|Home | About | Journals | Submit | Contact Us | Français|
Escherichia coli initiates the SOS response when single-stranded DNA (ssDNA) produced by DNA damage is bound by RecA and forms a RecA-DNA filament. recA SOS constitutive [recA(Con)] mutants induce the SOS response in the absence of DNA damage. It has been proposed that recA(Con) mutants bind to ssDNA at replication forks, although the specific mechanism is unknown. Previously, it had been shown that recA4142(F217Y), a novel recA(Con) mutant, was dependent on RecBCD for its high SOS constitutive [SOS(Con)] expression. This was presumably because RecA4142 was loaded at a double-strand end (DSE) of DNA. Herein, it is shown that recA4142 SOS(Con) expression is additionally dependent on ruvAB (replication fork reversal [RFR] activity only) and recJ (5′→3′ exonuclease), xonA (3′→5′ exonuclease) and partially dependent on recQ (helicase). Lastly, sbcCD mutations (Mre11/Rad50 homolog) in recA4142 strains caused full SOS(Con) expression in an ruvAB-, recBCD-, recJ-, and xonA-independent manner. It is hypothesized that RuvAB catalyzes RFR, RecJ and XonA blunt the DSE (created by the RFR), and then RecBCD loads RecA4142 onto this end to produce SOS(Con) expression. In sbcCD mutants, RecA4142 can bind other DNA substrates by itself that are normally degraded by the SbcCD nuclease.
The SOS response is a coordinated response of Escherichia coli at the level of transcription to DNA damage (10, 18, 26). RecA initiates this response by binding to single-stranded DNA (ssDNA) produced by DNA damage and serving as an allosteric effector for auto-proteolysis of the LexA transcriptional repressor. The transcription of at least 40 genes is increased during the SOS response (16). Some of the induced genes include recA, ruvAB, dinI, and recX. The RecA protein also plays a central role in recombinational repair and homologous recombination (8, 11, 22). It participates in all three processes through its ability to polymerize on ssDNA to create a RecA-DNA filament.
Several proteins are known to either help RecA load onto different DNA substrates or regulate the stability of the RecA-DNA filament (reviewed in reference 12). One of the loading complexes is RecBCD. This complex has the ability to load onto a double-strand end (DSE) of DNA. It processes the DNA using helicase, exonuclease, and Chi recognition activities to produce a region of ssDNA with a 3′ end. RecBCD then loads RecA onto this ssDNA, creating a RecA-DNA filament (reviewed in reference 14). Another complex, RecFOR, loads RecA onto ssDNA coated with single-stranded DNA binding protein (SSB) at a gapped DNA substrate (36, 38). After production, DinI stabilizes and RecX destabilizes RecA filaments (29, 37). RecFOR antagonizes the destabilization activity of RecX both in vivo and in vitro (28, 30).
There are certain mutants of recA that turn on SOS expression in the absence of external DNA damage. These are called recA constitutive [recA(Con)] mutants (reviewed in reference 33). While the specific mechanism of how this type of mutant induces SOS is presently unknown, it is thought that it occurs through RecA binding to ssDNA on the lagging strand at the replication fork. To begin to test this idea, the dependence of proteins known to be important for RecA loading and stability was tested for two recA(Con) mutants. One mutant (recA730), showed no dependence for any gene tested, while a second mutant [recA4142(F217Y)] was dependent on recBCD, recFOR, and the type of medium in which the cells were grown (28). For recA4142, it was hypothesized that the role of RecFOR was to regulate the activity of RecX to destabilize the filament and that the role of RecBCD was to load RecA4142 onto a DSE.
The observation that the loading of RecA4142 in log-phase cells is recBCD dependent and that almost all cells in the population have constitutive SOS [SOS(Con)] expression suggests that RecA4142 may not simply bind to ssDNA on the lagging strand at the replication fork. One can suggest several different hypotheses to explain this observation. One is that almost all cells have a DSE or DSEs where RecBCD can load RecA4142 to cause SOS(Con) expression. These could be formed at a stalled replication fork where the nascent leading and lagging strands anneal in a process called replication fork reversal (RFR) (40). A reversed replication fork is very similar, if not identical, in structure to a recombinational intermediate called a Holliday junction (27). Thus, RFR is often catalyzed by the RuvAB proteins [as in rep, holD, and dnaE(Ts) mutants], but can also occur independently of RuvAB in some special mutants [i.e., priA, dnaN(Ts), and dnaB(Ts) mutants] (reviewed in references 35, 39, and 40). Another hallmark of RFR is that the reaction is ruvC independent. This is because only the ability to form and branch migrate the Holliday junctions is required for RFR (not the ability to resolve them). Another hypothesis is that RecBCD has a yet undiscovered activity that allows RecA loading at substrates other than DSEs. While this seems unlikely, it is a formal possibility and may be particular for RecA4142. A third possibility is that is double-strand breaks (DSBs) occur in every cell and that RecBCD uses its well established activities for loading RecA4142 at these sites. This DSB (vis a vis one produced by an I-sceI cut ), however, must not absolutely require RecBCD for repair because recA4142 recB mutants are viable.
One way to test between these three ideas is to ask what gene products are necessary for SOS(Con) expression in a recA4142 mutant. The first hypothesis (loading during RFR) would be ruvAB dependent and ruvC independent, while the second two would be ruvAB independent. In this report, it is shown that SOS(Con) expression is dependent on RuvAB (and only the activity of RuvAB needed to reverse replication forks) and not ruvC. This piece of data better supports the first model, but does not exclude the latter two. SOS(Con) expression is also shown to be dependent on two exonucleases, RecJ and XonA. To accommodate these additional dependencies, it is further proposed that RecJ and XonA process the ends so that RecBCD can load RecA4142 onto RuvAB reversed forks. Lastly, it is shown that the absence of the SbcCD nuclease allows SOS(Con) expression, even when RecBCD, RuvAB, RecJ, and RecF are absent. It is proposed that SbcCD nuclease degrades some substrate that RecA4142 can bind in the absence of traditional RecA loading factors.
All bacterial strains used in this work are derivatives of E. coli K-12 and are described in Table Table1.1. The protocol for P1 transduction has been described elsewhere (44). All P1 transductions were selected on 2% agar plates containing either minimal or rich media. Where appropriate, plates also contained the following antibiotics at these final concentrations: tetracycline, 10 μg ml−1; chloramphenicol, 25 μg ml−1; or kanamycin, 50 μg ml−1. All transductants were purified on the same type of medium on which they were selected.
Cells were prepared for microscopy and other measurements as described in detail with examples elsewhere (28, 31). Cultures were grown in 56/2 minimal medium (44) until mid-log phase (optical density at 600 nm [OD600] of 0.3 to 0.4). Cells were concentrated 10-fold in 56/2 buffer. Approximately 3 to 5 μl was loaded onto fresh agarose pads, and a coverslip was applied. The agarose pads were prepared using a protocol from P. Levin (25). Microscopy was carried out by using an epifluorescent Nikon E600 microscope. An ORCA-ER-cooled charge-coupled device (CCD) camera (Hamamatsu) and Openlabs software (Improvision) were used for all image acquisition. The exposure time was 100 to 250 ms. Approximately nine fields (three on three different days) containing calibration beads were photographed. A phase-contrast image and a fluorescent image of each field were taken. The softwares Openlab 5.0 and Volocity 4.0 (Improvision, Inc.) were used to measure the amount of fluorescence and cell size in individual cells. Calibration of the fluorescence intensity was set by calibration beads [InSpeck Green (505/515) microscope image intensity calibration kit 2.5 μm I-7219 from Molecular Probes]. The relative fluorescence intensity (RFI) value of an individual cell is calculated by dividing the average calibrated pixel value of a particular cell by the average calibrated pixel value of a strain containing Δattλ::sulAp Ωgfp-mut2 (SS996). The RFI values of the population of cells from all three experiments (typically 1,000 to 3,000 cells) are combined and binned according to their RFI. The percentage of cells with a particular RFI is calculated and plotted. The average RFI for each experiment is also calculated. The average for the three experiments and their uncertainties are reported next to the plots in the figures.
In this work, all strains contain a sulAp-gfp reporter gene inserted at the attλ site. Justification of the use of this construct as a reporter for SOS expression as well as how the measurements are done and reported have been explained in detail elsewhere (28, 31).
To begin to test between the three models mentioned above, ruvAB6203::tet was combined with recA4142. Note that all recA4142 strains discussed here also have recAo1403. This is necessary so that all cells in a population have high levels of SOS(Con) expression (28). Figure Figure11 shows that strains missing RuvAB revealed a large decrease in SOS(Con) expression, suggesting that RuvAB is required for SOS(Con) expression. This result is consistent with the first hypothesis suggesting replication fork reversal as a mechanism to produce the DSE. A further test of this model is that SOS(Con) expression should also be ruvC independent. Figure Figure11 shows that a ruvC mutation did not affect the level of SOS(Con) expression. These observations support the hypothesis that RFR may be a mechanism important for high levels of SOS(Con) expression in recA4142 mutants. When ruvABC::cat and ruvA60::Tn10 were added to the recA4142 mutants, the double mutants revealed similar results to when ruvAB6203::tet was added (data not shown).
To further test if RuvAB and its ability to catalyze RFR were important for high SOS(Con) expression, two additional tests were performed. The first was to test if the addition of pGB2-RuvAB to the ruvAB mutant would restore a high level of SOS(Con) expression. Figure Figure22 shows that this plasmid complemented the ability of the strain to produce high levels of SOS(Con) expression (Fig. (Fig.2A)2A) and the UV sensitivity (Fig. (Fig.2B)2B) of the strain as compared to that of the pGB2 vector-only control. The second test used a novel ruvB(H198Y) mutant that is deficient in RFR but maintains the ability to participate in recombination and DNA repair (3, 24). If RFR is responsible for high SOS(Con) expression in the recA4142 mutant, then addition of the plasmid should not complement the decrease in SOS(Con) expression, but it should complement the increase in UV sensitivity. If RFR were not the required activity for RuvAB to produce high levels of SOS(Con) expression, then one would expect to see full complementation of both the SOS(Con) expression and UV sensitivity. Figures 2A and B show that the addition of pGB2-RuvAB(H198Y) (24) complements the UV sensitivity of the strain, but not the low levels of SOS(Con) expression. From these results, it is concluded that RuvAB's ability to catalyze RFR is required for high levels of SOS(Con) expression in the recA4142 mutant.
Other work suggests that the RecG helicase may also catalyze replication fork reversal (32, 42). To test if RecG was required for high levels of SOS(Con) expression in the recA4142 mutant, recG6200::tet was combined with recA4142. Figure Figure11 shows that this mutation did not change the level of SOS(Con) expression in the recA4142 mutant. From this, we conclude that RecG is not required for high levels of SOS(Con) expression in the recA4142 mutant.
As the data presented above support the first model that DSEs may be generated during RFR, other experiments were attempted to further define other gene products that may be important in the loading of RecA4142 by RecBCD at a DSE produced during RuvAB-catalyzed RFR.
RecBCD loads preferentially onto fairly blunt DSEs of DNA (43, 45). Since the DSE produced during RFR by the annealing of the nascent leading and lagging stands of DNA may not be blunt (depending on where the DNA polymerases stop), it is possible that the ends produced may need to be processed before RecBCD can load RecA. This processing may be accomplished by exonucleases. It had been shown previously that xthA, xseAB, xni, tatD, and exoX encode proteins whose activities degrade DNA substrates used by RecBCD to load RecA (6, 28). Therefore, it was not expected that any of these enzymes would be required for SOS(Con) expression in the recA4142 mutant. However, two other exonucleases, RecJ (5′→3′ exo) and SbcB (also called XonA) (3′→5′ exo), have been shown not to degrade RecBCD/RecA substrates using RecA-green fluorescent protein (GFP) focus formation as an assay (6, 7). Therefore, it was possible that either or both of these proteins might be required for blunting the emerging duplex during RFR so that RecBCD could load RecA4142 to produce SOS(Con) expression. To test this, recJ and sbcB null mutations were combined with recA4142. Figure Figure11 shows that recJ and sbcB mutations individually decreased SOS(Con) expression dramatically. Expression of recJ from a plasmid (courtesy of S. T. Lovett) in the recJ mutant and sbcB from a plasmid in the sbcB mutant showed complete reversal of the phenotype (data not shown). This suggests that both of these exonucleases are needed to blunt the ends so that RecBCD can load onto the DSE.
As mentioned above, it had been shown that recFOR mutations, like recBCD mutations, decrease SOS(Con) expression in recA4142 (28). The reason for the decrease in the recFOR mutant, however, was due to an inability to antagonize RecX activity since a recX mutation could suppress the decrease seen in the recFOR mutant (but not the recBCD mutant). Since recJ mutations decrease SOS(Con) expression as much as recFOR or recBCD mutations and RecJ often acts in a pathway with RecFOR, it was possible that RecJ participated in antagonizing RecX like RecFOR. To test this, recX and recJ mutations were combined in a recA4142 strain. The recX recJ strain showed no increase in SOS(Con) expression relative to the recJ strain alone (Fig. (Fig.1).1). It is concluded that RecJ is not involved in antagonizing RecX activity and is likely to help process the DNA substrate required by RecBCD to produce SOS(Con) expression in a recA4142 strain.
The RecJ exonuclease and RecQ helicase are often thought to interact when processing DNA for the loading of RecA (9, 19). It was therefore of interest to test if RecQ is also required for high levels of SOS(Con) expression. To test this, recQ6209::tet was combined with recA4142. Figure Figure11 shows that the absence of RecQ leads to a 30% decrease in SOS(Con) expression. This suggests that RecQ plays a minor role in RFR that leads to the loading of RecA4142 to produce SOS(Con) expression.
The combination of sbcB and sbcC null mutations is known to suppress the loss of RecBCD enzyme and load the RecA enzyme via the RecFOR pathway of recombination (4). This suggests that sbcBC mutations may restore SOS(Con) expression in the recA4142 recBCD strain. To begin to test this hypothesis, the recA4142 recBCD sbcBC strain was constructed. Figure Figure33 shows that all cells have a high level of SOS(Con) expression. This supports the initial hypothesis that the RecFOR pathway genes could load RecA4142 during RFR to produce SOS(Con) expression. If true, then mutations in ruvAB or recF should decrease SOS(Con) expression in the recA4142 recBCD sbcBC mutant. Note that recF mutations should cause the decrease for two reasons: its ability to load RecA and its ability to antagonize RecX. Contrary to expectation, SOS(Con) expression remained high in both the ruvAB and recF4115 derivatives (Fig. (Fig.3).3). Control experiments were performed to show that the ruvAB and recF4115 mutations had indeed been introduced into the strain by the extreme UV sensitivity they cause (data not shown). recO and recR derivatives of both the recBCD sbcBC mutant and the sbcBC mutant were constructed. The recR derivatives behaved very similarly to the recF mutants. The recO mutants showed slightly less SOS(Con) expression than the recFR mutants. Still, about 80% of the population had high levels of SOS(Con) expression (data not shown). The reason for this small difference is not clear.
To explain this, it was hypothesized that the sbcC null mutation suppressed the absence of sbcB and or recBCD for SOS(Con) expression in the recA4142 strain. If true, then when sbcC null mutations are added to either sbcB or recBCD single null mutants, the double-mutant strain should have high SOS(Con) expression. Figure Figure33 shows all cells in sbcB sbcC and recBCD sbcC mutants have a high level of SOS(Con) expression. Furthermore, it is shown that addition of an sbcC null mutation to a recA4142 mutant with either an ruvAB, recF, or recJ mutation also restores the high levels of SOS(Con) expression (Fig. (Fig.33).
Since SbcC forms a nuclease complex with SbcD, it was tested if an sbcD null mutation has the same ability as the sbcC mutation to suppress the requirement for sbcB (xonA), recBCD, ruvAB, recF, and recJ mutations for SOS(Con) expression in a recA4142 mutant. All sbcD mutants behaved as their sbcC counterparts (data not shown). Lastly, as expected, the ability of the sbcC mutation to suppress the low levels of SOS(Con) expression in the recA4142 sbcC recJ double mutant is recessive to wild type as the addition of a plasmid with sbcCD (courtesy of D. Leach) returns the strain back to low levels of SOS(Con) expression, as seen for a recA4142 recJ mutant (data not shown).
It is tentatively concluded that the SbcCD nuclease degrades a DNA substrate that RecA4142 could use to provide SOS(Con) expression. It does not require any of the standard RecA loading enzymes (i.e., RecBCD or RecFOR) to load onto this substrate to yield SOS(Con) expression.
This paper provides a mechanistic study of the way one particular recA(Con) mutant, the recA4142 strain, causes SOS(Con) expression. This is important because previously it was assumed that all recA(Con) mutants behaved similarly, with the mutant recA protein binding to ssDNA on the lagging strand template at a replication fork. This was probably due to observations with one well-studied mutant, RecA730, that could compete better in vitro with SSB for ssDNA than the wild type (15, 23). It is now well established that recA(Con) mutants can have different mechanisms for SOS(Con) expression and need not be confined to binding ssDNA on the lagging strand of a replication fork, although that may be true for some mutants. Of the three models proposed at the beginning of this study for how RecA4142 causes RecBCD-dependent SOS(Con) expression, this report supports the first model: that RecBCD loads RecA4142 at a reversed replication fork. The data that SOS(Con) expression is dependent on RuvAB, but not RuvC, and that pRuvAB(H184Y) is able to restore UV resistance to a ruvAB mutant, but not its ability to produce SOS(Con) expression, are consistent with this model and not the other two. No experiments were done here, however, that excluded the latter two models. It was further shown that SOS(Con) expression in the recA4142 mutant was xonA and recJ dependent. To accommodate these additional observations, the model (Fig. (Fig.4)4) was expanded to suggest that these exonucleases blunt the end of the newly formed duplex of DNA so that RecBCD can load RecA.
Figure Figure44 suggests that there may be two ways that RecA4142 can produce a stable RecA-DNA filament that can affect LexA autoproteolysis. The first is that it binds to ssDNA produced by Chi-stimulated activities. If this complex is long lived, it may be sufficient to drive down the concentration of LexA. However, given the fact a homologous duplex is in close proximity, it is likely that the RecA-ssDNA will strand invade the homologous duplex, producing a second structure where RecA may be bound to dsDNA. Since dsDNA can also act as a cofactor for LexA cleavage in vitro (M. Cox, personal communication), it is possible that this structure is stable in a recA4142 mutant and could continue catalyzing LexA cleavage until removed, possibly by helicases or by the next round of DNA replication. It is envisioned that wild-type RecA will go through a similar action but will not be stable on the dsDNA after the recombination event. This is the hypothesized difference between RecA4142 and the wild type and the reason why the wild type does not produce SOS expression constitutively in cells where RFR may occur.
Lastly, it was shown that SbcCD mutants provide an alternate substrate for RecA4142 to produce SOS(Con) expression that is independent of recBCD, recFOR, recJ, and ruvAB. A simple hypothesis for this mechanism is that since this enzyme has a variety of nucleolytic activities, it could destroy a potential RecA substrate. Interestingly, RecA4142 in an sbcC mutant now resembles RecA730 in wild-type cells not requiring any loading functions for SOS(Con) expression (28). However, RecA4142 does require RecA loading functions for DNA repair (and presumably recombination): both recA4142 recBC sbcBC recF and recA4142 recBC sbcBC strains have high SOS(Con) expression, but only the recA4142 recBC sbcBC recF strain is slow growing and very UV sensitive (data not shown). Other more complicated models are also possible to explain the SOS(Con) expression in the sbcC strains.
How often do reverse replication forks occur when cells are grown in log phase? This has been a difficult question to answer because assays for RFR are indirect, typically requiring that the cells be perturbed in some way (i.e., the addition of a rep or DNA replication mutation or some dose of a DNA-damaging agent) and, depending on the assay, that RecBCD also be absent. At face value, the data suggest that RFR occurs in most cells at some point during the cell cycle since almost all recA4142 cells have high levels of SOS(Con) expression. This suggestion has caveats, however. The first is that RecA4142 may facilitate RFR by causing DNA damage. This seems unlikely since recA4142 mutants are Rec+ and UVr like the wild type (28). Hence, it is probable that RecA4142 only takes advantage of the presence of a reversed fork. The second is the reporter for RFR is GFP that is very stable and has a half-life longer than the cell cycle (1). Therefore, the fluorescence is likely to persist through more than one cell cycle.
It is curious that both SbcB and RecJ exonucleases are needed to process the reversed fork so that RecBCD can load RecA4142. It is easy to imagine that the nascent leading strand would almost always be ahead of the nascent lagging strand, so that SbcB would be needed more often to resect the nascent leading strand and that RecJ would be needed less often to resect the nascent lagging strand. One possibility to explain why both proteins are needed equally is that the two form a complex to process the DSE for RecBCD. In support of this hypothesis, previous work on the networking interactions between E. coli proteins has shown a significant interaction between RecJ and SbcB (5).
The ability of recF to maintain the high levels of SOS(Con) expression in recA4142 strains was dependent on the allele of sbcC in the strain. If sbcC was wild type, then RecF (and RecOR) were required to antagonize the action of RecX. As noted before, the ability of RecX to dramatically decrease SOS(Con) expression was unexpected because it occurred in minimal media and produced such a large effect (until then, RecX effects were only see in rich media and produced subtle changes [28, 37]). If sbcC is deleted, then recF is not needed for the high levels of SOS(Con) expression. Either some other protein is available to antagonize RecX, or the substrate to which RecA4142 is now bound is different and is no longer susceptible to the destabilizing effects of RecX. In other words, RecX may only be able to destabilize RecA4142 at a specialized structure like a reversed replication fork and not at all structures where RecA4142 is bound on the DNA. Other models may also be possible.
Lastly, this work may be used to try to understand the paradoxical observation that on one hand, recA4142 cells required RecBCD for high SOS(Con) expression, suggesting that RecBCD was loading RecA at a DSB, and on the other hand, that the recA4142 recBCD double mutant was viable. In this work, data are shown that support the hypothesis that a reversed replication fork provides the DSE to which RecBCD loads RecA4142. Hence, one no longer has to hypothesize a DSB on which RecBCD loads RecA. Does this solve the paradox, however, of why the recA4142 recBCD mutant is viable? We think it does not, because if RFR is occurring in the recA4142 strain, then the production of the DSE is still known to be lethal in a recBCD mutant (note that although cleavage of the reversed fork by RuvC does produce a DSB, ruvC mutations do not rescue the rep recBCD synthetic lethality ). Therefore, the recA4142 recBCD double mutant should still be inviable (but it is not). One way out of this dilemma is to suggest that there is a RecBCD-independent process (via a presently unknown mechanism and possibly specific to recA4142 cells) that resets the fork and is fast compared with RuvABC binding and RuvC endonucleolytic action (for evidence for the reset of the reversed forks prior to cleavage by RuvABC, see references 17 and 40). Another way around this dilemma is to hypothesize that the number of RFRs that occur in a “compromised” strain [i.e., rep or dnaE(Ts)] is greater than the number of RFRs that occur in the recA4142 strain and that this is the reason for the viability. Future research may be able to test these ideas.
This work was supported by AI059027 from the National Institutes of Health.
We thank Benedicte Michel and David Leach for useful discussions during the writing of the manuscript and Benedicte Michel and Kevin Griffith for reading the manuscript before publication and offering further suggestions.
Published ahead of print on 19 March 2010.