Interestingly, it was recently found that modification of chromosomal DNA is not essential for discriminating self against foreign DNA (Makovets et al. 1999). Under a variety of conditions, unmethylated target sequences can be generated in the chromosomal DNA and, surprisingly, such DNA is not recognized as foreign DNA. The reason for this is an additional protection system in which the protease ClpXP prevents the cleavage of unmethylated chromosomal DNA. Thus, in the absence of ClpXP protease, the unmethylated bacterial chromosome is attacked by EcoKI and restricted. This protection by ClpXP protease is preferentially biased toward the bacterial chromosome and not toward incoming unmethylated phages (Doronina and Murray 2001).

The EcoKI enzyme, a member of the type IA R–M system, is encoded by three genes: hsdR, hsdM, and hsdS. It consists of two HsdR subunits, two HsdM subunits, and one HsdS subunit (R₂M₂S₁). The holoenzyme modifies hemi-methylated DNA via its methyltransferase activity, restricts unmethylated DNA, and leaves methylated DNA untouched. A smaller complex (M₂S₁) also exists, but it has only the methyltransferase activity. The HsdS subunit is responsible for target-site recognition, which is why both complexes and both activities respond to the same nucleotide sequence (Murray 2000). Restriction begins when EcoKI binds to an unmodified target sequence in the presence of S-adenosylmethionine and ATP. The target sequence is bipartite and asymmetric. In an ATP-dependent process, the enzyme moves the surrounding DNA toward itself while remaining bound to the target site. DNA cutting occurs far from the recognition site and is triggered when two EcoKI complexes collide (Studier and Bandyopadhyay 1988).

The protease ClpXP acts on the EcoKI complex which has started restriction and it degrades the HsdR subunit. Degradation prevents further translocation and endonuclease activity of the EcoKI enzyme (Makovets et al. 1999; Doronina and Murray 2001). Specific degradation of the HsdR subunit also occurs when Escherichia coli acquires a new type I R–M system (Makovets et al. 1998). This means that the presence of unmodified target sequences on the chromosome makes the cell phenotypically restriction deficient, an effect known as restriction alleviation (RA).

Cells treated with agents that damage DNA, such as UV light, nalidixic acid, or 2-aminopurine, have all been shown to induce restriction alleviation (Bertani and Weigle 1953; Day 1977; Thoms and Wackernagel 1982, 1984; Hiom and Sedgwick 1992; Kelleher and Raleigh 1994). In addition, it was shown that some mutants like dam, topA, mutD, rnhA, and recG have permanently (constitutively) alleviated EcoKI restriction (Efimova et al. 1988; Kelleher and Raleigh 1994; Makovets et al. 1999; Blakely and Murray 2006). Since restriction alleviation was shown to be dependent on the ClpXP protease in all these cases (Makovets et al. 1999; Blakely and Murray 2006), it was proposed that DNA damage leads to generation of unmodified target sequences by different mechanisms. One is recombination dependent and occurs when the progress of DNA replication is blocked due to lesions or breaks in the DNA template. Regions of unmethylated DNA could be generated by recombining two hemi-methylated DNA strands (Murray 2000). Alternatively, due to the mutagenic activity of 2-aminopurine, new unmodified target sequences could be occasionally created (Makovets et al. 1999). It has been recently proposed that initiation of a new replication fork at R-loops could also generate unmodified DNA (Blakely and Murray 2006).

Restriction alleviation was first observed and genetically best characterized in E. coli K-12 cells irradiated with UV light (Bertani and Weigle 1953; Day 1977). UV-induced RA is specific for the type IA restriction system (EcoKI) and does not affect type II (EcoRI) or type III (EcoPI) R–M systems (Thoms and Wackernagel 1982). As mentioned above, UV-induced RA is clpXP dependent (Makovets et al. 1999) but it also requires functional recA, recBC, and recF genes (Day 1977; Thoms and Wackernagel 1982, 1984), de novo protein synthesis, and sufficient time for expression (Thoms and Wackernagel 1982). Since UV-induced RA could not be induced in lexA (lex-1) and recA mutants (Day 1977), it was proposed that alleviation of restriction due to UV irradiation is another of the SOS functions. This hypothesis was dismissed later when it was shown that genetic control of RA differs from the regulation of SOS functions (Thoms and Wackernagel 1984). The SOS response is required for restriction alleviation after UV irradiation, but strains that have constitutive SOS induction such as recA730 or lexA55 do not alleviate restriction constitutively and express RA only after UV irradiation (Thoms and Wackernagel 1984; Hiom and Sedgwick 1992). Thus, DNA damage in the host DNA is critical for UV-induced RA. RA also depends on the umuCD genes, which are involved in mutagenic DNA repair (Hiom et al. 1991; Hiom and Sedgwick 1992). The necessity for DNA damage, recombination genes, and SOS response for UV-induced RA supports the proposed idea that unmethylated DNA can be generated by DNA repair via homologous recombination or by creating new unmethylated target sequences through increased frequency of mutations (Makovets et al. 1999; Murray 2000).

From the data above it can be concluded that the restriction alleviation of unmodified λ-phage in UV-treated cells is due to inactivation of the restriction activity of the EcoKI enzyme if double-strand unmethylated target DNA is generated in the host chromosome. Therefore, we propose that restriction alleviation can be looked upon as an indirect measure of recombination and/or replication that generates unmethylated double-strand DNA (dsDNA). In other words, mutations in genes that block essential steps of DNA repair that are necessary for generation of unmethylated dsDNA would not alleviate restriction. The purpose of this study was to examine the effects of various rec mutations on the survival of unmodified λ-phage to elucidate the molecular mechanism of restriction alleviation after UV irradiation. Our data are consistent with a recent study, which demonstrated that RA is induced as a response to unmethylated target sequences produced by homologous recombination, R-loop formation, and DNA synthesis (Blakely and Murray 2006). Our results strongly suggest that recombination intermediates (D-loops), generated during the repair of dsDNA breaks (DSBs), and replication recovery after UV irradiation are critical for UV-induced RA.

Bacterial strains and bacteriophages: The strain E. coli C600, which lacks the EcoKI restriction–modification system, was used for preparation of unmodified phage stock λvirC (designated as λvir.0) (Salaj-Šmic et al. 1997). Phages λvirC, grown on the E. coli AB1157 () strain, are modified and designated λvir.K. The plating efficiency of unmodified phage λ was used to examine the activity of the EcoKI restriction system of the various bacterial strains presented in Table 1. λ-Phage stocks were prepared by the standard plate method. Mutant bacterial strains were made by P1vir transduction and selected for the appropriate antibiotic resistance (Miller 1992).

TABLE 1 Strains used in this study

Media and growth conditions: High-salt Luria broth (LB) medium (10 g bacto-tryptone, 5 g yeast extract, 10 g NaCl, and water was added to 1000 ml) was used for growth of cells and the plating of phage λ (Miller 1992). Solid media for plates was supplemented with 16 g of agar, or 8 g/liter for soft agar. All experiments were done with exponentially growing cells (~1–2 × 10⁸ bacteria/ml, OD₆₅₀ = 0.4) at 37° in LB medium.

UV irradiation of cells: Ten milliliters of midlog phase cells were centrifuged (6000 × g, 6 min, 4°) and resuspended in the same volume of λ-buffer (10 mm Tris, 10 mm MgSO₄, pH 7.2). A 1-ml aliquot was taken to measure the zero-time value of RA (unirradiated sample) and the remaining culture was gently stirred during irradiation at room temperature by a 30-W Philips low-pressure Hg germicidal lamp at a distance of 1 m with the incident dose of 2.5 (J/m²/sec). The thickness of the irradiated layer of suspension was <0.7 mm. After irradiation, cells were pelleted, resuspended in the same volume of fresh LB medium, and incubated with aeration for the duration of the time course (3 hr). A 0.1-ml sample was taken before and after UV irradiation, diluted, and plated immediately for cell survival (Miller 1992). Each strain was irradiated with a UV dose of 150 J/m². UV dose was determined with a VLX-3W UV dosimeter (Bioblock, Illkirch, France).

Restriction alleviation assay: UV-induced RA was measured as described previously (Čogelja-Čajo et al. 2001) with some modifications. During post-irradiation incubation, at appropriate times, 1-ml aliquots were taken, centrifuged, and resuspended in 0.2 ml of LB medium supplemented with 50 mm MgSO₄. Bacteria were infected with an unmodified virulent mutant of phage λvir.0. The multiplicity of infection was <0.1 and adsorption of phages on bacteria was ~99.7%. After adsorption, 15 min at 37°, infected bacteria were plated for infective centers on an untreated overnight indicator strain AB1157. RA was expressed as the efficiency of plating unmodified phage λ onto UV-irradiated bacteria relative to that onto unirradiated cells. RA₀ value represents the initial plating efficiency of unmodified phage λ on unirradiated cells relative to phage titer on strain C600 at time 0, while RA_max value represents the maximal RA on UV-irradiated cells relative to phage titer strain C600 (usually reached after 2 hr of post-irradiation incubation).

UV-induced RA is dependent on the UvrA protein and primosome activity of PriA helicase: To examine the role of the rec gene products in UV-induced RA, we first determined experimental conditions for RA in wild-type cells. A simple test to measure restriction alleviation is based on measuring the plating efficiency of unmodified λ-phage on E. coli cells. The phage genome is a substrate for EcoKI cleavage, and unmodified λ-phage show increased survival if plated on cells previously treated with DNA-damaging agents such as UV light. A typical curve showing a temporary increase of plating efficiency of unmodified λ-phage in UV-irradiated wild-type cells (i.e., restriction alleviation) during 3 hr of post-irradiation incubation is shown in Figure 1. The optimal restriction alleviation under our conditions was observed at a UV dose of 150 J/m² for the wild-type strain, which is a stronger UV dose than used by others (Thoms and Wackernagel 1984; Kelleher and Raleigh 1994). We applied this UV dose in all experiments, except where mentioned, to expose each mutant to the same DNA damage; i.e., the same UV dose ensures the same number of pyrimidine dimer formation in excision-proficient cells. In addition to λvir.0, RA was determined in highly UV sensitive strains with modified phage λvir.K. The plating efficiency of modified phage was always ~1 (data not shown), implying that the results obtained with unmodified phage λvir.0 did not require correction for eventually reduced cell capacity for phage propagation. The maximal value of RA for wild-type cells was achieved after 120 min of post-UV incubation at 37°, which is similar to previous studies (Thoms and Wackernagel 1984; Čogelja-Čajo et al. 2001). For clarity, all results obtained in this study are summarized in Table 2. Table 2 contains data on the efficiency of plating λvir.0 on uninduced cells relative to the phage titer on strain C600 at time 0 (RA₀); the efficiency of plating λvir.0 on UV-induced cells, giving maximum RA relative to the phage titer on strain C600 (RA_max); the ratio of UV-induced RA_max relative to that on uninduced RA₀ (RA); and cell survival for all strains tested. As shown in Figure 1 and Table 2 for wild-type cells, RA₀ = 0.00015, RA_max = 0.14, and RA = ~1000. According to results obtained for wild-type and mutants used in this study, we considered that restriction alleviation following UV irradiation is induced if the RA_max value was ~>0.01 or RA was ~10. Taking into account the extent of standard deviation, we consider that the basal level of restriction alleviation is increased if the RA₀ value is at least four times higher compared to wild-type RA₀ value, but never reaches 0.01. The mutants that showed UV-induced RA_max < 0.01, and therefore a smaller RA value (~5–10), were said to have partially UV-induced RA (Table 2).

Figure 1.— UV-induced restriction alleviation is abolished in clpP and uvrA mutants during 3 hr of post-irradiation incubation. Strains wild type (wt;♦), clpP (), hsdR clpP () and uvrA (•) were irradiated with a UV dose of 150 J/m (more ...)

TABLE 2 Induction of restriction alleviation in E. coli mutants

In addition to wild-type cells, we wanted to confirm the earlier observation that RA in UV-irradiated cells is protease ClpXP dependent and to examine whether UV-induced RA is dependent on presence of the HsdR subunit of the EcoKI restriction enzyme. Namely, it is known that RA induction is dependent on the ClpXP protease, which is necessary for the degradation of the HsdR subunit and subsequent restriction-deficient phenotype (Makovets et al. 1999). Until now, it has not been shown that the HsdR subunit of EcoKI is involved in UV-induced RA. Our results are presented in Figure 1. As expected, cells mutated in clpXP genes did not alleviate restriction; i.e., maximal restriction was observed. Figure 1 shows low plating efficiency of unmodified λ in UV-irradiated clpP or clpX strains through 3 hr of post-irradiation incubation (RA_max ~ 0.00004; and data not shown for clpX mutant). This result confirms the requirement for ClpXP protease in UV-induced RA. In contrast, the maximal plating efficiency of unmodified λvir.0 or maximal RA was observed in a hsdR clpP mutant where the HsdR subunit is missing (RA_max ~ 1; Figure 1; Table 2), indicating its involvement in UV-induced RA.

It is known that the RecBCD enzyme is involved in the UV-induced RA phenomenon (Thoms and Wackernagel 1984), implying that unmethylated target sequences could be produced by recombination. Since the RecBCD enzyme must bind to free dsDNA ends that are blunt or nearly blunt to initiate recombination, we wanted to test whether UV-induced RA depends on DSBs produced during excision repair of pyrimidine dimers. DSBs can be introduced by excision repair of two closely spaced photoproducts on opposite DNA strands (Bonura and Smith 1975; Sedgwick 1975). Alternatively, the encounter of a replication fork with nicks created by excision repair in the DNA template can also result in blunt-end DSBs (Hanawalt 1966).

To test this possibility, we measured the survival of unmodified λ-phage in a uvrA mutant. The UvrA protein loads UvrB protein onto a damaged DNA site after which UvrC binds to UvrB, resulting in a UvrBC-incision complex. Therefore, mutation in the uvrA gene blocks the incision step of the excision repair of UV-induced lesions and other DNA damage (Van Houten 1990). In contrast to a previous result where a modest effect was observed (RA ~ 60) (Thoms and Wackernagel 1982), UV-induced RA was completely abolished in a uvrA mutant (RA_max = 0.00016; RA ~ 1). This result suggests that generation of unmethylated dsDNA is dependent on the excision repair protein UvrA (Figure 1; Table 2). The reason for this difference could be due to either different uvrA alleles or different UV doses used. The same experiment was repeated by using more appropriate UV doses for excision-repair-deficient mutants: 3, 10, 30, and 60 J/m². However, with each UV dose we observed the same effect: no UV-induced restriction alleviation (data not shown). Since we used a uvrA mutant inactivated with a Tn10 insertion, there is the possibility that uvrA1 is not a null mutant in all enzymatic activities of the UvrA protein (see discussion).

The introduction of DSBs during UV-induced RA could be the result of both incision and replication of DNA. Broken replication forks (indirectly induced DSBs) are known to be repaired by RecBCD and RecA via a recombination intermediate D-loop, onto which the PriA helicase can bind and load the DnaB helicase and the remainder of the replisome (Sandler and Marians 2000). The PriA protein possesses three biochemical activities: 3′ → 5′ DNA helicase, ATPase, and the specific activity for primosome assembly. If DSBs are indeed processed via formation of D-loops, then PriA would be required for UV-induced RA.

To test whether priA is needed for UV-induced RA, we used a sfiA11 priA2 double mutant because the sfiA mutation reduces the filamentous phenotype of a priA2 null mutant (Nurse et al. 1999). As expected, UV-induced RA was very low in a sfiA priA2 double mutant (RA_max = 0.00019; RA ~ 1), compared to the sfiA control (RA > 1000) (Figure 2; Table 2). We next tested which activity of the PriA is required for UV-induced RA. We repeated the same experiment using the priA300 mutant that lacks ATPase and helicase activities but is capable of catalyzing primosome assembly in vitro, and the priA2 dnaC810 double mutant. The latter mutant has a suppressor mutation in dnaC810, which enables DnaB loading directly onto replication forks without PriA (Liu et al. 1999). UV-induced RA was restored in the priA300 mutant but to a reduced level (RA_max = 0.0025; RA = 14; Figure 2; Table 2). A similar result was obtained with the priA2 dnaC810 mutant (RA_max = 0.0035; RA = 8). These results are consistent with our idea that PriA is required for UV-induced RA, particularly for its primosome assembly activity. The helicase activity of PriA is also needed for maximal effect on UV-induced RA induction, while DnaC was not as efficient for substituting PriA in DnaB-loading activity. Perhaps inappropriate loading of the DnaB helicase on a D-loop (Xu and Marians 2003) is the reason for the less efficient RA induction observed when the suppressor mutation dnaC810 is present.

Figure 2.— UV-induced restriction alleviation is dependent on PriA protein. Strains wt (♦), sfiA11 (), sfiA11 priA2 (), priA300 (□), and priA2 dnaC810 () were irradiated with a UV dose of 150 J/m2 and incubated for 3 hr. (more ...)

UV-induced RA in recF pathway mutants: The replication fork might encounter an unrepaired pyrimidine dimer, leaving the lesion in a single-strand gap (SSG) at the stalled fork. After some period of time, DNA synthesis resumes and the bacteria continues to divide. Resumption of replication following UV irradiation is dependent upon the RecFOR pathway (Courcelle et al. 1997, 1999). In wild-type cells, the RecFOR complex, the product of recF, recR, and recO genes, is involved in replacement of the single-stranded DNA binding (SSB) protein coating single-strand DNA (ssDNA) with the RecA protein and in stabilization of the RecA filaments at an arrested replication fork (Courcelle et al. 1999, 2003; Courcelle and Hanawalt 2001; Morimatsu and Kowalczykowski 2003; Chow and Courcelle 2004). Other proteins in the RecFOR pathway (RecQ helicase and RecJ nuclease) are shown to degrade nascent lagging strands, producing more ssDNA from the original SSG (Courcelle and Hanawalt 1999; Courcelle et al. 2003). The RecQ protein, a 3′ → 5′ helicase, and RecJ, a 5′ → 3′ exonuclease, act together to process a dsDNA break or gap to generate the 3′ single strand (Harmon and Kowalczykowski 1998). The biochemical activity of another protein of the RecF pathway, RecN, is unknown but it is probably required for RecBCD-dependent repair of dsDNA breaks (Lloyd et al. 1983; Wang and Smith 1986).

Since the RecFOR complex is involved in replication and recombination processes, we wanted to confirm its role in UV-induced RA. A previous study has shown that UV-induced RA is absent in a recF mutant after 90 min of post-irradiation incubation (Thoms and Wackernagel 1984; Figure 3A). Since the induction of the SOS response is delayed in recF/O/R mutants (Thoms and Wackernagel 1987; Hegde et al. 1995; Whitby and Lloyd 1995), we asked whether a longer post-irradiation incubation of 4 hr would induce RA in these mutants. Figure 3A and Table 2 show that the maximal expression of RA was achieved after 4 hr instead of the usual 2 hr of post-irradiation incubation, but the effect was rather small: recF (RA_max = 0.0018; RA ~ 8), recR (RA_max = 0.0011; RA = 4), and recO (RA_max = 0.00137; RA = 8). This result shows that reduced UV-induced RA can be expressed in recF/O/R mutants, but it is delayed in a manner similar to the SOS response. It can be concluded that UV-induced RA is partially dependent on RecFOR proteins.

Figure 3.— UV-induced restriction alleviation in recFOR pathway mutants. (A) Strains wt (♦), recO (□), recF (), and recR (Δ) were irradiated with a UV dose of 150 J/m2 and incubated for 6 hr. (B) Strains wt (♦), recJ (Δ), (more ...)

We next characterized UV-induced RA in other recF pathway genes: recJ, recQ, and recN. As shown in Figure 3B and Table 2, UV-induced restriction alleviation was normally expressed in these mutants and RA_max was >0.1.

RecA-loading activity of the RecBCD enzyme is required for UV-induced RA: Since a functional RecBCD enzyme is required for UV-induced RA, it was of interest to test which of the RecBCD enzyme activities are required for RA induction. RecBCD enzyme is a heterotrimer composed of RecB, RecC, and RecD subunits. It plays a central role in the major pathway of recombination and DNA repair of double-strand breaks in E. coli. It is a multifunctional enzyme regulated by the octamer sequence Chi, which stimulates recombination (for a review see Kowalczykowski et al. 1994). Three biochemical activities of the RecBCD enzyme are essential for recombination: 5′ → 3′ exonuclease, helicase, and RecA-loading activity (Amundsen and Smith 2003; Ivančić-Baće et al. 2003). In a previous study, we have shown that RA is normally induced in a recD mutant and have concluded that the helicase activity of the RecBCD enzyme is required for UV-induced RA (Čogelja-Čajo et al. 2001). To test specifically for the requirement of RecA-loading activity for UV-induced RA, we used the recB1080 mutant (Anderson et al. 1999; Jockovich and Myers 2001). This mutant has a point mutation in the C-terminal portion of the RecB subunit. Consequently, the RecB1080CD form of the enzyme is nuclease deficient and is unable to load RecA protein onto ssDNA, while it retains functional helicase activity (Yu et al. 1998a,b; Anderson et al. 1999; Wang et al. 2000). In this mutant, it is possible to distinguish RecA-loading activity from the helicase activity of RecBCD enzyme.

The recB1080 single mutant showed a reduced level of UV-induced RA (RA_max < 0.01; RA ~ 5), indicating that UV RA is partially induced and that only fully functional RecBCD is required for efficient RA. When both mechanisms of RecA loading were abolished (Ivančić-Baće et al. 2003), as was the case in the recB1080 recO double mutant, there was almost no UV-induced RA (RA_max = 0.0002; RA ~ 1; Figure 4A). As expected, the RecA-loading activity is required for UV-induced RA. To further confirm that RecA-loading activity is required for UV-induced RA, we measured the plating efficiency of unmodified λ-phage in a recB1080 recD double mutant. It is known that RecB1080C(D⁻), an enzyme produced by recB1080 recD cells, possesses RecA-loading activity due to inactivation of the RecD subunit, an inhibitor of RecA loading (Amundsen et al. 2000). As a control, the results for a recD single mutant are included in Figure 4A. RA following UV irradiation was induced in both the recD (RA_max = 0.28; RA > 1000) and the recB1080 recD mutant (RA_max = 0.011; RA ~ 60) (Figure 4A). Taken together, these results indicate that RecA-loading activity by the RecBCD enzyme is required for UV-induced RA.

Figure 4.— UV-induced restriction alleviation is abolished when RecA-loading activity is inactivated (A). Strains (A) wt (♦), recB1080 (), recB1080 recO (□), recB1080 recD (), and recD () were irradiated with a UV dose (more ...)

Constitutive RA: In our previous experiments, it was shown that RA can be induced in response to DNA damage by UV light. However, better survival of unmodified phage λvir.0 at the zero time of the experiment (increased RA₀ value without UV irradiation) can be seen in some mutants. This so-called constitutive RA was reported for type I EcoKI, but not for a type II system, and may occur for type III systems (Efimova et al. 1988). RA was also observed in dam, topA, mutD, rnhA, and recG mutants (Figure 2; Table 2; Efimova et al. 1988; Kelleher and Raleigh 1994; Makovets et al. 1999; Blakely and Murray 2006). These mutants accumulate DSBs (dam and topA) (Wang and Smith 1986; Kouzminova et al. 2004), have increased frequency of mismatches (mutD) (Echols et al. 1983), or enable R-loops to persist and initiate DNA replication (rnhA and recG) (Kogoma 1997). Constitutive RA is suppressed in all these mutants by the clpXP mutation (Makovets et al. 1999; Blakely and Murray 2006). These mutants also share other common features, such as (i) constitutively induced SOS response (Peterson et al. 1985; Lloyd and Buckman 1991; Kogoma et al. 1993; Slater et al. 1994; Kouzminova et al. 2004), (ii) unscheduled initiation of chromosomal replication (Kogoma 1997; Kouzminova et al. 2004), and (iii) recombination proficiency (i.e., they have functional RecBCD and RecFOR pathways of recombination). The reason for constitutive RA in these mutants is most likely due to modulated initiation of replication on either R-loops (rnhA and recG mutants) or oriC (dam mutant) (Bakker and Smith 1989; Kogoma 1997; Blakely and Murray 2006) than to accumulation of DSBs. This conclusion is supported by the observation that a dam mutH double mutant that does not accumulate DSBs still exhibits constitutive RA (Efimova et al. 1988; Kelleher and Raleigh 1994).

In this study, we found a new class of mutants that show constitutive RA activity; i.e., the basal level of restriction alleviation (RA₀) is at least four times greater compared to the wild-type RA₀ value. As shown in Figure 4A and Table 2, constitutive RA was observed in a recB1080 mutant (RA₀ = 0.00065), but not in a recD mutant, which is also nuclease deficient. It is known that recombination in recD mutants, which lacks the nuclease activity in RecBCD, is partially dependent on the RecJ nuclease (Lloyd et al. 1988; Lovett et al. 1988; Lloyd and Buckman 1991; Ivančić-Baće et al. 2005), so we were interested in testing whether the lack of 5′ → 3′ nuclease activity of this enzyme would enhance constitutive RA in a recD background. The results of constitutive and UV-induced RA in the single recD or the double-mutant recD recJ are presented in Figure 4, A and B. In agreement with our expectations, a recD recJ mutant exhibited a higher RA₀ value (0.0017) compared to wild type (0.00015) and recD (0.00025). Due to constitutive RA expression, recD recJ showed modest UV-induced RA (RA = 8), although its RA_max is relatively high (0.014). A recD single mutant had a lower RA₀ value probably due to preservation of residual 5′ → 3′ exonuclease activity of the RecBCD enzyme and the presence of a functional RecJ nuclease. As expected, the single recJ mutant did not show constitutive RA due to functional nuclease activity of the RecBCD enzyme (Figure 3B). However, the nuclease-deficient recB1080 recJ double mutant also did not show constitutive restriction alleviation (data not shown). The reason for this difference is probably due to lack of a constitutive SOS response in the recB1080 recJ double mutant (Ivančić-Baće at al. 2006).

To complete the genetic requirements for constitutive and induced restriction alleviation after UV irradiation, we have also done experiments with the mutants recG, ruvABC recG, ruvABC, and dam (Table 2; Figure 4B). The RuvAB and RecG proteins are helicases that catalyze branch migration, and the RuvC protein is a nuclease that resolves Holliday junctions in the late stages of homologous recombination (West 1996). RecG is also a junction-specific RNA/DNA helicase, which dissociates the R-loop and catalyzes branch migration of the Holliday junction in the reverse direction (Whitby et al. 1993; Hong et al. 1995; Vincent et al. 1996). Dam methyltransferase (DamMT), encoded by the dam gene, is an enzyme that methylates GATC sequences in E. coli DNA (for a review see Løbner-Olesen et al. 2005). As expected and in agreement with data from the literature (Efimova et al. 1988; Kelleher and Raleigh 1994; Blakely and Murray 2006; Table 2), constitutive RA was observed in recG (0.0047), ruvABC recG (0.0029), and dam (0.0048) mutants, and not in a ruvABC mutant (0.00014). Restriction alleviation was induced in all these mutants after UV irradiation (Figure 4B), since RA_max value was > 0.01 (Table 2).

A current interpretation for RA induction is that DNA damage induces loss of restriction activity due to generation of unmethylated target sequences within the bacterial chromosome (Makovets et al. 1999). It was proposed that unmethylated target sequences can be generated from replication associated with homologous recombination following DNA damage (Makovets et al. 1999; Murray 2000; Blakely and Murray 2006). In agreement, we provide evidence that a similar model can be applied to bacteria damaged by UV irradiation. In addition to a requirement for ClpXP, RecA, RecBCD, and induction of the SOS response, we show that UV-induced RA is also dependent on the excision repair protein UvrA, the RecA-loading activity of the RecBCD enzyme, the primosome activity of PriA, and is partially dependent on RecFOR proteins. We also show that RA is not dependent on recN, recJ, recQ, recG, and ruvABC.

The RecBC(D) enzyme was shown to be essential at an early stage during signal generation for alleviation of restriction, but it was not clear if the enzyme was also directly involved in the RA process due to experimental limitations (Thoms and Wackernagel 1984). It was shown that restriction alleviation after UV irradiation was blocked in a recB mutant, and also in an extragenic suppressor strain of the recBC defect (recB21C22 sbcB12 strain), which is recombination proficient, or in the double-mutant lexA55 recB, which has a constitutive SOS response. On the basis of these results it was concluded that the RecBCD enzyme is essential for EcoKI restriction alleviation (Thoms and Wackernagel 1984). However, RecBCD also has a destructive role because it degrades λDNA fragments to acid-soluble products after EcoKI cleavage (Simmon and Lederberg 1972). It also preserves the integrity of the bacterial chromosome after the cleavage of host DNA by a type I R–M system in the absence of efficient RA (Makovets et al. 2004). Similarly, it was proposed that RecBCD defends the host chromosome against restriction by the type II R–M system (Handa et al. 2000).

Strains that accumulate DSBs such as dam, rnhA, and topA express constitutive SOS induction and constitutive restriction alleviation. However, mutations that prevent repair-mediated breaks (mutH) or constitutive SOS induction (recA430) in a dam mutant do not suppress constitutive RA (Efimova et al. 1988; Kelleher and Raleigh 1994). On the other hand, a recG mutant does not accumulate DSBs but does express constitutive SOS induction and constitutive RA by EcoKI (Lloyd and Buckman 1991; McCool et al. 2004; Blakely and Murray 2006). Another interesting example is a priA mutant that has constitutive SOS induction but does not induce constitutive RA (Table 2; Figure 2). Therefore, in agreement with previous studies (Thoms and Wackernagel 1984; Hiom and Sedgwick 1992), we conclude that SOS induction itself is required to alleviate restriction but is not sufficient to induce RA. SOS induction possibly indicates that DNA replication is blocked, i.e., that single-strand gaps are generated at stalled replication forks.

It is known that DNA damage is also essential for restriction alleviation. There are at least two major types of DNA damage. If the fork encounters a DNA single-strand nick or gap, the replication fork will collapse, creating a DSB. If an unrepaired lesion is encountered, the lesion is left in a DNA gap at the stalled fork. Numerous studies have demonstrated that replication restart can proceed in a number of ways that are completely dependent upon several genes, mostly associated with homologous recombination (for a review see Kuzminov 1999). When a cell is subjected to high doses of UV light or a chemical mutagen, DNA lesions transiently block replication, which causes an induction of the SOS response (Setlow et al. 1963; Sassanfar and Roberts 1990). At a UV dose of 50 J/m², replication restart (recovery) is seen 30 min after the DNA damage is introduced (Courcelle et al. 2003). The recovery of replication requires RecA, RecF, RecO, RecR, UvrA, and PriA proteins (Masai et al. 1994; Courcelle et al. 1997, 1999; Rangarajan et al. 2002). Interestingly, none of the mutants in these genes is able to alleviate restriction following UV irradiation. On the other hand, RecBCD is not required for replication recovery but is essential for restriction alleviation as mentioned above, probably due to its major role in DSB repair. On the basis of these observations, we summarized all the data and listed the requirements for constitutive RA and restriction alleviation following UV irradiation in various mutants in Table 3. Table 3 shows that both replication recovery and functional DSB repair are required for UV-induced RA. In this regard, we also note that constitutive RA requires constitutive SOS induction in addition to functional replication recovery and DSB repair. The constitutive SOS response probably reflects abundant endogenous lesions in these mutants.

TABLE 3 Requirements for constitutive and UV-induced RA in E. coli

A uvrA mutant is not able to excise and remove UV-induced lesions, so it fails to recover replication although it shows elevated levels of strand exchange (Courcelle et al. 2003). This result indicates that one of the reasons for the absence of restriction alleviation in a uvrA mutant is the lack of replication recovery after UV irradiation. This could explain the different results obtained by us and a previous study regarding the uvrA effect (Thoms and Wackernagel 1982). Recovery of DNA synthesis was reported to occur in a uvrA6 mutant that is excision repair deficient (Kogoma 1997 and references therein), but not in a uvrATn10 mutant (Courcelle et al. 1999). In a previous study (Thoms and Wackernagel 1982), the uvrA1 mutant was used (i.e., uvrA6 mutant), whereas in our research we used the uvrATn10 mutant. However, it should be stressed that a second role of excision repair in UV-induced RA is in the appearance of dsDNA ends, which are the result of replication fork collapse.

The RecFOR complex is known to facilitate the loading of RecA protein onto SSB-coated gaps (Morimatsu and Kowalczykowski 2003). It was shown that replication fails to recover in mutants lacking recF, recO, or recR gene products (Courcelle et al. 1999; Rangarajan et al. 2002). In these mutants, the DNA lesions are removed by excision repair but nascent strands of the disrupted replication fork are not protected and are degraded by the action of RecQ helicase and RecJ nuclease (Courcelle et al. 1997; Courcelle and Hanawalt 1999). We observed small and delayed induction levels of UV-induced RA in recF/O/R mutants, which are probably dependent on the RecBCD function, similarly to SOS induction following UV irradiation. Taken together, these results indicate that recF/O/R mutations affect both replication recovery and activation of RecA for SOS induction following UV irradiation (Rangarajan et al. 2002). Both of these functions are required for successful UV-induced RA, which explains the small and delayed restriction alleviation.

Another mutant that fails to recover replication after UV irradiation is recA (Rangarajan et al. 2002). Since RecA protein is involved in the central steps of any recombination process and SOS induction, this result is not surprising and it is obvious that RA following UV irradiation cannot be induced (Day 1977).

The PriA protein is a 3′ → 5′ DNA helicase and the specificity protein for primosome assembly (Xu and Marians 2003 and references therein). It was demonstrated that priA mutants are defective in homologous recombination (Kogoma et al. 1996; Sandler et al. 1996), DSB repair (Kogoma et al. 1996), and both forms of stable DNA replication (Masai et al. 1994). This is why priA null mutants have reduced cell viability, defective cell division, and increased sensitivity to DNA damage, constitutive SOS induction, and recombination deficiency (Nurse et al. 1991; Kogoma et al. 1996). In agreement with our observation, the lack of replication recovery and recombination deficiency are strong arguments for the absence of RA induction following UV irradiation.

On the other hand, the second group of mutants listed in Table 3 are those that have constitutive RA (dam, recG, ruv recG, recB1080, and recD recJ). Among these mutants, only the recD recJ double mutant does not express constitutive SOS. The simplest explanation for constitutive RA in this mutant would be that less nascent ssDNA is degraded during the processing of dsDNA ends. Therefore, creation of 3′ ssDNA recombinogenic filaments occurs more frequently. Accordingly, a recD single mutant had a lower RA₀ value probably due to the action of the RecJ nuclease (Table 2).

Finally, recN and ruvABC represent a third group of mutants (Table 3). These mutants express a constitutive SOS response (Asai and Kogoma 1994; O'Reilly and Kreuzer 2004), but do not show constitutive RA. This indicates that unmethylated dsDNA does not accumulate in these mutants, supporting the finding that processing of Holliday junctions is not required for restriction alleviation (Blakely and Murray 2006).

On the basis of these and previous results, we provide arguments that replication restart at a D-loop is the mechanism for generation of unmethylated dsDNA that might contain the EcoKI target sequence following UV irradiation, as shown in our model in Figure 5. The strongest argument is the simultaneous requirement for DNA damage (SOS induction), replication recovery, and DSB repair (or functional RecBCD) for UV-induced RA (Table 3). Invasion of unmethylated ssDNA (created during DSB repair) into a homologous hemi-methylated dsDNA region would generate unmethylated dsDNA at D-loops (Murray 2000; Blakely and Murray 2006; Figure 5) stabilized by PriA activity. Binding of PriA to D-loops promotes replication fork assembly and replication recovery as shown by biochemical and in vivo studies (McGlynn et al. 1997; Liu and Marians 1999; Liu et al. 1999; Rangarajan et al. 2002). RecG was shown to dissociate junctions, i.e., to disrupt D-loop structures and destabilize R-loops by removing RNA (Whitby et al. 1993; Hong et al. 1995; Vincent et al. 1996). If unmethylated dsDNA is formed at a D-loop, then generation of unmethylated dsDNA would be decreased by RecG, which disrupts D-loop and R-loop formation. The high RA₀ value in the recG mutant supports this prediction (Figure 4B; Blakely and Murray 2006). The need for RecA-loading activity of the RecBCD enzyme also argues that a recombinogenic filament is required for UV-induced RA.

Figure 5.— A model for the generation of unmethylated dsDNA in UV-irradiated cells. dsDNA ends can be created either directly by excision repair proteins or after replication of a nicked template (a). Two recombining DNA molecules involved in recombinational repair (more ...)

Another alternative explanation for the mechanism of restriction alleviation has been recently proposed for type III and partially type I restriction systems. By using single-infecting phage conditions, it was shown that type I and type III, but not type II, restriction was alleviated by homologous recombination functions of a Rac prophage, i.e., by RecE and RecT proteins (Handa and Kobayashi 2005). It has been proposed that DNA replication of infecting phage could take place before a type III (or type I) restriction enzyme complex meets another enzyme to cleave DNA. If two daughter copies of the phage genome carry breaks at different loci, then RecET-mediated homologous recombination would be able to reconstitute one intact copy from them (Handa and Kobayashi 2005). However, further work is necessary to test the impact of RecET-mediated recombination on UV-induced RA, since the RecET effect is seen only when RecBCD nuclease is inactivated and in the presence of Rac prophage. These genetic requirements are different from those for UV-induced RA.

Before we proceed to outline our model for generation of unmethylated dsDNA, we would like to stress that only a small fraction of λ infections may undergo restriction alleviation. Up to 75% of the infecting nonmodified λDNA is converted to acid-soluble material (Thoms and Wackernagel 1982). The reason for the small fraction of λ survival could be that unmethylated dsDNA that would stimulate alleviation of restriction occurs rarely as the result of DNA repair on the chromosome, in a small fraction of UV-irradiated cells. Interestingly, it has been recently demonstrated that higher order DNA structure has an enormous effect on the activity of the type I restriction–modification enzymes (Keatch et al. 2004). DNA in the nucleoid is condensed and coated with nonsequence-specific ligands whereas foreign DNA is relatively naked and in a random coil conformation. The naked form of DNA is a good substrate for translocation and cleavage, while translocation on the nucleoid DNA is inefficient and the ClpXP protease can inactivate the enzyme. Thus, the difference in DNA conformation can explain why unmodified target sites in phage λ do not stimulate the RA response.

We propose that the replication fork breaks when it encounters a nick generated during the excision repair in one of the template strands or directly by the excision repair of two closely spaced photoproducts on opposite DNA strands in the chromosome. This dsDNA end has to be processed by the RecBCD and RecA proteins to promote homologous pairing and strand exchange with an intact sister hemi-methylated duplex. Strand exchange creates a D-loop where two newly synthesized unmethylated DNA strands anneal, thus creating unmethylated dsDNA that might contain the target sequence for the EcoKI enzyme. The PriA protein then targets this D-loop and replication is possibly stimulated by the RecFOR proteins (McGlynn et al. 1997; Nurse et al. 1999; Xu and Marians 2003). Subsequent resolution of the Holliday junction at the D-loop by the RuvABC resolvase would restore a replication fork and fix the unmethylated dsDNA fragment. A similar model for generation of unmethylated dsDNA during the repair of broken replication fork was proposed by Foster (1998) and Blakely and Murray (2006). Finally, a form of DNA replication called inducible stable DNA replication (iSDR), which is suggested to occur during DSB repair, also partially requires RecFOR and strongly requires RecBC (Kogoma 1997). Initiation of iSDR requires DNA damage, D-loop formation by RecA and RecBC, the primosome assembly activity of PriA, and SOS induction (Kogoma 1997), which are similar to the requirements for UV-induced RA. On the other hand, constitutive stable DNA replication that is induced in rnhA and recG mutants is similar to constitutive RA (Blakely and Murray 2006). SOS induction is dispensable for iSDR in recD mutants (Kogoma 1997), which explains the constitutive RA observed in a recD recJ double mutant. Adaptive mutagenesis is another interesting phenomenon where creation of adaptive mutations also depends on recombination functions. It has been suggested that iSDR is also involved in adaptive mutation (Kogoma 1997). Thus, recombination-dependent generation of unmethylated dsDNA is a potentially dangerous event: it can either cause DNA cleavage by restriction–modification enzymes or induce spontaneous mutations.

We thank Robert G. Lloyd (University of Nottingham), N. Murray (University of Edinburgh), and I. Matić (Necker Université) for providing us with bacterial strains. We are also grateful to Robert G. Lloyd and Mary Sopta (Ruđer Bošković Institute) for critical reading of the manuscript. This work was supported by the Croatian Ministry of Science (grant 0098070) and the Academic Links and Interchange Scheme program supported by the Croatian Ministry of Science and the British Council.