DNA polymerase I acts in translesion synthesis mediated by the Y-polymerases in Bacillus subtilis


  • Stéphane Duigou,

    1. Laboratoire de Génétique Microbienne, Domaine de Vilvert, INRA, 78352 Jouy en Josas Cedex, France.
    Search for more papers by this author
    • Department of Molecular Microbiology, Tufts University School of Medicine and The Howard Hughes Medical Institute, Boston, MA 02111, USA.

  • S. Dusko Ehrlich,

    1. Laboratoire de Génétique Microbienne, Domaine de Vilvert, INRA, 78352 Jouy en Josas Cedex, France.
    Search for more papers by this author
  • Philippe Noirot,

    1. Laboratoire de Génétique Microbienne, Domaine de Vilvert, INRA, 78352 Jouy en Josas Cedex, France.
    Search for more papers by this author
  • Marie-Françoise Noirot-Gros

    Corresponding author
    1. Laboratoire de Génétique Microbienne, Domaine de Vilvert, INRA, 78352 Jouy en Josas Cedex, France.
    Search for more papers by this author

*E-mail Marie-Francoise.Gros@jouy.inra.fr; Tel. (+33) 1 34 65 25 19; Fax (+33) 1 34 65 25 21.


Translesion synthesis (TLS) across damaged DNA bases is most often carried out by the ubiquitous error-prone DNA polymerases of the Y-family. Bacillus subtilis encodes two Y-polymerases, Pol Y1 and Pol Y2, that mediate TLS resulting in spontaneous and ultraviolet light (UV)-induced mutagenesis respectively. Here we show that TLS is a bipartite dual polymerase process in B. subtilis, involving not only the Y-polymerases but also the A-family polymerase, DNA polymerase I (Pol I). Both the spontaneous and the UV-induced mutagenesis are abolished in Pol I mutants affected solely in the polymerase catalytic site. Physical interactions between Pol I and either of the Pol Y polymerases, as well as formation of a ternary complex between Pol Y1, Pol I and the β-clamp, were detected by yeast two- and three-hybrid assays, supporting the model of a functional coupling between the A- and Y-family polymerases in TLS. We suggest that the Pol Y carries the synthesis across the lesion, and Pol I takes over to extend the synthesis until the functional replisome resumes replication. This key role of Pol I in TLS uncovers a new function of the A-family DNA polymerases.


Chromosome replication proceeds with high fidelity and processivity on an undamaged DNA template. However, the replicative DNA polymerases are most often unable to incorporate a nucleotide across a non-instructive damaged base. Translesion DNA synthesis (TLS) is mediated by specialized DNA polymerases, able to replicate across a damaged site. The vast majority of TLS enzymes belong to the Y-family of DNA polymerases that display a low fidelity of DNA synthesis, a low processivity, and are devoid of 3′−5′ proofreading activity (Friedberg et al., 2001). As a consequence of these properties, the fidelity of Y-family polymerases depends on the specific DNA lesion and the specific polymerase considered. Crystallographic studies revealed that they adopt the ‘right hand’-like structure of the classical DNA polymerases, with a more flexible active site, provided by shorter ‘thumb’ and ‘fingers’, which permits accommodation of various distorting lesions (Ling et al., 2001; Zhou et al., 2001). The Y-polymerases also possess an additional C-terminal little finger (LF) domain that contributes to the overall enzymatic properties of the Y-polymerases (Boudsocq et al., 2004).

Revp1 and Polη in yeast, together with polymerase IV (Pol IV) and Pol V in Escherichia coli, are the prototypical members of this family (Ohmori et al., 2001). Pol IV and Pol V are part of the SOS regulon, and are differentially involved in mutagenesis. Pol IV mediates untargeted mutagenesis in vivo, and is required for adaptive mutagenesis (Kim et al., 1997; Wagner and Nohmi, 2000; McKenzie et al., 2001). In the presence of RecA, the UmuD′2C protein complex (Pol V) is the major polymerase that catalyses the efficient bypass of ultraviolet light (UV)-induced DNA lesions in vitro (Tang et al., 2000; Pham et al., 2002; Fujii et al., 2004). Bacillus subtilis possesses two Y-family members, Pol Y1 and Pol Y2 encoded by the yqjH and yqjW genes respectively. Similar to Pol IV, Pol Y1 promotes untargeted mutagenesis when overproduced (Duigou et al., 2004), and acts in adaptive mutagenesis (Sung et al., 2003). However, unlike Pol IV, Pol Y1 expression is not SOS-controlled (Duigou et al., 2004). The second Y-polymerase, Pol Y2, is expressed only upon induction of the SOS response and is required for UV-induced mutagenesis. When overproduced, Pol Y2 also promotes untargeted mutagenesis (Duigou et al., 2004).

Translesion DNA synthesis can be also carried out by a repair polymerase which does not belong to the Y-family. In yeast, the B-family, Pol ζ, cooperates with other Y-polymerases to accomplish TLS (Johnson et al., 2000; 2001). In E. coli, the B-family Pol II is responsible for −2 deletion bypass products (Becherel and Fuchs, 2001). In the Gram-positive bacteria B. subtilis and Streptococcus pyogenes, the type C replicative polymerase DnaE, which is essential for the elongation step of chromosome replication (Dervyn et al., 2001), is also an error-prone TLS enzyme capable of extending mismatched termini in vitro (Bruck et al., 2003; Le Chatelier et al., 2003). Thus, depending on the particular conformation of the damaged primer end, TLS can be achieved in an error-free or error-prone manner by single, or a combination of polymerases.

The access of one or several specialized polymerases to the stalled replication fork implies a transient switch from the high-fidelity replicative DNA polymerase to the low-fidelity translesion DNA polymerase. During lesion bypass, the processivity clamp has been proposed to act as a ‘tool-belt’ to recruit the translesion DNA polymerases at the lesion site and to co-ordinate their activities (Pages and Fuchs, 2002). For most polymerases, DNA synthesis activity is increased by physical interaction with the clamp (Wagner et al., 2000; Haracska et al., 2002; Vidal et al., 2004), consistent with the presence of β-binding motifs in their primary structure (Dalrymple et al., 2001). In E. coli, the interaction of the replicative Pol III and of the TLS enzymes Pol IV and Pol V with the β-clamp homodimer is mediated by a hydrophobic pentapeptide motif (Bunting et al., 2003; Lopez De Saro et al., 2003; Burnouf et al., 2004). In B. subtilis, a β-binding signature was found to be important for Pol Y1- and Pol Y2-mediated untargeted mutagenesis (Duigou et al., 2004).

In E. coli, the A-family DNA polymerase I (EcPol I) was shown to interact with the β-clamp (Lopez de Saro and O’Donnell, 2001). EcPol I plays an essential role in filling DNA gaps that arise during DNA replication, recombination and repair (Kornberg and Baker, 1992). Recently, EcPol I was also found to be involved in chromosome termination in vivo (Markovitz, 2005). Structural characterization and biochemical studies of several prokaryotic Pol I enzymes established an organization in three functional domains consisting of an N-terminal domain associated with a 5′−3′ exonuclease activity, a central domain mediating the proofreading 3′−5′ exonuclease activity, and a C-terminal domain responsible for the polymerase activity (for a review, see Patel et al., 2001). The 5′−3′ exonuclease activity of EcPol I is required to remove RNA primers during the processing of Okazaki fragments and to remove damaged nucleotides during UvrABC-dependent excision repair, while the polymerase activity is filling the resulting gaps. The 3′−5′ exonuclease activity acts in proofreading by catalysing the excision of misincorporated nucleotides from the nascent strand (Kornberg and Baker, 1992). However, despite an overall structural similarity among the A-family polymerases, various members such as Pol I from Thermus aquaticus, Bacillus stearothermophilus and S. pneumonia lack the proofreading activity associated with the central domain (Tindall and Kunkel, 1988; Diaz et al., 1992; Aliotta et al., 1996). This deficiency results from the absence of critical residues that are involved in co-ordination of the divalent cations required for 3′−5′ exonuclease activity (Kiefer et al., 1998).

Here, we report that in B. subtilis Pol I (BsPol I) the 5′−3′ exonuclease function is essential for cell survival, whereas the polymerase activity is dispensable. Interestingly, we found that BsPol I belongs to a Gram-positive subfamily of the A-polymerases harbouring a vestigial and non-functional 3′−5′ exonuclease active site. Because the functionality of the 3′−5′ exonuclease was shown to affect the processing of DNA lesions by DNA polymerases (Eckert and Opresko, 1999), we investigated the role of the BsPol I in TLS mutagenesis.

We found that Pol I was catalytically required for most of the UV-induced mutagenesis mediated by Pol Y2. Pol Y1-dependent untargeted mutagenesis was also found to require the polymerase function of Pol I. Correlating with these genetic interactions, Pol I physically contacts Pol Y1 and Pol Y2 in a yeast two-hybrid assay. In addition, ternary interactions between Pol I, Pol Y1 and the β-clamp DnaN were detected in a yeast three-hybrid assay, suggesting the formation of a tripartite complex. Our results reveal that, B. subtilis TLS mutagenesis is mediated by a bipartite dual polymerase mechanism, in which the A-family DNA Pol I assists the Y-family DNA polymerases Pol Y1 and Pol Y2. These findings illustrate a striking difference between TLS mechanisms in the Gram-negative and Gram-positive bacteria and point to a new function of Pol I.


Bacillus subtilis Pol I functional domains

Alignment of the BsPol I sequence with orthologues from various eubacterial Pol I revealed that BsPol I belongs to the Gram-positive subfamily of A-polymerases with a vestigial 3′−5′ exonuclease active site (Fig. S1). Indeed, the amino acid residues critical for binding the metal ions necessary to catalyse the 3′−5′ exonuclease reaction (Derbyshire et al., 1995) are not conserved in BsPol I. A lack of 3′−5′ exonuclease activity has been established for the close relative B. stearothermophilus Pol I (Aliotta et al., 1996).

The BsPol I is entirely dispensable, as attested by our ability to construct and propagate a ΔpolA strain in rich media. Both the wild-type and the ΔpolA strains formed colonies of similar size, and had comparable growth rates in LB and in minimal media (data not shown), thus attesting that BsPol I is dispensable for normal growth. This is in contrast with EcPol I where the polymerase or the 5′−3′ exonuclease catalytic domains of the enzyme are required for growth in rich media (Joyce and Grindley, 1984). However, as in E. coli, the absence of BsPol I caused a profound decrease in the ability to repair UV- and methyl methane sulphonate (MMS)-induced DNA damages (see Table 1). These results are in agreement with the previous characterization of BsPol I-deficient mutants (Villani et al., 1974). The ypcP gene of B. subtilis encodes a putative 5′−3′ exonuclease which is homologous to the N-terminal part of BsPol I (38% identity). Disruption of ypcP in a polA+ background did not affect the strain viability (Table 1). However, the combination of ypcP and ΔpolA mutations was lethal (Table 1). The viability, as well as the survival to UV and MMS treatments, was restored to a level similar to that of the polA+ strain by the presence of an ectopic copy of polA in the ΔpolA ypcP strain. These results suggest that the co-lethality results from the lack of the essential 5′−3′ exonuclease activity.

Table 1.  Genetic characterization of Pol I catalytic mutant.
with ypcP a
survival (%)c
survival (%)d
  • a

    . Recipient strains harbouring various polA alleles were transformed with a donor DNA carrying pMUTIN::ypcP (from BSF2204), and a control DNA carrying pMUTIN::ypdP (from BSF2211) respectively. As donor DNAs are polA+, erythromycin-resistant transformants from polA-deficient recipient strains were tested for the presence of the

  • Δ polA (PhleoR, MMS sensitivity) or of Δ polA amyE:: polAcat (PhleoR, SpcR, MMS sensitivity). The ratio of transformation efficiency by the ypcP donor relative to the control is indicated.

  • b

    . The plasmid pIL253, which is a pAM

  • β

    1 derivative, requires the Pol I activity to replicate in B. subtilis (Bruand et al., 1993). The transformation efficiency by pIL253 (EmR) was normalized to that obtained with a control DNA for each recipient strain. Transformation efficiencies are expressed in number of transformants per

  • µ

    µg of DNA.

  • c

    . Survival after UV treatment at 60 J m−2. Values for polA+ and

  • Δ polA are from Fig. 1A.

  • d

    . Survival to methyl-methane sulphonate treatment (350 ng ml−1). The ratio MMSR/viable is indicated.

  • e

    . Assays were performed in the presence of 1% xylose.

polA+  1   5 × 10412.521
ΔpolA< 0.003< 10 0.04 4 × 10−3
ΔpolA amyE:: polAe  1   1.6 × 104 4.9830
ΔpolA amyE:: polAcate  0.1< 10 0.05 3 × 10−3

To investigate the contribution of the BsPol I polymerase activity to the viability of a ypcP strain, a BsPol I catalytic mutant was constructed. The Asp834 residue, which is invariant in all DNA polymerases (Delarue et al., 1990), was substituted by an Ala residue, yielding to polAD834A (Pol Icat). The corresponding Asp residue was shown to be essential for polymerase activity in the B. stearothermophilus Pol I large fragment (Asp 830) (Kiefer et al., 1997). The polAD834A strain was functionally characterized for its sensitivity to MMS and UV, its capacity to support Pol I-dependent plasmid replication (Bruand et al., 1993), and for its co-lethality with ypcP (Table 1). As expected, BsPol Icat was deficient for the initiation of plasmid pAMβ1 replication, and for the repair of DNA damages caused by MMS and UV treatments, confirming that it was polymerase-deficient. However, the finding that the polAD834A could be combined with a ypcP null mutation whereas ΔpolA could not suggested that 5′−3′ exonuclease activity was retained in the BsPol Icat mutant. Taken together, these results established that the 5′−3′ exonuclease activities of BsPol I and YpcP are redundant and essential for cell survival, whereas the BsPol I polymerase activity is dispensable.

BsPol I polymerase activity is required for UV-induced mutagenesis

The observation that BsPol I did not have 3′−5′ exonuclease proofreading activity raised the interesting possibility that it could participate in mutagenesis in Gram-positive bacteria. A ΔpolA strain exhibited a clear sensitivity to UV irradiation relative to the wild-type strain (Fig. 1A). Survival of the ΔpolA was moderately reduced (fivefold) at 40 J m−2, and dropped sharply (about 100-fold) at 60 J m−2. In all subsequent experiments, the UV doses were adjusted to 40 J m−2 in the ΔpolA strains and to 60 J m−2 in all the polA+ strains in order to maintain survival constant (5–10%). Mutagenesis was examined by scoring rifampicin-resistant (RifR) mutants, as described in Experimental procedures. Exposure to UV of the wild-type strain increased the proportion of RifR mutants about 15-fold over the background level (Fig. 1B). The proportion of RifR cells dropped sevenfold in the irradiated ΔpolA strain (P < 0.0001), suggesting that Pol I could act in UV-induced mutagenesis. Located downstream of polA on the B. subtilis chromosome, the mutM, ytaF and ytaG genes are likely to form an operon. The phleomycin-resistance cassette replacing polA contains an outward constitutive PsacB promoter that allows the expression of the phleoR as well as that of the three downstream genes. The last gene ytaG encodes a dephospho-CoA kinase, which is essential for cell survival (Kobayashi et al., 2003). As the ΔpolA::phleo strain grew as wild type, we concluded that the expression of the operon was not grossly affected. However, mutM, encodes 8-oxo Guanine (8-oxoG)-DNA glycosidase. In E. coli, mutM protects the cell from the mutagenic effects of 8-oxoG (Lu et al., 2001). To rule out that the reduced mutagenesis observed in the ΔpolA strain could result from a polar effect on mutM expression, a copy of polA was placed ectopically at the amyE locus. In this strain, survival to UV as well as UV-induced mutagenesis was restored to the wild-type level (P = 0.32), indicating that the ectopic copy of polA is functional, and that polA is the only gene from the operon required for mutagenesis at UV lesions.

Figure 1.

Effect of polA on UV-induced mutagenesis.
A. UV survival of polA and yqjW mutant strains. Strains JJS149 (WT), JJS96 (ΔpolA), JJS93 (ΔyqjW ) and JJS163 (ΔpolA ΔyqjW ) were subjected to UV irradiation, as described in Experimental procedures, and the proportions of surviving cells were plotted as a function of the UV doses. Results are the mean values of at least four independent experiments with standard deviations.
B. Effect of polA on UV-targeted mutagenesis. The proportions of RifR cells were determined with (dark bars) or without UV exposure (light bars). The proportion of surviving cells was maintained between 5 and 10%, subjecting the polA+ strains JJS149 (WT), JJS93 (ΔyqjW) and JJS157 (ΔpolA polA+) to 60 J m−2, and the ΔpolA strains JJS96 (ΔpolA), JJS163 (ΔpolA ΔyqjW) and JJS187 (ΔpolA polAcat) to 40 J m−2. In strains JJS157 and JJS187, the deficiency in polA was complemented by wild type and mutant copy derivatives of polA, respectively, inserted at the amyE locus. For each condition, at least eight independent measurements were performed, and the results were analysed by a LSmeans (Least Square means) statistical analysis. The mean values with corresponding standard deviation are also indicated in Table S3 in Supplementary material.

We examined the UV-induced mutagenesis in a strain expressing the polAD834A derivative, which encodes the catalytically deficient polymerase BsPol Icat. The proportions of RifR were compared in a ΔpolA strain carrying an ectopic copy of the polAD834A gene or of the wild-type polA (Fig. 1B). UV-induced mutagenesis remained at background level in the polAD834A strain indicating that the polymerase activity of BsPol I is required for UV-induced mutagenesis.

BsPol I acts in Pol Y2-mediated UV mutagenesis

In B. subtilis, Pol Y2 is responsible for the vast majority of the UV-induced mutagenesis (Duigou et al., 2004). Indeed, the proportion of RifR mutants was reduced 8.4-fold in the ΔyqjW strain compared with wild type (Fig. 1B). The decrease in mutagenesis observed in the ΔpolA strain was not significantly modified in the double mutant ΔpolA ΔyqjW (P = 0.41), suggesting that the two genes might act in the same pathway.

The overproduction of Pol Y2 from a chromosomal gene placed under the control of the IPTG-inducible Pspac promoter was shown to increase both untargeted and UV-induced mutagenesis (Duigou et al., 2004). We examined whether this stimulation required Pol I activity. Overproduction of Pol Y2 increased spontaneous mutagenesis similarly (approximately sixfold, P = 0.3) in the presence and in the absence of Pol I (Fig. 2A), suggesting that Pol I is not involved in Pol Y2-mediated untargeted mutagenesis. Upon exposure to UV irradiation, the mutagenesis increased 195-fold in a strain overproducing Pol Y2 relative to an isogenic strain overexpressing the polymerase-deficient Pol Y2cat (Fig. 2B). In the absence of BsPol I, the UV-induced mutagenesis dropped 10-fold relative to the polA+ strain upon overproduction of Pol Y2 (Fig. 2B), followed by a further drop to background levels upon overproduction of Pol Y2cat. In the absence of IPTG, the residual activity of the Pspac promoter enabled a low level of production of Pol Y2 (Duigou et al., 2004). This low level of Pol Y2 triggered UV mutagenesis, which was abolished in the presence of Pol Y2cat, and was dependent on the presence of BsPol I (Fig. 2B). Thus, these results indicate that BsPol I is required for the main part of the UV mutagenesis mediated by Pol Y2.

Figure 2.

Involvement of Pol I in Pol Y2-mediated untargeted and UV mutagenesis. Pol Y2 and Pol Y2cat and Pol Y2intN mutant derivatives were overproduced from the corresponding Pspac-yqjW constructs integrated at the chromosomal locus.
A. Pol I is not involved in Pol Y2-mediated untargeted mutagenesis. The proportions of RifR cells were determined in the presence (dark bars) or absence of IPTG (light bars), in the polA+ and the ΔpolA strains. The mean values were calculated from at least eight independent experiments (see also Table S3).
B. Pol I is involved in Pol Y2-mediated UV mutagenesis. Cultures were irradiated with the appropriate UV doses to maintain survival constant for all strains, as described in Fig. 1B. The proportions of RifR cells were determined in the presence (dark bars) or absence of IPTG (light bars). The mean values were calculated from at least eight independent experiments (Table S3).
C. In vivo expression levels of Pol Y2 and mutated derivatives in the polA+ and ΔpolA backgrounds. The immunodetection assays were performed as described in Experimental procedures, using anti-FLAG antibody. The signals corresponding to Pol Y2 were normalized to the amount of proteins loaded on the gels.

In the absence of BsPol I, UV mutagenesis remained stimulated 10-fold upon Pol Y2 overproduction, thus revealing a minor Pol Y2-dependent and Pol I-independent mutator activity. This fraction of UV mutagenesis could be caused by an excess of Pol Y2 acting on undamaged DNA, or could correspond to an activity of Pol Y2 on UV-damaged DNA. To distinguish between these two possibilities, we used the fact that Pol Y2-mediated mutagenesis on undamaged DNA required the integrity of the [QL(S,D)LF]β-binding consensus located at the C-terminal part of Pol Y2, whereas UV-induced mutagenesis did not (Duigou et al., 2004). As expected, overproduction of Pol Y2 and of Pol Y2intN, which carries the double substitutions L360A and L362A in the β-binding motif, stimulated UV mutagenesis to a similar extent in a polA+ strain (Fig. 2B). Interestingly, overproduction of Pol Y2 and Pol Y2intN also caused a roughly similar level of UV mutagenesis in a ΔpolA strain. The twofold increase observed in the presence of Pol Y2intN could be considered as statistically significant, although not highly (P = 0.01). We concluded that the β-binding consensus was not required for the polA-independent mutator activity of Pol Y2, suggesting that this activity did not result from a mutagenesis activity taking place on undamaged DNA. The intracellular levels of Pol Y2 and its mutated derivatives were similar in all the wild-type and ΔpolA background, as attested by immunodetection (Fig. 2C). Taken together, these results reveal two pathways of Pol Y2 activity in UV mutagenesis. The major pathway requires BsPol I polymerase function, whereas the minor pathway is independent of the presence of BsPol I.

BsPol I is required for the spontaneous mutagenesis mediated by Pol Y1 overproduction

Overproduction of Pol Y1, using a Pspac::yqjH construct carried by a multicopy plasmid, results in a sevenfold increase of spontaneous mutagenesis (Duigou et al., 2004). We examined whether BsPol I was required for this spontaneous mutagenesis mediated by Pol Y1. The proportion of RifR cells in the population was determined in the presence and in the absence of IPTG, in the wild type and in a ΔpolA genetic background (Fig. 3). Upon Pol Y1 overproduction, spontaneous mutagenesis increased approximately sixfold in the polA+ strain, whereas it remained at the background level in a ΔpolA strain (Fig. 3). This result suggests that BsPol I is involved in the spontaneous mutagenesis mediated by Pol Y1 overproduction. Placing an ectopic copy of polA in the ΔpolA background restored the mutagenesis mediated by Pol Y1 overproduction to wild-type level (P = 0.45). However, the expression of BsPol Icat failed to restore mutagenesis in the ΔpolA strain. Thus, the DNA polymerase catalytic activity of BsPol I is required for Pol Y1-mediated spontaneous mutagenesis.

Figure 3.

Pol I requirement in Pol Y1-dependent untargeted mutagenesis. Pol Y1 was overproduced from a multicopy plasmid carrying the Pspac-yqjH construct. The proportions of RifR cells were determined in strains harbouring the Pspac-yqjH carrying plasmid (pSD67), or the Pspac control plasmid (pDG148ΔpBR), in the presence (dark bars) or absence of IPTG (light bars). The various genetic backgrounds tested were described in Fig. 1B. The mean values were calculated from at least eight independent experiments (Table S3).

BsPol I physically interacts with Pol Y1 and Pol Y2

The involvement of BsPol I in Pol Y1- and Pol Y2-mediated mutator activities raises the possibility that BsPol I might establish physical contact with the two Y-polymerases. To test this hypothesis, we examined the potential binary interactions between BsPol I, Pol Y1 and Pol Y2, using the yeast two-hybrid assay. Full-length proteins, as well as protein fragments, were expressed as bait [binding domain (BD)] and prey [activated domain (AD)] fusions (Fig. 4). The Pol Y1 and Pol Y2 ‘N-terminal’ fragments included the ‘palm-finger’ and part of the ‘thumb’ structural domains of the DNA polymerases, while the ‘C-terminal’ part corresponded to the LF domain. Full-length Pol Y2 expressed as a bait exhibited a self-activation phenotype and could not be used in this assay. Full-length Pol I as well as a truncated N-terminal fragment (from aa 1–574) encompassing the 5′ exonuclease domain, and a C-terminal fragment (from aa 294–880) encompassing the polymerase domain were used both as preys and as baits. Also, a prey corresponding to an internal 174-aa-long specific interaction domain (SID) of Pol I was used. This domain of Pol I was originally identified in a genome-wide two-hybrid screen with HolB, one component of the clamp loader machinery (Noirot-Gros et al., 2002). Alignment with the three-dimensional structure of BsPol I revealed that the Pol ISID is exposed to the outside of the polymerase covering parts of the thumb and the vestigial proofreading domains (see Supplementary material, Fig. S2), and thus could mediate protein interaction.

Figure 4.

Pol I interacts with both Pol Y1 and Pol Y2.
Protein–protein interactions were detected using the yeast two-hybrid assay. The indicated proteins or protein fragments were expressed as baits (BD, Gal4 DNA binding domains fusion, in green) and/or as preys (AD, Gal4 activation domain fusion, in red). The diploid cells were tested for their ability to express the ADE2 interaction reporter. The Pol Y1 C- terminal fragments previously identified for their ability to interact with DnaN (Noirot-Gros et al., 2002) were used as positive controls for interaction.

Baits and prey fusions were combined in yeast diploid cells, and their ability to interact was tested as described in Experimental procedures. The control interactions between BD–DnaN and the AD–Pol Y1Cter were obtained as expected. However, no interaction between BD–DnaN and the full-length AD–Pol Y1 was detected, possibly due to an inaccessibility of the LF interacting domain. Remarkably, although the N- and C-terminal fragments of Pol I and the full-length Pol I expressed as preys failed to reveal an interaction, the AD–Pol ISID interacted with the BD–Pol Y1 and the BD–Pol Y1Nter fusion proteins. It also interacted with the BD–Pol Y2Nter fusion (Fig. 4). No interaction was detected between BD–DnaN and the AD–Pol Y2 full-size and LF subdomains. Hence, BsPol I established physical contact with the DNA polymerase domains of Pol Y1 and Pol Y2. These interactions correlated well with the finding that, genetically, BsPol I was required for the mutagenesis mediated by the two Y-polymerases.

Pol Y1 forms a ternary complex with BsPol I and DnaN

In the two-hybrid assay, Pol Y1 N- and C-terminal parts interacts with Pol I and with β (DnaN) respectively. This configuration of the interaction domains raises the possibility that Pol Y1, BsPol I and DnaN could form a ternary protein complex. To test this hypothesis, interactions between these three proteins were assayed using the yeast three-hybrid system. The capacity of various baits to interact with the BsPol ISID prey was tested in the presence of a third protein partner. Ternary interactions were revealed when the interaction phenotypes were expressed conditionally to the presence of a specific third protein (3HB), and binary interactions were detected independently of the presence of a third protein (Fig. 5A).

Figure 5.

Pol I, Pol Y1 and DnaN form a ternary protein complex.
A. Ternary protein interactions between Pol Y1 (full-size, N- and C-terminal domains), DnaN and the Pol ISID domain were assayed by a three-hybrid strategy in yeast. Protein or protein fragments were fused with Gal4 BD domain (green), or untagged with Gal4 sequences (3HB, red). Colonies containing the different BD/3HB combinations were mated with the strain expressing the AD–Pol ISID prey fusion and with a control strain expressing the Gal4 AD prey. The diploid cells were tested for their ability to express the ADE2 interaction reporter. The binary interactions, which are independent of the presence of the third 3HB partner, appeared as a row of colonies (lines 3 and 4). The ternary interactions are indicated by red arrows (2b and 2c). No expression of the ADE2 reporter was observed with the control prey (data not shown), indicating that interactions were specific of AD–Pol ISID. Models of the binary and ternary protein complexes revealed in this assay are also shown. (I) binary interaction between AD–Pol I (pink) and BD–Pol Y1 (green); Gal4 AD and BD domains are represented in purple (dark and light respectively); (II) ternary protein complex involving PolA, Pol Y1 and DnaN (β); (III) ternary protein complex between AD–Pol I and the BD–DnaN–DnaN dimer (grey and light blue respectively).
B. Illustration of the physical (black arrows) and functional (red arrows) interactions detected in this study between Pol Y1, Pol Y2, Pol I and DnaN.

The binary interactions obtained between BD–Pol Y1 (full-size and N-terminal fragments) and AD–Pol ISID (Fig. 5A, I) were expected from the observation in Fig. 4. Interestingly, BD–DnaN interacted with AD–Pol ISID provided that either 3HB–Pol Y1 or 3HB–DnaN were coexpressed in the cell. These findings revealed that two distinct ternary complexes were formed in yeast. The first complex, DnaN/Pol Y1/BsPol I, could be formed via the simultaneous binding of DnaN and BsPol I to the C- and N-terminal domains of Pol Y1 respectively (Fig. 5A, II). The second ternary complex could be formed between BsPol I and a DnaN homodimer (Fig. 5A, III). Indeed, the presence in the yeast cell of an unmodified form of the DnaN protein (3HB) together with the BD–DnaN and AD–Pol ISID fusions triggered an interaction phenotype (Fig. 5A, III). Because the DnaN–BsPol I interaction was not detected in a two-hybrid assay, this finding suggests that the binding of BsPol I to the β-clamp takes place with a DnaN dimer (Fig. 5A, III). Such a dimer likely does not form with the BD–DnaN fusion, possibly due to steric hindrance by the BD domain.


BsPol I assists the two B. subtilis Y-DNA polymerases Pol Y1 and Pol Y2 in TLS

In the Gram-positive bacterium B. subtilis, we established that the A-family polymerase BsPol I is required for mutagenesis mediated by the two Y-polymerases, thus revealing a major difference with the Gram-negative bacterium E. coli, which has been the paradigm for TLS studies in bacteria. We showed that the UV-induced mutagenesis depended entirely on the polymerase catalytic activities of both BsPol I and Pol Y2, suggesting that both polymerases are acting in TLS at UV lesions. This conclusion was supported by the finding that upon Pol Y2 overproduction, the vast majority (> 90%) of the observed UV mutagenesis still depended on the presence of BsPol I. However, a minor part of the UV mutagenesis (< 10%) occurs in a BsPol I-independent manner. Similarly, the spontaneous mutagenesis mediated by Pol Y2 overproduction does not required BsPol I. The Pol Y2 spontaneous mutagenesis and UV-induced mutagenesis can be distinguished upon their requirement for binding to the β-clamp, as mutation of key residues in the Pol Y2 β-binding motif does not affect UV mutagenesis but abolishes the spontaneous mutagenesis (Duigou et al., 2004). We found that the alteration of the β-binding consensus in Pol Y2 did not affect this Pol I-independent fraction of UV mutagenesis. Thus, this mutagenesis did not arise from Pol Y2 activity on undamaged DNA but rather resulted from Pol Y2 activity at some UV-induced DNA lesions. To account for these results, we propose that BsPol I assists Pol Y2 in TLS taking place at the majority of UV lesions. For a minority of UV-damaged sites, the Pol Y2 protein could either perform TLS alone or be assisted by a yet unknown DNA polymerase. Remarkably, BsPol I was also required for the spontaneous mutagenesis mediated by Pol Y1 overproduction. The catalytic activities of both enzymes were necessary suggesting that they act in a co-ordinated manner to fix mutations during replication on undamaged DNA template.

The significance of these findings in the field of TLS and mutagenesis could be evaluated relative to the E. coli paradigm where the study of the SOS response mechanism has guided DNA mutagenesis research. In E. coli, two Y-family DNA polymerases, Pol IV and Pol V, are involved in mutagenesis and lesion bypass. Pol V is able to copy through a UV-damaged DNA template in an error-prone manner, and thus is responsible for the majority of the UV-induced mutagenesis. Importantly, it was reported by numerous studies carried out both in vivo and in vitro that, in the presence of RecA, Pol V acts alone to insert bases across UV lesions and to extend the lesion termini during bypass (Tang et al., 2000; Pham et al., 2002; Fujii et al., 2004). Pol IV was found to be involved in untargeted and adaptive mutagenesis, but its biological role remains unclear. Importantly, the Pol IV-mediated mutagenesis was found to be independent of Pol I and Pol II (Wagner and Nohmi, 2000). The concerted action of two DNA polymerases in E. coli has been observed in the bypass of some bulky lesions in particular DNA sequence contexts (Wagner et al., 2002). The requirement of Pol I for Pol Y-mediated spontaneous and UV mutagenesis in B. subtilis is in contrast with the major one-enzyme process described in E. coli. In eukaryotes, TLS is often promoted by the concerted action of two polymerases, one acting as an ‘inserter’ and the other one as an ‘extender’. Here, we provide evidence that a multipolymerase mechanism involving both a Pol Y enzyme and the DNA polymerase Pol I is the major pathway for TLS in B. subtilis. Together, our results indicate that TLS in B. subtilis is mechanistically different from that in E. coli, and is likely conserved in many other Gram-positive bacteria.

Implications of the BsPol I catalytic properties for the TLS-assisted process

Bacillus subtilis Pol I belongs to the Gram-positive subfamilly of A-polymerases, which harbours a vestigial 3′−5′ exonuclease active site and thus is deficient for proofreading activity. This observation supports the proposed role of BsPol I in assisting the TLS reaction catalysed by the Y-DNA polymerases. Indeed, proofreading activity is expected to oppose translesion DNA synthesis by preventing the stable incorporation of a mispaired nucleotide opposite DNA lesions (Sagher et al., 1994; Khare and Eckert, 2002). In E. coli, Pol I, which is proficient for 3′−5′ proofreading activity, does not participate in spontaneous or in UV-targeted mutagenesis in vivo (Wagner and Nohmi, 2000). However, a successful bypass event requires the resumption of DNA synthesis by switching back to the replicative polymerase. It was recently established that E. coli Pol V synthesizes a TLS patch of a length sufficient to prevent the sensing of the distortion at the lesion site by the 3′−5′ exonuclease-proficient replicative polymerase, and thus favour extension, rather than proofreading degradation from the primer termini (Fujii and Fuchs, 2004). Here, the requirement of BsPol I in Pol Y1- and Pol Y2-mediated mutagenesis suggests that the Y-DNA polymerases might not generate a primer terminus past the lesion that can be directly extended by the replicative enzyme. In this model, the role of BsPol I could be to extend these termini sufficiently so that the replicative polymerase can elongate them. An alternative possibility is that BsPol I could carry out the translesion synthesis and generate the lesion termini that would be further extended by Pol Y1. Although this latter hypothesis can not be ruled out, it appears unlikely as the increase in mutagenesis relies on overproduction Pol Y1, but not on that of BsPol I (data not shown), suggesting that Pol Y1 acts first in introducing the mutation.

We found that the redundant 5′−3′ exonuclease activity between polA and ypcP was essential for cell survival. Furthermore, a DNA polymerase-deficient BsPol I derivative enabled the survival of a ypcP-disrupted strain, indicating that the polymerase activity of BsPol I is dispensable for cell survival. Pol I polymerase function was also found to be dispensable in S. pneumoniae (Diaz et al., 1992). However, the polymerase activity was required for the repair of MMS- and UV-induced DNA damages, and for the initiation of plasmid replication. Together, with its role in the maturation of Okazaki fragments (Tamanoi et al., 1977), Pol I is a multifunctional enzyme involved in several pathways of DNA replication and repair. Our finding that BsPol I is a key player in translesion synthesis represents a novel function for the A-family DNA polymerase. This role in TLS can likely be generalized to the Pol I of the Gram-positive bacteria, which are devoid of proofreading exonuclease activity.

Correlation of the physical and functional interactions between BsPol I, the Y-polymerases and the β-clamp

Using the yeast two- and three-hybrid assays, physical interactions between B. subtilis proteins acting in TLS were detected. An interaction between the BsPol I and DnaN was detected in a three-hybrid assay only in the presence of an untagged copy of DnaN. The formation of a heterodimer between the bait BD–DnaN and the untagged DnaN could create an interface between the N-terminal domain of one clamp subunit and the C-terminal domain of the other. This interface could be the site for BsPol I binding. Furthermore, the Pol ISID domain contains the DnaN-interacting signature (see Supplementary material). In E. coli, Pol I was found to interact in vitro with the β-clamp (Lopez de Saro and O’Donnell, 2001).

BsPol I also interacted specifically with the N-terminal half of Pol Y2. The formation of a potential Pol I–Pol Y2 complex in B. subtilis correlates well with the genetic requirement of Pol I in Pol Y2-mediated TLS at UV lesions. The presence of the conserved β-binding motif in the Pol Y2-LF domain, which is necessary for Pol Y2-mediated untargeted mutagenesis, also suggests that a physical interaction between DnaN and Pol Y2 takes place (Duigou et al., 2004). However, we failed to detect this interaction in the yeast two- and three-hybrid assays. Possibly, the binding of Pol Y2 to DnaN might require an additional factor.

In addition, BsPol I appears to interact with the Pol Y1 N-terminal domain. A three-hybrid assay revealed that DnaN, BsPol I and Pol Y1 could form a ternary complex. The notion that such a complex forms in B. subtilis is supported by the genetic requirement for BsPol I and the interaction with DnaN in Pol Y1-mediated mutagenesis. The strong correlation between functional and physical interactions (Fig. 5B) reflects the network of protein interactions within the TLS machinery, in which the different DNA polymerases would act in co-ordinated fashion at the lesion site.

The role of Pol I in mutagenesis mediated by the Y-polymerases in B. subtilis

Here it is shown that in B. subtilis, TLS is a bipartite dual polymerase process involving a Y-family polymerase and an A-family polymerase (Pol I). The propensity of Pol Y enzymes to accommodate distorted DNA templates within their active sites makes these polymerases the likely candidates to generate mutations. In addition, we observed that overproduction of Pol I did not increase mutagenesis, suggesting that Pol I is likely not responsible for the misincorporation. From these two arguments, we propose that the Pol Y enzyme acts first, generating a mismatched primer termini that is extended by Pol I. This hypothesis is supported by the genetic data and reflects the possible roles of the protein–protein interactions.

The role of Pol I in Pol Y1-mediated spontaneous mutagenesis upon overproduction

We propose that overproduction of Pol Y1 in the cell facilitates its accessibility to stalled replication forks. Indeed, under normal growth conditions, and in the absence of exogenous DNA-damaging treatment, the progression of the replisome can be impeded by the presence of an obstacle such as a protein–DNA complex, or by an endogenous DNA lesion, thus causing a partial disassembly of the replication machinery (Michel et al., 2004). The binding of Pol Y1 to the β-clamp could provide access to the DNA primer template termini. In the absence of DNA damage, replication by the error-prone Pol Y1 could lead to misincorporation. In the absence of intrinsic proofreading activity, the subsequent elongation of the mismatch primer termini by Pol Y1 would be impeded. Then, by interacting with Pol Y1, BsPol I could be recruited to the misaligned primer termini and extend it, resulting in the fixation of the mutation. Pol Y1 could also bind to the β-clamp at the replication fork arrested at a site of endogenous DNA damage. Pol Y1 could insert a nucleotide opposite to the damaged base but could not extend the lesion termini any further. Then, BsPol I could be brought to the terminus and extend it. Thus, the role of BsPol I in Pol Y1-mediated spontaneous mutagenesis would be to extent primer template termini resulting from Pol Y1 mutagenic activity. In this model, the two enzymes would act sequentially in a catalytic manner.

The role of Pol I in Pol Y2-mediated UV mutagenesis

Upon encounter of a UV lesion, the replicative polymerase would dissociate and be replaced by the specialized Pol Y2 polymerase. Pol Y2 could be recruited to the lesion by RecA (Duigou et al., 2004), as described for E. coli Pol V (Pham et al., 2002), and achieve translesion synthesis across the UV lesion. We propose that Pol Y2 could incorporate a nucleotide across the UV-induced photoproduct but could not extend the lesion terminus to the length adequate for elongation by the replicative polymerase. Then, BsPol I would be recruited to the lesion terminus by interacting with Pol Y2, and extend the lesion terminus.

In our model, the bypass of DNA lesion is mediated by the sequential action of a Y-polymerase and of a A-family polymerase. In such scheme, a productive bypass would require the co-ordination of BsPol I with Pol Y2 (or Pol Y1) activities to synthesize a DNA tract of a sufficient length after the lesion to allow the replicative polymerase to resume replication. Whether the proofreading-deficient BsPol I is mutagenic during this process is not yet known. However, we observed that in the absence of Pol Y2, UV-induced mutagenesis was abolished, suggesting that the Pol I action in the filling of the gaps generated by the UvrABC excinuclease during nucleotide excision repair (NER) is error-free, as described in E. coli. Furthermore, we observed that the overexpression of Pol I did not enhance UV mutagenesis, suggesting that it is not responsible for misincorporation.

The observation that a small proportion of UV mutagenesis occurs in a BsPol I-independent manner suggests that, when present in excess, Pol Y2 may have some ability to extend sufficiently past a distorting UV lesion. Interestingly, BsPol I is not involved in spontaneous mutagenesis mediated by Pol Y2 overproduction. This suggests that, unlike Pol Y1, Pol Y2 may be able to replicate past the mismatch resulting from erroneous incorporation.

It was previously reported that BsPol I interacted with the two replicative DNA polymerases PolC and DnaE, as well as with HolB, a subunit of the clamp loader complex (Noirot-Gros et al., 2002). All these interactions involved the same internal SID domain of BsPol I, comprising half of the vestigial 3′−5′ exonuclease and part of the thumb domains. These interactions suggest that Pol I could link several components of the replication machinery in B. subtilis. A recent study suggested that the mouse Rev1 protein may function as a scaffold or a docking site to mediate polymerase switching and/or to deliver the appropriate TLS enzyme at the primer template lesion terminus (Guo et al., 2003). Possibly, Pol I could also play a similar scaffolding role in B. subtilis.

Experimental procedures

Strains and plasmids

Strains and plasmid constructs and their relevant genotypes are described in detail in Table S1 in Supplementary material. Standard techniques were used for strain construction and for plasmid transformation (Sambrook et al., 1989; Harwood, 1990).

UV and MMS survival assays

Several independent overnight cultures (> 8) were grown exponentially (up to OD 0.2–0.5) at 37°C in LB medium supplemented with appropriate antibiotics. Cultures were rinsed in Spizizen minimal medium (MM10) and were subjected to UV exposure in Petri dishes with gentle stirring. Serial dilutions were then plated on LB. The surviving fraction was determined after 20 h of incubation at 37°C. MMS sensitivity was tested by plating serial dilutions of exponential growing cultures onto LB and LB supplemented with MMS (350 ng ml−1). The ability of the cells to form colonies was determined after 24 h at 37°C.

Rifampicin resistance mutation assay

The proportion of RifR cells was determined from at least eight independent cultures as described previously (Duigou et al., 2004). To measure UV mutagenesis, exponentially growing cells were subjected to UV irradiation at the indicated doses (40 J m−2 or 60 J m−2, for the polA+ or the ΔpolA strain backgrounds respectively). The irradiated cells were resuspended in rich medium, and incubated overnight. These overnight cultures were used for determination of the proportion of RifR cells. The relative fitness of B. subtilis RifR mutants cells relative to wild-type rifampicin-sensitive (RifS) cells was evaluated as described in Table S2 in Supplementary material. The data were analysed with a non-parametric LSmeans (Least-Squares means) statistical test (SAS/STAT 6.2). The proportion of mutants are significantly different when P < 0.01.

Yeast two-hybrid and three-hybrid methodology

The B. subtilis polA, yqjH and yqjW full-size and truncated open reading frames (ORFs), as well as DnaN ORF, were polymerase chain reaction (PCR) amplified from strain 168 genomic DNA, and fused in frame with (i) the Gal4 BD into the bait vector pGBDU-C1 and (ii) the Gal4 AD into the prey vector pGAD-C1, between restriction sites SmaI and SalI (see Fig. 4 for the domains delineation). The resulting constructs were isolated from E. coli and transformed into yeast PJ69-4α (for the pGBDU derivatives) and PJ69-4α (for the pGAD derivatives) haploid strains, using URA3 and LEU2 as selective markers respectively (James et al., 1996). The pGAD-PolASID prey plasmid was isolated from a previous screening for HolB interacting partners (Noirot-Gros et al., 2002). The two-hybrid assays were performed according to a previously described mating strategy (Noirot-Gros et al., 2002). Positive protein–protein interaction between the bait and a prey was detected by the ability of the cells to grow onto plates of synthetic complete medium depleted for leucine, uracil and histidine (SC-LUH) and for leucine, uracil and adenine (SC-LUA).

In addition, the different yqjH fragments were transferred into the three-hydrid vector p3HB(Trp1) between EcoRI and SalI restriction site (Noirot-Gros et al., 2002). The constructions were isolated from E. coli and co-transformed into a PJ69-4α yeast strain carrying a resident pGBDU fusion vector, using both URA3 and TRP1 as selection markers. The resulting URA+ TRP+ strains were then mated with PJ69-4α strains containing the pGAD-PolASID and the empty pGAD-C1 vector, and diploids were selected on synthetic complete media depleted for leucine, uracil and tryptophan (SC-LUW). Interaction phenotypes were tested by replica-plating the diploids onto selective plates SC-LUWH and SC-LUWA, as described previously (Dervyn et al., 2004).

Immunodetection assays

The N-terminally FLAG-tagged proteins were detected in cell extracts using an anti-FLAG M2 antibody (♯F 3165 from Sigma-Aldrich) as previously described (Duigou et al., 2004).


We are grateful to B. Michel, M.-A. Petit and P. Lewis for stimulating discussions and critical reading of this manuscript, and to Rut Carballido-Lopez, E. Lechatelier and P. Polard for helpful comments. We are also indebted to P. Bellenand for performing extensive statistical analysis of our data.