Introduction

DNA is an inherently stable and inert molecule when in duplex form. However, for its biological functions to be fulfilled it needs to be temporarily and site-specifically activated. Activation is achieved by means of creating a single-stranded (ss) intermediate, which enables DNA metabolic processes such as transcription, replication, homologous recombination and repair etc.

During homologous recombination ssDNA is a substrate onto which a recombinase protein (RecA, RadA, Rad51 (Dmc1) from bacteria, archaea and eukaryotes, respectively) is loaded, which then catalyzes an invasion of a homologous duplex DNA, giving rise to a joint molecule structure, named D-loop [1]. In that way DNA gets modified through exchanges with homologous sequences, but also preserved as this process repairs DNA lesions such as double strand breaks (DSBs) and impaired replication forks [1].

Cellular helicases and nucleases produce 3′-ending ss tails when processing DSBs, a process known as a DNA end resection, while internal ss regions, gaps, usually originate from replication defects [1]. When ssDNA appears in a cell, it is bound by a specific protein designated SSB in bacteria and RPA in eukaryotes. These proteins orchestrate action of many other proteins on ssDNA, while preventing recombinase polymerization [2]. This recombination inhibition is overcome by the action of dedicated proteins called recombination mediator proteins (RMPs), which catalyze an exchange of ss-binding protein for its cognate recombinase protein [3]. The main RMPs in bacteria are RecFOR and RecBCD/AddAB proteins [4]. While RecFOR proteins catalyze an SSB-RecA exchange on gapped DNA at the ssDNA-dsDNA junction [5], RecOR complex binds to an SSB-ssDNA structure and mediates an exchange of SSB for RecA [6]. RecBCD/AddAB class of enzymes binds with high affinity to a flush dsDNA end and unwinds it with its fast and processive helicase activity (reviewed in [7]). The enzyme degrades the unwound strands, until it encounters a specific octanucleotide sequence named χ. Interaction with χ changes RecBCD, which ceases degradation of the 3′-ending strand while increasing degradation of its 5′-ending complement. Concomitantly, a χ-modified RecBCD facilitates RecA loading onto thus created 3′-ending tail [8]–[10].

There are some situations in E. coli in which RecFOR proteins catalyze RecA polymerization onto ssDNA originating from a dsDNA end. This happens in a recBC null mutant that is also deficient in exonucleases exonuclease I (ExoI) and SbcCD, which degrade 3′-ending tails. Here, RecQ helicase produces ss tails and RecFOR proteins enable RecA loading onto the 3′-ending one, while its complementary strand is degraded by a RecJ exonuclease [1]. Another example is a “hybrid” recombination pathway [11], which occurs in mutants devoid of RecBCD’s nuclease and RecA loading activities due to a single amino acid change in an active center of the unique nuclease and RecA loading domain of the RecB subunit (recB1080 and recB1067 mutations) [12]–[15]. Here, intact DNA binding and helicase activities of RecB¹⁰⁸⁰CD (and RecB¹⁰⁶⁷CD) enzyme are coupled with an exonuclease activity of RecJ exonuclease (degrading the 5′-ended unwound strand) and a RecA loading activity of RecFOR proteins (onto the unwound 3′-ending tail) to promote homologous recombination, as well as UV and γ-survival [14], [16].

Because of its essential role in virtually every organism, homologous recombination is a well conserved and regulated process. When regulation of homologous recombination is compromised, an aberrant, illegitimate recombination ensues, occurring between regions of limited, or of no homology [17]. Illegitimate recombination leads to a genome instability, which causes a poor viability in viruses, bacteria and eukaryotes, and also cancer, sterility and many other illnesses in vertebrates.

DNA helicases of the RecQ family are critical for preserving genome integrity, hence their classification as “caretakers” of the genome [18]. RecQ helicases are highly conserved in evolution, both functionally and structurally, from bacteria to humans [19]. In unicellular organisms one family member is found in each organism (e.g. RecQ in bacteria and Sgs1 in budding yeast, Rqh1 in fission yeast), whereas there are as many as five human analogs: RECQL, RECQ4, RECQ5, WRN and BLM [19].

Biochemically, they show 3′–5′ polarity of DNA unwinding, with respect to the strand that binds the enzyme [20]. RecQ binds to a variety of substrates including: three and four way junctions (Holliday junctions), both the 3′ and 5′ ssDNA overhangs, a duplex DNA both internally and at the end, and, notably, it binds to recombination intermediates, joint molecules in Kappa and D-loop formation [21]. Such a broad substrate specificity allows RecQ helicase to act both presynaptically to promote homologous recombination in bacteria (as described above) as well as in humans [22] and yeast [23], and postsynaptically to decatenate junctions and disrupt nascent joint molecules, especially those arising from aberrant, nonhomologous joint exchanges [21], [24], [25].

A physiological effect of RecQ activity is seen in humans where deficiency in RecQ helicases WRN, BLM and RECQ4 give rise to recessive disorders Werner syndrome, Bloom syndrome and Rothmund-Thompson syndrome (as well as RAPADILINO and Baller-Geller syndromes), respectively [18]. All these disorders are characterized by a cancer predisposition and an increased genomic instability, the latter also being a phenotype in yeast Sgs1 mutants, along with accelerated ageing [18].

However, considering the adversity of effects rendered by inactivating a single out of five RecQ analogs in humans, it is quite remarkable that a complete RecQ deficiency in wild-type (wt) E. coli has barely detectable effects. Although it causes an increase in illegitimate recombination [26], it has no effect on cell viability, DNA repair and homologous recombination [27], [28].

The aim of this study was to characterize the factors responsible for such different phenotypes of RecQ deficient E. coli and humans. The recombinogenic substrates, dsDNA ends, are processed in wt E. coli exclusively by RecBCD enzyme which, for having no eukaryotic analogue, we considered the most likely differentiating factor in a recombination regulation between E. coli and humans. Since RecBCD is unique among DSB resection machines for possessing a recombinase loading activity, we made use of the aforementioned mutant RecB¹⁰⁸⁰CD and RecB¹⁰⁶⁷CD enzymes, exhibiting only a helicase activity, to uncouple creation of a long, reactive 3′-ending overhang from RecA polymerization onto it, and hence to make a DSB resection in E. coli more similar to that in eukaryotes. Inactivation of RecQ helicase in recB1080 and recB1067 mutants indeed caused a sickly phenotype consisting of a decreased cell viability and a reduced DNA repair and homologous recombination efficiency. At the same time, an illegitimate recombination frequency strongly increased in a recB1080 recQ mutant. The effects of the recQ mutation were exacerbated by a xonA mutation, inactivating a processive exonuclease, ExoI, that degrades 3′-ending tails [29], suggesting that the long 3′-ending tails produced by the mutant RecBCD are shifted from homologous recombination reactions into the nonproductive, aberrant DNA exchanges. We conclude that in E. coli the recombinogenic 3′ tails are channeled into productive reactions and away from aberrant ones by the coupled helicase and RecA loading activities of RecBCD enzyme and, upon their uncoupling, by the concerted action of RecQ helicase and ExoI.

Results

RecQ Helicase and ExoI are Required for DNA Repair in recB Mutants Deficient in Nuclease and RecA Loading Activities

By using P1 transduction we introduced a recQ deletion into recB1080 and recB1067 mutants, as well as in wt strain, and assessed their DNA repair capacity. DNA repair was evaluated by measuring UV and γ-survival of the mutants. As shown in Figure 1, the recB1080 ΔrecQ mutant was considerably more sensitive to both UV and γ-irradiation than the parental recB1080 strain. At the highest UV and γ-dose tested, the survival of the double mutant was about 10 fold lower than that of the parent strain. A Δ recQ derivative of the recB1080 mutant was about two fold more radiosensitive than recB1080, when comparing survival slopes (Figure 1B). The same effect was observed in recB1067 background (not shown). Although our results are somewhat similar to those of a previous study, the conclusion of that study (that RecQ is “not needed for RecB1080CD-mediated recombination”) [16] is opposite to our conclusions. As expected, and in accord with previous studies [27], [28], RecQ inactivation did not produce any effect on UV and γ-survival of otherwise wt cells (Figure 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. ΔxonA and ΔrecQ mutations impair DNA repair in UV irradiated (A) and γ-irradiated (B) recB1080 mutant cells, which lack nuclease and RecA loading activities of RecBCD.

Fraction survival is given as a fraction of the unirradiated control. Symbols: (▪) recB1080; (▴) ΔxonA recB1080; (▾) ΔrecQ recB1080; (♦) ΔxonA ΔrecQ recB1080; (•) recB268::Tn10.

https://doi.org/10.1371/journal.pone.0039030.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. DNA repair in UV irradiated (A) and in γ-irradiated (B) wild-type bacteria is mostly unaffected by ΔxonA and ΔrecQ mutations.

Fraction survival is given as a fraction of the unirradiated control. Symbols: (▪) AB1157 rec⁺; (▴) ΔxonA; (▾) ΔrecQ; (♦) ΔrecQ ΔxonA.

https://doi.org/10.1371/journal.pone.0039030.g002

In order to determine the role of the unwound 3′ tail processing in recB1080, we inactivated ExoI, the most powerful exonuclease that acts on 3′ overhangs in E. coli [29]. Much like a recQ deletion, an ExoI inactivation rendered a resultant recB1080 ΔxonA mutant substantially more sensitive to UV and γ-irradiation compared to the parental recB1080 mutant (Figure 1). Essentially the same results were obtained with a ΔxonA derivative of recB1067 mutant. It was also notably more sensitive to UV and γ-irradiation than its parental strain (data not shown). On the other hand, a ΔxonA derivative of a wt strain (AB1157) was as UV resistant as its parental strain, while being slightly more radiosensitive than AB1157 (Figure 2), indicating that ExoI function is required much more for DNA repair in the nuclease and RecA loading deficient recB mutants than in wt bacteria.

Next, we inactivated genes coding for RecQ and ExoI in recB1080 and wt strains and noticed opposite effects. Whereas in the wt background no effect was seen on both UV and γ-survival (Figure 2), a recB1080 ΔrecQ ΔxonA mutant was extremely sensitive to both agents, almost as a recB null mutant DE101 (Figure 1). As expected, a ΔxonA ΔrecQ recB1067 mutant was also almost completely devoid of DNA repair (data not shown), suggesting that DNA repair in mutants deficient in RecBCD’s nuclease and RecA loading activities depends heavily on the functions of ExoI and RecQ proteins. The synergism of ExoI and RecQ suggests that they act in parallel, overlapping pathways in that background.

Conversely, a pSQ211 plasmid, carrying a recQ⁺ gene, rendered the ΔrecQ recB1080 mutant as resistant to UV irradiation as the single recB1080 mutant, while having no effect on UV survival of the ΔxonA recB1080 mutant (data not shown). When introduced into the triple ΔxonA ΔrecQ recB1080 mutant, the plasmid complemented its UV sensitivity only partially, to the level of double mutants (data not shown). These results suggest that the RecQ helicase supplied in trans complements only a missing RecQ helicase but not an ExoI activity in the recB1080 mutant background.

RecQ Helicase and ExoI are not Required for DNA Repair in Nuclease Deficient but RecA Loading Proficient recB Mutants

When the RecB¹⁰⁸⁰CD and RecB¹⁰⁶⁷CD mutant enzymes, devoid of both nuclease and RecA loading activities [14], [30], lack the RecD subunit (in recD derivatives of recB1080 and recB1067), they gain the ability to constitutively load RecA protein, while remaining nuclease deficient [14], [30]. We used the recD recB1080 and recD recB1067 mutants to determine which of the two missing RecBCD functions is/are bypassed with the functions of ExoI and RecQ helicase. When either single ΔxonA or ΔrecQ mutations, or both of them, were introduced into a recD derivative of the recB1080 mutant, they did not affect its UV and γ-survival (Figure 3). The same behavior was observed in the recD recB1067 mutant; introduction of either the ΔxonA or the ΔrecQ or both of these mutations did not influence its UV and γ-survival (data not shown). These results corroborate earlier findings that recQ [31] or xonA [32] mutation does not affect UV and γ-survival in recD derivatives of recB nuclease deficient strains. An individual ΔxonA or ΔrecQ mutation, as well as their combination, also did not change UV and γ-survival of RIK144, a recD derivative of the wt strain (not shown). These results show that ExoI and RecQ helicase are not required for DNA repair in the nuclease deficient but RecA loading proficient recD recB mutants; meaning that RecB¹⁰⁸⁰C-catalyzed RecA loading at DNA ends renders the two enzymes dispensable.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Inactivation of xonA and recQ genes does not affect UV radiation survival (A) and γ-survival (B) of nuclease deficient but RecA loading proficient recB1080 recD mutants.

Fraction survival is given as a fraction of the unirradiated control. Symbols: (▪) recB1080 recD; (▴) recB1080 recD ΔxonA; (▾) recB1080 recD ΔrecQ; (♦) recB1080 recD ΔrecQ ΔxonA.

https://doi.org/10.1371/journal.pone.0039030.g003

RecQ Helicase and ExoI are Required for Maintaining Cellular Viability in Nuclease and RecA Loading Deficient recB Mutants

Exogenous genotoxic agents (e.g. UV and γ) are thought to pose only a minor, occasional threat to a cell, wherein the majority of DNA damage stems from a continuous action of endogenous genotoxic agents such as reactive oxidative species and various replication impediments [1]. The ability of E. coli to cope with an endogenously caused DNA damage is assessed genetically by determining its viability. A cell is considered viable when able to give rise to a colony.

A colony-forming ability of the nuclease and RecA loading deficient recB1080 mutant was about one third lower than that of the wt strain, and was further decreased by one third and half upon introducing ΔrecQ and ΔxonA mutations, respectively (Table 1). ΔxonA and ΔrecQ mutations exhibited a synergistic effect rendering the ΔxonA ΔrecQ recB1080 mutant poorly viable. Its viability of about 13% was about 8 and 5 fold lower than that of the wt and the recB1080 mutant, respectively (Table 1), meaning that, on average, only one out of eight ΔxonA ΔrecQ recB1080 cells was able to give rise to a colony. As a consequence, this mutant grew slowly in a liquid medium and on agar plates. As expected, a colony-forming ability of the ΔxonA ΔrecQ recB1067 mutant was similar to that of the recB1080 ΔxonA ΔrecQ (data not shown). The viability of the triple ΔxonA ΔrecQ recB1080 mutant increased to about 32% when harboring the RecQ-producing plasmid.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Relative cellular viability and conjugational recombination proficiency of ΔrecQ and ΔxonA derivatives of wild-type, recB1080, recD and recD recB1080 strains.

https://doi.org/10.1371/journal.pone.0039030.t001

The observed low colony-forming ability of the triple mutant might be due either to its low cellular viability or a filamentation. In order to differentiate between these two possibilities, we examined ΔxonA ΔrecQ recB1080 mutant cells microscopically and observed only a few filaments in their population (not shown). We conclude, therefore, that the low colony-forming ability of the triple mutant reflects its low cellular viability.

On the other hand, cellular viability of wt, recD and recD recB1080 mutants (as well as that of recD recB1067, not shown), was not affected by ΔxonA and ΔrecQ mutations (Table 1). These results are in accord with previous reports showing that recQ inactivation does not render wt and recB1080 recD cells less viable [27], [31], and also a xonA mutation does not affect a viability of recD and recB1067 recD mutants [32].

Overall, these results suggest that gene products of xonA and recQ are required for maintaining cellular viability of the recB mutants deficient in nuclease and RecA loading activities, but not in their recD derivatives (deficient only in nuclease activity) nor in wt strain and its recD derivative.

Synergistic Action of RecQ Helicase and ExoI Suppresses Illegitimate Recombination in Nuclease and RecA Loading Deficient recB1080 Mutant

During λ phage infection, a viral genome gets integrated into an E. coli chromosome between gal and bio genes, thus establishing a lysogenic mode of infection. Upon λ prophage induction, its genome is excised from a bacterial chromosome in a generally precise reaction. However, the excision is sometimes aberrant, caused by an illegitimate recombination event, and hence a neighboring part of the bacterial genome is packed into λ’s head instead of part of its own genome [34]. When during a λcI857 prophage thermoinduction excised phage carries a part of the bio gene containing a χ sequence, part of its genome coding for red and gam functions is deleted. Such a λbio phage, excised by illegitimate recombination, gives rise to a Spi⁻ phenotype [35]. That means it is capable of growing on a P2 lysogenic E. coli, whereas wt λ is not. Therefore, the frequency of Spi⁻ phage arising from λcI857 prophage induction represents a frequency of illegitimate recombination [35].

We measured the appearance of Spi⁻ phage in wt and recB1080 backgrounds. Very low frequency of Spi⁻ phage originating from wt cells was increased about 13 fold in recQ mutant (Table 2), which is close to the lower limit of the range observed earlier [26]. Inactivation of ExoI increased appearance of Spi⁻ phage only slightly (Table 2). In our assay, therefore, we did not observe an effect of ExoI on illegitimate recombination in wt bacteria, in contrast to earlier studies (using different assays) which found that ExoI suppresses illegitimate recombination in that background [36], [37]. A double recQ xonA mutant produced Spi⁻ phage at a rate similar to a recQ single mutant, i.e. it was increased about 18 fold (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Illegitimate recombination is suppressed in recB1080 mutant by synergistic activities of RecQ helicase and ExoI exonuclease.

https://doi.org/10.1371/journal.pone.0039030.t002

The nuclease and RecA loading deficient recB1080 mutant yielded somewhat higher number of Spi⁻ phage, frequency of which was increased about five fold compared to wt strain (Table 2). This suggests that processive helicase activity of the mutant RecB¹⁰⁸⁰CD enzyme, devoid of nuclease and RecA loading activities, leads to an increased illegitimate recombination. Inactivation of RecQ helicase in recB1080 mutant increased the Spi⁻ phage frequency about 20 and 90 fold, compared to its parental and wt strain, respectively (Table 2). About the same effect was observed in a xonA derivative of the recB1080 mutant (Table 2). Furthermore, when both recQ and xonA mutations were introduced into recB1080 strain, their synergistic effect was observed and Spi⁻ phages were produced about 85 fold more often than in the parental strain, and about 390 fold more frequent than in wt strain.

We also determined a Spi⁻ phage yield in a recD derivative of recB1080, wherein RecB¹⁰⁸⁰C enzyme exhibits coupled helicase and RecA loading activities. The slightly increased rate of a Spi⁻ phage occurrence in recB1080 recD strain was further increased about 40 fold by the recQ mutation (Table 2). On the other hand, the xonA mutation did not exert any effect on a Spi⁻ phage production in the recB1080 recD genetic background. The recB1080 recD recQ xonA mutant produced somewhat less Spi⁻ phage than the recB1080 recD recQ strain (Table 2). The recB1080 recD strain is therefore similar to the wt strain by the way of RecQ suppressing illegitimate recombination and ExoI not having any effect on it. In contrast, illegitimate recombination in the recB1080 mutant is suppressed by synergistic action of RecQ and ExoI. However, the fact that illegitimate recombination is still increased in that mutant despite intact activities of RecQ and ExoI, suggests that their combined suppressing capacity is exceeded.

In an earlier study [35] an illegitimate recombination frequency was increased in UV irradiated E. coli. We irradiated wt and recB1080 strains as well as their recQ, xonA and recQ xonA derivatives in order to determine the overall cellular capacity for illegitimate recombination. The UV irradiated cells indeed gave rise to more Spi⁻ phage than their unirradiated counterparts. The increase in a Spi⁻ phage production ranged from 7.2 to 48.9 fold in the wt background and from 2.8 to 118.5 fold in the recB1080 background (Table 2). Surprisingly, the recB1080 recQ xonA mutant, suffering the highest spontaneous illegitimate recombination, had the lowest Spi⁻ phage frequency increase upon UV irradiation (just 2.8 fold) (Table 2), suggesting that a limit of illegitimate recombination occurrence in E. coli, or in our assay sensitivity, is met. Strains exhibiting a lower rate of spontaneous illegitimate recombination often had the higher increase upon UV irradiation, coming closer to a maximal illegitimate recombination frequency of about 4×10⁻⁷ (Table 2). This saturation point is about 1100 fold above the level in unirradiated wt strain.

Overall, our results suggest that when a highly processive and fast helicase activity of RecBCD enzyme is uncoupled from its RecA loading activity, a long 3′ tail hence produced is very prone to engaging in aberrant, nonproductive exchanges, thus leading to illegitimate recombination and away from homologous recombination events. This activity is counteracted by a 3′–5′ helicase activity of RecQ and a 3′–5′ exonuclease activity of ExoI, which thereby enable homologous and suppress illegitimate recombination in the recB1080 background.

Discussion

RecBCD, a multifunctional bacterial enzyme, belongs to the common DNA-end resecting machinery consisting of a helicase-nuclease tandem. A combination of a helicase and a 5′–3′ exonuclease resects DSBs in: bacteria (RecBCD, RecQ/RecJ), archaea (HerA/NurA), yeast (Sgs1/Dna2), Xenopus laevis (xWRN/xDNA2), humans (BLM/EXOI), etc. [38]. However, RecBCD is unique for having a recombinase loading activity, which, together with the enzyme’s helicase and nuclease activities, produces a RecA nucleofilament in a highly coordinated manner.

In order to make a DSB resection in E. coli more similar to that in eukaryotes and archaea, we made use of the recB1080 and recB1067 mutants, coding for the RecB¹⁰⁸⁰CD and RecB¹⁰⁶⁷CD enzymes, respectively, that possess a fast and processive helicase activity, but are devoid of nuclease and RecA loading activities (as no difference was observed between the two mutants, recB1080 will be discussed hereafter, referring also to recB1067). A helicase/exonuclease pair RecB¹⁰⁸⁰CD/RecJ, operative in that mutant [14], is therefore functionally analogous to budding yeast Sgs1/Dna2, Xenopus laevis xWRN/xDNA2 and human BLM/EXOI helicase/exonuclease pairs that catalyze a long-range DSB resection step during homologous recombination repair in those organisms [38].

The RecB¹⁰⁸⁰CD/RecJ pair is not as efficient as RecBCD since recB1080 mutant exhibits reduced cell viability, an impaired DNA repair and homologous recombination, but an increased illegitimate recombination. We show here that RecQ helicase is required for maintaining cell viability, DNA repair and homologous recombination in the recB1080 mutant, but also for suppressing illegitimate recombination. These results indicate a postsynaptic, disruptase role for RecQ in controlling the fidelity of recombination exchanges, i.e. funneling the recombinogenic 3′ tails produced by RecB¹⁰⁸⁰CD into productive DNA exchanges and away from the aberrant ones. This kind of behavior mimics that observed in eukaryotes, where RecQ analogues were shown to both promote homologous recombination and suppress illegitimate recombination and genomic instability that may lead to cancer [18].

We also inactivated ExoI, the principal E. coli 3′–5′ exonuclease [29], in order to increase longevity of the recombinogenic 3′ tails, and show that the resultant recB1080 xonA strain displays similar pleiotropic effects as its recQ counterpart. ExoI and RecQ act in synergy since the triple recB1080 recQ xonA mutant is poorly viable and is almost completely devoid of homologous recombination and DNA repair, yet with ∼400 fold increased illegitimate recombination, compared to the wt strain. In that strain 3′ overhangs are directed into aberrant, nonproductive exchanges (increased ∼400 fold); instead of in productive reactions (decreased ∼30–100 fold).

Since the shift from homologous to illegitimate recombination was not observed in recQ and xonA derivatives of wt and recB1080 recD cells, exhibiting helicase/nuclease/RecA loading and helicase/RecA loading activities respectively, we conclude that this shift in the recB1080 mutant is due to a fast and processive helicase activity of RecB¹⁰⁸⁰CD, lacking a coordinated RecA polymerization activity. Hence, the differences in phenotypes of E. coli and humans deficient in RecQ and its eukaryotic analogs, respectively, are not due to differences among these enzymes but rather to differences in recombination substrates metabolisms in these organisms. A robust RecBCD action at DSBs, which fixes them efficiently through homologous recombination, does not leave much space for the RecQ function in wt E. coli. Accordingly, when RecBCD function is affected by an overexpression of RecET proteins in otherwise wt cells, DSB processing becomes increasingly aberrant, leading to increased illegitimate recombination. In these cells RecE nuclease, along with RecT, RecJ and RecFOR proteins, promotes illegitimate recombination while RecQ suppresses it (reviewed in [17]).

The synergism of a 3′–5′ helicase activity of RecQ and a 3′–5′ exonuclease activity of ExoI in the recB1080 mutant, which we revealed in this study, mimics the activities of human and X. laevis WRN and xWRN RecQ analogues, respectively, which are both 3′–5′ helicases and 3′–5′ exonucleases [18]. They act in both homologous recombination initiation and in preventing aberrant exchanges [39], [40], and when WRN protein lacks helicase and exonuclease activities, the cancer-prone Werner syndrome ensues [41], which is characterized by an increased genomic instability and reduced homologous recombination [41].

The evolutionary conservation of a RecQ family function is exemplified by the ability of human BLM protein to partially suppress illegitimate recombination in a recQ mutant of E. coli, as assessed by the λ Spi⁻ assay [42], which we used in this study too. This suppression indicates that substrate specificity is conserved from RecQ to BLM, and that illegitimate recombination events detected by the λ Spi⁻ assay are similar enough to those occurring in human cells to fall into range of BLM substrate specificity. This should not be surprising since illegitimate recombination events observed in that assay stem from resected DSBs (3′ overhangs) caused by replication impairment [17], which is a common challenge to genome integrity in living organisms [1].

In fact, when comparing an illegitimate recombination pathway (IR) observed in the λ Spi⁻ assay in E. coli with eukaryotic DSB repair pathways, one can notice that the IR bears resemblance with microhomology-mediated end joining (MMEJ). Both pathways are initiated by DSB resection, giving rise to 3′ overhangs, which align broken ends by an end-joining reaction dependent on microhomologies (∼4 nt for MMEJ and ∼9 nt for IR) and a ligase function [17], [43]. Both pathways are recombinase independent and are suppressed by homology dependent repair, which is regulated at a DSB resection step [43], [17]. Although known for some time, MMEJ has only recently been appreciated as the “major mechanism for chromosome translocations, and possible other rearrangements, in mammalian cells” [43]. Recurrent chromosome translocations are found in many hematological, epithelial malignancies, also in solid tumors etc. [43]. A key factor in generating these tumorigenic rearrangements appears to be the extent of DNA resection [43].

The conservation of RecQ helicase family functions, DNA damage inducing mechanisms, as well as aberrant recombination intermediates between E. coli and humans, allows the recB1080 mutant and the λ Spi⁻ assay to make up a suitable bacterial model for studying RecQ helicase family functions, especially that of WRN and BLM proteins, in order to elucidate the molecular bases of genomic instability and its malignant phenotypes.

Our results show that DSB resection is a critical factor determining the balance of homologous and illegitimate recombination in E. coli and, hence, its genome stability. Remarkably, when E. coli DSB resection becomes mechanistically similar to that in eukaryotes (by its uncoupling from a DNA recombinase polymerization step), the role of RecQ increases and also becomes similar to that of its eukaryotic counterparts in a manner of propagating homologous and preventing illegitimate recombination events.

Another analogy between E. coli and eukaryotic DSB resection arises from this study, namely the adverse effect of increase in its processivity on DNA repair efficiency and genome stability. Just as RecBCD, the RecB¹⁰⁸⁰CD enzyme is a fast (∼1–2 kbp s⁻¹) and very processive helicase [7] (≥50 kbp/binding event, our estimation from the fact that wt RecBCD efficiently unwinds a 50 kbp long χ-less F plasmid fragment and a part of chromosomal DNA before encountering a χ site during Hfr-conjugation, [44]). Since in vitro the enzyme is not inactivated upon interaction with χ (as is its wt counterpart), but just slowed down [13], its processivity is certainly higher than that of the wt enzyme. Furthermore, as RecB¹⁰⁸⁰CD lacks any nuclease activity, it produces full length overhangs (their length being in range of tens of kb), whereas RecBCD produces just a post-χ 3′ tail whose length is likely several kb at most. Lower viability, decreased homologous recombination efficiency and increased sensitivity to genotoxic agents by the recB1080 mutant, compared to the wt cells, thus correlates well with an increased processivity of DSB resection in that mutant.

A long (≥50 kb) 3′ tail produced by the RecB¹⁰⁸⁰CD in a matter of seconds is long-lived, thus posing a threat for a cell, and a great challenge for ExoI with its ∼275 nt s⁻¹ [45] rate of 3′-tail resection. ExoI trimming is required to reduce its length, especially since RecFOR-mediated RecA polymerization, active in that strain [16], is shown to take place at the ss-ds DNA junction, i.e. at the 5′ end of that tail [5]. As RecA needs to be present at the 3′ tip of ssDNA in order to make a productive plectonemic joint molecule [46], a long 3′ overhang makes RecA polymerization from 5′ towards 3′ end a lingering process that inefficiently compete with nonproductive, aberrant DNA exchanges, thus shifting recombination from predominantly homologous to illegitimate.

Similar to E. coli, an increased DSB resection processivity in eukaryotes leads to sensitivity to DNA damage and genomic instability [38]. For instance, hyperactive Sae2-S267E and CTIP-T847E nucleases in yeast and humans, respectively, which process dsDNA ends with an increased processivity (leaving longer 3′tails), cause an increased ionizing radiation sensitivity and an increased genome instability, leading to gross chromosomal rearrangements [47], [48]. Also, inactivation of a BRCA1-RAP80 complex, which restricts a DSB resection in humans, confers increased ionizing radiation sensitivity and increased genome instability to HeLa cells [49]. An aberrant DSB resection in all these instances impairs balance of DSB repair pathways in yeast and human cells [47]–[49].

Another consequence of an increased dsDNA end resection by a “runaway” helicase activity needs to be taken into an account. Namely, it is related to an SSB protein pool in a cell. There are about 6000 SSB molecules in an E. coli cell [50], which comprise 1500 tetramers that can complex about 50–100 kb of ssDNA (1 tetramer per either 35 or 65 nt, [50]). If RecB¹⁰⁸⁰CD unwinds 50 kbp of DNA duplex (which is a quite conservative estimate), then 200 kb of ssDNA is produced in a cell during conjugation (i.e. two unwound strands at two ends of a transferred DNA). That means that during conjugation, at least 100–150 kb of an unwound DNA is free of SSB and thus increasingly reactive, and moreover, more resistant to ExoI and RecJ trimming, which is stimulated by SSB (see below). In that respect, exonuclease activities of ExoI and RecJ serve not just to decrease the length of the unwound overhangs, but also to replenish the exhausted SSB pool in a cell, thus enabling their repeating binding to the SSB-free unwound ssDNA. This argument also provides a rationale for coupling of a vigorous nuclease activity to a fast and processive helicase activity of RecBCD enzyme.

We propose a model for a DSB resection in the E. coli recB1080 genetic background (Figure 4). As opposed to the situation in wt strain (A), wherein RecBCD enzyme unwinds and degrades both unwound strands upon interaction with a χ site, after which it only degrades the 5′ ending strand while polymerizing RecA onto the 3′ ending strand, thus creating a RecA nucleofilament; in recB1080 mutant (B), RecB¹⁰⁸⁰CD helicase is active. Its fast and processive helicase activity produces long 3′ and 5′ overhangs, onto which SSB molecules are bound. The SSB interacts with ExoI and RecJ ssExos and stimulates their trimming of 3′ and 5′ tails, respectively [51], [52]. Since exonuclease activities of ExoI and RecJ are slower and less processive than the RecB¹⁰⁸⁰CD helicase activity, unwound tails are quite stable, which enables them to get involved into nonproductive, aberrant DNA exchanges dependent on a limited homology, thus leading to illegitimate recombination events (ii). However, these aberrant reactions are counteracted by 3′ overhang trimming by ExoI and by a disruptase activity of RecQ helicase, enabling ExoI to further degrade the tail, thus directing it into homologous recombination reactions (i). Homologous recombination is effected by RecFOR-dependent RecA polymerization onto SSB-ssDNA filament, which occurs at an ssDNA-dsDNA junction, i.e. at the 5′ end of the tail. This localization is unfavorable for a RecA filament to catalyze a productive plectonemic joint, and instead an unstable paranemic structure ensues [46]. We propose that this unstable structure is either disrupted by RecQ, as shown for WRN and BLM proteins [18], [40], or by some other factors, thus allowing ExoI and RecA to deliver the RecA nucleofilament to the 3′ tip of the tail and hence direct it into a productive exchange. An alternative RecA polymerization reaction, the one catalyzed by a RecOR complex [6], seems unlikely since recF, recO and recR mutations exert mutually undistinguishable phenotypes in recB1080 strain [16]. In the recB1080 recQ xonA mutant (C), the absence of ExoI function increases the 3′ overhang’s longevity (although ExoVII, SbcCD and ExoX can still act on it); while simultaneously RecQ deficiency increases its reactivity. As a result, nonproductive, aberrant DNA exchanges are stimulated, leading to both RecA independent illegitimate recombination (ii) and RecA dependent paranemic joint molecules (i).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. A model for a DSB resection in E. coli.

Red and gray dots bound on ssDNA tails symbolize RecA and SSB proteins, respectively. Designations (i) and (ii) are assigned to pathways leading to RecA dependent and RecA independent (illegitimate) exchanges, respectively. Hatched arrows represent the nonobligatory, sporadic reaction steps. Details of the reactions are described in the text.

https://doi.org/10.1371/journal.pone.0039030.g004

Materials and Methods

Strains, Growth Conditions and Media

Designations, genotypes and references/construction of the E. coli K-12 strains used in this study are listed in Table 3. Most of them were AB1157 derivatives and were constructed by P1 transduction, as described earlier [53]. Transductants that acquired new allele(s) were selected by new antibiotic resistance(s) they obtained. The genotype recBD1080A describes a RecB mutant protein with an amino acid aspartate substituted with an alanine at position 1080 by a point mutation [12], [14]. A plasmid pSQ211 carries the recQ⁺ gene under a T7 polymerase promoter and overproduces RecQ enzyme upon induction with IPTG [21].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Bacterial strains used in this study.

https://doi.org/10.1371/journal.pone.0039030.t003

Bacteria were grown in Luria-Bertani (LB) broth and on LB plates [53] at 37°C. Broth and plates used for growing of the mutant strains marked with antibiotic resistance were supplemented with appropriate antibiotics: 20 µg ml⁻¹ chloramphenicol, 40 µg ml⁻¹ kanamycin, 50 µg ml⁻¹ ampicillin and 10 µg ml⁻¹ tetracycline. When a strain contained two or three antibiotic markers, antibiotic concentrations were adjusted accordingly.

A fresh overnight culture from a single colony was diluted 100 fold in fresh LB medium supplemented with appropriate antibiotics, when required. Then the culture was grown at 37°C until it reached an optical density at 600 nm (OD₆₀₀) of 0.3. All the experiments were done with log-phase cultures at an OD₆₀₀ of ∼0.3. Unless otherwise stated, each presented value is a mean of at least three independent experiments ± standard deviation.

Acknowledgments

The authors wish to thank Noreen Murray, Susan Lovett and Richard Myers for bacterial strains and Stephen Kowalczykowski for the plasmid. We are grateful to Nikola Paić for his help with manuscript preparation.