pH 9.5) and 10 mM MgCl 2 and washed in water to stop the reaction.
Surface Plasmon Resonance-A Pharmacia Biosensor BIAcore instrument was used for protein-protein binding analysis. All buffers were filtered before use. Streptavidin (30 l, 200 g/ml) in coupling buffer (10 mM sodium acetate (pH 4.5)) was injected over a CM5 sensor chip at 5 l/min to capture streptavidin to the carboxymethyl dextran matrix of the chip by NHS/EDC coupling reaction (30 l of mix) as described (Olson et al., 1995). Unreacted N-hydroxysuccinimide ester groups were inactivated using 1 M ethanolamine-HCl (pH 8.0). Typically, this reaction immobilizes about 5000 response units (RU) of streptavidin. The biotinylated ␣ deletions were then injected over the immobilized streptavidin in HBS buffer. Typically, 500-1000 RU of each ␣ deletion bound to the immobilized streptavidin. Binding analysis of the ␣subunit was performed in HKGM buffer at 20°C at varying concentrations. Kinetic parameters were determined using the BIAevaluation 2.1 software.
Gap-filling Polymerase Assay-This assay was performed as described by McHenry and Crow (1979). It is used to measure the minimal polymerase activity of the DNA polymerase III core in an assay that does not require interaction with other holoenzyme subunits. The reaction was initiated by the addition of the ␣ subunit to a solution (25 l) containing 32 mM HEPES (pH 7.5), 6.4 mM DTT, four dNTPs (100 cpm/pmol dNTPs), 10 mM MgCl 2 , and 1 g of activated calf thymus DNA. One unit is defined as the amount of enzyme catalyzing the incorporation of 1 pmol of dNTPs/min at 30°C.
Other Methods-Protein concentrations were determined by the method of Bradford (1976). SDS-polyacrylamide gel electrophoresis was performed as described by Laemmli (1970).

RESULTS
Deletion of the ␣ Subunit-To identify the NH 2 -terminal limit of sequences required to form the -binding domain of ␣, we constructed a series of deletions of the ␣ subunit using the polymerase chain reaction (Fig. 3). All deletion proteins were FIG. 2. Fusion peptide region of vectors. A, NH 2 -terminal fusion vector with either T7 promoter (pET11-N) or P A1 (pDRK-N(M)). The DNA sequence represents the fusion peptide region encoding a short sequence that is biotinylated in vivo (indicated as bold amino acids) (Schatz, 1993), a hexahistidine sequence, and a thrombin cleavage site. Fusions to the NH 2 terminus of proteins can be achieved using cloning sites: PstI, AvrII, DraIII, SphI (P A1 vector only), KpnI, or BamHI (P T7 vector only). The vector with P A1 contains lacI q instead lacI. S/D and T indicate a Shine-Dalgarno site and transcriptional terminator, respectively. The vertical arrow indicates the thrombin cleavage site. B, COOH-terminal fusion vector. Using restriction endonucleases of XbaI, NcoI, NotI, DraIII, or SpeI, the COOH terminus of a protein can be fused to the fusion peptide. Two tandem stop codons were introduced at the end of the fusion peptide.
To identify the COOH-terminal end of sequences required to form the -binding domain, plasmids that express COOH-terminal ␣ deletion proteins were constructed. The NcoI-DraIII fragment of pHN4 (Kim and McHenry, 1996) was cloned into the pDRK-C vector at the corresponding sites to create pDRK-CdnaE, which does not produce an ␣ fusion protein (Fig. 1B). This plasmid was digested with SpeI and StuI, and the large fragment was ligated to PCR products digested with SpeI and StuI to produce plasmids pA1-C0 and pA1-C1, which express proteins ␣C⌬0 and ␣C⌬48 (Fig. 3B).
Expression of the ␣ Deletion Proteins-Based on expression and solubility of proteins, we selected a panel of ␣ deletion constructs for expression in E. coli: ␣N⌬1, ␣N⌬60, ␣N⌬240, ␣N⌬360, ␣N⌬705, and ␣N⌬859 from the P A1 , and ␣N⌬542, ␣N⌬812, ␣C⌬0, and ␣C⌬48 from the T7 promoter. E. coli strain HMS174 was used for the P A1 -based plasmids, and BL21(DE3) for T7 promoter-based plasmids to express the deletion proteins. Cells expressing the ␣N⌬1, ␣N⌬60, and ␣N⌬240 deletions were grown and induced at 37°C, but other deletion strains were grown and induced at room temperature except cells expressing ␣N⌬812 (15°C) because they were insoluble when expressed at 37°C. The expressed fusion proteins were detected by a biotin blot as described under "Experimental Procedures." Fig. 4 shows the expression pattern of the ␣N⌬1 deletion protein. Based on a densitometric scan of the gel, ␣N⌬1 represented ϳ5% of the total protein. Biotin blotting (lane 3) revealed a high-molecular-weight protein corresponding to the ␣N⌬1 deletion protein and the biotin carboxyl carrier protein of acetyl-CoA carboxylase (BCCP), the only other biotinylated protein in E. coli (Samols et al., 1988).
Purification of the ␣ Deletion Proteins-Two special properties of the fusion peptide enabled purification of the ␣ deletion proteins to near homogeneity in one or two chromatographic steps. The hexahistidine sequence fused to the ␣ sequences specifically interacts with Ni 2ϩ ions chelated to column resins, and biotinylated deletion proteins were further purified by exploiting the high affinity interaction between biotin and avidin. The ␣N⌬1, ␣N⌬60, ␣N⌬240, and ␣N⌬542 proteins were purified to near homogeneity by Ni 2ϩ -NTA ion chelating chromotography alone as determined by SDS-polyacrylamide gel analysis (Fig. 5). ␣N⌬360, ␣N⌬812, ␣C⌬0, and ␣C⌬48 proteins required further purification by monomeric avidin chromatography, resulting in nearly homogeneous preparations (Fig. 5). Protein ␣C⌬48 was expressed at a low level requiring a 270fold purification. For ␣N⌬812, only 30 g was obtained from 21 g of lysate protein. The protein was catalytically inactive, precluding accurate determination of yields, but if yields of the soluble fraction of protein were the same as ␣C⌬0 and ␣C⌬48, a 30,000-fold purification was achieved, illustrating the power of two affinity tags in one fusion peptide in facilitating difficult purifications. Table II summarizes the purification for ␣N⌬1, ␣N⌬812, ␣C⌬0, and ␣C⌬48 deletion proteins. The gap-filling polymerase assay indicated that ␣N⌬1 and ␣C⌬0 were fully active, whereas the ␣C⌬48 deletion retained ϳ85% of the gap filling activity of wild-type ␣. Purification of the other deletion products was monitored by gel electrophoresis because they showed no detectable polymerase activity in the gap-filling assay. Twelve mg of ␣N⌬60, 15 mg of ␣N⌬240, 21 g of ␣N⌬360, and 6 mg of ␣N⌬542 were purified from 1500, 2000, 960, and 390 mg of total proteins from cell lysates, respectively. ␣N⌬705 (3.2 mg) and ␣N⌬859 (1.4 mg) were partially purified from 470 and 530 mg of cell lysate proteins, respectively. The   FIG. 3. A, strategy used to construct deletions in NH 2 -terminal fusion proteins of the ␣ subunit. PCR primers (2276, 2278, 2279, 2280, 2281, 2286, 600, or 700) containing a 12-nucleotide noncomplementary sequence with a PstI site are followed by an 18-nucleotide sequence complementary to the first 6 dnaE codons of the intended fusion protein. A second primer (114, 48, or 2752) was used to generate the desired PCR product. PCR was carried out in a Minicycler (MJ Research) for 35 cycles of denaturation at 94°C, annealing at 52°C and extension at 72°C in a buffer (20 mM Tris-HCl (pH 8.75), 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 2 mM MgCl 2 , 0.1% Triton X-100, and 100 g/ml) containing two primers (each 1 M), DNA template (200 ng), four dNTPs (each 200 M), and Pfu DNA polymerase (2.5 units). The resulting PCR product was cleaved with PstI and either AflII for ␣N⌬1, ␣N⌬60 and ␣N⌬240, BglII for ␣N⌬360 and ␣N⌬542, or StuI for ␣N⌬705, ␣N⌬812 and ␣N⌬859, and ligated to the corresponding fragment of pDRK-NdnaE or pET11-NdnaE. The fusion region immediately preceding ␣ was Leu-Gln preceded by the 4 glycines shown in Fig. 2. B, strategy used to construct deletions in COOH-terminal fusion proteins. PCR primers (2287 and 2288) containing a noncomplementary sequence with a SpeI site are followed by an 18-nucleotide sequence complementary to the last 6 codons of the intended fusion protein. A second primer (47) was used to generate the desired PCR product. The resulting PCR product was cleaved with SpeI and StuI and cloned into the corresponding sites of pDRK-CdnaE or pET11-CdnaE. The fusion region immediately following ␣ was Thr-Ser, followed by the 4 glycines shown in Fig. 2. purity of all purified proteins was determined by 10% SDSpolyacrylamide gel electrophoresis (Fig. 5). The degree of biotinylation of ␣N⌬1 was determined to be 80% using immobilized avidin AffinityPak column analysis. 4 Fusion Peptide Does Not Block Interactions with Other Subunits-Holoenzyme reconstitution assays of two full-length ␣ fusion proteins were performed as described  to determine whether the fusion peptide interferes with the interaction of ␣ and other accessory subunits required for full holoenzyme activity. ␣N⌬1 and ␣C⌬0 had a specific activity of 2.2 ϫ 10 6 and 2.4 ϫ 10 6 units/mg, respectively, comparable to the activity of wild-type ␣ (2.9 ϫ 10 6 units/mg). 5 Thus, the fusion peptide did not block interactions between full-length ␣ and other subunits required in the holoenzyme reconstitution assay.
␣-Binding Analysis Using Surface Plasmon Resonance-Kinetic parameters for ␣binding were determined using the ␣N⌬1 deletion protein immobilized to a sensor chip in a BIAcore. Streptavidin was chemically coupled to the chip and ␣N⌬1 was immobilized via the biotin-streptavidin interaction. Three different concentrations (34 nM, 68 nM, and 170 nM) of the subunit were used for binding analysis. The off-rate (k off ) was calculated by injection of saturating amounts of over the ␣N⌬1 fusion to minimize reassociation and was used in calculations of the on-rates (k on ) at all three concentrations of ( Table III). dissociated very slowly from ␣ with a half-life of ϳ3.4 h (Fig. 6). The dissociation constant (K D ) of ␣binding was determined to be approximately 70 pM. BIAcore analysis of ␥-␣ or ␥-complex (␥ 4 ␦␦Ј␥)-␣ binding revealed no detectable interaction between ␣ and ␥ (data not shown).
Mapping the Binding Domain of the ␣ Subunit-The binding domain of the ␣ subunit was mapped by BIAcore binding analysis using a series of deletions from the NH 2 or COOH terminus. All ␣ deletion proteins were individually immobilized to the streptavidin chip, and the binding with the subunit was analyzed. ␣ fusions with 60, 240, 360 and 542 amino acids deleted from the NH 2 terminus all bound tightly to the subunit (Table IV), with K D values in the range of 0.19 -0.39 nM. A further ␣ deletion of 812 amino acids from the NH 2 terminus yielded a protein that interacted weakly with with a 5% binding stoichiometry. It is possible that most of the ␣N⌬812 deletion protein has an unfolded structure that is unable to bind to and a low level of properly folded proteins results in substoichiometric binding to . However, we are unable to distinguish this possibility from a low-level nonspecific interaction with an unfolded ␣ COOH-terminal domain. Two other plasmids pA1-N6 and pA1-N7 were constructed to express larger and smaller ␣ deletion proteins (␣N⌬859 and ␣N⌬705) in hope of obtaining better behaved proteins. Both ␣N⌬705 and ␣N⌬859 deletion proteins were partially purified from Ni 2ϩ -NTA column chromatography and final purification was carried out on the BIAcore sensor chip by a specific interaction between biotinylated protein and immobilized streptavidin. Neither of the two deletions bound to the subunit on BIAcore.
The COOH-terminal fusion, ␣C⌬0, bound almost as tightly (K D ϭ 0.37 nM) as the other NH 2 -terminal deletion proteins (Table IV). The ␣C⌬48 deletion protein lost binding activity to the subunit, but the polymerase domain was functionally intact as indicated by its nearly full polymerase activity in the gap-filling assay (Table II). Further deletions of ␣ from the 4 Purified ␣N⌬1 was loaded onto the immobilized avidin (tetramer) column (1-ml prepacked column from Pierce) at 5-fold less column capacity. All biotinylated proteins bound tightly to the column, and unbiotinylated proteins flowed through. The extent of biotinylation was calculated by analyses of load, flow-through, and bound fractions of ␣N⌬1 in activity and gel electrophoresis. 5 For purposes of convenient comparison, the specific activity of ␣ was calculated using the protein concentration determined by Bradford (1976). The true specific activity of ␣ using 280 was 6.6 ϫ 10 6 units/mg (Kim and McHenry, 1996).

FIG. 4. Expression and detection of ␣N⌬1 fusion protein. E. coli
strain HMS174 containing plasmid pA1-N0, which expresses the ␣N⌬1 deletion protein, was grown as described under "Experimental Procedures." Just before induction and 3 h later, 1.5 ml of the liquid culture was removed and spun in a microcentrifuge for 5 min at room temperature. The cell pellet was resuspended in 105 l/A 600 of SDS sample buffer (50% glycerol, 10 mM DTT, 0.005% bromphenol blue, and 1% SDS), boiled for 10 min, and spun at room temperature for 10 min. Each sample (20 l) was loaded onto a 10% SDS-polyacrylamide gel.  Laemmli (1970) and 1 g of each purified protein or 13 and 10 g of total protein from ␣N⌬705 and ␣N⌬859, respectively, was loaded into the gel and separated overnight at a constant 65 V. The asterisk indicates the ␣N⌬705 and ␣N⌬859 proteins. The gel was stained with Coomassie Brilliant Blue overnight and destained in a solution of 10% methanol and 10% acetic acid to visualize proteins. The lane on the left contains protein molecular size standards.
COOH terminus showed no interaction with the subunit (data not shown). The stoichiometric ratio of the -␣ interaction was about 1:1 (Table IV), consistent with the assigned stoichiometry of pol IIIЈ in solution (␣ 2 2 ) (McHenry, 1982;Studwell-Vaughan and O'Donnell, 1991). Presumably, two immobilized ␣s come together to bind a dimer. 6 Superose 12 Gel Filtration of ␣N⌬705 and -To examine the possibility that ␣N⌬705 and interact at levels too low or with such slow binding kinetics that no binding is detected on the BIAcore, we tested the binding of ␣N⌬705 to by Superose 12 gel filtration chromatography. ␣N⌬542, the last deletion protein that still had binding activity, was used as a positive control. ␣N⌬542 and 4 alone were detected at fractions 25-27 and 19 -21, respectively (Fig. 7, A and B). The elution position of ␣N⌬542 protein shifted to near the 4 7 peak at fractions 19 -21 when the mix of ␣N⌬542 and was applied to the gel filtration column (Fig. 7C), indicating that ␣N⌬542 bound tightly to . When ␣N⌬705 alone was injected into the column, it eluted in a broad range (fractions 16 to 28), giving two separate peaks (Fig. 7D). The first peak might be an unfolded or aggregated form of ␣N⌬705 eluting in the excluded volume, while the second peak may represent the properly folded form. The mix of ␣N⌬705 and also eluted broadly, but two separate peaks were not detected and the low molecular weight ␣N⌬705 was shifted to a higher molecular weight (Fig. 7E). These data suggest that ␣N⌬705 could be shifted by the interaction with , although the partial unfolding or aggregation of ␣N⌬705 complicate interpretation of the data.
48-Amino Acid Deletion from the COOH Terminus Eliminates Binding Activity of ␣-␣C⌬48 was unable to interact FIG. 6. A sensorgram of ␣N⌬1binding on BIAcore. The binding assay was performed as described under "Experimental Procedures." Streptavidin was chemically immobilized to the sensor chip, and ␣N⌬1 was captured by the streptavidin-biotin reaction, yielding about 750 RUs. Three different concentrations of the subunit (34 nM, 68 nM, and 170 nM) were injected over ␣N⌬1 for 6 min at 5 l/min. The background resulting from injection of buffer alone was subtracted from the data before plotting. k off was determined between 600-1900 s.   with the subunit (Table IV). Thus, ␣C⌬48 could not be dimerized by for the coordinated DNA synthesis of both leading and lagging strands. When this protein was analyzed in holoenzyme reconstitution assays using complex ( 4 ␦␦Ј␥) , which requires interactions with other subunits, ␣C⌬48 possessed only 7% of the activity of wild-type ␣. Interestingly, when ␥-complex (␥ 4 ␦␦Ј) replaced complex in this assay, ␣C⌬48 retained about 40% activity of wild-type ␣.
Nevertheless, ␣C⌬48 showed nearly full polymerase activity in gap-filling assay (Table II). Thus, the COOH terminus of the ␣ subunit is required for full reconstitution of an active replication complex. DISCUSSION We report here a novel method to map the terminal limits of the binding domains of a protein in the BIAcore using a series of deletion proteins fused to a 13-amino acid consensus sequence (Schatz, 1993) that is efficiently biotinylated in vivo in the presence of 10 M biotin. The high-affinity binding of streptavidin or avidin with biotin (Green, 1975) can be exploited to immobilize proteins on a sensor chip coupled with streptavidin, providing a ϳ90 Å tether to the chip and minimizing steric interference. Unlike the immobilization resulting from NHS/EDC coupling reaction which uses lysine residues of the ligand protein, the targeted protein is immobilized homogeneously in an oriented manner. This permits a clean kinetic analysis from a uniform population. The ability to analyze fusions with both protein termini permits determination, by comparison, of whether either fusion interferes with specific interactions.
The presence of a fusion peptide that interacts with two affinity matrices provides a powerful and rapid purification method, even for proteins that are expressed at very low levels. For most proteins, Ni 2ϩ -NTA chromatography provides homogeneous protein. The only naturally biotinylated protein in E. coli, the biotin carboxyl carrier protein, flows through Ni 2ϩ -NTA columns, providing preparations that can be further purified, when necessary, on reduced affinity monomeric avidin columns. Use of monomeric instead of native avidin permits bound protein to be eluted with 5 mM biotin under mild conditions. For some types of analyses, the avidin chromatography step can be skipped, since the streptavidin-BIAcore chip also provides a purification, as non-biotinylated proteins wash off during the chip preparation stage and should not be present during the binding analysis. In this report, we exploited this technique for two ␣ derivatives, ␣N⌬705 and ␣N⌬859, that were expressed at very low levels.
We mapped the binding domain of the 1160-amino acid ␣ subunit using eight proteins containing deletions from the NH 2 terminus and two proteins containing deletions from the COOH terminus. ␣ proteins containing amino-terminal deletions up to 542 amino acids showed high-affinity binding to the subunit and displayed very similar off-rates (ϳ 3.4 h half-life). Deletion of 705 or more residues from the amino terminus yielded proteins that showed no or limited interactions with . Limited interaction, with only a 5% stoichiometric interaction, was observed between and ␣N⌬812. Two additional fusions, ␣N⌬705 and ␣N⌬859, designed with deletions that flank either side of the 812 residue deletion, did not show interaction in the FIG. 7. Superose 12 gel filtration. A 24-ml Superose 12 FPLC gel filtration column was equilibrated with buffer E. A 200-l volume of each sample was injected into the column at 0.1 ml/min. Forty fractions of 0.5 ml were collected. ␣N542 (0.35 nmol) and (1.4 nmol as monomer) were mixed in buffer E and incubated at room temperature for 15 min before injection. The mixture of ␣N705 (0.35 nmol, assumed 20% purity) and (1.4 nmol) was incubated at room temperature for 15 min or on ice overnight before injection. Fractions (70 l) were separated by 10% SDS-polyacrylamide electrophoresis, and the gel was either stained with Coomassie Brilliant Blue (A) or subjected to a biotin blot (B-E) as described under "Experimental Procedures." only (A), ␣N⌬542 only (B), -␣N⌬542 (C), ␣N⌬705 only (D), and -␣N⌬705 (E) were gel-filtered. BIAcore, either due to deletion of part of the -binding site or due to deletions of sequences required for stable folding of the -binding domain. Further analysis of the ␣N⌬705 deletion protein by Superose 12 gel filtration indicated that a significant portion of ␣N⌬705 was excluded, consistent with an unfolded and/or aggregated conformation. Addition of to ␣N⌬705 shifted the lower Stokes' radius peak, presumably properly folded ␣N⌬705, to a higher molecular weight complex, suggesting a limited, weak interaction. These results suggest that residues beyond 542 in the 705-812 range stabilize a domain involved in binding but that the actual binding site is COOHterminal to these residues. This conclusion is supported by the abrogation of binding by deletion of only 48 residues from the carboxyl terminus.
From these results, we conclude that the COOH-terminal region of the ␣ subunit is involved in binding. This result is consistent with the retention of auxiliary subunit independent gap-fillingpolymeraseactivityby␣C⌬48butthelossof-complexdependent holoenzyme activity. Photocrosslinking experiments (Reems et al., 1995) demonstrated that holoenzyme in initiation complexes involved interactions with the polymerase between 13 nucleotides at the 3Ј-primer terminus and additional contacts at position Ϫ18 and Ϫ22 for the DnaX and the ␤ subunits, respectively. Because of the evidence for interactions between ␣ and DnaX and ␤ that do not map to DNA contacts, we proposed a model where an appendage of ␣, distal from the polymerase active site, contacts these auxiliary subunits (Reems et al., 1995). The present studies would suggest that the appendage is the carboxyl-terminal end of ␣. The special features of a replicative polymerase derive, in part, from its ability to interact with other replication proteins. The biotin tagging deletion analysis approach should be valuable for analyzing other important interactions between DNA polymerase III holoenzyme subunits and should be generally applicable for analyzing other multiprotein assemblies as well.