|Accueil | Aperçu | Revues | Soumettre | Nous Contacter | English|
Actively dividing cells perform robust and accurate DNA replication during fluctuating nutrient availability, yet factors that prevent disruption of replication remain largely unknown. Here we report that DksA, a nutrient-responsive transcription factor, ensures replication completion in Escherichia coli by removing transcription roadblocks. In the absence of DksA, replication is rapidly arrested upon amino acid starvation. This arrest requires active transcription, and is alleviated by RNA polymerase mutants that compensate for DksA activity. This replication arrest occurs independently of exogenous DNA damage, yet it induces the DNA damage response and recruits the main recombination protein RecA. This novel function of DksA is independent of its transcription initiation activity, but requires its less studied transcription elongation activity. Finally, GreA/B elongation factors also prevent replication arrest during nutrient stress. We conclude that transcription elongation factors alleviate fundamental conflicts between replication and transcription, thereby protecting replication fork progression and DNA integrity.
Accurate and processive DNA replication is crucial for the preservation of genome integrity. DNA replication has three phases- initiation, elongation and termination. Elongation of DNA replication is highly susceptible to disruptions leading to genome instability (Aguilera and Gomez-Gonzalez, 2008; Branzei and Foiani, 2009; Mirkin and Mirkin, 2007; Wang et al., 2007). How replication elongation remains processive during changing external environment conditions and conflicting cellular processes remains an important unresolved question.
DNA replication and RNA transcription occur on the same DNA template and have an inherent potential to interfere with each other (Mirkin and Mirkin, 2007). In vitro, DNA replication can be slowed significantly by encounters with RNA polymerase (RNAP) (Elias-Arnanz and Salas, 1999; Liu and Alberts, 1995). In vivo, replication is blocked by strong transcription at multiple sites (Azvolinsky et al., 2009; Deshpande and Newlon, 1996). Several factors have been shown to promote replication fork progression through these natural impediments, including the Rrm3 helicase in yeast (Torres et al., 2004), the Rep, DinG and UvrD helicases in E. coli (Boubakri et al., 2009), and the THO/TREX complex that acts at the interface between transcription and mRNA export in yeast (Wellinger et al., 2006). In their absence, transcription can pose a significant barrier to replication, which may result in loss of genome integrity (Boubakri et al., 2009; Torres et al., 2004; Tourriere and Pasero, 2007). None of these factors deal directly with the RNAP-DNAP collision, and it remains to be understood how the transcription machinery acts when encountering oncoming replication, and whether transcription barriers can become deleterious upon unfavorable environmental conditions such as starvation.
Nutritional starvation is frequently encountered by bacteria, and can affect both replication initiation (Ferullo and Lovett, 2008) and elongation (Wang et al., 2007). In the widely studied E. coli strain K-12, the rate of replication elongation varies by more than two-fold when cells are growing in different nutrient conditions (Bipatnath et al., 1998; Michelsen et al., 2003), but the reason for this variation is unknown. Starvation also induces a profound change in global transcription, including inhibition of rRNA and tRNA synthesis and induction of stress and stasis survival genes. This response is mediated via the synthesis of the nucleotide guanosine (penta)tetraphosphate, also called (p)ppGpp, and requires the transcription initiation factor DksA (Barker et al., 2001; Cashel et al., 1996; Paul et al., 2004; Paul et al., 2005). DksA is known to interact with the ‘secondary channel’ of RNAP to change the kinetics of transcription initiation (Paul et al., 2004; Perederina et al., 2004). Interestingly, DksA is also found to have an effect on resistance to DNA damage by ultra-violet light (UV) and genotoxic agents (Trautinger et al., 2005). Recently, CarD, an essential protein of the pathogenic bacterium Mycobacterium tuberculosis, which is upregulated by DNA damage, has been shown to restore partial growth of an E. coli dksA mutant, and is upregulated by DNA damage (Stallings et al., 2009). How nutrient-responsive transcription factors such as CarD/DksA maintain genome integrity remains enigmatic.
Here, we reveal that DksA prevents transcription from interfering with replication upon nutrient stress. During starvation, replication elongation is stalled throughout the genome in ΔdksA cells, even in the absence of external DNA damaging agents. This replication block is due to stalled transcription complexes, since inhibiting transcription abolishes this replication arrest. The arrested replication forks recruit the recombination protein RecA and induce the SOS DNA damage response. We found that in contrast to its well-known function in transcription initiation with (p)ppGpp, DksA alone prevents transcription from interfering with replication by acting directly on RNAP elongation complexes. In addition to DksA, several transcription factors including GreA, GreB (TFIIS homologs in eukaryotes) and TraR also promote replication fork progression through transcription roadblocks. Our results reveal a novel pathway for dealing with the transcription/replication conflict at the time of nutritional stress.
We monitored genome-wide replication in E. coli using genomic microarrays (Breier et al., 2005; Khodursky et al., 2000). As outlined in Fig. 1A, we synchronized DNA replication in a population of cells using a temperature-sensitive allele of the replication protein DnaC (dnaC2) (Carl, 1970). After replication initiation, cells were treated and DNA was purified and hybridized on microarrays against reference DNA from cells with fully replicated chromosomes, to obtain the genome-wide dosage profile. In each cell, the genes near the replication origin (oriC) should be enriched two-fold compared to genes near the terminus (terC), with the replication forks located in between (see Fig. 1B for schematics). The actual microarray profile (Fig. 1C) has a more gradual transition from double to single gene dosages due to an inherent stochasticity in initiation timings and rates of replication elongation. We define the average distance of the replication forks from oriC as x, and the variability of the transition to be Δx, as described in Experimental Procedures.
We examined the effect of amino acid starvation on replication elongation in wild-type and ΔdksA cells. Starvation was induced by the standard treatment with serine hydroxamate (SHX), an inhibitor of serinyl tRNA charging, which immediately shuts down cell growth (Tosa and Pizer, 1971) (Fig. 2A, C). We synchronized wild-type and ΔdksA cells, treated them with SHX 5 min. after the synchronized initiation of replication, and examined their gene dosage profiles 25 min. after initiation (Fig. 1A). In wild-type cells, the rate of replication fork progression was not significantly affected by SHX treatment, with both forks having progressed to ~0.82 Mb (Δx=0.07 and 0.09 Mb, with and without SHX respectively) (Fig. 2B). This agrees with previous reports that replication elongation is not significantly affected by amino acid starvation in E. coli (Ferullo and Lovett, 2008; Maaloe and Hanawalt, 1961).
In contrast, in the absence of DksA, replication fork progression is drastically affected by starvation. Whereas in untreated ΔdksA cells, replication forks moved to ~0.75 Mbp (Δx=0.08 Mbp), similarly to wild-type cells; in SHX-treated ΔdksA cells, replication forks progressed only ~0.49 Mbp, with an increase in the variability of fork speeds (Δx=0.12 Mbp) (Fig. 2D). Further measurement 20 min. later showed little fork progression in SHX-treated ΔdksA cells (Fig. S1). The observed reduction in elongation rate associated with the increase in the variability of fork positioning suggests that replication forks stall in many, although not all, starved ΔdksA cells.
We verified that in asynchronously growing ΔdksA cells, DNA replication is also blocked by starvation. Employing the diphenylamine colorimetric assay that measures the total amount of DNA (Bipatnath et al., 1998), we first established that, as expected, DNA content in untreated wild-type and ΔdksA cells increased exponentially with time (Fig. 2E, F). Upon SHX treatment in wild-type cells, the increase in DNA content gradually stopped, due to a lack of new initiation of replication while ongoing elongation is completed (Ferullo and Lovett, 2008; Levine et al., 1991; Schreiber et al., 1995) (Fig. 2E). In contrast, in ΔdksA cells, the increase in DNA content stopped rapidly (Fig. 2F), confirming that starvation results in near-complete inhibition of replication in the absence of DksA. We also measured DNA replication by incorporation of tritiated thymidine (3H-thy) into the trichloroacetic acid-precipitable fraction (Fig. 2G). Upon SHX treatment in wild-type cells, the rate of 3H-thy incorporation decreases rapidly due to an inhibition of replication initiation (Ferullo and Lovett, 2008; Levine et al., 1991; Schreiber et al., 1995) and a change in the equilibration of 3H-thy with the intracellular TTP pool (Neuhard and Nygaard, 1987)(Cashel, personal communication). The remaining 10-20% of 3H-thy incorporation in wild-type cells corresponds to the unperturbed replication elongation observed in our microarray experiments. Importantly, in ΔdksA cells, 3H-thy incorporation was completely inhibited, indicating blockage of elongation. Therefore, thymidine incorporation can be used as a reliable assay for comparing DNA replication in the presence and absence of DksA, despite the accompanying changes in nucleotide pools.
We verified that the observed replication arrest is not due to the presence of hydroxamate. We took advantage of the observation that unlike SHX, arginine hydroxamate (RHX) does not induce amino acid starvation in E. coli (Fig. 2A, C) (Tosa and Pizer, 1971), and monitored DNA replication in cells treated with RHX (Fig. 2G). Unlike SHX, RHX treatment did not inhibit replication in either wild-type or ΔdksA cells (there is less increase of incorporation in ΔdksA cells than in wild-type cells due to slower growth), demonstrating that the dramatic reduction in replication rate upon SHX treatment is not due to hydroxamate, but due to amino acid starvation.
Arrested replication forks can have different fates depending on the modes of arrest. Forks stalled by depletion of the dNTP substrate or by DNA polymerase inhibitors recruit the recombination protein RecA, whereas forks that undergo a regulated arrest do not (Wang et al., 2007). The recruitment of RecA facilitates recombination, and activates error-prone DNA polymerases (Cox, 2007). Therefore, we monitored the recruitment of RecA using a recA-gfp fusion allele (Renzette et al., 2005) (Fig. 3) to examine whether the stalled replication forks in ΔdksA cells invoke a cellular response of replication fork repair and/or recombination.
In agreement with previous observations, we observed RecA foci in a fraction of cells at cell poles (the storage structure for RecA) or nucleoid-associated positions (due to disruptions in replication) (Renzette et al., 2005). We found nucleoid-associated RecA foci in ~27% of dksA+ cells, similar to previous observations (Renzette et al., 2005). In ΔdksA cells, nucleoid-associated RecA foci formation is elevated to ~52% (Fig. 3A). We note, as others have (Ishii et al., 2000; Magnusson et al., 2007; Yamanaka et al., 1994), that ΔdksA cells are longer than dksA+ cells (Fig. 3C, D). To rule out the possibility that the increased percentage of RecA foci is due to the increased length of ΔdksA cells, we calculated the average foci number per cell length, which is 0.08/μm for dksA+ cells but 0.13/μm for ΔdksA cells (Fig. 3B). Therefore, regardless of the method of calculation, ΔdksA cells have more RecA foci than dksA+ cells, indicating elevated replication disruptions even under untreated conditions.
We then induced amino acid starvation and observed a dramatic difference between dksA+ and ΔdksA cells. Only ~13% of starved dksA+ cells have nucleoid-associated RecA foci (Fig. 3A, E). Even after we accounted for the shortened cell lengths (to ~0.05/μm), no increase was observed compared to 0.08/μm for untreated dksA+ cells. We observed a decrease of RecA foci instead, perhaps because replication has completed without initiating a new round upon starvation. In contrast, nearly all starved ΔdksA cells (92%) have RecA foci with many cells containing multiple foci (the density of foci is ~0.32/μm) (Fig. 3A, F). This result supports the conclusion that DNA replication is highly disrupted in ΔdksA cells subjected to nutrient stress, leading to the recruitment of RecA.
RecA recruitment can potentially turn on the SOS response by inducing auto-cleavage of LexA, the repressor of SOS response genes. We found that the level of LexA is not affected by starvation in wild-type cells, but rapidly falls in ΔdksA cells indicating LexA cleavage (Fig. 4A, B). The cleavage of LexA partially depends on the RecB, RecF and RecO proteins (Fig. 4A). The RecBCD and RecFOR pathways process double-stranded DNA ends and single-stranded DNA gaps to load RecA, respectively (Dillingham and Kowalczykowski, 2008; Kuzminov, 1995). Thus, the replication forks stalled by starvation are disrupted, creating replication-dependent DNA ends and gaps in ΔdksA cells.
We verified, using microarray-based gene expression profiling, that SHX treatment induces a strong SOS response in ΔdksA cells. As expected, the transcription response of wild-type cells is very similar to previously published microarray results (Durfee et al., 2007) (Fig. S2). Importantly, all known members of the SOS regulon, such as sulA, umuD and recA, are mildly induced upon starvation in wild-type cells, but highly induced in ΔdksA cells (Fig. 4C). On the other hand, DNA repair genes such as recBCD, which are not regulated by the SOS response, do not follow the same pattern (Fig. S2).
The SOS response is also mildly elevated in untreated ΔdksA cells compared with dksA+ cells. This was verified with a single-cell reporter created by fusing an SOS-inducible promoter (sulA) to gfp (Pennington and Rosenberg, 2007). Using fluorescence microscopy (Fig. 4D) and flow cytometry (not shown), we quantified the degrees of the SOS response in ΔdksA and dksA+ cells by their average fluorescence intensities. The background autofluorescence was obtained using a constitutively repressing lexA allele, lexA3(Ind−). We found that ΔdksA cells have significantly higher spontaneous SOS induction than dksA+ cells. These results indicate that ΔdksA cells, even without starvation, experience a chronic SOS induction.
ΔdksA cells are known to be auxotrophic for certain amino acids (Brown et al., 2002), we examined whether this amino acid requirement is related to the observed induction of sulA, which encodes the SOS-induced cell division inhibitor. We found that deletion of sulA does not affect the plating efficiency of ΔdksA cells on minimal medium. On the other hand, abolishing the induction of the SOS response via lexA3(Ind−) mutation, which disables the ability of ΔdksA cells to cope with DNA damage, results in 50 fold further loss of plating efficiency on minimal medium (Table S1). Therefore, the amino acid requirement of ΔdksA is not caused by the chronic sulA induction, but rather a consequence of replication fork collapse.
We found that replication arrest during amino acid starvation depends on transcription. After treating ΔdksA cells with rifampicin (Rif), which abolishes transcription, replication is no longer arrested by starvation (Fig. 5A). Addition of Rif also abolished LexA cleavage upon starvation of ΔdksA cells (Fig. 4A). Therefore, disruption of replication is due to barriers created by transcription. In addition, since inhibition of transcription alleviates the replication blockage in ΔdksA cells, de novo protein synthesis is not required to prevent replication arrest upon starvation.
How does DksA prevent replication from being blocked by the transcription barrier? We found that DksA prevents replication arrest by acting directly on transcription. We took advantage of the fact that many mutants of rpoB and rpoC, encoding the β and β’ subunits of RNAP, can bypass the requirement for DksA in transcription control (Rutherford et al., 2009). We tested one such mutant, rpoB111 (P564L), annotated as rpoB* for this work (Murphy and Cashel, 2003; Zhou and Jin, 1998) (Table S1), and found that it is sufficient to prevent replication arrest even in the absence of DksA. Using thymidine incorporation (Fig. 5B), total DNA content (Fig. 5C, D compared to Fig. 2E, F) and microarrays (Fig. S3A), we consistently observed that the rpoB* allele restores replication in a ΔdksA background. rpoB* also abolished LexA cleavage in a ΔdksA background (Fig. 4A), confirming that DksA prevents disruption of replication via its effect on RNAP.
We found that similar to several previously reported rpoB alleles (Trautinger et al., 2005), the rpoB* allele also confers UV resistance in the absence of dksA. We verified that ΔdksA cells are more sensitive to UV light in a ΔruvB background, and found that the rpoB* mutation compensates for the lack of DksA in UV resistance (Fig. 5E). These data imply that DksA confers resistance to DNA damage via its effect on RNAP.
DksA is best known for its concerted action with the small nucleotide (p)ppGpp (Paul et al., 2004). Upon amino acid starvation, (p)ppGpp is rapidly produced by the enzyme RelA and along with DksA, alters the transcription of many genes. (p)ppGpp/DksA-mediated changes of transcription might be required for preventing replication arrest upon starvation. If so, then ΔrelA cells should also exhibit replication arrest due to failure to produce (p)ppGpp in response to amino acid starvation. However, we found that both the ΔrelA strain, and a ΔrelA ΔspoT strain which completely fails to produce (p)ppGpp ((p)ppGpp0), continued replication upon amino acid starvation (Fig. (Fig.5F,5F, S3B,C). This is not due to compensatory mutations in RNAP, as we verified that cells did not acquire suppressors by assaying for failure to grow on minimal medium. Therefore, DksA promotes replication elongation even in the absence of (p)ppGpp during nutrient stress.
Finally, we examined whether the lack of DksA leads to elevated levels of (p)ppGpp, which in the Gram positive bacterium Bacillus subtilis has been shown to impede replication fork progression (Wang et al., 2007). We verified that starvation induces (p)ppGpp levels in a comparable manner to wild-type cells (Brown et al., 2002; Paul et al.), although basal (p)ppGpp levels in untreated ΔdksA cells are higher (Fig. S3D), likely because (p)ppGpp concentrations are inversely proportional to growth rate, and ΔdksA cells grows more slowly than wild-type cells. Therefore, the effect of DksA on replication upon starvation is not from an increase in (p)ppGpp levels.
DksA is a well-characterized transcription initiation factor (Aberg et al., 2008; Blankschien et al., 2009a; Paul et al., 2004; Perederina et al., 2004). We tested whether DksA promotes replication processivity by affecting transcription initiation, albeit independently of (p)ppGpp. We took advantage of the fact that the effect of DksA on transcription initiation requires one or both of the two conserved aspartic acid residues (D71 and D74) near the tip of its coiled-coil domain (Blankschien et al., 2009a; Perederina et al., 2004). The DksA mutant with both aspartic acid residues mutated to asparagines (D71N and D74N), named DksANN, can no longer inhibit transcription from rRNA promoters (Fig. 6A). However, DksANN can still prevent replication arrest upon starvation (Fig. 6B), indicating that the role of DksA during transcription initiation is not necessary for its effect on replication. DksANN also partially rescues the UV sensitivity of ΔdksA ΔruvB cells, indicating that the effect of DksA on transcription initiation is not necessary for its role in UV-resistance (Fig. 6C). In addition, DksANN is sufficient to reduce the SOS response of ΔdksA cells (Fig. 6E). Cell lengths are also reduced, suggesting replication disruption contributes to the defect of ΔdksA cells in cell division (Ishii et al., 2000; Magnusson et al., 2007; Yamanaka et al., 1994), at least partially by affecting SOS response (Fig. 6D, F).
In summary, the effect of DksA on transcription elongation, rather than initiation, is crucial for processivity of replication and resistance to DNA damage.
In vitro, DksA is known to affect transcription elongation by preventing transcriptional pausing (Perederina et al., 2004). Importantly, DksANN is shown to prevent transcriptional pausing similarly to wild-type DksA (Perederina et al., 2004), despite the loss of its effect on transcription initiation. Therefore, DksA might facilitate replication by affecting transcriptional pausing. To test whether transcriptional pausing affects replication, we examined rpoB alleles with different transcription processivities including rpoB8 (Q513P), rpoB2(H526Y) and rpoB3595 (S522F) (Jin and Gross, 1988). The rpoB2 allele has less transcriptional pausing activity than wild-type rpoB (Landick et al., 1990; McDowell et al., 1994), and the rpoB3595 allele has a faster rate of transcription elongation (Jin et al., 1992). We found that both the rpoB2 and rpoB3595 alleles suppress replication arrest observed in starved ΔdksA cells (Fig. 6G). The rpoB8 allele, on the other hand, has an elevated level of transcriptional pausing (Fisher and Yanofsky, 1983; Landick et al., 1990), and is extremely sick when combined with ΔdksA (doubling time>150min. at 37°C).
These results suggest that DksA prevents transcriptional pausing from interfering with DNA replication. In addition to DksA, the transcript cleavage factors GreA and GreB also interact with the RNAP secondary channel and prevent prolonged transcriptional pausing (Artsimovitch et al., 2000; Marr and Roberts, 2000; Orlova et al., 1995). We found that although the loss of either GreA or GreB did not result in a replication block, loss of both eventually resulted in near-complete replication blockage upon starvation (Fig. 6H). Interestingly, over-expression of GreA compensates for the lack of DksA in preventing disruption of replication (Fig. 6I). In addition, TraR, another secondary channel protein found on conjugative plasmids (Blankschien et al., 2009b), can also compensate for the lack of DksA (Fig. S4A). Interestingly, removal of all three factors-greAB and dksA, results in significant decrease of replication progresssion even in the absence of starvation, as inhibition of transcription via addition of Rif can enhance replication immediately even in the absence of starvation (Fig. 6J). We conclude that certain kinds of factors (DksA, GreAB, TraR) can alleviate the replication block, and apparently DksA and the Gre factors are not interchangeable for this purpose: the cell needs both.
Not every factor that affects stalled transcription complexes promotes replication elongation upon starvation. Removal of the transcription repair coupling factor Mfd does not lead to starvation-induced replication arrest (Fig. S4B), and over-expressing the enzyme RnaseH which removes R-loops, has only a slight effect in preventing replication arrest in the absence of DksA (Fig. S4C, Table S1). While these factors are not involved in preventing transcription from blocking DNA replication during starvation as DksA/GreAB do, they accompany DksA/GreAB as an extended class of factors that maintain processive DNA replication and genome integrity.
In this study, we show that transcription factors interacting with the secondary channel of RNA polymerase (DksA, TraR, GreA and GreB) prevent replication from being disrupted by transcription, and this effect is strongly amplified during amino acid starvation. Our work reveals new insights of conflict and conflict-resolution between replication and transcription, by revealing that a class of transcription factors mediates these conflicts and guarding the genome against instability during nutritional stress.
In bacteria, DNA replication takes place during most of the cell cycle, or continuously during fast growth. Encounters of replication forks with damaged (nicked) DNA templates lead to replication fork collapse (Kuzminov, 1995). Even in the absence of exogenous DNA damage, replication is disrupted in more than 15% of E. coli cells (Cox et al., 2000; Renzette et al., 2005), some of which leads to induction of the SOS response (Pennington and Rosenberg, 2007). The causes of such replication disruption are not clear and our findings highlight the contribution of interference by transcription. The conflict between replication and transcription is not apparent (Skarstad et al., 1986) because it is kept in check by factors such as DksA. In the absence of DksA, even without stress, cells exhibit a chronic DNA damage response (Fig. (Fig.3,3, ,4),4), indicating that a subpopulation of replication forks is disrupted. This explains why deletion of dksA is synthetically lethal with deletion of priA, a factor required for replication restart (Mahdi et al., 2006). Removing DksA and GreA/B slows replication elongation significantly during normal conditions (Fig. 6J), highlighting the importance of these factors. We found that starvation strongly elevates the conflict between replication and transcription. Replication fork progression is almost entirely stopped in starved ΔdksA cells, inducing a full DNA damage response. Cells also exhibit an increased dispersion of replication fork positions (Fig. 2D) probably due to stochastic transcription barriers.
DksA and other factors might have important functions in preventing replication arrest in other stressful situations, with different factors acting primarily under different circumstances. It was shown that (p)ppGpp could remove RNAP arrays stalled at damaged DNA, potentially providing access of the lesion to the DNA repair machinery and preventing fatal collisions of transcription complexes with replisomes. It was postulated that DksA, GreA and Mfd act similarly to (p)ppGpp at DNA lesions (Trautinger et al., 2005). Here, we observed lesion-independent, formidable transcription roadblocks during starvation, obtained direct evidence for replication blockage, and showed that it can be prevented by DksA, TraR or GreA/B. On the other hand, removal of (p)ppGpp or Mfd does not lead to starvation-induced replication arrest (Fig. S3, S4). We propose that DksA, GreA/B and TraR prevent the transcription-replication conflict during different stress conditions than Mfd and (p)ppGpp, but all constitute the currently identified group of transcription-replication mediators. Similarly, the recently identified CarD factor in Mycobacterium tuberculosis, which partially compensates for loss of DksA in E. coli and is DNA damage-inducible (Stallings et al., 2009), might also directly play a role in avoiding replication arrest. This class of transcription factors may prove to be large and its further characterization should have important implications for understanding fundamental DNA metabolic processes.
In E. coli, DksA prevents rapid replication arrest upon amino acid starvation, whereas in B. subtilis, replication is rapidly arrested in wild-type cells under amino acid starvation (Wang et al., 2007). Despite the apparent similarities, these replication arrests are different in nature. First, in E. coli, (p)ppGpp does not inhibit replication elongation significantly; whereas in B. subtilis, the inhibition is mediated by (p)ppGpp, most likely by targeting the replication enzyme primase. Second, in E. coli, starvation-arrested replication forks recruit the recombination protein RecA, indicating that the arrested forks are disrupted; whereas in B. subtilis, RecA recruitment is not elevated upon amino acid starvation, suggesting that the replication arrest is non-disruptive (Wang et al., 2007). The proposed concepts underlying such ‘accidental’ versus ‘natural’ replication arrests (Bidnenko et al., 2002) may reflect fundamental differences in the lifestyles of these organisms during feast and famine.
DksA is absent in B. subtilis. Previous studies have strongly suggested that the effect of transcription on replication is directional (Brewer, 1988; French, 1992; Liu and Alberts, 1995; Mirkin and Mirkin, 2005; Pomerantz and O’Donnell, 2008). B. subtilis has a 75% bias towards genes being transcribed and replicated co-directionally, compared to only ~55% for E. coli (Rocha, 2004). Reversing this bias over a genomic segment or rrn operons lead to impairment of replication and delay in cell cycle progression (Srivatsan et al., 2010), suggesting that B. subtilis minimizes the transcription/replication conflict partly via genome organization. In E. coli, similar reversions impair replication and cell proliferation only in the absence of DNA repair helicases (Boubakri et al., 2009), suggesting that diverse strategies are used by different organisms to overcome the challenge of replication over transcription.
Our work supports the idea that DksA uses a different mechanism for altering transcription elongation complexes than for transcription initiation, and highlights the physiological importance of DksA during transcription elongation. DksANN is a separation-of-function mutant that allowed us to show that the role of DksA in transcription elongation. DksANN lacks the regulatory activity during transcription initiation, yet it still prevents blocks to DNA replication, and attenuates the SOS response, UV-sensitivity, and filamentation of ΔdksA cells (Fig. 6). In addition, the differential effects of DksA and (p)ppGpp on replication during starvation attest to the conflict and suggest an additional function of DksA beyond transcription initiation.
Our work demonstrates that transcription is a potent barrier to replication elongation upon amino acid starvation. We also provide significant new information about a class of proteins that cells use to prevent collisions between the transcription and replication machineries from compromising genome integrity, even in the absence of external DNA damage. Details about the mechanism(s) involved are unresolved and the following issues will need to be addressed by future studies: First, the precise mechanism by which DksA alters transcription complexes remains to be defined, and exactly how alteration of transcription prevents disruption of replication will have to be worked out. DksA might prevent transcription stalling, or destabilize stalled transcription elongation complexes. Second, how amino acid starvation affects transcription elongation remains unknown. One possibility is that starvation leads to a redistribution of the RNAP from rRNA promoters to the rest of the chromosome (Jin and Cabrera, 2006). This rapid redistribution might create congestion of transcription flux, blocking replication and requiring DksA for its prevention, but this mechanism is unlikely since a lack of (p)ppGpp has no effect on replication arrest. Alternatively, starvation induces (p)ppGpp or lowers NTP, which might directly inhibit transcription elongation (Kingston et al., 1981; Krohn and Wagner, 1996; Vogel and Jensen, 1994). Finally, amino acid starvation stalls translation by depleting charged tRNAs, which uncouples translation from transcription. Coupling of transcription-translation is proposed to preclude R-loop formation (Gowrishankar and Harinarayanan, 2004) or alter RNA folding, unmasking a pausing signal normally hidden by the process of translation. However, R-loops are unlikely to play a major role in the observed replication arrest (Fig. S4C).
The physical nature of the replication barrier may have multiple origins: backed-up arrays of stalled RNAP (Trautinger et al., 2005), direct physical interaction with the head-on transcription machinery (Mirkin and Mirkin, 2005), backtracked paused RNAP (Artsimovitch and Landick, 2000; Komissarova and Kashlev, 1997), topological barriers to replication created by the effect of transcription on the supercoiling of DNA (Liu and Wang, 1987; Olavarrieta et al., 2002) and RNA secondary structure. It remains to be elucidated how transcription impedes replication upon starvation.
Regardless of the details of the mechanism of the replication block, it has become clear that replication elongation is highly susceptible to nutrient availability and perhaps other types of environmental stress. DksA and other transcription factors robustly maintain ongoing replication progression by preventing conflicts with transcription. The connection between replication complexes and the cellular environment is likely to be far more extensive than previously appreciated. Further study will broaden our understanding of how cells protect DNA replication from stress and maintain genome integrity.
All E. coli strains used are derivatives of MG1655 or W3110. Standard growth, transformation, and transduction procedures were used (Miller, 1992). Deletion derivatives of various genes were constructed by phage transduction from the Keio collection (Baba et al., 2006). Strains are described in Supplemental Table S2. Unless indicated, the strains were grown in M9 medium supplemented with 0.2% glucose and 0.4% casamino acids at 37°C with shaking at 250 rpm.
Replication was synchronized using the temperature-sensitive dnaC2 mutant (Carl, 1970) that fails to initiate replication at the non-permissive temperature (42°C). Cells were collected and DNA was purified as described (Breier et al., 2005). DNA samples were labeled as described (Wang et al., 2007) and hybridized to Agilent oligo-arrays following the Agilent oligo-aCGH procedure.
Analysis was performed using Agilent Feature Extraction software. The ratios of the fluorescence intensity of the sample versus the fully replicated pre-initiation reference were averaged for each gene position, and were smoothed by moving average (window size 100-150) to obtain the gene dosage profile. To calculate the average distance of the replication forks from oriC (x) and the variability of the positions (Δx), we fitted the gene dosage profile of each replichore similar as described (Breier et al., 2005) except that a binomial distribution was used instead of a linear transition with p proportional to probability of unblocked replication fork progression in each increment of time, and n proportional to time after treatment. x is obtained from the mean (x = n * p) and Δx is obtained by taking the square root of the variance of the distribution var = n * p * (1-p).
Cultures were grown to OD600~0.2, left untreated or treated with 0.5mg/ml SHX. 2.5ml samples were collected at each time point and DNA content was evaluated as described (Bipatnath et al., 1998).
Cells were grown to mid-log phase. Labeling was done by mixing 5 μl of 3H-thy (80 mCi/mol) (Perkin Elmer) with 200 μl of culture for 2 min. The amount of radioactivity incorporated was determined as described (Wang et al., 2007).
Cultures were labeled and loaded on PEI cellulose plates as described (Schneider et al., 2003). Plates were developed in 1.5 M KH2PO4 (pH = 3.4), exposed to a Storage Phosphor Screen and scanned using a GE Typhoon scanner.
Overnight cultures were diluted 1:1000 in LB and grown to early log phase. Dilutions were plated and cells were immediately irradiated using a Stratalinker UV crosslinker. Irradiated cells were incubated in the dark at 32°C and colony forming units were scored after 48 hours.
Cells were viewed with a Zeiss Axiovert 200 equipped with a 100x phase contrast objective. For visualization of the nucleoid, cells were fixed in 70% ethanol, and DAPI (4, 6-diamidino-2-phenylindole) was added (0.1 μg/ml) before visualization. Images were analyzed using AxioVision software (Zeiss).
Samples were loaded on a 15% SDS/PAGE gel, transferred to a Hybond ECL membrane (GE) and blotted with a rabbit polyclonal anti-LexA antibody (Fisher) (1:6000). Membranes were blotted with a goat anti-rabbit antibody (1:6000), treated with ECL reagents (GE), exposed on film and quantified with ImageJ.
Cells were grown to OD600~0.3 and treated with SHX (0.5 mg/ml) for 40 min. Cultures were mixed with 1/8 volume of ice-cold stop solution (5% phenol, 95% ethanol) and harvested. RNA was extracted using QIAGEN RNeasy kit. Relative mRNA levels were determined by hybridization to Agilent oligo-arrays, and data were analyzed using Genepix software as described (Britton et al., 2002). Results are presented as average values of three independent experiments with error bars showing standard errors of the mean.
We thank Michael Cashel, Marc Drolet, Patrick Piggot, Susan Rosenberg and Steve Sandler for strains; Adam Breier, Michael Cashel, Herman Dierick, Tamas Gaal, Alasdair Gordon, Rick Gourse, Phil Hastings, Greg Ira, Allison Kriel, Bob Landick, Jeong Hyun Lee, Susan Rosenberg, Wilma Ross, Jonathan Weissman and the anonymous reviewers for critical comments. CH is supported by a Human Frontier Young Investigator Grant. JDW is supported by a Welch Research Grant (Q-1698) and a Public Health Service grant (1R01GM084) from NIH.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Accession Number Microarray data are available in the NCBI GEO database with accession number GSE19742 (http://www.ncbi.nlm.nih.gov/geo/).