Functional fine-tuning between bacterial DNA recombination initiation and quality control systems.

Homologous recombination (HR) is crucial for the error-free repair of DNA double-strand breaks (DSBs) and the restart of stalled replication. However, imprecise HR can lead to genome instability, highlighting the importance of HR quality control. After DSB formation, HR proceeds via DNA end resection and recombinase loading, whereas helicase-catalyzed disruption of a subset of subsequently formed DNA invasions is thought to be essential for maintaining HR accuracy via inhibiting illegitimate (non-allelic) recombination. Here we show that in vitro characterized mechanistic aberrations of E. coli RecBCD (resection and recombinase loading) RecQ (multifunctional DNA-restructuring helicase) mutant enzyme variants, on one hand, cumulatively deteriorate cell survival under certain conditions of genomic stress. On the other hand, we find that RecBCD and RecQ defects functionally compensate each other in terms of HR accuracy. The abnormally long resection and unproductive recombinase loading activities of a mutant RecBCD complex (harboring the D1080A substitution in RecB) cause enhanced illegitimate recombination. However, this compromised HR-accuracy phenotype is suppressed in double mutant strains harboring mutant RecQ variants with abnormally enhanced helicase and inefficient invasion disruptase activities. These results frame an in vivo context for the interplay of biochemical activities leading to illegitimate recombination, and underscore its long-range genome instability effects manifest in higher eukaryotes.

Introduction

Homologous recombination (HR) is a biological process that occurs in all kingdoms of life, and provides the most efficient means for the error-free repair of DNA double-strand breaks (DSBs) and the restart of stalled DNA replication [1,2]. HR is mediated by the resection of DSB ends, producing 3’ single-stranded (ss) DNA overhangs that are substrates for the subsequent formation of recombinase nucleoprotein (RNP) filaments. RNP filaments catalyze invasion into, and homology search within, an undamaged homologous DNA template, leading to the formation of joint DNA molecules (JMs). Subsequent DNA synthesis and various pathways of JM processing ultimately lead to the restoration of the information content of the damaged DNA molecule by using information from the undamaged template [2,3].

In E. coli, RecBCD is the main enzyme complex catalyzing DSB end resection and coupled RecA recombinase loading, resulting in efficient production of RNP filaments on 3’ ssDNA overhangs [3]. Amino acid substitutions in the nuclease/RecA loading domain of the RecB subunit (e.g., D1080A, denoted as recB1080) lead to enhanced helicase (dsDNA unwinding) activity of the RecBCD complex, associated with compromised nuclease and RecA loading activities[4-6]. As a result, the mutant RecBCD complex produces abnormally long 5’ and 3’ overhangs that can only be partially filled with RecA [4]. Thus, the recB1080 genotype is associated with a specific, well-characterized alteration in the initiation (resection/RNP formation) mechanism of bacterial HR [4, 6].

RecQ helicase, a member of the alternative RecF recombination pathway, has been implicated to play multiple important roles assisting DNA replication and HR-based repair [7]. In conjunction with RecJ exonuclease, RecQ can catalyze DNA end resection during the repair of single-strand gaps (SSGs) in DNA. Moreover, the ‘invasion disruptase’ function of RecQ is thought to be important for the quality control (i.e., accuracy) of HR, as RecQ can unwind short base-paired DNA intermediates formed between non-allelic DNA sequences that could lead to illegitimate recombination (IR) [7,8]. The dsDNA unwinding activity of RecQ also contributes to the restart of stalled replication and the completion of chromosomal replication [9-11]. The latter function of the eukaryotic homologs of RecQ is indicated by severe hereditary diseases associated with hyperrecombination, genome instability, accelerated aging and high cancer predisposition [12-14].

In our recent work we showed that RecQ exhibits characteristic complex biochemical activities besides the elementary dsDNA unwinding activity [15]. RecQ displays short-range repetitive shuttling (unwinding and rezipping) along DNA segments as well as directed disruption of DNA strand invasions, HR intermediates thought to be the key targets for HR quality control [15]. In addition, we identified engineered RecQ variants that perform enhanced dsDNA unwinding but specifically lack the complex shuttling and disruptase activities (see below). We showed that these properties together are detrimental for HR accuracy, and lead to IR frequencies elevated significantly over the wild-type (WT) level [15].

The above described specific, well-characterized alterations of RecBCD and RecQ activities provide an unprecedented opportunity to probe the poorly understood, but potentially crucially important, functional interplay between these enzymes in the repair of DNA lesions and in HR quality control. In the present study we quantitatively tested the phenotypic effects of these alterations, and their combination, on cell growth/survival under normal and genotoxic conditions, besides measuring HR accuracy in untreated cells. We find that the specific RecBCD and RecQ functional defects have cumulative effects on the survival of certain forms of genomic stress. Surprisingly, however, we show that the compromised HR accuracy resulting from aberrant RecBCD function is significantly ameliorated by specifically perturbing, but not by abolishing, RecQ activities. These results suggest that the activities of different HR enzymes have been adaptively co-optimized to their specific natural (WT) functional environment in order to serve efficient DNA lesion repair on one hand, and long-range genome stability (HR accuracy) on the other. The importance of a similar balance in higher eukaryotes is reflected in the association of similar functional aberrations with genome instability and cancer predisposition syndromes [16-18].

Results

Design and generation of E. coli strains

By using the RedET recombination system, we generated the following mutant E. coli strains (using the MG1655 strain as WT) expressing modified protein variants driven by their respective native promoters (). recB1080 harbors a RecB protein with amino acid (aa) substitution D1080A, which has been shown to cause enhanced helicase activity of the RecB1080CD complex, combined with impaired nuclease and RecA-loading activities [5]. Thus, RecB1080CD performs modified DSB processing that results in abnormally long 5’ and 3’ ssDNA overhangs that are only partially filled with RecA [6]. The recQ* strain harbors the Y555A substitution within RecQ’s Helicase-and-RNaseD-C-terminal (HRDC) domain, abolishing the ssDNA interaction of this domain [19]. In the recQ-dH and recQ-dWH strains the protein is C-terminally truncated after aa 523 and 413, respectively, thereby lacking the HRDC (recQ-dH) and winged-helix (WHD) plus HRDC domains (recQ-dWH) [15, 20].The RecQ* and RecQ-dH proteins exhibit enhanced DNA unwinding activity, but lack the shuttling activity characteristic of RecQ-WT, and RecQ-dH is further severely impaired in its directed invasion disruption activity [15, 20]. RecQ-dWH is a weak helicase that can be considered a loss-of-function variant for all DNA-restructuring activities [15]. The ΔrecQ strain is a null mutant expressing no RecQ helicase [15]. In addition, we generated double mutant strains for recB1080 and each recQ gene variant to assess the interplay between RecBCD and RecQ activities.

In the experiments described below, we quantitatively determined how the above mutations affect cell growth and division under stress-free conditions, cell survival under different forms of genomic stress, and HR accuracy under stress-free conditions.

The recB1080 mutation, but not recQ defects, impairs cell division under stress-free conditions

Previously we found that recQ mutations do not markedly impair E. coli culture growth under stress-free conditions [15]. In the present study we monitored cell growth in more frequent (5-min) OD600 (optical density at 600 nm) sampling intervals, allowing quantitative analysis of growth curves based on the modified Gompertz model () [21]. We also tested the effects of recQ mutations in the recB1080 background (). One-way ANOVA analysis of the maximal extent and rate of growth indicated that the recB1080 recQ-dWH double mutant and recQ-dH single-mutant strains had a slightly but significantly elevated growth extent and lowered rate, respectively, over several other assessed strains (). Apart from these findings, the analysis revealed no systematic effects of the mutations under these conditions.

RecB and RecQ defects do not affect cell growth, whereas the recB1080 mutation markedly impairs cell division under stress-free conditions.

(A) Growth (OD600) curves of indicated E. coli strains, fitted by the modified Gompertz model (solid lines, ) (21). Means ± SE for n = 5 biological replicates are shown. Error bars are within symbols where not visible. Inset: cell growth monitored via changes in CFU/ml over time (means ± SE for n = 3). Determined parameters and analysis results are shown in and . (B) Example images of methylene blue stained cells (see also A panel in ). (C) Fraction of the total cell area found in large elongated structures (with individual area > 4 μm2; see also Panel D in ). Percentages for the number of large cells (> 4 μm2) and other determined parameters are listed in . Means ± SE for n = 3 biological replicates are shown. Asterisks indicate significant difference from the WT value (ANOVA, Tukey’s post-hoc test, p < 0.05).

To further probe potential defects in cell division, we determined the size distribution and morphology of cells by light microscopic analysis following methylene blue staining (). Strikingly, the recB1080 mutation resulted in the appearance of large elongated cell structures (with areas larger than 4 μm2, compared to the average WT cell size around 1 μm2) that were not seen in strains with intact RecB protein. In all strains of recB1080 background, about 15–20% of the total cell area, contributed by about 3–4% of the total number of cells, was found in these aberrant structures (). Aside from this effect, we found no significant differences between any strains regarding the average size of cell populations below 4 μm2 (Panel B in ).

To assess whether the appearance of large cells in recB1080-based strains significantly influences the OD600 profiles (), we also determined the titer of colony forming units (CFU/ml) during growth of WT and recB1080 ΔrecQ strains (). We found no significant differences between the strains regarding growth rate and amplitude (), indicating that the OD600 curves are not markedly influenced by the observed differences in cell size distribution.

In summary, the above experiments show that recQ and recB1080 mutations do not have systematic effects on cell culture growth under stress-free conditions. However, the recB1080 mutation, but not recQ defects alone, selectively impairs the segregation of cells. The effects of recB1080 and recQ mutations on cell growth and division are non-cumulative under stress-free conditions.

RecQ and RecB functional defects exert cumulative effects on interstrand DNA crosslink repair

Nitrofurantoin (NIT) causes the formation of interstrand DNA crosslinks and oxidative DNA damage, and also inhibits translation in bacterial cells [22,23]. The ΔrecQ and ΔrecB mutations have been shown to sensitize E. coli cells to this antibiotic [24]. Previously we showed that the helicase-competent RecQ* and RecQ-dH constructs, but not the helicase-deficient RecQ-dWH protein, are capable of compensating NIT sensitivity to the WT level [15]. In the present study we found that the recB1080 mutation causes NIT sensitization similar to that seen for ΔrecQ (). Importantly, in the recB1080 background, the ΔrecQ and recQ-dWH mutations (but not the dsDNA unwinding-competent recQ* and recQ-dH variants) caused significant further NIT sensitization (). These results show that both the normal functioning RecBCD and the DNA unwinding activity of RecQ are necessary for efficient repair of NIT-induced DNA damage, and the effects of the functional deficiency of the two repair enzymes are cumulative.

RecQ and RecB functional defects exert cumulative effects on the repair of nitrofurantoin-induced DNA damage.

(A) Dose-dependent NIT survival of E. coli strains (means ± SE), fitted by a standard dose-response model (). Symbols for low survival values at high NIT doses (all measured up to 10 μg/mL) are mostly obscured by those of other strains with similar values. (B) Fitted log LD50 values of strains. Means ± SE of individual best-fits to biological replicates are shown. Asterisks indicate significant difference from the WT value (ANOVA, Tukey’s post-hoc test, p < 0.05). Double asterisks indicate further significant difference from values with single asterisk. Multiple WT controls are shown alongside mutant values determined in the same experimental run. Data for non-recB1080 strains are from ref. [15]. Sample sizes and determined parameters are listed in .

Loss of RecQ function further sensitizes recB1080 cells to UV-induced lesions

Previously we and others found that mutations in the recQ gene do not affect the sensitivity of E. coli cells to UV irradiation [5, 15, 25, 26]. In the present study we found that, contrary to RecQ deletion, the recB1080 mutation drastically reduces the survival of E. coli cells upon UV radiation () in line with previous results [5,26]. Importantly, in the recB1080 background, the lack of RecQ DNA unwinding activity (ΔrecQ, recQ-dWH) caused further reduction of UV survival, whereas this effect was not seen in strains expressing unwinding-competent RecQ constructs (recQ*, recQ-dH) (). These results show that, in contrast to NIT sensitivity where the recB1080 and ΔrecQ mutations individually have effects of the same magnitude (), normal RecBCD functions are more important than RecQ functions for the survival of UV-induced DNA lesions (). Nevertheless, the contribution of RecQ’s DNA unwinding activity to survival of UV-induced damage becomes manifest in the recB1080 background ().

Loss of RecQ function further sensitizes recB1080 cells to UV-induced lesions.

(A) UV survival (mean ± SE) of E. coli strains. Means ± SE for n = 5 biological replicates are shown. Low survival values (< 10−4) for recB1080 recQ-dWH and recB1080 ΔrecQ strains at UV doses above 10 J/m2 could not be accurately determined. Due to this condition and the generally low UV survival values measured in the recB1080 background, the standard dose-response model () could not be applied. (B) Log relative survival values at 10 J/m2 UV irradiation (means ± SE). Asterisks indicate significant difference from the WT value (ANOVA, Tukey’s post-hoc test, p < 0.05). Double asterisks indicate further significant difference from each of the recB1080 recQ* and recB1080 recQ-dH strains (for recB1080 recQ-dWH), or from each of the recB1080, recB1080 recQ* and recB1080 recQ-dH strains (for recB1080 ΔrecQ). Multiple WT controls are shown alongside mutant values determined in the same experimental run. Data for non-recB1080 strains are from ref.[15].

RecQ defects compromise the survival of recB1080 cells under replication stress

To determine how RecBCD and RecQ functions contribute to cell growth under replication stress, we recorded the time-dependent OD600 profiles of E. coli strains in the presence of 100 mM hydroxyurea (HU) (). HU is known to deplete cellular dNTP pools and also to cause an increase in the intracellular concentration of reactive oxygen species (ROS) [27,28]. The OD600 profiles showed an initial increase even in the presence of HU, but a noticeable drop in OD600 was observed after 200 min in recB1080-based strains (). To assess the basis of this optical behavior, we determined the time-dependent survival of strains in 100 mM HU via monitoring CFU/ml counts, and applied a linear model for the time dependence of log-transformed relative survival values to aid statistical comparison between strains (). While both the WT and ΔrecQ strains showed stagnation over time, the recB1080 single mutant and recB1080 recQ* double mutants showed a significant time-dependent reduction in survival. Interestingly, the recB1080 recQ-dH mutant a showed survival profile that was restored to the WT level. The survival of the recB1080 recQ-dWH and recB1080 ΔrecQ double mutants was even lower than that of other strains, indicated by statistically significant differences in most pairwise comparisons (). Taken together, the data indicate that, in contrast to the loss of RecQ helicase function alone, the recB1080 mutation compromises cell survival under replication stress, and RecQ loss-of-function further sensitizes cells to replication stress in the recB1080 background.

Loss of RecQ function compromises the survival of recB1080 cells under replication stress.

(A) Time-dependent OD600 profiles (ln (OD600(t)/OD600(t0)), means ± SE) of indicated E. coli strains in the presence of 100 mM HU. (B) Time-dependent HU (100 mM) survival profiles of indicated E. coli strains, monitored via changes in CFU/ml (means ± SE). Solid lines show linear regression best-fits to the time dependence of log-transformed relative survival data. Results of statistical comparison of best-fit slopes are shown in .

RecQ helicase variants with enhanced processivity ameliorate RecB1080-induced impairment of recombination accuracy

The above results showed how recB1080 and/or various recQ mutations affect cell survival, division and growth under stress-free and genome-damaging stress conditions. To quantify how the same mutations affect the accuracy of HR, we applied the well-established Spi– λ phage assay that allows the determination of illegitimate recombination frequencies (IRF) [29]. Upon lytic induction, λ phage is normally excised from the genome of lysogenic E. coli strains by site-specific recombination. However, IR events produce phage variants that lack the red and gam genes and are therefore able to form plaques on P2 lysogenic E. coli strains (Spi−phenotype), thereby allowing their selection against WT phages.

We performed IRF measurements on strains harboring WT and mutant recQ variants in the recB1080 background under stress-free conditions, and synthesized the new results with our previous data on strains with WT recB background [15]. One-way ANOVA analysis (with Tukey’s post-hoc test) of data of all strains measured in the absence of UV irradiation showed significantly elevated IRF for the recQ-dWH and recB1080 single-mutant, as well as for the recB1080 recQ-dWH and recB1080 ΔrecQ double mutant strains, compared to the WT strain (). (The IRF of recB1080 recQ-dWH was also significantly elevated over that of recB1080 recQ*.) Importantly, however, the IRF of the recB1080 recQ* and recB1080 recQ-dH double mutant strains was not significantly elevated over that of the WT strain (). This result indicates that the recQ* and recQ-dH mutations, resulting in RecQ helicase variants with enhanced DNA unwinding processivity but reduced invasion disruptase activity [15], suppress the elevated UV-free IRF phenotype of the recB1080 mutation.

RecQ helicase variants with enhanced processivity ameliorate RecB1080-induced impairment of recombination accuracy.

UV-free IRF values of E. coli strains (mean ± SE) are shown. Asterisks indicate significant pairwise differences (ANOVA, Tukey’s post-hoc test, p < 0.05). Data for non-recB1080 strains are from ref. [15]. Determined parameters, sample sizes and statistical p values are listed in .

UV irradiation has been shown to cause a large increase in IRF [5, 8, 15]. Previously we found that, upon UV irradiation (50 J/m2), the recQ* and recQ-dH single mutant strains show significantly increased IRF as compared to the WT strain, suggesting that the enhanced DNA unwinding activity of the RecQ* and RecQ-dH proteins, in the absence of efficient invasion disruption, impairs HR accuracy under these conditions [15]. Unlike that for recQ mutants in WT recB background [15], the UV-induced IRF in recB1080-harboring strains could not be reliably determined due to the very low survival of these strains even at low UV doses (cf. ). A comparison of ΔrecQ and recB1080 mutation-induced effects on IRF found in the present study with those in earlier reports is provided in SI ( legend).

Taken together, the IRF data reveal that in the absence of genomic stress (i.e., no UV irradiation), the aberrant RecB1080 activity leads to increased IRF. Strikingly, this IRF-enhancing effect of RecB1080 is abolished in double mutants expressing RecQ variants with enhanced helicase but compromised disruptase activities (recB1080 recQ*, recB1080 recQ-dH; )—even though the same RecQ variants in WT recB background caused increased IRF under UV-induced genomic stress leading to more frequently stalled replication and increased DSB formation () [15]. These results point to a functional antagonism between the applied specific RecBCD and RecQ mechanistic alterations (see below).

Discussion

Recent biochemical/biophysical investigations of the DNA-restructuring activities of RecBCD and RecQ enzymes yielded important insights into their mechanisms of action, and precisely defined the mechanistic perturbations characteristic of their engineered variants [5, 15, 25, 26, 30]. Specifically, the enhanced helicase but deficient nuclease and RecA-loading activities of the RecB1080CD complex lead to the formation of abnormally long, partially RecA-filled 5’ and 3’ DNA overhangs [6,31] (upper row in ), while the enhanced (highly processive) DNA unwinding but compromised DNA invasion disruption and/or shuttling activities of the RecQ* and RecQ-dH variants lead to impaired suppression of IR events [15]. In the present study, we quantitatively tested the in vivo cellular effect of these specific perturbations and the functional interplay between the above-mentioned genome maintenance enzymes.

Functional fine-tuning between RecBCD and RecQ enzymes.

Upper row: DSB processing by the RecBCD-WT complex (RecBCD protein not shown) results in RecA-loaded 3’ DNA overhangs. In contrast, RecB1080CD produces aberrant long, partially RecA-filled 5’ and 3’ overhangs. Middle row: The complex DNA unwinding, directed invasion disruption and shuttling activities of the RecQ-WT enzyme (triangle) enable efficient disruption of IR invasions (green) in the recBCD-WT, but low efficiency (yellow) in the recBCD background. Lower row: The enhanced (highly processive) DNA unwinding, but compromised directed invasion disruption and/or shuttling activities of RecQ* and RecQ-dH enzymes (triangle) are inefficient (yellow) for IR disruption in the recBCD-WT background. However, these activities become efficient (green) in the recBCD background where HR is more likely proceed through joint DNA molecules of aberrant structure.

Our data support that recQ mutations do not systematically affect E. coli cell growth under stress-free conditions, either in the recB-WT or in the recB1080 background (). However, in line with earlier studies pointing to the role of RecBCD in late stages of replication and in the induction of DNA damage-dependent SOS response that halts cell division [32, 33], we detected significant filamentation in recB1080 cells, which was unaffected by their recQ genotype (). The recB1080 mutation was shown to cause a marked increase in the level of SOS induction even in the absence of genotoxic agents, whereas the ΔrecQ knockout did not influence SOS induction in either the recB-WT or the recB1080 background [30]. Taken together, these results underscore the linkage between elevated SOS induction and filamentation (cell division defects) in recB1080 cells. Thus, the resection of dsDNA ends, performed mainly by RecBCD in WT cells, appears to be catalyzed by the aberrant RecB1080CD enzyme in conjunction with the RecJ exonuclease in recB1080 cells, while RecQ may only have a minor role in DSB resection. Accordingly, unlike ΔrecQ, the recJ knockout led to significantly lowered SOS induction in recB1080 cells [30]. recB1080-induced SOS response and filamentation may reflect multiple possible underlying mechanisms. (i) Aberrant dsDNA end processing in recB1080 cells may lead to aberrant RecA.ssDNA nucleoprotein structures that cannot initiate HR efficiently, and the persistence of these structures may trigger SOS response (30). (ii) The formation of aberrantly joined DNA molecules during HR-based replication restart, DSB repair or in late stages of replication, which cannot be resolved by RecQ helicase, may halt chromosome segregation and lead to filamentation. (iii) Due to its enhanced helicase activity, RecB1080CD may excessively unwind DNA structures, thereby facilitating aberrant DNA processing and, in turn, DNA damage.

We monitored cell survival or growth in the presence of three different genotoxic agents (NIT, UV irradiation, HU) to probe the interplay between RecBCD and RecQ activities during DNA damage repair (Figs –). NIT addition mostly causes interstrand DNA crosslinks, and also generates oxidative DNA damage [22,23]. UV irradiation predominantly leads to the formation of intrastrand DNA crosslinks and, to a lesser extent, 8-oxoguanine and oxidized pyrimidines [34]. Inter- and intrastrand crosslinks caused by NIT and UV, respectively, can both frequently lead to replication stalling and DSB formation [35,36]. HU inhibits ribonucleotide reductase and thus depletes cellular dNTP pools, causing general replicational stress. In addition, HU increases membrane permeability, leading to increased intracellular concentration of reactive oxygen species, causing additional DNA damage [27, 28]. Interestingly, the relative severity of the effects of recB1080 mutation and RecQ DNA unwinding deficiency (ΔrecQ, recQ-dWH) were different in the case of different genome-damaging agents (Figs –). Both of the recB1080 and ΔrecQ (or recQ-dWH) single mutations significantly sensitized cells to NIT (). However, only recB1080, but not ΔrecQ or recQ-dWH alone, impaired the UV survival of cells (, in line with ref. [5]). Similarly, unlike the ΔrecQ mutation, the recB1080 mutation caused reduced survival in HU, and this effect was exacerbated by recQ loss-of-function in the recB1080 background ().

Two important general conclusions emerge from the above experiments. First, the effects of recB1080 and RecQ deficiency (ΔrecQ or recQ-dWH) were mostly cumulative (i.e. double mutant phenotypes were generally more severe than any single mutant phenotype), showing that both the RecBCD and RecQ enzymes significantly contribute to various forms of DNA repair. Second, the RecQ* and RecQ-dH constructs fully compensated for RecQ-WT function in most cases, showing that shuttling during DNA unwinding and the directed invasion disruptase function (present in RecQ-WT, absent in RecQ* and RecQ-dH) [15] are dispensable in terms of the efficient repair of DNA lesions, either in the recBCD-WT or the recB1080 background.

Interestingly, while ΔrecQ reduces UV survival in the recB1080 (; in line with ref. [5]) and also in the ΔrecBCD [37] background, recQ deficiency has no effect on UV survival in either recBCD-WT () or recB1080 recD [5] backgrounds where the RecA loading capability of the RecBCD complex is retained. Together these findings show that RecQ functions confer a significant advantage in DNA lesion repair only when the RecA loading function of RecBCD is compromised. These findings are in line with the genome instability phenotypes of RecQ helicase deficiencies in eukaryotes where, in the absence of a RecBCD homolog, the DSB resection and recombinase loading activities are not coupled in a single enzyme complex [12,38,39].

The ΔrecB mutation was found to significantly sensitize E. coli cells to HU in a ruvABC-WT dependent manner (27). The authors proposed that, upon replication fork stalling, the regression of the fork leads to the formation of a single Holliday junction in the absence of significant resection (ΔrecB), and this structure is then converted to a DSB by RuvC. In line with these earlier results, we found reduced HU survival for recB1080, as compared to the WT strain (further exacerbated by recQ loss-of-function, ), indicating that compromised resection in the recB1080 mutant strain leads to enhanced DSB formation.

Besides probing the efficiency of HR-based DNA repair via genotoxic agents, we also tested the interplay between RecBCD and RecQ activities in HR quality control, measured through the frequency of IR [29]. IR events have been proposed to occur through local base pairing between short homologous regions of non-allelic DNA sequences [8,29]. In line with earlier studies [5], we found that the recB1080 mutation alone leads to increased IRF in the absence of UV irradiation (), presumably because at free DNA ends the RecB1080CD enzyme generates long, partially RecA-filled 5’ and 3’ ssDNA overhangs that are prone to form IR initiations (). Intriguingly, however, we found that in the recB1080 recQ* and recB1080 recQ-dH double mutants, the IRF was not significantly higher than that in the WT strain (Figs ; see also below). The specific suppression of RecB1080-induced IR by RecQ* and RecQ-dH is remarkable in the light of the finding that the activities of the same RecQ variants lead to increased IRF (over WT) when DSB formation is facilitated by UV irradiation in the recBCD-WT background () [15]. Besides compromised IR suppression (), these highly processive RecQ variants are likely to support enhanced DSB resection and RecA loading under these conditions.

Taken together, the above findings reveal a precise functional fine-tuning between HR initiation and quality control systems of WT E. coli cells, and provide an in vivo context for the characterized complex biochemical activities of the WT and mutant RecBCD and RecQ proteins () [5,15,20,26]. In the recBCD-WT background, the directed invasion disruption and shuttling activities of RecQ-WT enable efficient and selective disruption of IR invasions, whereas the enhanced DNA unwinding but compromised directed disruption and shuttling activities of RecQ* and RecQ-dH are counter-productive for IR suppression (). However, RecQ-WT is less productive in disrupting IR events resulting from the aberrant long overhangs produced by RecB1080CD, while the aberrant RecQ* and RecQ-dH activities become productive for IR suppression in this context (). RecBCD has thus been evolutionarily optimized for efficient resection/recombinase loading under normal conditions (low concentration of DNA ends), whereas the complex activities of RecQ are optimized for the suppression of harmful consequences of genomic stress in WT cells (imprecise HR at elevated DNA breakage/replication fork stalling). The fact that RecQ aberrations exert only mild effects on cell survival (short-term effect) but significantly compromise HR accuracy (long-term effect) is in accordance with the genome instability and cancer predisposition symptoms of human RecQ helicase deficiencies including Bloom’s, Werner’s, Rothmund-Thompson and RAPADILINO syndromes [13,39,40].

Methods

Bacterial strains and modifications

recQ and recB mutant strains (derivatives of MG1655 denoted as WT) were constructed using the Red/ET Recombination Kit (GeneBridges, Heidelberg, Germany) (). ΔrecQ is a recQ knock-out mutant. In place of the WT recQ gene, the recQ-dH and recQ-dWH strains harbor recQ genes in which the regions 1569–1830 bp (resulting termination at aa 523) and 1239–1830 bp (resulting in termination at aa 413), respectively, have been deleted. In place of the respective WT enzymes, the recQ* and recB1080 strains express the proteins harboring the Y555A and D1080A substitution, respectively. Point mutations were generated by QuikChange II XL site-directed mutagenesis (Agilent Technologies).

Growth curves

OD600-based growth curves were recorded by inoculating aliquots of overnight E. coli clonal cultures into Luria-Bertani (LB) medium to a volume of 200 μl and OD600 = 0.01, followed by incubation at 37°C with aeration and monitoring their OD600 value in 5-min intervals in a Synergy H4 Hybrid Multi-Mode Microplate Reader instrument. Time-dependent changes in CFU/ml were followed by plating 5-μl aliquots of serial dilutions from cultures grown as described above. CFUs were counted after overnight growth on LB plates at 37°C. Curves were analyzed based on the modified Gompertz model () [21].

Cell staining and microscopy

For visualization, 10 μl of clonal cell cultures (OD600 = 0.7) were spread onto microscope slides, fixed in flame, stained with methylene blue (0.3 g/L) for five minutes, washed with distilled water and dried. Images were taken using a Nikon Japan Planflour microscope with objective 100x/1.30 Oil DIC H/N2. For cell area analysis, 10 images were randomly selected for each biological replicate (3 replicates for each genotype). Cell aggregates and overlapping cells were observed in regions of high cell density and in those around long filaments (present in all strains with recB1080 background, cf. A panel in ). Images were converted to 8-bit grayscale images using ImageJ. Background correction of images was performed using the built-in IsoData automated threshold detection algorithm. Particle detection and area measurements were performed on background-corrected images using the built-in particle detection algorithm of ImageJ (0–infinity area detection, roundness range 0–1). Particles at image edges, as well as visually identified aggregates and overlapping cells were excluded from statistical analysis. Thus, the calculated cell area fractions occupied by large cells () are slightly underestimated for strains where these structures appeared (i.e., for all strains with recB1080 background). However, this condition does not affect the conclusion that the large cell fractions are significantly higher in each of these strains than in those with WT recB background.

UV and NIT survival assays

In UV irradiation survival assays, cells were grown clonally to 4 x 108 cells/ml at 37°C in LB medium. Aliquots of cultures were diluted, spread onto LB plates and irradiated with different UV doses (at 254 nm) using an UVP CX-2000 crosslinker. Following overnight growth at 37°C, colonies were counted to quantify survival. NIT survival assays were performed as described for UV survival assays, except that UV irradiation was omitted and cells were spread and grown on LB plates containing different concentrations of NIT.

Hydroxyurea survival assays

These assays were performed as described for growth curve monitoring, except that 100 mM HU was included in the medium.

λ Spi- phage assay

The frequency of illegitimate recombination of λ phage was measured using the λ Spi−assay developed by Ikeda et al. [29]. For lysogenization, WT and mutant E. coli cells were grown to 7 x 108 cells/ml at 30°C in T-broth (1 w/v % bacto tryptone (Difco), 0.5 w/v % NaCl, 10 mM MgSO4, 0.2 w/v % maltose). λcI857 phage, isolated from HI1165 cells after heat shock induction of the lytic phase, was added to the cells (at multiplicity of infection = 2) and incubated for 30 min at 30°C. Cells were then spread onto T-plates (T-broth with 1.2 w/v % agar) and incubated at 30°C overnight. Single colonies were picked, amplified and checked for lysogeny by heat shock induction of the lytic phase (42°C for 15 min, followed by shaking (250 rpm) at 37°C and monitoring decrease in OD600). Three independent lysogens were generated for each genotype.

To analyze the formation of λ Spi−phages, lysogenic strains () were grown to 4 x 108 cells/ml in λYP broth (1 w/v % bacto tryptone (Difco), 0.1 w/v % yeast extract (Oxoid), 0.25 w/v % NaCl, 0.15 w/v % NaHPO4, 0.018 w/v % MgSO4). To enter the lytic phase, λcI857 prophage was heat induced by incubation of the culture at 42°C for 15 min. The culture was then incubated at 37°C for 2 hours with aeration, and centrifuged to isolate the phages from cell debris. The total phage titer and the titer of λ Spi−phages was determined by phage infection of YmeI and WL95 (P2 lysogen) E. coli cells, respectively (). Mixtures of (diluted) aliquots of the phage suspension and YmeI or WL95 cells were incubated at room temperature for 30 min, mixed with λ top agar (1 w/v % bacto tryptone, 0.5 w/v % NaCl, 0.4 w/v % agar; for YmeI) or λ trypticase top agar (1 w/v % trypticase peptone (Difco), 0.5 w/v % NaCl, 0.4 w/v % agar; for WL95) and spread onto 1.2% agar plates (otherwise identical to the respective top agar). Plaques were counted after overnight incubation at 37°C. The burst size, calculated by dividing the titer of total phages by the titer of infective centers, was in the range of 18–145. The frequency of illegitimate recombination (IRF) was obtained by dividing the titer of λ Spi−phages by the titer of total phage.

Data analysis and statistics

Mean ± SE values are reported in this study for the specified number of biological replicates (individual clones). Statistical significance (p < 0.05) was determined by ANOVA followed by the Tukey post-hoc test. For fitted parameters, statistical analysis was performed on the collection of best-fit values to individual datasets. UV survival and IRF values showed log-normal distribution, and thus the statistical analysis was performed on log-transformed values. Fitting and data analysis was performed using OriginLab 8.0 (Microcal Corp.).

Modified Gompertz equation for bacterial growth curves (Fig 1A).

(DOCX)

Click here for additional data file.

Standard dose-response equation used to fit NIT survival (Fig 2A).

(DOCX)

Click here for additional data file.

Cell growth parameters determined in the absence of genomic stress.

Parameters (panel A, growth amplitude (A); panel B, growth rate (μM)) were determined from best-fits to individual growth curves, based on the modified Gompertz equation (). Means ± SE for n = 4–10 are indicated. Results of ANOVA analysis are summarized in . Parameters determined from CFU counts () were A = 3.18 ± 0.23 and μM = 0.013 ± 0.004 min-1 for WT, and A = 3.65 ± 0.33 and μM = 0.025 ± 0.013 min-1 for recB1080 ΔrecQ (differences between strains are not significant).

(TIF)

Click here for additional data file.

Cell size distribution of E. coli strains.

(A) Example images of methylene blue stained cells.

(B-C) Cell size distribution histograms (area < 4 μm2) for E. coli strains.

(D) Scatter graph of cells with area > 4 μm2.

(TIF)

Click here for additional data file.

E. coli strains used in this study.

(PDF)

Click here for additional data file.

Statistical p values of growth parameter comparisons.

Shown are probability (p) values resulting from one-way ANOVA analysis (Tukey’s post-hoc test) of A and μm values () obtained in best-fits to individual growth curves (averaged data shown in ). Significant differences (p < 0.05) are highlighted in red.

(PDF)

Click here for additional data file.

Morphological and NIT survival parameters.

Significant difference from WT values is indicated by asterisks (one-way ANOVA, Tukey post-hoc test, p < 0.05). Double asterisk indicates further significant difference from values with single asterisk. Means ± SE values are shown for sample sizes (n, biological replicates (individual clones)) specified below. N.a., not applicable.

a Gaussian means are shown for cell populations (± SE for n = 3 biological replicates) with individual cell area < 4 μm2, with the total number of analyzed cells in parentheses.

b Specified as % of total cell area (± SE for n = 3 biological replicates) in cells with individual area > 4 μm2. Values in parentheses show the percentages for the number of large cells (> 4 μm2) within the assessed population.

c Determined using . (p (Hill slope coefficient) values could not be robustly determined.) Statistics for individual best-fits to biological replicates are shown, with n in parentheses. Values for strains not harboring the recB1080 mutation are from ref. [2] (S1 Refs).

d WT control run in parallel with strains of recB1080 background.

(PDF)

Click here for additional data file.

Statistical p values for comparisons between HU survival profiles.

Shown are probability (p) values resulting from one-way ANOVA analysis (Tukey’s post-hoc test) for the linear slopes of time-dependent log-transformed HU survival profiles (). Significant differences (p < 0.05) are highlighted in red.

(PDF)

Click here for additional data file.

Illegitimate recombination frequencies (IRF).

Values were determined by using the λ Spi−assay (3). Means ± SE are shown, with sample numbers (biological replicates) in parentheses. Results of one-way ANOVA analysis of UV-free log IRF values are summarized in . Values for non-recB1080 strains are from ref. [2] (). The UV-induced IRF values for recQ* and recQ-dH were significantly elevated (p < 0.05) over that of the WT strain [2] ().

N. d., not determined. Unlike those for recQ mutants in WT recB background [2] (), the UV-induced IRF values in recB1080-harboring strains could not be reliably determined due to the very low survival of these strains even at low UV doses (cf. ).

The magnitude and the cumulative nature of the IRF increase elicited by the ΔrecQ mutation and UV irradiation (~10-100-fold for each effect, ~100-1000-fold for the combined effect) was similar in our studies (this work and (2)) to those determined in earlier works [2-7] (). However, in the absence of UV irradiation, we found a larger effect for the recB1080 mutation than that reported in ref. [7] (), and thus we observed no further IRF elevation in recB1080 ΔrecQ. We note that the mean IRF values determined in our studies are generally lower than those reported in the mentioned studies. This difference probably originates mostly from the fact that we applied the statistical analysis to log-transformed IRF values because the original values showed a log-normal spread. The applicability of this transformation is further substantiated by the fact that the average values collected from different studies themselves show a logarithmic spread. We also used higher sample numbers (n = 9–19) than those in previous studies (n = 2–4).

(PDF)

Click here for additional data file.

Statistical p values of illegitimate recombination frequencies (log IRF values).

Shown are probability (p) values resulting from one-way ANOVA analysis (Tukey’s post-hoc test) of log IRF values determined in the absence of UV irradiation (). Significant differences (p < 0.05) are highlighted in red.

(PDF)

Click here for additional data file.

Supporting information references.

(DOCX)

Click here for additional data file.