Competitive processivity‐clamp usage by DNA polymerases during DNA replication and repair

We have recently demonstrated that the C‐terminal 20 residues of the α‐subunit are required for functional interaction with β (López de Saro et al., 2003). To define the critical residues that contribute to the α‐β interaction, we designed overlapping N‐biotinylated peptides spanning the C‐terminal region of α and assayed them for interaction with the β‐clamp by attaching them to streptavidin‐coated microplates and then probing them for ability to bind ³²P‐β. Radiolabeled β used in these studies contained a six‐residue C‐terminal extension that can be phosphorylated with protein kinase A and [³²P]γ‐ATP (Stukenberg et al., 1994; Kelman et al., 1995). We found that peptides corresponding to the C‐terminal 20 amino acids interact with the β‐clamp, as did a peptide lacking the last residue, but deletion of two residues from the C‐terminus abolished the interaction (Figure 1A). Hence, the penultimate F‐residue of the α‐peptide is required for interaction with β.

Next, we used a protein shift assay (López de Saro and O'Donnell, 2001) to determine whether the C‐terminal α‐peptide (α1141–1160) complexed to streptavidin in solution is sufficient to bind to ³²P‐β and to produce a mobility shift in a native polyacrylamide gel. The results showed that streptavidin does indeed produce a slower‐mobility complex, presumably consisting of streptavidin·biotin‐peptide bound to ³²P‐β (Figure 1B, lane 4). We also observed that Pol III core retards the mobility of ³²P‐β in a native gel (Figure 1C, compare lanes 1 and 2). If the polymerase·β interaction requires the α C‐terminal sequences, the α1141–1160 peptide may be expected to disrupt this interaction, thus displacing the ³²P‐β from core polymerase, and the results showed that this is in fact the case (Figure 2C, lane 6). To demonstrate specificity of this interaction further, heterologous clamps were tested for displacement of the Pol III core·³²P‐β interaction. The addition of excess yeast PCNA clamp or phage T4 gp45 clamp did not displace core from ³²P‐β (Figure 1C, lanes 4 and 5, respectively). Nor did the α‐peptide bind to human ³²P‐PCNA or inhibit the binding of ³²P‐PCNA to the cell‐cycle regulator p21^WAFI (data not shown). As expected from their ability to compete with DNA polymerase for β, the β‐binding peptides α1140–1159 and α1141–1160 were inhibitory to Pol III core when added to a replication reaction using primed M13mp18 ssDNA as template; core requires β for synthesis of this substrate (Figure 1D).

Similar binding assays as described above were performed on peptides spanning the N‐terminal region of α and the region of the protein from residue 812 to 991 (gray boxes in the key of Figure 1A), with negative results (data not shown). Since the interaction could require a specific conformation not provided by 20‐mer peptides, our results do not exclude the possibility that additional sites of contact between α and the β‐clamp could be present in this area, as suggested by the studies of Kim and McHenry (1996). On the other hand, our results do not detect the interaction observed by Dalrymple et al. (2001), in which a sequence corresponding to region 918–926 of α (sequence IGQADMFGV), fused onto a reporter protein, resulted in an interaction with the β‐clamp using a yeast two‐hybrid assay. Peptides spanning this region do not show binding to the β‐clamp using the assays and conditions described here (data not shown).

To identify residues at the C‐terminus of the α‐subunit of Pol III that are critical for interaction with β, we performed an alanine scan analysis of the α1141–1160 peptide. These peptides were used in three different assays: (i) direct binding of ³²P‐β to biotinylated peptide immobilized in wells of a microtiter plate (Figure 2A); (ii) ability to displace Pol III core from a ³²P‐β·core complex observed in a native polyacrylamide gel (Figure 2B); and (iii) ability to displace ³²P‐β from δ‐subunit immobilized in wells of a microtiter plate (Figure 2C). The δ‐β displacement assay is based on the previous finding that binding of α and δ to β is sterically exclusive, and that the same single amino‐acid changes on β that abolish interaction with α also abolish interaction with δ, suggesting that α and δ share a common site on β (Naktinis et al., 1996). This conclusion is supported here by the finding that peptide α1141–1160 competes ³²P‐β off the δ subunit (Figure 2C).

While most alanine mutants of the α1141–1160 peptide retain full activity for binding ³²P‐β, mutation of residues Q₁₁₅₄, L₁₁₅₇ and F₁₁₅₉ to alanine clearly diminished the activity of the peptides in all three assays. Interestingly, peptide D1160A showed no retention of ³²P‐β on the microtiter plate (Figure 3A), but was almost as effective as wild‐type peptide in competition assays (Figure 3B and C). Since deletion of residue D₁₁₆₀ (peptide α 1140–1159) in the experiment shown in Figure 1A did not impair binding to ³²P‐β, we suggest that the presence of alanine at the position of D₁₁₆₀ lowers the affinity of peptide for the β‐clamp. The large concentration of peptide used in the competition assays of Figure 1B and C may overcome the deficiency.

In light of the experiments described above, we deleted the seven C‐terminal residues of α that contain the three critical β‐interactive residues identified in the alanine scanning experiments. This C‐terminal deletion will be referred to here as αΔ7. We then labeled the β‐clamp with the fluorescent 488 Oregon Green (OG) probe at the single exposed Cys333, and used fluorescence to determine the affinity of the labeled β‐clamp for wild‐type α and for αΔC7. Cys333 is on the opposite face of β from where α binds, and OG‐labeled β^OG was as active as wild‐type β in DNA synthesis activity assays with γ‐complex and core (data not shown). Titration of α and αΔ7 into β^OG reveals that αΔC7 has a 10‐fold reduced affinity for the β‐clamp (Figure 3A; K_D of 1.16 μM versus 0.11 μM at 100 mM NaCl). The residual binding of αΔC7 for the β‐clamp could be due to additional interaction points between the two proteins, perhaps in regions distinct from the C‐terminal region. Alternatively, C‐terminal residues immediately upstream could participate in binding β via protein backbone contacts with the clamp, as in the p21·PCNA complex (Gulbis et al., 1996). This possibility is supported by our earlier study of αΔ20 (deletion of the C‐terminal 20 residues), which binds β 100‐fold weaker than wild‐type α (López de Saro et al., 2003). In this case, the upstream residues that contribute peptide backbone contacts to β would not be detected by the alanine scanning experiments described above. This residual affinity of αΔC7 for the β‐clamp was clearly salt dependent (Figure 3B), suggesting that the contribution of the last seven residues has a strong hydrophobic component, consistent with the requirement of L₁₁₅₇ and F₁₁₅₉ in the interaction of α with β.

DNA synthesis assays performed with αΔC7 reconstituted with the ϵ and θ subunits to form Pol III (αΔ7) core showed that its activity in a β‐clamp‐dependent replication assay is considerably lower than that of wild‐type core (Figure 3C). This result supports the conclusion that the extreme C‐terminal seven residues of α are important to functional interaction with β. Both core complexes displayed similar activities in a β‐independent extension assay using gapped DNA as a substrate (not shown), indicating that αΔC7 is not impaired in its catalytic activity.

During lagging strand synthesis, the clamp loader (γ‐complex) loads β‐clamps onto multiple RNA primers. The K_D of γ‐complex for β in solution is 8 nM, ∼30‐fold lower than that of the Pol III core·β complex (250 nM). However, when β is on DNA this affinity is reversed, resulting in a K_D value for the Pol III core·β of <5 nM (Naktinis et al., 1996) and a reduced affinity of the γ‐complex for β (Turner et al., 1999; Ason et al., 2000). This DNA‐modulated competition between clamp loader and polymerase for β generates a molecular switch that assures an ordered sequence of events during the assembly of the replicase at a primed site (Ason et al., 2003). A previous study on the site of interaction of δ subunit (of γ‐complex) with β, and α subunit (of Pol III core) with β, demonstrated that both proteins interact with the clamp near the C‐terminus, which is near the hydrophobic pocket between β domains II and III (Naktinis et al., 1996). Here, as shown in Figure 2C, we find that the 20‐mer peptide derived from the C‐terminus of α displaces δ from ³²P‐β. This result indicated that the α and δ proteins bind the same site on β, rather than competing by steric occlusion, in which two large proteins bind different sites on β, but are too close for both α and δ to bind β at the same time.

The ability of the α C‐terminal peptide to compete δ from β suggests that this peptide may inhibit the clamp loading reaction. Indeed, the peptide inhibits loading of the β‐clamp on DNA by γ‐complex, but a related peptide mutated in a single residue (Q1154A) does not (Figure 4). These results further support the conclusion that the site of interaction of δ and α with β is the same, namely the hydrophobic pocket between domains II and III of the β ring, which is near the β C‐terminus.

The α C‐terminal residues identified by alanine scanning as being essential for interaction with β appear to be present in the C‐terminal sequences of Pol II and Pol IV, although the spacing is somewhat different (Figure 5A). A peptide corresponding to the C‐terminus of Pol IV has been shown recently to interact with β directly, and a C‐terminal deletion of this Y‐family polymerase no longer functions with β (Lenne‐Samuel et al., 2002). The experiment shown in Figure 5B (upper panel) demonstrates that immobilized peptides derived from the C‐terminus of either Pol II or Pol IV interact with ³²P‐β and retain ³²P‐β in ELISA plates. If these Pol II and Pol IV peptides bind β at the same site, they should be capable of competing ³²P‐β off δ protein. The experiment shown in Figure 5B (lower panel) illustrates that this prediction is valid. Immobilized δ retains ³²P‐β in the well of a microtiter plate, and the addition of C‐terminal peptide derived from α, Pol II or Pol IV displaces ³²P‐β from the immobilized δ. Further, the Pol II and Pol IV proteins also displace Pol III core from the Pol III core·β complex in the native PAGE assay, as shown in Figure 5C. The 20‐mer peptides derived from the C‐terminus of Pol II and Pol IV also inhibit β‐dependent replication by Pol III core (data not shown).

The alignment of the C‐terminal tails of various polymerases reveals slight differences in the presence and/or spacing of the residues that are important for α to bind β (see Figure 5A). To assess the importance of these variations to the relative affinity to β, we compared the Pol II and Pol III C‐terminal peptides for their effectiveness in disrupting polymerase·clamp complexes. In Figure 5D (upper panel) we titrated the 20‐mer peptide of Pol II into the ³²P‐β·Pol III core complex, and in the lower panel, the C‐terminal Pol III α 20‐mer peptide was titrated into ³²P‐β·Pol II complex. The results demonstrate that the peptides have similar efficiencies in disrupting the polymerase‐clamp complex.

We next determined the apparent K_D of Pol III core·β (51 ± 9 nM) and Pol II·β (120 ± 23 nM) using β^OG (Figure 5E, upper panels). To compare the relative affinity of the C‐terminal 20 residues of both polymerases for β, we used synthetic peptides labeled at their N‐terminus with rhodamine (TAMRA); the apparent K_D values calculated from titration measurements were 3.2 μM for the α peptide and 3.5 μM for the Pol II peptide, revealing that despite the differences in sequence, both peptides bind to the clamp with very similar affinities (Figure 5E, lower panels). Calculation of the free energy of the polymerase·β interaction indicates that ∼75% of the interaction energy between Pol III and β resides in the C‐terminal 20 residues (ΔG = 9.94 kcal/mol for Pol III·β, ΔG = 7.50 kcal/mol for Pol III peptide·β) as does 79% of the interaction energy between Pol II and β (ΔG = 9.45 kcal/mol for Pol II·β, ΔG = 7.44 kcal/mol for Pol II peptide·β).

Our previous studies have shown that the β‐clamp can stimulate the Klenow fragment of DNA polymerase I in synthesis on SSB‐coated, singly primed M13mp18 ssDNA (López de Saro and O'Donnell, 2001). However, the C‐terminus of the Klenow fragment lacks a sequence containing the β‐binding residues of the Pol III α‐subunit. Consistent with this, we detected no binding to the β‐clamp of a C‐terminal 20‐residue peptide derived from the Klenow sequence (data not shown). However, the Pol III α subunit peptide (α 1141–1160) did inhibit β‐stimulated DNA synthesis by Pol I Klenow fragment (Figure 6A). In this experiment the β‐clamp was loaded onto the primed DNA before the peptide and Pol I (Klenow fragment) were added. Hence, the observed inhibition is likely the result of disruption of the Pol I·β contact by the peptide. Further, ³²P‐β is retained in wells by immobilized Klenow fragment (Figure 6B), suggesting that the site on β that is bound by the α C‐terminal peptide and the Pol I Klenow fragment is the same, and that both polymerases interact with the clamp in a competitive manner.

Finally, we turned our attention to Pol V of E.coli, a heterotrimeric polymerase of the Y‐family with the subunit structure UmuD′₂UmuC (Woodgate et al., 1989; Bruck et al., 1996) that is involved in translesion synthesis (Tang et al., 2000). UmuD′ (115 amino acids) is produced by RecA‐induced proteolytic removal of an N‐terminal peptide from UmuD (139 amino acids). We examined 20‐mer peptides derived from the N‐ and C‐terminal sequences of UmuD and UmuC for binding to the β‐clamp, with negative results. We then fixed the full UmuD, UmuD′, UmuC and Pol V complex (UmuD′₂UmuC) to ELISA plates and probed the immobilized protein with ³²P‐β. As shown in Figure 6C, only UmuC and Pol V complex showed an interaction with ³²P‐β, and these interactions were disrupted by the addition of the Pol III α C‐terminal peptide (α 1141–1160). These results indicate that the main connection between Pol V and β is mediated by the single UmuC subunit, and that it binds to β in the same location as Pol III and the other DNA polymerases. Consistent with this, Dalrymple et al. (2001) have reported, on the basis of sequence similarity, a putative binding site on UmuC for the β‐clamp that is situated ∼62 residues from the C‐terminus, and more recently it has been shown that removal of these residues abolishes the lesion bypass ability of Pol V (Becherel et al., 2002). This sequence (QLNLF) resembles the motifs present in the other E.coli DNA polymerases. An interaction between UmuD and β has been reported using crosslinking assays (Sutton et al., 1999). Despite identification of a UmuC‐β interaction, we cannot exclude the possibility that UmuD or UmuD′ may interact with β in a fashion undetected under the conditions of this report.

This report demonstrates that E.coli Pol II, III and IV interact with the β‐clamp via the C‐terminal tail of these polymerases. It has been hypothesized that attachment via a flexible C‐terminal extension could allow for the clamp and polymerase to rotate with respect to one another, perhaps aiding the 3′ terminus of DNA to switch from the polymerization to the exonuclease site (Shamoo and Steitz, 1999). Some degree of flexibility would also be advantageous if two different polymerases were to bind the same clamp concurrently, as some models have proposed. Since the β‐clamp is a homodimer, two hydrophobic pockets could potentially be available for binding two proteins. The arrangement of the subunits places both sites on the same face of the ring, which may create steric constraints that preclude binding of more than one protein at a time. No conclusive evidence has been presented in favor of or against this possibility, but the alternative model, one involving sequential binding by different proteins, is also possible. In this case, different proteins would compete for attachment to the clamp, and traffic flow on β could be regulated by other factors, such as DNA structure or post‐translational modifications of the clamp.

Competition of proteins for the same clamp has been documented in the case of the γ‐complex and the Pol III core, which switch their affinity for β‐clamps sequentially during the first steps in processive DNA synthesis (illustrated in Figure 7A). The switch is regulated by whether DNA is inside the β‐ring or not (Stukenberg et al., 1994). When β is in solution (i.e. not on DNA), γ‐complex binds β tighter than core, but when DNA goes through β, core develops a much greater affinity for β and takes the ring over from the clamp loader (Ason et al., 2003). A second switch, also dependent on DNA structure, occurs at the end of processive Okazaki‐fragment synthesis. In this case Pol III core is released from the β‐clamp only when synthesis is complete, allowing core to rapidly recycle to a β‐clamp at a new primed site for the synthesis of the next Okazaki fragment (O'Donnell, 1987; Stukenberg et al., 1991). In a recent study we have demonstrated that the C‐terminus of the α‐subunit is also a target for binding by the τ‐subunit of the clamp loader (López de Saro et al., 2003). The τ‐subunit senses DNA structure, and when it lacks ssDNA template, τ binds the C‐terminus of Pol III core and displaces it from β and DNA as illustrated in Figure 7A (Leu et al., 2003). This 'processivity switch' is mediated by direct competition between a DNA‐sensing protein (τ) and the β‐clamp for a common site on the Pol III α‐subunit and reveals an elegant mechanism by which polymerase trafficking on clamps can be connected to the different structures of the DNA substrate and product.

Analogous competition mechanisms for clamps regulated by DNA structure may be discovered for other clamp‐related reactions. For example, the clamp left by Pol III core at the completed Okazaki fragment may be sequentially used by Pol I and DNA ligase for RNA primer removal and sealing of the nick, respectively (see key to Figure 7B). Perhaps Pol I binds β tightest when RNA is present, but loses affinity for β when the processing job is complete. Ligase may interact most favorably with β at the nicked site created by Pol I action. Besides polymerases and ligase, MutS has also been shown to bind β (López de Saro and O'Donnell, 2001). Hence, either before or following Okazaki‐fragment processing, β‐clamps left on DNA may serve as markers of newly synthesized DNA for genome surveillance by repair proteins or other DNA modifying proteins before being recycled from DNA for use at other locations.

The SOS response in E.coli includes the induction of three polymerases; Pol II, Pol IV and Pol V (Goodman, 2002). It has been suggested that these polymerases are called into action to bypass DNA damage during chromosomal replication, and that their diversity reflects the variety of damaged DNA that can be encountered (Goodman and Tippin, 2000; Sutton and Walker, 2001; Wagner et al, 2002). The finding that they interact with the β‐clamp on the same locus as Pol III core suggests that they compete with core for β at sites of DNA damage (see Figure 7C). The levels of Pol V (UmuC subunit) in the cell rise to ∼200 copies per cell upon induction (Woodgate and Ennis, 1991), ∼10‐fold higher than the intracellular level of Pol III core, which is present at a concentration of ∼20 copies per cell (Baker and Kornberg, 1992).

A combination of DNA structural elements or signals from the stalled replication fork could be responsible for the selection of the polymerase that resides with β at a 3′ terminus at any given time. However, the fact that the repair polymerases and Pol III core compete for the same site on the clamp suggests that control of polymerase selection may be exercised at the level of their affinity for the clamp. In the case of Pol V, action with β requires RecA, which is also induced in the SOS response (Pham et al., 2001, 2002). Pol III core, on the other hand, is inhibited by RecA. Hence, one may expect that Pol III core, stalled at a DNA lesion, may be displaced by RecA, and at the same time the presence of RecA and β would recruit Pol V (Figure 7C) (Pham et al., 2001). After lesion bypass and disassembly of the RecA filament (due to inherent instability of RecA), Pol V would lose affinity in the absence of RecA and dissociate from β (Pham et al., 2001). Pol III core could then reassociate with β to resume chain extension.

Despite the lack of sequence similarity between Pol II, III and IV of E.coli, they have evolved similar C‐terminal sequence motifs that bind to β. The sequence motif consists of the pattern QxxL(x)F. This sequence is somewhat reminiscent of the PCNA consensus binding motif (Qxx[I/L]xxFF) (Warbrick, 2000) and the β binding motif (QLxLF) suggested by Dalrymple et al. (2001). This is even more remarkable in view of the fact that β and PCNA share no sequence similarity, despite having a common fold. Clamp‐binding motifs could either be ancient and reflect a common ancestor, or may have occurred as a result of convergent evolution.

The finding that the clamp loader and all the polymerases present in the E.coli cell interact with the β‐clamp at the same location suggests that this site may be an attractive target for the development of DNA replication inhibitors. This is especially true considering that many proteins interact with the same hydrophobic pocket on β. Overcoming drug binding to this pocket would likely impair its binding to several proteins at once, thus making the appearance of antibiotic‐resistant variants less likely.

HPLC‐purified, N‐terminal biotinylated 20‐mer peptides corresponding to N‐ and C‐terminal sequences of Pol III α, Pol II and Pol IV were purchased from Bio‐synthesis, Inc.; N‐terminal biotinylated 20‐mer peptides corresponding to internal regions of the α‐protein and the Ala‐scanning peptide series shown in Figure 2 were purchased from Chiron Technologies; they were synthesized using Multipin peptide technology, which leaves a diketopiperidine linker at the C‐terminus of the peptide. Streptavidin‐coated, 96‐well microtiter plates were from Costar, and non‐streptavidin plates (MaxiSorp) were from Nalge Nunc, Inc. Streptavidin was from Sigma Co. Labeled nucleotides were from Dupont, New England. Nuclear and unlabeled nucleotides were from Pharmacia‐LKB.

Proteins were purified as described previously: α, ϵ, θ (Studwell and O'Donnell, 1990) and β (Kong et al., 1992). ³²P‐β was labeled using γ‐³²P ATP as described previously (Kelman et al., 1995). Pol III core was reconstituted from isolated subunits, then purified from unbound proteins as described (Studwell‐Vaughn and O'Donnell, 1991).

Microplate assays were performed as described previously by Yuzhakov et al. (1999). N‐terminally biotinylated peptides (100 pmol) were added to streptavidin‐coated 96‐well plates in 30 μl PBST buffer [0.01 M phosphate‐buffered saline pH 7.2 (PHS) supplemented with 0.1% v/v Tween 20] and incubated for 1 h at 23°C. They were then washed three times by the addition and removal of 100 μl PBST buffer. ³²P‐β was added (90 nM) directly to wells in a volume of 30 μl and incubated for 1 h, after which each well was washed three times with 50 μl PBST buffer and the plates were analyzed using a PhosphorImager (Molecular Dynamics).

MaxiSorp plates were used in 96‐well plate assays in which other proteins were immobilized. Proteins (100 pmol of either δ, Pol I, Pol IV, UmuC, UmuD or UmuD′) were added directly to the plate in 30 μl of water and the plates were dried at 23°C. Plates were then blocked with 5% non‐fat dry milk in PBST for 1 h and washed. ³²P‐β (90 nM in 30 μl PBST buffer) was then added and the plates analyzed using a PhosphorImager.

Native gel electrophoresis assays using ³²P‐β were performed as described by López de Saro and O'Donnell (2001). Reactions (15 μl) contained 20 mM Tris‐HCl (pH 7.5), 0.1 mM EDTA, 4% glycerol, 50 μg/ml bovine serum albumin (BSA), 100 mM NaCl, 5 mM DTT, 200 nM Pol III core (as indicated) and 90 nM ³²P‐β (150 d.p.m./fmol). When present, either of the following components were added: 4 μg E.coli β, 4 μg human PCNA, 4 μg T4gp45 or 25 pmol α1141–1161 peptide. Reactions were incubated at 37°C for 5 min before loading 4 μl onto a 4% native polyacrylamide gel. Electrophoresis was performed using 40 mM Tris‐acetate (pH 8.0), 1 mM EDTA, at 4°C and developed at 17 mA for 90 min. Gels were dried and detection of ³²P‐β was performed using a PhosphorImager.

DNA synthesis assays were performed using M13mp18 phage circular ssDNA (1.7 nM) primed with a DNA 30‐mer. Reactions included 0.42 μM ssDNA‐binding protein (as tetramer), 2 nM γ‐complex (γ₃δδ′χψ) 30 nM β (as dimer), and either 25 nM α, Pol III core or Pol I Klenow fragment, as indicated, in 25 μl 20 mM Tris‐HCl (pH 7.5), 0.1 mM EDTA, 4% glycerol, 40 μg/ml BSA, 5 mM DTT, 8 mM MgCl₂, 0.5 mM ATP, 60 μM each of dGTP, dATP and dCTP, and 20 μM [α‐³²P]dTTP. When present, peptide was added at the concentrations indicated prior to shifting to 37°C. Reactions were incubated for 30 s (Pol III) or 30 min (Pol I Klenow fragment) at 37°C and stopped with 25 μl of 1% SDS, 50 mM EDTA, then analyzed for nucleotide incorporation by spotting onto DE81 filter papers as described previously (Studwell and O'Donnell, 1991).

Clamp loading assays of ³²P‐β onto singly primed M13mp18 ssDNA by γ‐complex were performed as described for replication reactions, but in the absence of DNA polymerase or dNTPs. Reaction volumes were 100 μl of 20 mM Tris‐HCl (pH 7.5), 0.1 mM EDTA, 4% glycerol 40 μg/ml BSA, 5 mM DTT, 8 mM MgCl₂, 0.5 mM ATP and incubation was for 5 min at 37°C. Reactions were then applied to a 5 ml BioGel A15M column (BioRad) pre‐equilibrated with 20 mM Tris‐HCl (pH 7.5), 0.1 mM EDTA, 4% glycerol, 40 μg/ml BSA, 5 mM DTT, 8 mM MgCl₂ and 100 mM NaCl. Chromatography was performed at 22°C and fractions of 180 μl were collected and analyzed by liquid scintillation counting.

The β‐subunit can be uniquely labeled at Cys333 using maleimide derivatives (Griep and McHenry, 1988). β (3 mg) was labeled using OG 488 maleimide (Molecular Probes) in 1 ml of 50 mM potassium phosphate (pH 7.5) and 100 mM NaCl. The OG maleimide (90 nmol) was dissolved in 100 μl DMSO, then 80 μl was added to β with gentle stirring at 4°C, followed by overnight incubation at 4°C in the dark. β^OG was separated from unreacted reagent on a 50 ml column of BioGel 6 and contained ∼0.91 molecules of OG per β monomer as determined from protein absorbance at 280 nM (ϵ280 = 14 890/M/cm) and OG at 490 nm (ϵ491‐76 000/M/cm).

Titration of wild‐type α or core αΔC7 into β^OG was performed as follows. Reactions contained 50 nM β₂^OG and 1.5 μM α in 60 μl 20 mM Tris‐HCl (pH 7.5), 0.5 mM EDTA, 1 mM DTT and 0–250 mM NaCl (as indicated) on ice. Titrations of Pol III core or Pol II were performed in a similar buffer supplemented with 100 mM NaCl. Reactions were shifted to 22°C for 15 min then 51 μl was placed in a 3 X 3 mm cuvette. Excitation was at 490 nm and emission was monitored from 500–600 nm in a PTI Quantamaster spectrofluorimeter. Fluorescence emission at 517 nm was used for analysis. Additional titrations of wild‐type α into high concentrations of β₂^OG (250 nM and 500 nM) showed a α:β₂ =1:1 stoichiometry. As a consequence, data points were fit according to the model, A + B → AB using the Origin software (Microcal, Inc.). Reactions using rhodamine (TAMRA)‐labeled peptides contained 1 μM peptide and β‐subunit as indicated. Excitation was at 490 nm, and the emission was monitored from 500–600 nm. Fluorescence emission at 517 nm was used for analysis as above.