During Translesion Synthesis, Escherichia coli DinB89 (T120P) Alters Interactions of DinB (Pol IV) with Pol III Subunit Assemblies and SSB, but Not with the β Clamp

ABSTRACT

Translesion synthesis (TLS) by specialized DNA polymerases (Pols) is an evolutionarily conserved mechanism for tolerating replication-blocking DNA lesions. Using the Escherichia coli dinB-encoded Pol IV as a model to understand how TLS is coordinated with the actions of the high-fidelity Pol III replicase, we previously described a novel Pol IV mutant containing a threonine 120-to-proline mutation (Pol IV-T120P) that failed to exchange places with Pol III at the replication fork in vitro as part of a Pol III-Pol IV switch. This in vitro defect correlated with the inability of Pol IV-T120P to support TLS in vivo, suggesting Pol IV gains access to the DNA, at least in part, via a Pol III-Pol IV switch. Interaction of Pol IV with the β sliding clamp and the single-stranded DNA binding protein (SSB) significantly stimulates Pol IV replication and facilitates its access to the DNA. In this work, we demonstrate that Pol IV interacts physically with Pol III. We further show that Pol IV-T120P interacts normally with the β clamp, but is impaired in interactions with the α catalytic and εθ proofreading subunits of Pol III, as well as SSB. Taken together with published work, these results provide strong support for the model in which Pol IV-Pol III and Pol IV-SSB interactions help to regulate the access of Pol IV to the DNA. Finally, we describe several additional E. coli Pol-Pol interactions, suggesting Pol-Pol interactions play fundamental roles in coordinating bacterial DNA replication, DNA repair, and TLS.

IMPORTANCE Specialized DNA polymerases (Pols) capable of catalyzing translesion synthesis (TLS) generate mutations that contribute to bacterial virulence, pathoadaptation, and antimicrobial resistance. One mechanism by which the bacterial TLS Pol IV gains access to the DNA to generate mutations is by exchanging places with the bacterial Pol III replicase via a Pol III-Pol IV switch. Here, we describe multiple Pol III-Pol IV interactions and discuss evidence that these interactions are required for the Pol III-Pol IV switch. Furthermore, we describe several additional E. coli Pol-Pol interactions that may play fundamental roles in managing the actions of the different bacterial Pols in DNA replication, DNA repair, and TLS.

INTRODUCTION

Faithful replication of cellular DNA is critically important for cell health and viability. While organisms have evolved highly accurate DNA polymerases (Pols) responsible for replicating their genomes, cellular DNA nevertheless undergoes frequent damage from a wide variety of sources (reviewed in references 1 and 2). The resulting DNA lesions can block DNA replication (reviewed in reference 2). Even though most of these lesions are accurately repaired, allowing for replication to proceed, some evade repair and are instead managed by DNA damage tolerance mechanisms, including DNA template switching and translesion DNA synthesis (TLS) (reviewed in reference 2). Template switching is largely accurate and involves the stalled Pol using the newly synthesized lesion-free daughter DNA strand as an alternate template to circumvent the lesion and continue replication. In contrast, TLS involves replication past the lesion by one or more specialized Pols and can be accurate or mutagenic, depending on the lesion and the Pol(s) involved.

The Escherichia coli dinB-encoded Pol IV is a distributive enzyme that, like other Y family Pols, lacks an intrinsic 3′→5′ exonuclease proofreading activity (3). Transcription of dinB is regulated as part of the global SOS response (4, 5). Thus, following replication-blocking DNA damage, the RecA-facilitated autodigestion of the LexA repressor protein leads to a more than 10-fold increase in cellular Pol IV protein levels (6). Pol IV catalyzes the bypass of several DNA minor groove adducts, including N²-furfuryl-dG (N²-dG) (7) and N³-methlyadenine (3m-dA) (8), as well as the oxidative lesion 8-oxo-7,8-dihydro-2′-deoxyguanosine 5′-monophosphate (8-oxo-dG) (9). Compared with the bacterial replicase, Pol IV displays a reduced fidelity when replicating undamaged DNA in vitro (10). Consistent with this finding, results of genetic experiments using a collection of lacZ alleles developed in the Miller lab that permit detection of defined mutations (11, 12) demonstrated that Pol IV has a propensity to catalyze frameshift mutations in homopolymeric sequences, as well as GC→AT transitions and GC→TA and AT→TA transversions when replicating undamaged DNA in vivo (13).

We and others have used Pol IV as a model to understand how E. coli coordinates the potentially mutagenic actions of TLS Pols with accurate replication catalyzed by the bacterial replicase Pol III holoenzyme (Pol III HE) (14–20; reviewed in reference 21). Pol III HE is comprised of three subassemblies: two Pol III core (Pol IIIαεθ) complexes for simultaneous synthesis of leading and lagging strands, each of which consists of the Pol IIIα catalytic and Pol IIIεθ proofreading subunits; two homodimeric ring-shaped β clamps, which act to tether Pol III core to the DNA, significantly increasing its processivity; and one DnaX clamp loader complex (τ₂γδδ’ψχ) (22), which loads the β clamp onto primed DNA in an ATP-dependent manner and tethers the leading and lagging Pol III core complexes to each other to stabilize the Pol III HE complex (reviewed in references 23 and 24). In contrast with the multimeric structure of Pol III, Pol IV is comprised of a single 351-amino-acid polypeptide that folds into two discrete domains, referred to here as the Pol IV catalytic domain (Pol IV^CD; residues 1 to 230) and the Pol IV little finger domain (Pol IV^LF; residues 243 to 351), which are connected via a short linker sequence (25) (Fig. 1A). Pol IV replication activity is stimulated by the β sliding clamp (26) and the single-stranded DNA (ssDNA) binding protein (SSB) (27). As with Pol III core, the β clamp tethers Pol IV to the DNA to increase its processivity and functions as a mobile scaffold on which Pol III and Pol IV can exchange places to gain sequential access to the DNA (14, 16–18). Based on structural and biochemical results, Pol IV^LF makes two distinct contacts with the β clamp: the C-terminal six residues of Pol IV (³⁴⁶QLVLGL³⁵¹) bind a hydrophobic cleft present in each clamp protomer, while residues ³⁰³VWP³⁰⁵ of Pol IV bind the rim (E93, L98) of the adjacent clamp protomer (14, 26, 28) (Fig. 1A and B; see Fig. S1 in the supplemental material). The Pol IV^LF-β rim interaction is weak (14) (equilibrium dissociation constant [K_D], ≥1.3 μM), but nevertheless helps to regulate Pol IV function by pulling it away from the face of the clamp (Fig. 1B; Fig. 1S), preventing its access to the DNA (28). This interaction is also required for Pol III-Pol IV switching (14, 15, 18, 29). In contrast with Pol IV^LF, a previous study demonstrated that Pol IV^CD failed to interact with the β clamp as measured using a pull down assay (30). The Pol IV-SSB interaction also stimulates the replication activity of Pol IV (27) and may help to recruit Pol IV to replication forks and/or ssDNA gaps (31).

FIG 1.

Models of the E. coli Pol IV-β clamp and S. solfataricus Dpo4-PCNA complexes. (A) Stick cartoon of the E. coli Pol IV protein. Residues comprising Pol IV^CD (residues 1 to 230), which contains the palm (residues 1 to 10 and 74 to 165 in red), fingers (residues 11 to 73 in cyan), and thumb (residues 166 to 242 in blue) domains and Pol IV^LF (residues 243 to 351 in green), which interacts with the rim and cleft of the β clamp, are indicated. Approximate locations of active-site residues (D8, D103, and D104), residues that interact with the rim (³⁰³VWP³⁰⁵) and cleft (³⁴⁶QLVLGL³⁵¹) of the β clamp, and residue T120 are indicated. (B) In silico model of the E. coli Pol IV-β clamp complex. The palm, fingers, thumb, and little finger domains are colored as in panel A, and the position of residue T120 (yellow) is indicated. (C) In silico model of the S. solfataricus Dpo4-PCNA complex. The Dpo4 palm (residues 1 to 10 and 78 to 166), fingers (residues 11 to 77), thumb (residues 167 to 243), and little finger (residues 244 to 341) domains are colored as in panel A. (D) In silico model of the E. coli Pol IV-β clamp complex in which the Pol IV^CD (residues ³⁵RERR³⁸ in magenta) contacts the back of the β clamp (residues ²⁵¹NPDKH²⁵⁵, S311, and M339 in orange), similar to how Dpo4 contacts PCNA1. Pol IV domains are colored as in panel A, and the position of Pol IV residue T120 (yellow) is indicated.

The dinB89 allele, which encodes a Pol IV mutant bearing a threonine 120-to-proline substitution within the catalytic domain (Pol IV-T120P) (Fig. 1A and B), was identified based on its inability to impede E. coli growth when expressed at levels ∼4-fold higher than SOS-induced levels using a low-copy plasmid (15). Our subsequent biochemical and genetic analyses of this mutant protein demonstrated that it was proficient in the replication of both undamaged and damaged DNA in vitro, despite the failure of dinB89 mutant strain to tolerate methyl methanesulfonate (MMS), which generates DNA lesions, including 3m-dA, that Pol IV bypasses accurately (8). Importantly, this MMS-sensitive phenotype correlated with the inability of the T120P mutant to support exchange with Pol III core on a β clamp assembled at the 3′ end of a primer-template junction in vitro (15). Since the components in this in vitro assay were limited to Pol IV^CD (or Pol IV^CD-T120P), Pol III HE (Pol IIIαεθ, β clamp, DnaX clamp loader complex), and a primed DNA template, we hypothesized that the T120P mutation impaired an as yet unidentified interaction of Pol IV with Pol III core and/or the β clamp that is required for granting access of Pol IV to the DNA. The findings discussed in this report extend this published work with Pol IV-T120P by correlating defects in its interactions with key partner proteins with its failure to switch with Pol III in vitro and to tolerate MMS-induced DNA damage in vivo (15). Our results demonstrate that Pol IV-T120P interacts normally with the β clamp, but is impaired in interactions with the α catalytic and εθ proofreading subunits of Pol III, as well as SSB. Taken together with published work (15), these results support the view that Pol III and SSB help to regulate the access of Pol IV to primed DNA. We further demonstrate that Pol-Pol interactions are commonplace in E. coli, suggesting they play fundamental roles in coordinating the function of the different Pols in DNA replication, DNA repair, and TLS.

RESULTS

Goals of this study.

While the T120P mutation in Pol IV fails to affect its ability to catalyze TLS, it nevertheless impedes the ability of Pol IV to gain access to the DNA to catalyze TLS via a Pol III-Pol IV switch (15). Since the components of our in vitro Pol III-Pol IV switch assay were limited to a primed DNA template, Pol IV^CD (or Pol IV^CD-T120P), and Pol III HE ([Pol IIIαεθ]₂[β clamp]₂[τ₂γδδ′ψχ]) (15), we hypothesized that the failure of Pol IV^CD-T120P to switch with Pol III core for access to the DNA was the result of its inability to participate in one or more previously unidentified interactions with the β clamp and/or Pol III core; importantly, Pol IV has not yet been demonstrated to interact physically with the Pol III core complex. The goal of the work discussed in this report was to test this hypothesis by measuring interactions of Pol IV and Pol IV-T120P with the β clamp and the α catalytic and εθ proofreading subunits of Pol III. In addition, since the Pol IV-SSB interaction contributes to Pol IV function (27), we asked whether the T120P mutation impaired the interaction of Pol IV with SSB, possibly contributing to the in vivo TLS defect of this Pol IV mutant (15). Since at the inception of this project we knew that Pol IV contacts multiple surfaces on the β clamp (28, 32), we anticipated that it may be difficult to detect potential effects of the T120P mutation on Pol IV-partner protein interactions without a sensitive and quantitative method for measuring their interactions. We therefore analyzed these protein-protein interactions by using two optical methods that are both sensitive and quantitative: bio-layer interferometry (BLI) and surface plasmon resonance (SPR). Furthermore, in select cases, results were further supported using a native electrophoretic mobility shift assay (EMSA).

The catalytic domain of Pol IV interacts with the β clamp.

E. coli Pol IV is comprised of two distinct domains: a catalytic domain (Pol IV^CD) corresponding to the N-terminal 230 residues and a little finger domain (Pol IV^LF) comprised of the C-terminal 109 residues (Fig. 1A). E. coli Pol IV is commonly accepted to interact with the β clamp entirely through its Pol IV^LF domain (28, 30) (Fig. 1B; Fig. S1). In contrast, the orthologous Sulfolobus solfataricus Pol IV ortholog, Dpo4, contacts its cognate clamp, PCNA (specifically the PCNA1 subunit), through both its catalytic (Dpo4^CD) and little finger (Dpo4^LF) domains (33). Dpo4^LF interacts with PCNA1 in a manner that is similar to how Pol IV^LF contacts the β clamp (Fig. 1B and C; and Fig. S1). Dpo4 makes two additional contacts with PCNA1 via its thumb and finger domains (Fig. 1C) (33). In light of the high degree of functional and structural similarity between E. coli Pol IV and S. solfataricus Dpo4 (25, 34), we used bio-layer interferometry (BLI) to determine whether E. coli Pol IV^CD interacts with the β clamp. The full-length Pol IV and Pol IV^LF were analyzed in parallel as positive controls. In the X-ray crystal structure of the Pol IV^LF-β clamp complex (PDB no. 1UNN), Pol IV^LF simultaneously binds to the clamp cleft as well as the rim of the adjacent clamp protomer (28) (Fig. 1B; Fig. S1). We therefore used the two-site binding model to interpret the BLI results, yielding K_D values of 0.74 and 22.7 nM for the Pol IV-β clamp interaction and 8.68 and 238 nM for the Pol IV^LF-β clamp interaction (Table 1). Consistent with Pol IV^LF binding to the β clamp more weakly compared with Pol IV, we observed a Pol IV^CD-β clamp interaction. The results did not fit the one-site binding model, but they did fit well to the two-site binding model, yielding K_D values of 945 and 2,827 nM (Table 1).

TABLE 1.

Interaction of Pol IV with the β clamp^a

Interactionb	Expt	KD (nM)	Avg KD, nM (range)
			KD 1	KD 2
Pol IV-Hisβ	1	KD1 = 1.22	0.74 (0.97)	22.7 (4)
		KD2 = 24.7
	2	KD1 = 0.25
		KD2 = 20.7
Pol IV-HisβBack	1	KD1 = 1.84	1.84 (0.01)	32.8 (4.70)
		KD2 = 30.4
	2	KD1 = 1.83
		KD2 = 35.1
Pol IV-T120P–Hisβ	1	KD1 = 2.70	3.06 (0.72)	12.2 (0.70)
		KD2 = 11.8
	2	KD1 = 3.42
		KD2 = 12.5
Pol IVCD-Hisβ	1	KD1 = 968	945 (46.0)	2,827 (1,170)
		KD2 = 3,410
	2	KD1 = 922
		KD2 = 2,244
Pol IVCD-HisβBack	1	KD1 = 731	720 (22.3)	2,008 (396)
		KD2 = 1,810
	2	KD1 = 709
		KD2 = 2,206
Pol IVCD-T120–Hisβ	1	KD1 = 948	837 (224)	1,592 (756)
		KD2 = 1,970
	2	KD1 = 726
		KD2 = 1,214
Pol IVLF-Hisβ	1	KD1 = 9.02	8.68 (0.68)	238 (40.0)
		KD2 = 218
	2	KD1 = 8.34
		KD2 = 258

^aSee Table S1 for the k_a (on rate), k_d (off rate), χ² (likelihood of no relationship), and R² (goodness of fit) values for these interactions.

^bThe indicated His-tagged protein (ligand) was captured on a HIS1K, biosensor and interactions were measured using BLI.

As an independent method, we used surface plasmon resonance (SPR) (see Fig. S2A in the supplemental material). The shape of the rise (i.e., the first half of each curve that describes the on-rate, or k_a, of the interaction) and decline (i.e., the second half of the curve that describes the off-rate, or k_d) of the SPR sensorgrams for the Pol IV-β clamp and Pol IV^CD-β clamp interactions with a low analyte (Pol IV or Pol IV^CD) concentration differed from those with high concentration (Fig. 2A and B), suggesting both Pol IV and Pol IV^CD contact multiple sites on the clamp. Based on their analysis using the two-site model, Pol IV bound to the β clamp, with K_D values of 13.5 and 224 nM compared with 567 and 2,109 nM for Pol IV^CD-β (Fig. 2C). We attribute the differences in K_D values measured by BLI compared with SPR observed here and later in this report to the fact that analyte can rebind to ligand during the dissociation stage of BLI, but not SPR, possibly leading to underestimates of the off-rates (k_d), the differences in how the His-tagged analytes adsorb to the BLI biosensor tips compared with SPR biosensors, and differences in the algorithms used to extract kinetic constants from the BLI results and SPR sensorgrams. Irrespective of these differences, results of the BLI and SPR experiments demonstrate that Pol IV^CD interacts with the β clamp, albeit weakly. Pol IV^LF binds the rim of the β clamp with a similarly weak affinity (14) (K_D, ≥1.3 μM). While the β rim interaction is dispensable for Pol IV replication, it is required for Pol III-Pol IV switching, and may also contribute to the recruitment of Pol IV to DNA, as well as replication forks (14, 29, 35). Thus, the Pol IV^CD-β clamp interaction may act as a foothold to help recruit Pol IV to the Pol III-β clamp complex for a Pol III-Pol IV switch or alter the conformation that full-length Pol IV adopts when bound to the clamp.

FIG 2.

Pol IV^CD interacts with the β clamp. Representative SPR sensorgrams for the (A) β^His-Pol IV and (B) β^His-Pol IV^CD interactions. The cartoon (inset to panel A) summarizes the approach used to measure these interactions with SPR. (C) Summary of the kinetic parameters describing the β^His-Pol IV and β^His-Pol IV^CD interactions. Instead of χ² (goodness of fit), the ClampXP 3.50 software provides residual sum of squares (goodness of fit) values, which were each <10% of the respective R_max, confirming the specificity of each interaction. See Fig. S2 for the k_a (on-rate) and k_d (off-rate) values for these interactions.

To better understand how Pol IV^CD interacts with the β clamp, we used the S. solfataricus Dpo4-PCNA1 structure (PDB no. 3FDS) to model an E. coli Pol IV-clamp complex in which Pol IV^CD contacts the β clamp in a manner similar to that in which Dpo4^CD contacts PCNA1 (33) (Fig. 1D). As a test of this in silico model, we measured the interaction of Pol IV with a mutant form of the β clamp (β^Back) in which each of the seven residues encompassing the surface of the clamp involved in the hypothesized interaction (²⁵¹NPDKH²⁵⁵, S311, and M339) was substituted for with alanine (Fig. 1D). The β^Back mutant behaved indistinguishably from the wild-type clamp during purification, suggesting the amino acid substitutions failed to significantly affect its structure. Furthermore, our finding that the β^Back mutant was comparable with the wild-type β clamp for interaction with both full-length Pol IV and Pol IV^CD (Table 1) suggests that E. coli Pol IV^CD contacts the β clamp in a manner that is different from how S. solfataricus Dpo4^CD contacts PCNA1 (Fig. 1C).

The T120P mutation fails to disrupt the β clamp-Pol IV interaction.

Based on molecular modeling, residue T120 of Pol IV is well removed from the corresponding region in Dpo4 that binds its clamp (Fig. 1C). However, since Pol IV^CD and Dpo4^CD appear to interact differently with their respective clamps, we asked whether the T120P mutation affected the Pol IV-β clamp interaction. The results of BLI experiments indicated that Pol IV-T120P bound the β clamp, with K_D values of 3.06 and 12.2 nM, which, after taking into consideration the range between experimental results, are similar to those observed for the Pol IV-β clamp interaction (0.74 and 22.7 nM) (see Table 1). Since Pol IV contacts multiple clamp surfaces (28, 32), we considered the possibility that the binding of Pol IV^LF to the clamp may conceal an effect of the T120P mutation. To address this possibility, we measured the interaction of Pol IV^CD-T120P with the clamp. K_D values for Pol IV^CD-T120P (837 and 1,592 nM) were similar to the values observed for Pol IV^CD (945 and 2,827 nM) (Table 1). Based on these results, we conclude that the T120P mutation fails to significantly affect the interaction of Pol IV with the β clamp.

Pol IV interacts with the α catalytic and εθ proofreading subunits of the Pol III replicase.

The results discussed above, taken together with our previous finding that the in vitro catalytic activity and processivity of Pol IV-T120P in the presence of the β clamp were comparable to those of Pol IV (15), indicate that the inability of Pol IV-T120P to displace Pol III core from the clamp assembled at the 3′ end of a primer template was not the result of an impaired interaction between Pol IV-T120P and the β clamp. Since Pol III was the only other protein in our in vitro switch assay (15), we used an EMSA to determine whether Pol IV interacts with Pol III core. Because Pol IV is highly basic, when loaded alone in native PAGE, it migrates out of the top of the gel toward the cathode and is therefore undetected by Western blotting (Fig. 3, lane 1). However, when complexed with the acidic β clamp, Pol IV migrates into the gel, toward the anode, and is detected by Western blotting using anti-Pol IV polyclonal antibodies (Fig. 3, lane 6). A similar result was observed when Pol IV was mixed with the Pol III core complex (Pol IIIαεθ) (Fig. 3, lane 7). Analysis of the separate Pol III core components in this assay indicated that Pol IV interacts with both the Pol IIIα catalytic and Pol IIIεθ proofreading subunits (Fig. 3, lanes 8 to 9). As controls, we failed to detect Pol IV in our β clamp and Pol III protein preparations (Fig. 3, lanes 2 to 5). To obtain quantitative information on these interactions, we used SPR (see Fig. S4A in the supplemental material). Similar to the Pol IV-β clamp interaction, the shape of the SPR sensorgrams for both the Pol IV-Pol IIIα and Pol IV-Pol IIIεθ interactions with a low analyte concentration differed from those with a high concentration (Fig. 4A and B), suggesting Pol IV contacts multiple sites on each Pol III subunit. Consistent with this, the SPR data fit well to the two-site binding model, with K_D values of 384 and 779 nM for Pol IV-Pol IIIα and 422 and 3,040 nM for Pol IV-Pol IIIεθ (Fig. 4E).

FIG 3.

Pol IV interacts with the α catalytic and εθ proofreading subunits of the Pol III replicase. The ability of Pol IV to interact physically with the β clamp, Pol III core (Pol IIIαεθ), Pol IIIα, and Pol IIIεθ was measured by EMSA. An empty lane was left between each sample containing Pol IV and the β clamp, Pol III core (Pol IIIαεθ), Pol IIIα, or Pol IIIεθ (shown as lanes 6 to 9), resulting in their different spacing compared with controls (lanes 1 to 5). The position of Pol IV identified by Western blotting using rabbit anti-Pol IV polyclonal antibodies is indicated.

FIG 4.

Pol IV-T120P is impaired in interactions with Pol IIIα and Pol IIIεθ. Shown are representative SPR sensorgrams for the (A) Pol IIIα^His-Pol IV, (B) Pol IIIεθ^His-Pol IV, (C) Pol IIIα^His–Pol IV-T120P, and (D) Pol IIIεθ^His–Pol IV-T120P interactions. (E) Summary of the kinetic parameters describing the Pol IIIα^His-Pol IV, Pol IIIα^His–Pol IV-T120P, Pol IIIεθ^His-Pol IV, and Pol IIIεθ^His–Pol IV-T120P interactions. Instead of χ² (goodness of fit), the ClampXP 3.50 software provides residual sum of squares (goodness of fit) values, which were each <10% of the respective R_max, confirming the specificity of each interaction. See Fig. S4 for a cartoon depiction of the SPR method used, as well as the k_a (on-rate) and k_d (off-rate) values for these interactions.

To better understand how Pol IV interacts with Pol III, we asked if Pol IV^CD or Pol IV^LF was sufficient for the interaction with Pol IIIα and/or Pol IIIεθ. Based on the results of BLI experiments, both Pol IV^CD and Pol IV^LF interacted with Pol IIIα and Pol IIIεθ, although we were unable to obtain reliable values for the kinetic constants describing the Pol IIIεθ interactions (Table 2, footnote a). BLI results fit best to the two-site interaction model. Interestingly, compared with Pol IV^CD, Pol IV^LF bound more tightly to Pol IIIα (K_D1 = 59.0 nM and K_D2 = 819 nM compared with K_D1 = 5.02 nM and K_D2 = 15.7 nM, respectively) (Table 2). Taken together, these results support the conclusion that Pol IV makes multiple contacts with each of these Pol III subunits.

TABLE 2.

Interaction of Pol IV with the Pol IIIα catalytic and Pol IIIεθ proofreading subunits^a

Interactionb	Expt	KD (nM)	Avg KD, nM (range)
			KD 1	KD 2
Pol IV-Pol IIIαHis	1	KD1 = 11.8	13.9 (4.10)	52.6 (4.80)
		KD2 = 50.2
	2	KD1 = 15.9
		KD2 = 55.0
Pol IV-T120P–Pol IIIαHis	1	KD1 = 5.97	4.88 (2.18)	62.1 (29.9)
		KD2 = 77.0
	2	KD1 = 3.79
		KD2 = 47.1
Pol IVCD-Pol IIIαHis	1	KD1 = 68.7	59.0 (19.5)	819 (362)
		KD2 = 1,000
	2	KD1 = 49.2
		KD2 = 638
Pol IVCD-T120P–Pol IIIαHis	1	KD1 = 215	216 (2)	3,790 (40)
		KD2 = 3,810
	2	KD1 = 217
		KD2 = 3,770
Pol IVLF-Pol IIIαHis	1	KD1 = 4.59	5.02 (0.85)	15.7 (8.20)
		KD2 = 19.8
	2	KD1 = 5.44
		KD2 = 11.6

^aSee Table S2 in the supplemental material for the k_a (on rate), k_d (off rate), χ² (likelihood of no relationship), and R² (goodness of fit) values for these interactions.

^bThe indicated His-tagged protein (ligand) was captured on a HIS1K biosensor, and interactions were measured using BLI. Interactions were identified between Pol IIIεθ^His and Pol IV, Pol IV^CD, and Pol IV^LF, but we were unable to obtain reliable numbers describing the kinetics of their interactions.

Pol IV-T120P is impaired in interactions with Pol III.

In light of the results discussed above, we used our EMSA to ask whether Pol IV-T120P was impaired in interaction with Pol IIIα. As shown in Fig. S3 in the supplemental material, Pol IV and Pol IV-T120P each interacted similarly with the β clamp and Pol IIIα. Since our EMSA is qualitative, we used SPR to ask whether the T120P mutation disrupted the interaction of Pol IV with Pol IIIα and/or Pol IIIεθ (Fig. S4A). In contrast with wild-type Pol IV, the shapes of the SPR sensorgrams for both Pol IV-T120P–Pol III interactions were similar at all analyte concentrations examined, suggesting Pol IV-T120P makes a single contact with each Pol III subunit (Fig. 4C and D). Consistent with this suggestion, SPR sensorgrams fit well to the one-site binding model, yielding a single K_D for each interaction: 414 nM for Pol IV-T120P–Pol IIIα and 776 nM for Pol IV-T120P–Pol IIIεθ (Fig. 4E). These K_D values are comparable to the higher-affinity K_D observed for the interaction of wild-type Pol IV with each Pol III subunit (Fig. 4E), explaining why the Pol IV-T120P–Pol IIIα interaction was observed by EMSA (Fig. S3).

The results of BLI experiments support the conclusion that the T120P mutation affects interactions of Pol IV with Pol IIIα (Table 2), although as noted above, the binding constants were stronger. While we do not understand why the full-length Pol IV-T120P was comparable to Pol IV for interaction with Pol IIIα^His (K_D1 = 4.88 nM and K_D2 = 62.1 nM for Pol IV-T120P compared with K_D1 = 13.9 nm and K_D2 = 52.6 nM for Pol IV) (Table 2), Pol IV^CD-T120P was impaired in this interaction (K_D1 = 59 nm and K_D2 = 819 nM for Pol IV^CD compared with K_D1 = 216 nM and K_D2 = 3,790 nM for Pol IV^CD-T120P) (Table 2). Although we detected an interaction of Pol IIIεθ with Pol IV, Pol IV^CD, and Pol IV^LF, we were unable to obtain reliable values describing their kinetic constants (Table 2, footnote b). That Pol IV-T120P is impaired in interaction with both Pol IIIα and Pol IIIεθ suggests that two molecules of Pol IV act to displace Pol III as part of the Pol III-Pol IV switch, with one Pol IV contacting Pol IIIα and a second contacting Pol IIIεθ (see Discussion).

Pol IV-T120P is impaired in interaction with the single-stranded DNA binding protein.

Although our published in vitro work demonstrating the importance of Pol IV residue T120 to displacing Pol III core from a β clamp assembled at the 3′ end of a primed DNA substrate was performed in the absence of SSB (15), the Pol IV-SSB interaction nevertheless stimulates Pol IV replication and is presumably required for Pol IV replication in vivo (27). In light of our previous finding that Pol IV-T120P was impaired in tolerating MMS-induced DNA lesions in vivo (15), we questioned whether the T120P mutation also impaired the Pol IV-SSB interaction. We first measured the affinity of full-length Pol IV for SSB by using BLI. Since Pol IV reportedly makes multiple contacts with SSB (27), we used the two-site model to analyze the BLI results, yielding K_D values of 0.47 and 11.6 nM for the Pol IV-SSB interaction (Table 3). By comparison, the K_D values for the Pol IV-T120P–SSB interaction were 37.7 and 70.3 nM (Table 3), demonstrating that the T120P mutation impairs this interaction. The Loparo lab has described a similar negative effect of the T120P mutation on the Pol IV-SSB interaction (31).

TABLE 3.

Interaction of Pol IV with SSB^a

Interactionb	Expt	KD (nM)	Avg KD, nM (range)
			KD 1	KD 2
Pol IV-SSBHis	1	KD1 = 0.33	0.47 (0.27)	11.6 (1.40)
		KD2 = 10.9
	2	KD1 = 0.60
		KD2 = 12.3
Pol IV-T120P–SSBHis	1	KD1 = 42.5	37.7 (9.6)	70.3 (7.4)
		KD2 = 74.0
	2	KD1 = 32.9
		KD2 = 66.6
Pol IVCD-SSBHis	1	KD1 = 28.1	28.5 (0.70)	340 (52.0)
		KD2 = 314
	2	KD1 = 28.8
		KD2 = 366
Pol IVCD-T120P–SSBHis	1	KD1 = 2,190	1,845 (690)	3,095 (490)
		KD2 = 3,340
	2	KD1 = 1,500
		KD2 = 2,850
Pol IVLF-SSBHis	1, 2	NDc	ND	ND

^aSee Table S3 in the supplemental material for the k_a (on-rate), k_d (off-rate), χ² (likelihood of no relationship), and R² (goodness of fit) values for these interactions.

^bThe indicated His-tagged protein (ligand) was captured on a HIS1K biosensor, and interactions were measured using BLI.

^cND, no interaction was detected.

We next asked whether both Pol IV^CD and Pol IV^LF interacted with SSB, as previously reported (27). While we failed to detect a Pol IV^LF-SSB interaction, Pol IV^CD bound SSB substantially weaker than full-length Pol IV (Table 3), consistent with both Pol IV domains contributing to the interaction. Unexpectedly, the BLI results for the Pol IV^CD-SSB interaction fit poorly to the one-site binding model; however, the results fit well to the two-site model, yielding K_D values of 28.5 and 340 nM (Table 3). We next measured the affinity of Pol IV^CD-T120P for SSB. The results fit well to the two-site binding model, yielding K_D values of 1,845 and 3,095 nM. These BLI results, indicating the T120P mutation impedes the Pol IV^CD-SSB interaction, were confirmed by SPR (Fig. 5; see Fig. S5 in the supplemental material).

FIG 5.

Pol IV-T120P is impaired in interactions with SSB. Shown are representative SPR sensorgrams for the (A) Pol IV^CD-SSB^His and (B) Pol IV^CD-T120P–SSB^His interactions. (C) Summary of the kinetic parameters describing the SSB^His-Pol IV^CD and SSB^His–Pol IV^CD-T120P interactions. Instead of χ² (goodness of fit), the ClampXP 3.50 software provides residual sum of squares (goodness of fit) values, which were each <10% of the respective R_max, confirming the specificity of each interaction. See Fig. S5 for a cartoon depiction of the SPR method used, as well as the k_a (on-rate) and k_d (off-rate) values for these interactions.

Multiple E. coli Pols interact with each other in vitro.

Given the importance of the Pol IV-Pol III interaction to their exchange (15), we asked if other E. coli Pols interact. Using a combination of SPR and BLI, we analyzed each possible pairwise interaction between the five E. coli Pols. For these experiments, we used Pol V (UmuD′₂C) rather than the individual UmuD′₂ and UmuC subunits, and analyzed Pol IIIα and Pol IIIεθ separately instead of using the Pol III core complex; the remaining three E. coli Pols (Pol I, Pol II, and Pol IV) are each comprised of a single polypeptide. Our results, summarized qualitatively in Fig. 6A, identified 8 of the 14 possible Pol-Pol interactions. In addition to its interactions with Pol IIIα and Pol IIIεθ, Pol IV interacted with each of the remaining three E. coli Pols (Pol I, Pol II, and Pol V) (Fig. 6D and E and Table 4). In addition to Pol IV, Pol II interacted with both Pol IIIα and Pol IIIεθ, as well as Pol I (Fig. 6B and C and Table 4). Interestingly, the BLI and SPR results for each of these interactions, except Pol II-Pol I and Pol IV-Pol V, fit well to the two-site model, suggesting multiple contacts, with most of the K_D values falling in the nM-to-μM range (Table 4). The Pol II-Pol IIIα and Pol IV-Pol I interactions were modeled with subnanomolar K_D1 values. Based on results of gel filtration experiments, we previously described a Pol II-Pol III core interaction (36). Our finding that the level of the shift of free Pol II into the Pol II-Pol III complex was incomplete suggests that either Pol II has a reduced affinity for Pol III core relative to the individual Pol IIIα and Pol IIIεθ subassemblies, or Pol II may contact more than two sites on Pol IIIα, and as a result, the two-site interaction model used here incorrectly describes the kinetic values for these interactions. A similar situation may explain the K_D values for the Pol IV-Pol I interaction. Regardless, the reproducibility of the k_a and k_d values for these Pol-Pol interactions, taken together with the fact that 6 of the 14 possible Pol-Pol interactions were not detected, nevertheless supports the specificity of these Pol-Pol interactions in vitro.

FIG 6.

Summary of E. coli Pol-Pol interactions. (A) Cartoon summarizing the E. coli Pol-Pol interactions identified in this work. Red arrows represent interactions between TLS Pols (Pol II, Pol IV, Pol V), while black arrows represent interactions between TLS Pols (Pol II, Pol IV) and Pols involved in DNA replication and repair (Pol I, Pol III). Shown are representative SPR sensorgrams for the (B) Pol IIIα^His-Pol II, (C) Pol IIIεθ^His-Pol II, (D) Pol II^His-Pol IV, and (E) Pol V^His-Pol IV interactions. See Table 4 and Table S4 in the supplemental material for a complete summary of the kinetic values describing each of the different Pol-Pol interactions, including the Pol I^His-Pol II and Pol I^His-Pol IV interactions that are not included in this figure because they were measured using BLI.

TABLE 4.

Pol-Pol interactions^a

Interaction	Expt	KD (nM)	Avg KD, nM (range)
			KD 1	KD 2
Pol II-Pol IHisb	1	KD1 = 81.1	73.0 (16.3)	NAe
	2	KD1 = 64.8
Pol II-Pol IIIαHisc	1	KD1 = 0.15	0.13 (0.05)	1,064 (127)
		KD2 = 1,000
	2	KD1 = 0.10
		KD2 = 1,127
Pol II-Pol IIIεθHisc	1	KD1 = 59.7	76.7 (34)	113 (34.6)
		KD2 = 95.4
	2	KD1 = 93.7
		KD2 = 130
Pol IIHis-Pol IVc	1	KD1 = 640	648 (15)	3,084 (3,758)
		KD2 = 1,205
	2	KD1 = 655
		KD2 = 4,963
Pol IV-Pol IHisb	1	KD1 = 0.004	0.010 (0.011)	37.5 (7.3)
		KD2 = 41.1
	2	KD1 = 0.015
		KD2 = 33.8
Pol IV-Pol IIIαHisc,d	1	KD1 = 307	384 (457)	779 (153)
		KD2 = 786
	2	KD1 = 194
		KD2 = 852
	3	KD1 = 651
		KD2 = 699
Pol IV-Pol IIIεθHisc,d	1	KD1 = 279	422 (287)	3,040 (1,516)
		KD2 = 3,798
	2	KD1 = 566
		KD2 = 2,282
Pol IV-Pol VHisc	1	KD1 = 43.7	36.8 (13.7)	NA
	2	KD1 = 30.0

^aSee Table S4 for the k_a (on-rate) and k_d (off-rate) values for these interactions.

^bThe indicated His-tagged protein (ligand) was captured on a HIS1K biosensor prior to analyzing its interaction with the indicated untagged protein (analyte) using BLI. χ² values (likelihood of no relationship) ranged from 0.5474 to 1.070, while R² (goodness of fit) ranged from 0.9493 to 0.9972. Results of the Pol II-Pol I^His and Pol IV-Pol I^His interactions were analyzed using the 1:2 bivalent analyte model.

^cThe indicated His-tagged protein (ligand) was captured in the left channel of the SPR sensor surface by Penta-His antibody (Qiagen) that was attached to both the left and right channels prior to flowing the untagged protein (analyte) over both channels. Instead of χ² (goodness of fit), the ClampXP 3.50 software provides residual sum of squares (goodness of fit) values, which were each <10% of the respective R_max, confirming the specificity of each interaction.

^dThese values were reported in Fig. 4E and are included here for comparison with the other Pol-Pol interactions.

^eNA, not applicable.

DISCUSSION

Pol IV can be recruited to the DNA by switching places with a stalled Pol III via a poorly defined mechanism involving the β clamp ((14, 16–18, 20; reviewed in reference 21) and possibly SSB (27, 31). Our finding that the T120P mutation abrogated the ability of Pol IV^CD to displace Pol III core from the β clamp assembled at the 3′ end of a primed DNA template in vitro in an assay that omitted SSB (15) suggests that one or more previously unidentified Pol III-Pol IV and/or Pol IV-β clamp interactions disrupted by the T120P mutation are required for the Pol III-Pol IV switch. Since Pol IV-T120P (and Pol IV^CD-T120P) interacts normally with the β clamp (Fig. 2 and Table 1), and the components of our in vitro Pol III-Pol IV switching assay were limited to a primed DNA template, purified Pol IV^CD (or Pol IV^CD-T120P), and Pol III HE (15), we conclude that one or both of the impaired Pol IV-Pol III interactions described in this work (Fig. 4 and Table 2) are responsible for the Pol III–Pol IV-T120P switching defect. It is interesting that the T120P mutation interferes with both the Pol IIIα and Pol IIIεθ interactions. Residue T120 is one helical turn from the start of α-helix E, and its replacement with proline likely truncates the start of this helix, changing the structure of the short sequence in Pol IV between β-strand 6 and α-helix E. Thus, it remains unclear whether residue T120, or one or more residues in its immediate vicinity whose structure is altered by the T120P mutation, mediates the interaction of Pol IV with Pol IIIα and/or Pol IIIεθ.

How might Pol IV-Pol III interactions contribute to the Pol III-Pol IV switch? Based on results of structural analyses and molecular modeling, Pol III core is predicted to adopt distinct conformations on the β clamp, depending on whether it is actively replicating DNA, proofreading misinsertions, or stalled at a lesion (20, 37) (Fig. 7A). Moreover, our finding that Pol IV contacts multiple Pol III core surfaces suggests that Pol IV may interact differently with distinct conformations of the Pol III core-β clamp-DNA complex. For example, Pol IV may be recruited to sites of replication in part by interactions with the replicating and/or proofreading conformations of the Pol III core-β clamp-DNA complex. After Pol III stalls at a DNA lesion, the conformation of the Pol III core-β clamp-DNA complex changes, which may allow Pol IV to interact differently with the Pol III core-β clamp-DNA complex to displace Pol III core from the clamp to permit Pol IV access to the DNA for TLS (Fig. 7B). While the stoichiometry of the Pol III core-β clamp-Pol IV-DNA complex is unknown, the finding that a similar surface of Pol IV interacts with both the Pol IIIα and Pol IIIεθ subunits suggests that two Pol IV molecules contribute to Pol III-Pol IV switching, with one Pol IV contacting Pol IIIα and a second Pol IV contacting Pol IIIεθ in such a way as to displace Pol III core from the β clamp (Fig. 7B). Consistent with this scenario, molecular modeling suggests that when Pol III core adopts the stalled position, Pol IV may simultaneously interact with both the β clamp and Pol IIIα or Pol IIIεθ (Fig. 7C). Importantly, based on the modeling, the T120P mutation is positioned in such a way that it could interfere with both of these Pol IV-Pol III interactions. The possibility that two Pol IV molecules are required to displace Pol III core from the face of the β clamp is consistent with our finding that the β⁺/β^C heterodimeric clamp protein bearing one protomer with a functional cleft (β⁺) and a second protomer that lacks a functional cleft (β^C) fully supported Pol III-Pol IV switching in vitro (14). In this case, Pol IIIα bound the single cleft in β⁺/β^C, presumably leaving Pol IIIεθ to adopt a conformation similar to its hypothesized conformation when Pol III is stalled at a lesion (Fig. 7A). Thus, a single Pol IV may be sufficient to mediate a switch with Pol III bound to β⁺/β^C. However, a determination of whether both the Pol IV-Pol IIIα and Pol IV-Pol IIIεθ interactions are required, or whether the Pol IV-Pol IIIα or Pol IV-Pol IIIεθ interaction is on its own sufficient to displace Pol III from the β clamp, must await the identification of separation-of-function mutants impaired in the individual interactions. Finally, β clamp-DNA interactions contribute modestly to Pol IV function in vivo (38). Thus, in addition to serving as a mobile platform that tethers the different Pols to the DNA (reviewed in reference 21), dynamic interactions of the β clamp with Pol III core, Pol IV, and the DNA may play an active role in positioning Pol III and Pol IV relative to each other as well as the DNA template to facilitate access of Pol IV to damaged DNA for TLS. Current efforts are focused on determining whether the Pol IV^CD-β clamp interaction described here (Fig. 2 and Table 1) collaborates with one or more Pol III-Pol IV interactions to help recruit Pol IV to the DNA and/or enable a Pol III-Pol IV switch.

FIG 7.

A Model for Pol III-Pol IV switching. (A) In silico models for possible conformations of the Pol III core-β clamp-DNA complex during replication, proofreading, or when stalled at a DNA lesion (37). (B) A model for Pol III-Pol IV switching in which two molecules of Pol IV are recruited to a replicating or proofreading Pol III via interactions with the β clamp, Pol IIIα, and Pol IIIεθ. When Pol III adopts a stalled conformation, Pol IV contacts different surfaces of Pol IIIα and Pol IIIεθ, which contributes to the dissociation of Pol III from the β clamp. Pol IV then binds to the face of the β clamp and gains control of the 3′ end of the nascent DNA strand to catalyze TLS. Due to its distributive nature, Pol IV releases the DNA and dissociates from the β clamp after replicating several nucleotides past the lesion, allowing Pol III to regain control of the replication fork for continued replication. For simplicity, Pol IIIθ is omitted from this model. See the Discussion for further details. (C) In silico model of a possible conformation of the stalled Pol IIIαεθ-β clamp-DNA-Pol IV₂ complex. The position of Pol IV residue T120 (red) is indicated.

The Maki lab previously described roles for both Pol IV^CD and Pol IV^LF in the Pol IV-SSB interaction (27). While we failed to detect a Pol IV^LF-SSB interaction, full-length Pol IV bound to SSB more strongly than Pol IV^CD (Table 3), suggesting that residues in both Pol IV^LF and Pol IV^CD contribute to the Pol IV-SSB interaction. Although SSB is not required for Pol III-Pol IV switching in vitro (15, 18), Chang et al. recently concluded that the Pol IV-SSB interaction contributes to the recruitment of Pol IV to sites of replication for Pol III-Pol IV switching in vivo (31). That Pol IV uses a shared surface for interaction with SSB, Pol IIIα, and Pol IIIεθ suggests that these partner proteins compete with each other for interaction with Pol IV. Thus, in addition to recruiting Pol IV to sites of replication, SSB may also help to regulate the Pol III-Pol IV switch by competing with Pol III core for interaction with Pol IV, which is required for switching. While impaired Pol IV-Pol III core interactions clearly interfere with their switching in vitro (15), it seems likely that impaired Pol IV-SSB interactions are one cause of the MMS sensitivity of the dinB89 (Pol IV-T120P) mutant strain. However, the impaired Pol IV-Pol III core interactions may also contribute. Separation-of-function mutations that distinguish the different Pol III core and SSB interactions are required to define their respective contributions to Pol IV function in vivo.

In addition to the Pol IV-Pol III core interactions discussed above, we identified several other E. coli Pol-Pol interactions (Fig. 6 and Table 4). All three SOS-regulated TLS Pols (Pol II, Pol IV, Pol V) interact with each other through Pol IV. In addition, Pol II and Pol IV each interact with Pol I and Pol III (Fig. 6A). While these in vitro results must be corroborated by more detailed biochemical and genetic analyses before any conclusions may be made regarding their biological importance, there are at least three instances where certain pairs of these E. coli Pols are known to functionally interact in vivo. In the first example, Pol II corrects misinsertions catalyzed by Pol III in vivo (39). Thus, interactions of Pol II with Pol IIIα and/or Pol IIIεθ (36) (Fig. 6B and C and Table 4) may facilitate access of Pol II to replication forks, possibly via the Pol III-Pol II switch that we described previously, to enable extrinsic proofreading of Pol III errors by Pol II (29, 36). As a second example, the Pol II-Pol IV interaction (Fig. 6D and Table 4) may enable access to the DNA of Pol II following Pol IV replication/TLS. Consistent with this hypothesis, Pol II limits the frequency of mutations catalyzed by Pol IV in vivo (39). Such a switch could serve as a transitional step between Pol IV TLS and Pol III replication by providing the opportunity for Pol II to ensure that Pol IV replicated a sufficient length of DNA past the lesion in an accurate manner before returning control of the nascent strand to Pol III. This could provide a mechanism to limit futile cycles of Pol III recruitment and its subsequent displacement due to the Pol IIIεθ proofreading assembly detecting the DNA lesion and/or Pol IV misinsertions (40). As a third example of multiple Pols functionally interacting, Pol IV and Pol V are both required for tolerance of benzo(a)pyrene adducts in vivo (41). Thus, the Pol IV-Pol V interaction (Fig. 6E and Table 4) may contribute to a Pol IV-Pol V switch for efficient tolerance of this lesion. Likewise, interactions of Pol I with Pol II and Pol IV (Table 4) may facilitate access of these TLS Pols to lesions that are encountered by Pol I during repair synthesis. Consistent with the possible contribution of TLS Pols to DNA repair, Cohen et al. previously presented genetic evidence of a role for Pol IV in transcription-coupled nucleotide excision repair (42, 43). The identification of separation-of-function Pol mutants deficient for these individual interactions would represent important tools for investigating the biological importance of these novel Pol-Pol interactions.

MATERIALS AND METHODS

E. coli strains and plasmids.

E. coli strains are described in Table S5 in the supplemental material. The salient features of each plasmid directing the overexpression of the indicated recombinant protein(s) are described in Table S5. Plasmid pβBack^HMK, which expresses ^Hisβ^Back, was constructed by GenScript under a fee-for-service agreement. All other plasmids were constructed by subcloning into the indicated plasmid previously cloned NdeI-BamHI fragments corresponding to the coding sequence of the gene for the specified protein using standard cloning approaches (44, 45). E. coli strain DH5α (Table S5) was made competent using CaCl₂ (45) and was used as the host for high-efficiency transformation of plasmid ligation products. The correct sequence of each plasmid insert, as well as the NdeI and BamHI cloning sites, was confirmed by automated Sanger sequencing (Roswell Park Biopolymer Facility, Buffalo, NY, and Genewiz, South Plainfield, NJ).

Protein purification.

Pol V^His was overexpressed using E. coli strain RW644, while all other recombinant proteins were overexpressed using E. coli strain BL21(DE3) (Table S5). Pol I (46), Pol II (44), Pol IIIα (44), Pol IIIαεθ^His (35), Pol IV (14), Pol IV-T120P (15), Pol IV^CD (15), Pol IV-T120P^CD (15), Pol V^His (47), β clamp (48), β^HMK (referred to as β^His) (49), ^Hisβ^Back (14), and SSB^His (50) were each purified as described previously. Pol IV^LF was purified as described previously (14), and the N-terminal His₁₀ tag was removed from the purified Pol IV^LF using factor Xa (New England Biolabs, Inc.) according to the manufacturer’s recommendation. The untagged Pol IV^LF was separated from intact His₁₀-tagged Pol IV^LF and the cleaved His₁₀ tag by using HisTrap high-performance (HP) chromatography (GE Healthcare) as described previously (51). Pol I^His (pRM106), Pol II^His (pMKS146), Pol IIIα^His (pMKS147), and Pol IV^His (pRM114) were purified using HisTrap HP chromatography (GE Healthcare) as described previously (51). Pol IIIεθ^His was purified from pRM111, which encodes Pol IIIαεθ^His but expresses Pol IIIεθ^His at a level higher than Pol IIIα, using HisTrap HP chromatography (GE Healthcare) as described in reference 51, followed by gel filtration using a Superose 12 column (GE Healthcare) equilibrated in HBS-EP buffer (10 mM HEPES [pH 7.4], 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20) to resolve Pol IIIεθ^His from Pol IIIαεθ^His.

Molecular modeling.

The in silico model of the full-length Pol IV-β clamp complex was built by aligning Pol IV^LF in the full-length Pol IV X-ray crystal structure (PDB no. 4IR9) with Pol IV^LF in the Pol IV^LF-β clamp X-ray crystal structure (PDB no. 1UNN) using MacPyMOL (version 2.0.4; Molecular Graphics System, Schrödinger, LLC). DNA was added from the cryo-electron microscopy (cryo-EM) structure of the Pol IIIαετ_C in complex with the β clamp on DNA (5FKV) by grouping the clamp and the DNA substrate and overlaying them with the clamp in the Pol IV^LF-β clamp crystal (PDB no. 1UNN). The in silico model of the Dpo4-PCNA complex was built by aligning the trimeric PCNA ring structure (PCNA1-PCNA2-PCNA3) in the X-ray crystal structure (PDB no. 2NTI) with PCNA1 in the Dpo4-PCNA1-PCNA2 X-ray crystal structure using MacPyMOL. DNA was modeled as described above by overlaying and grouping the clamp and the DNA substrate together and overlaying them with PCNA (PDB no. 2NTI). The in silico model of the Pol IV-β clamp complex in which the Pol IV^LF interacts with the clamp rim and cleft while the Pol IV^CD interacts with the back of the β clamp was built by threading the Pol IV amino acid sequence through the PCNA1-bound structure of the homologous Sulfolobus solfataricus P2 DNA polymerase IV (Dpo4) from the Dpo4-PCNA1-PCNA2 X-ray crystal structure (PDB no. 3FDS) using SWISS-MODEL (https://swissmodel.expasy.org). To predict how the catalytic domain of Pol IV (Pol IV^CD; residues 1 to 230) interacts with the β clamp, we aligned the Pol IV C-terminal six residues (QLVLGL) in the Pol IV SWISS-MODEL with the corresponding sequence in the Pol IV^LF-β clamp X-ray crystal structure (PDB no. 1UNN) by using MacPyMOL. DNA was modeled as described above.

The in silico models for the open conformations of Pol IIIα representing possible proofreading and stalled conformations of the Pol IIIαεθ-β clamp-DNA complex were constructed by aligning the Pol IIIα sequence in the Pol IIIαετ_C-β clamp-DNA complex (PDB no. 5FKV) with the structure of the Thermus aquaticus Taq Pol (PDB no. 2HPI) as described previously (37). These conformational models differ with respect to the conformation of Pol IIIα and the position of Pol IIIε: we have modeled Pol IIIε bound to the cleft of the β clamp in the proofreading conformation, consistent with the reported importance of this interaction to the proofreading activity of Pol III (37), and have modeled Pol IIIα in an extended conformation with Pol IIIε free of (i.e., not bound to) the β clamp in the stalled conformation (20). Two Pol IV protomers were modeled onto the stalled Pol IIIαεθ-β clamp-DNA complex. The Pol IV protomer adjacent to Pol IIIα was aligned using 1UNN and the full-length Pol IV (4IR9) in such a way to contact the cleft and rim of the β clamp. The Pol IV adjacent to Pol IIIεθ was docked similarly, but was rotated out and off the clamp’s cleft to which Pol IIIε was bound without disrupting the Pol IV^LF-β clamp rim interaction. The cryo-EM structure of Pol IIIαετ_C-β-DNA complex (PDB no. 5FKV) is missing residues A⁹²⁸ to Q⁹⁴² of Pol IIIα and Pol IIIθ. Therefore, to model the full-length Pol IIIα protein, the full structure of Pol IIIα as described by Ozawa et al. (52) was extracted and superimposed onto the portion of Pol IIIα present in the cryo-EM complex (PDB no. 5FKV). Pol IIIθ was similarly modeled onto Pol IIIε as described by Ozawa et al. (52).

Electrophoretic mobility shift assay.

Purified proteins (3 μg each, unless otherwise stated) were incubated independently or together, as indicated, in 25 mM Tris-HCl [pH 8.5], 19.2 mM glycine, and 100 mM NaCl for 10 min at room temperature. Samples were mixed with 50% glycerol, loaded into the wells with 1% TAE–agarose gel (agarose plus 40 mM Tris, 20 mM acetic acid, 1 mM EDTA), and electrophoresed at constant voltage (100 V) for ∼2 h. After electrophoresis, the agarose gel was submerged in buffer A containing 0.1% SDS at 25°C for 30 min prior to transfer of proteins to a polyvinylidene difluoride (PVDF) membrane using a Hoefer SE 600 series electrophoresis chamber at constant amperage (50 mA) for 6 h at 4°C. After blocking of the PVDF membrane in 1× TS (50 mM Tris-HCl [pH 7.6], 150 mM NaCl) containing 0.05% Tween 20 and 2% nonfat milk, it was incubated with rabbit anti-Pol IV polyclonal sera (1:20,000) followed by a secondary goat anti-rabbit horseradish peroxidase (HRP)-conjugated antibody (Sigma; 1:50,000) as described previously (53). The anti-Pol IV polyclonal antibody was generated using a University at Buffalo IACUC-approved protocol and was described previously (53). The PVDF membrane was washed with 1× TS containing 0.05% Tween 20 prior to development with Clarity Western ECL enhanced chemiluminescence substrate (Bio-Rad) for 2 min. Pol IV was visualized using a ChemiDoc Imager (Bio-Rad).

Bio-layer interferometry.

BLI experiments were conducted at 25°C using a FortéBio Octet RED 96e instrument (Sartorius). All experiments were carried out in HBS-EP buffer (10 mM HEPES [pH 7.4], 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20) using FortéBio Penta-His (HIS1K) dip and read biosensors. Biosensor tips were hydrated for 10 min in HBS-EP buffer and allowed to equilibrate to obtain a stable baseline in HBS-EP buffer for 1 min prior to capturing the His-tagged ligand (300 to 1,000 nM). Following ligand capture, biosensors were quenched in SuperBlock (Thermo Fisher Scientific) and EZ-Link Biocytin (50 μM) (Thermo Fisher Scientific) for 5 min before being allowed to reach a stable baseline for 3 min. The untagged analyte (5 to 10,000 nM) was allowed to associate with individual biosensors for 3 min before monitoring of dissociation for 3 min. Binding kinetics (K_D, k_a, k_d) were calculated from binding curves using the FortéBio Data Analysis HT software (version 12.0.2.59; Molecular Devices, LLC). Reference biosensors were also utilized to measure background and nonspecific binding.

Surface plasmon resonance.

SPR experiments were conducted at 25°C using a dual-channel Reichert SR7500DC instrument. All experiments were carried out in HBS-EP buffer (10 mM HEPES [pH 7.4], 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20). Approximately 4,000 response units of bovine serum albumin (BSA)-free Penta-His antibody (Qiagen) were covalently captured on a 500,000-Da carboxymethyl dextran chip (Reichert, Inc.) via amine coupling to both channels according to the manufacturer’s recommendation. Approximately 1,000 RU of each His₆-tagged protein (ligand) was captured in the left channel only for 3 min at a flow rate of 25 μL/min, and 250 to 4,000 nM untagged protein (analyte) was introduced to both the left and the right channels for 1.5 min at a flow rate of 25 μL/min. Interaction of analyte with the right channel permitted the detection of nonspecific interactions between the analyte and the antibody, which were subtracted as background from the signal provided between the ligand (His-tagged protein) and analyte in the left channel. Interactions were analyzed using a kinetic titration approach in which increasing concentrations of the analyte were injected over both channels (54). Once the injections were completed, the chip surface was regenerated using a regeneration cocktail introduced to both channels for 45 s at a flow rate of 25 μL/min. The regeneration cocktail was composed of three solutions (solution A, 0.5 M oxalic acid, 0.05 M phosphoric acid, 0.05 M formic acid, and 0.05 M malonic acid, adjusted to pH 5.0 using NaOH; solution B, 0.46 M KSCN, 1.83 M MgCl₂, 0.92 M urea, and 1.83 M guanidine-HCl; and solution C, 20 mM EDTA) mixed at a 1:1:1 ratio as described previously (55). During regeneration, ligand and analyte were removed from the antibody surface prior to the analysis of another interaction. Kinetic constants (k_a, k_d, K_D) were determined using Scrubber 2.0c (Biologic Software Pty Ltd., Australia) and ClampXP 3.50 software provided by Reichert, Inc. (56).