The seqA defect in Escherichia coli increases the ori/ter ratio and causes chromosomal fragmentation, making seqA mutants dependent on recombinational repair (the seqA recA colethality). To understand the nature of this chromosomal fragmentation, we characterized ΔseqA mutants and isolated suppressors of the ΔseqA recA lethality. We demonstrate that our ΔseqA alleles have normal function of the downstream pgm gene and normal ratios of the major phospholipids in the membranes, but increased surface lipopolysaccharide (LPS) phosphorylation. The predominant class of ΔseqA recA suppressors disrupts the rfaQGP genes, reducing phosphorylation of the inner core region of LPS. The rfaQGP suppressors also reduce the elevated ori/ter ratio of the ΔseqA mutants but, unexpectedly, the suppressed mutants still exhibit the high levels of chromosomal fragmentation and SOS induction, characteristic of the ΔseqA mutants. We also found that colethality of rfaP with defects in the production of acidic phospholipids is suppressed by alternative initiation of chromosomal replication, suggesting that LPS phosphorylation stimulates replication initiation. The rfaQGP suppression of the seqA recA lethality provides genetic support for the surprising physical evidence that the oriC DNA forms complexes with the outer membrane.
Initiation in bacteria is accomplished by binding the DnaA initiator protein to the replication origin, oriC (Fig. 1). Since both the origin DNA and DnaA are constantly present in the cell, unscheduled initiation must be actively prevented. Two separate regulatory cycles in the initiation of DNA replication in Escherichia coli are known. In the DnaA cycle, unscheduled initiation is prevented by a transient inactivation of DnaA after initiation via Hda + DnaN-stimulated hydrolysis of the associated ATP (Fig. 1, left). DnaA is then ‘rejuvenated’ by exchanging its ADP for ATP in the presence of acidic phospholipids of the cell envelope. Finally, ATP-bound DnaA, with the help of DiaA (Ishida et al., 2004; Keyamura et al., 2007), binds to a fully methylated oriC DNA in preparation for origin firing (Fig. 1, centre). The operation of the DnaA cycle is grossly affected by titration of the free DnaA excess to the multiple chromosomal DnaA-binding sites (Christensen et al., 1999; Hansen et al., 2007), among which datA locus alone binds ∼60% of the protein (Kitagawa et al., 1996; 1998). In addition, DnaA binding in the promoter of its own gene autoregulates the overall amount of the protein in the cell (Atlung et al., 1985; Braun et al., 1985).
In parallel to the DnaA cycle, there is the oriC cycle, in which unscheduled initiation is prevented by sequestering the replication origin by SeqA (reviewed by Waldminghaus and Skarstad, 2009), a 21 kDa dimeric protein that forms spiral filaments in vitro (Guarnéet al., 2005) and a few foci in dividing cells in vivo (Hiraga, 2000) (Fig. 1, right). SeqA binds properly spaced pairs of hemi-methylated GATC sites in DNA and therefore is proposed to transiently bind and organize any newly replicated DNA (Brendler et al., 2000), as GATC sites throughout the chromosome remain hemi-methylated for about 2 min after replication fork passage (Campbell and Kleckner, 1990). The oriC DNA contains several properly spaced pairs of GATC sites (Kaguni, 2006) and binds SeqA strongly (Kang et al., 1999; Skarstad et al., 2000). Although SeqA itself does not have membrane association domains, the SeqA-bound replication origin is found associated with the membrane fraction until the oriC DNA becomes completely methylated (Slater et al., 1995; d'Alençon et al., 1999), which, in contrast to the rest of the chromosome, takes up to 13 min after the replication start (Campbell and Kleckner, 1990). Thus, hemi-methylated oriC remains inaccessible to DnaA, preventing unscheduled initiation (Fig. 1, right). Similar to the DnaA cycle, operation of the oriC cycle could be affected at the cellular level due to titration of SeqA by its multiple transient chromosomal contacts, although this aspect has never been experimentally addressed.
The seqA mutant cells have increased number of origins (von Freiesleben et al., 1994; Boye et al., 1996; Riber et al., 2006) and therefore are presumed to overinitiate chromosomal replication (reviewed in Waldminghaus and Skarstad, 2009). One benefit of finding a colethal combination, like seqA recA, is that it offers a strong selection for suppressors – mutants that return viability to colethal combinations. These suppressors frequently reveal specific damaging steps that poison the original colethal combinations. The seqA recA colethality proved to be readily suppressed, and we expected these suppressors to inactivate various positive factors in the initiation of chromosomal replication. Assuming that chromosomal fragmentation is the lethal event in the seqA mutants, we further expected that the depression of replication initiation by suppressors of the seqA recA colethality would also translate into reduced chromosomal fragmentation. We did isolate the expected suppressors in the potential initiation factors that reduced both the initiation and the chromosomal fragmentation (E. Rotman and A. Kuzminov, unpublished). However, the majority of suppressors of the seqA recA colethality were of a different kind, indicating that the lethality can in principle be relieved without decreasing chromosomal fragmentation.
Characterization of the ΔseqA allele
The seqA gene in the E. coli chromosome is upstream of pgm in a two-gene operon (Lu and Kleckner, 1994) (Fig. 2A). The Pgm protein is the phosphoglucomutase that catalyses the reversible transformation of glucose-1-phosphate (an intermediate in maltose, galactose and glycogen metabolism and a precursor to polysaccharide wall biosynthesis) into glucose 6-phosphate (the starting point of glycolysis and gluconeogenesis) (Joshi and Handler, 1964; Adhya and Schwartz, 1971; Lu and Kleckner, 1994). To avoid the two imperfections of the previously constructed ΔseqA::tet allele of Lu and Kleckner – an incomplete removal of the seqA ORF and the significant polar effect on pgm (Lu et al., 1994; Lu and Kleckner, 1994) – we precisely deleted the seqA ORF by replacing it with a kanamycin-resistance cassette (the ΔseqA20::kan allele) and later removed most of the insert (the ΔseqA21 allele) (Fig. 2A). Since the pgm mutants are sensitive to 1% SDS (Lu and Kleckner, 1994), we tested for any polar effects on pgm in our two ΔseqA alleles by plating E. coli on media supplemented with this detergent. We found that, while both the original Δpgm::tet and ΔseqA::tet mutants of Lu and Kleckner cannot grow on 1.5% SDS, both our ΔseqA mutants can, although they form much smaller colonies than wild-type cells (Fig. 2B and not shown), suggesting a change in the outer membrane.
We also tested for functional phosphoglucomutase activity by iodine staining. When a pgm mutant is grown on minimal media with galactose as the sole carbon source, maltodextrin accumulates as an amylose intermediate and is secreted by the cell. As a result, colonies of pgm mutants form a dark blue border when overlaid with a solution containing 0.1% iodine and 1% potassium iodide, whereas colonies of pgm+ cells do not (Adhya and Schwartz, 1971). Our ΔseqA20::kan and ΔseqA21 mutant colonies did not form a border, confirming their pgm+ status, whereas both the ΔseqA::tet and the Δpgm::tet mutants did (Fig. 2C). To avoid working with a double seqA pgm mutant, Lu and Kleckner complemented their experimental strain with the pgm+ gene on a plasmid (Lu et al., 1994). Indeed, when we tested the ΔseqA::tet pML14 combination with iodine staining, the blue border no longer appeared (Fig. 2C).
Inactivation of rfaQ, rfaG, or rfaP suppresses the seqA recA lethality
To understand the nature of chromosomal fragmentation in the ΔseqA mutants, we used insertional mutagenesis to isolate suppressors of the seqA recA synthetic lethality. The ΔseqAΔrecA double mutant was barely viable at 42°C (and completely dead at lower temperatures) (Fig. 3A), which disqualified the double deletion mutant for suppression analysis due to the anticipated high background of spontaneous suppressors. Since we [and others before (Lu et al., 1994)] noticed that the ΔseqA single defect was more severe at lower temperatures, we combined the ΔseqA allele with recA629(Cs), a cold-sensitive allele of recA, which is deficient at 28°C, but only moderately defective at 42°C (Knight et al., 1984). As expected, the double ΔseqA21 recA629(Cs) mutant grew well at 42°C, but failed to grow at 28–30°C (Figs 3B and 4A), although a M9-based medium or anaerobic conditions relieved the lethality slightly (Fig. 3B).
After insertional mutagenesis, we selected for colonies that were able to grow at 30°C and, using the kanamycin-resistance determinant of the insert, verified suppressors after P1 transduction into the original ΔseqA recA(Cs) double mutant. Strong suppressors were able to form visible colonies after 1 day incubation, while weaker ones took 2 days. Out of multiple suppressors isolated (E. Rotman and A. Kuzminov, unpublished), approximately three quarters took 2 days to grow and were highly mucoid. Sequencing identified 18 of these weak suppressors as insertions in rfaQ, rfaG and rfaP[also known as waaQ, waaG and waaP (Heinrichs et al., 1998)], the first three genes of the rfaQGPSBIJYZK operon (Fig. 4B), responsible for lipopolysaccharide (LPS) biosynthesis (Schnaitman and Klena, 1993). Another weak (mildly mucoid) suppressor was a single hit to gmhB, involved in the synthesis of an LPS precursor. LPS forms the outer leaflet of the outer membrane in Gram-negative bacteria, its core covering the surface of the cell as scales of protective armour (Fig. 4B, inset) (Vaara, 1992; Nikaido, 1996). All three rfa genes function in the biosynthesis of the LPS core (see below). While all three are weak suppressors, the rfaG and rfaP mutant colonies appear larger at 30°C due to the production of the slime capsule of colanic acid (Fig. 4A). Secretion of this polysaccharide is a response to the cell envelope stress and is characteristic of the rfaGP mutants at lower temperatures (Parker et al., 1992). To avoid problems due to this colanic acid capsule production, we handled our rfa mutants at 42°C.
The rfa inactivation is not in response to Pgm overproduction
Isolation of significant numbers of the rfa suppressors again raised a possibility of the pgm defect – this time of pgm overexpression – in our ΔseqA mutants. Indeed, too much of phosphoglucomutase is known to shift the balance of glucose metabolism in E. coli from glycolysis towards the production of UDP-glucose (Mao et al., 2006) which, in some bacteria (but not in others) translates into higher yield of polysaccharides (discussed in Boels et al., 2003). This may create cell envelope stress, which could be, somehow, alleviated by the rfa defect. To test for the possibility that the rfa suppressors are isolated in response to a change in pgm gene expression in our ΔseqA mutants, we selected suppressors of the triple recA629(Cs) ΔseqAΔpgm mutant, expecting no rfa hits if this explanation were true. However, the triple mutant was suppressed by the same rfa inserts at about the same frequency as the double ΔseqA recA mutant (not shown), indicating that the rfa suppressors are not in response to changes in expression of the phosphoglucomutase gene.
Gene specificity of the rfa suppression
Since insertions in rfaQ and rfaG would interfere with the rfaP expression, it was possible that the rfaP defect was the only real suppressor of the ΔseqA recA lethality. Reducing this possibility was the fact that most of our insertions in rfaQ and rfaG have their kanamycin-resistance gene co-oriented with the rfa operon (Fig. 4B), so no major disruption of the rfaP expression was expected at least in those cases. To clarify which of the three genes were contributing to the suppression, we constructed a precise deletion of all three genes and then verified that the triple ΔseqA recA629(Cs) ΔrfaQGP mutant is still suppressed and can grow at 30°C (Fig. 4C). We then complemented the triple mutant with different combinations of the rfaQ+, rfaG+ and rfaP+ genes on a low-copy-number plasmid (Fig. 4C). As expected, all three genes together abolished the suppression; moreover, the clones with the rfaQ or rfaG genes deleted, either singly or together, still mostly abolished the suppression (Fig. 4C). In contrast, inactivation of the single rfaP+ gene, or the rfaG+ and rfaP+ genes together restored the suppression (Fig. 4C), pointing to the rfaP status as the significant variable. Although we tentatively concluded that inactivation of rfaP alone is sufficient to suppress the ΔseqA recA lethality, we proceeded with further characterization of this suppression with all three mutations, to be on the safe side.
The rfaP and rfaG mutations compensate for the increased LPS phosphorylation in the seqA mutants
The E. coli LPS has three main components: the membrane anchor lipid A, the core region and the O-antigen (Fig. 5A). Laboratory E. coli strains, such as K-12 and its derivatives that we work with, lack their O-antigen and are said to have a ‘rough’ phenotype (non-smooth surface of colonies). If a strain also has a modified core region of its LPS, it is considered a ‘deep rough’ mutant, in which destabilization of the outer membrane results in sensitivity to detergents and hydrophobic antibiotics (Schnaitman and Klena, 1993). RfaQ adds the branch heptose III residue on heptose II of the core (Yethon et al., 1998), RfaG builds the first glucosyl group on heptose II (Parker et al., 1992), while RfaP phosphorylates heptose I (Parker et al., 1992; Yethon et al., 1998) (Fig. 5A). It should be noted that the rfaP defect prevents phosphorylations of heptose II (by RfaY), as well as addition of the heptose III branch to heptose II (Yethon et al., 1998), while the rfaG defect decreases phosphorylation of the core LPS in an unknown way (Yethon et al., 2000). Therefore, all three rfa mutants that we have isolated as suppressors of the seqA recA lethality are ‘deep rough’ mutants, although the rfaP mutant strains have the strongest defects, while the rfaQ mutant strains have the mildest defects of the three.
To verify that our rfaG and rfaP suppressors indeed decrease the LPS core phosphorylation, we grew ΔseqA21 recA629(Cs) and its rfaQ, rfaG and rfaP derivatives in the MOPS (reduced phosphate) minimal medium and labelled the cells with 32P-orthophosphoric acid for 5 min. Inorganic phosphorus is expected to incorporate primarily into RNA, DNA, LPS and phospholipids, but can also be found in polyphosphates (Brown and Kornberg, 2004). We employed an isolation/separation protocol that removed phospholipids and also hydrolysed RNA, leaving three species: DNA, LPS without ester-attached fatty acid tails and polyphosphates, which can be separated from each other (L. Amado and A. Kuzminov, unpublished). To this end, the material produced by the ‘total DNA extraction protocol’ was run in an alkaline agarose gel. Under these conditions, RNA is completely hydrolysed by the alkaline buffer, chromosomal DNA forms a band right below the wells, LPS forms a faster migrating oval below the DNA band, while polyphosphates form an even faster-migrating smear (Fig. 5B).
We found that, compared with the wild-type cells, rfaP single mutants do have lower phosphate content in their LPS (or a lower LPS/DNA ratio, which is less likely, but we did not distinguish between the two). Unexpectedly, we also found that seqA single mutants and the seqA recA double mutants have higher LPS-phosphate content (or a higher LPS/DNA ratio) (Fig. 5B), while the rfaQGP suppressors lower this content/ratio to either wild-type levels (rfaQ) or to the single rfaP mutant levels (rfaG and rfaP) (Fig. 5B and C). Thus, one of the proximal changes in the seqA mutants that leads to their synthetic lethality with recA could be this increased LPS phosphorylation, compensated by the rfaQGP defect.
The decrease in LPS phosphorylation in the rfa mutants should lead to the weakening of the outer armour of the cell (in which individual scales are linked together via the Mg2+-phosphate interactions), making mutants sensitive to anionic detergents (Vaara, 1992; Nikaido, 1996). In fact, the rfaG and rfaP mutants could not grow on LB, supplemented with 1% SDS, whereas the rfaQ mutants could grow, but more slowly than the seqA mutants, and only above 30°C (Fig. 5D and data not shown). Conversely, if the increased LPS phosphorylation is the real cause of the seqA recA inviability, shielding the extra negative LPS charge with magnesium should alleviate the lethality – and it indeed did so completely (Fig. 5E). Magnesium supplementation also eliminated mucoidy of the rfa mutants (Fig. 5E), suggesting that it is the reduced LPS phosphorylation and the resulting weakening of magnesium cross-linking in these mutants that causes cell envelope stress.
No change in phospholipids in the ΔseqA mutants
Since the rfa mutations, by reducing phosphorylation, decrease the negative charge of LPS, the recent demonstration that the seqA defect changes the phospholipid composition of the E. coli membranes (Daghfous et al., 2006), was also of interest. The three major phospholipids in E. coli are the zwitterionic (neutral) phosphatidylethanolamine (PE) and the two acidic species, phosphatidylglycerol (PG) and cardiolipin (CL) (Fig. 6A) (Cronan and Vagelos, 1972). The seqA defect was reported to dramatically decrease the fraction of PE, while increasing the fraction of both PG and CL (Daghfous et al., 2006), in effect, increasing the acidity of the phospholipid membranes. As mentioned in the Introduction (Fig. 1), in vitro, acidic phosholipids PG and especially CL stimulate the ADP→ATP exchange in the DnaA replication initiation protein, associated with the oriC DNA, thus recharging it for the new initiation round (Sekimizu and Kornberg, 1988; Crooke et al., 1992). The defect in production of PG in the E. coli pgsA mutant apparently inhibits replication initiation, because this lethal phenotype was once reported to be suppressed by inactivation of the rnhA gene (Xia and Dowhan, 1995). The rnhA defect also suppresses the replication initiation problems of the dnaA and ΔoriC mutants by permitting alternative replication initiations all around the chromosome (Kogoma and von Meyenburg, 1983). Therefore, if confirmed, the increased acidity of phospholipids in the seqA mutants would have increased the initiation potential, supporting an additional mechanism for replication overinitiation in the seqA mutants unrelated to the lack of the origin sequestration (Fig. 1, compare the DnaA cycle and the oriC cycle). This possible additional DnaA activation in seqA mutants would explain our isolation of the rfa suppressors of the seqA recA lethality as a compensation of this change in the overall membrane charge, especially so that the DnaA activation by acidic lipids is not specific for a particular lipid chemistry or even lipid structure (Castuma et al., 1993; Garner and Crooke, 1996). Accordingly, the rfa suppressors, by decreasing the charge of the outer membrane, could have reduced the overall membrane charge, shifting the ADP/ATP in the DnaA population towards the inactive DnaA species and thus, reducing unscheduled initiations.
We analysed the phospholipid composition in wild type and ΔseqA mutants of E. coli after labelling phospholipids with 32P and separating them in one-dimensional TLC (Fig. 6B) (Cronan, 1968). A typical phospholipid composition in wild-type cells is 80–85% PE, the remaining 15–20% being distributed between PG and CL (Cronan and Vagelos, 1972). We confirmed this composition for wild-type cells (Fig. 6C). However, we did not detect any significant differences between two different wild-type strains and their ΔseqA derivatives in phospholipid composition, although we did detect a significant increase in the acidic species as the cells moved from growth phase into the stationary phase (Fig. 6C). We conclude that the rfa suppressors are not compensating for any changes in the phospholipid composition.
Ori/ter ratio, SOS induction and chromosomal fragmentation in seqA rfa mutants
At the chromosomal level, the major difference between the seqA mutants versus wild-type cells is the greatly increased ori/ter ratio in the former (which is > 6 at 30°C, compared with < 2 in the seqA+ parent) (Fig. 7A). Thus, the expected general consequence of suppression of the seqA recA lethality is reduction of the high ori/ter ratio of the ΔseqA mutants (Fig. 7A). The ori/ter ratio analysis of the seqA rfa double mutants indeed showed a significant reduction in the ratio (to 3–4) for all three rfa suppressors, although not to the wild-ype levels (Fig. 7A, centre). A trivial reason for the decreased ori/ter ratio could be a slower growth of the rfaQGP mutants: as already mentioned, growing the ΔseqA recA629 double mutant on a minimal medium or in anaerobic conditions did relieve the lethality slightly (Fig. 3B). However, simply increasing the doubling time of the seqA mutants to match the doubling time of seqA rfa double mutants by supplementing the growth medium with substatic concentrations of tetracycline (0.05–0.1 μg ml−1) did not reduce the ori/ter ratio of the former (not shown). Thus, the rfa suppressors are likely to reduce the ori/ter ratio in the chromosomal DNA of seqA mutants directly.
The ΔseqA mutants have increased chromosomal fragmentation, detectable in the recBC-deficient background, which is presumed to be the ultimate reason for the seqA recA colethality (Kouzminova et al., 2004). Thus, we expected that the viability of the seqA recA rfa triple mutants would reflect decreased chromosomal fragmentation in the double seqA rfa mutants. To quantify chromosomal fragmentation, we introduced the rfa suppressors into the seqA recBC(Ts) double and recBC(Ts) single mutants. Surprisingly, we found no significant alleviation of the high chromosomal fragmentation levels in the ΔseqA21 recBC(Ts) mutant by the rfaQ, rfaG or rfaP suppressors (Fig. 7B, centre). At the same time, the rfa recBC(Ts) double mutants (seqA+) showed the recBC(Ts) levels of chromosomal fragmentation (Fig. 7B, right). We conclude that: (i) lowering of the ori/ter ratio does not necessarily translate into reduced chromosomal fragmentation and (ii) the rfaQGP defect allows the seqA recA mutants to tolerate the levels of chromosomal fragmentation that are apparently lethal for the RfaQGP+ strains.
The increased chromosomal fragmentation in the seqA mutants leads to a significant SOS induction (a transcriptional response of bacterial cells to chromosomal damage) (Kouzminova et al., 2004). Consistent with the high chromosomal fragmentation in the seqA rfa mutants, SOS induction in the double seqA rfa mutants was still high and was even higher than in the single ΔseqA mutant (Fig. 7C, centre), suggesting even more DNA damage. One possibility was that the rfa suppressors, while decreasing the SOS induction due to the seqA defect, caused some SOS response by themselves, which is not a reaction to double-strand breaks [absent in single mutants (Fig. 7B)]. This indeed was found to be the case for the rfaG and rfaP single mutants, but the magnitude of induction was not enough to account for the values in the seqA rfa double mutants (Fig. 7C, right). We conclude that the seqA and rfa defects have an additive effect on SOS induction, suggesting that the rfa suppressors do not decrease SOS in the seqA mutants but, instead, contribute their own induction.
Genetic analysis of the rfaP suppression of the seqA recA lethality
The increased number of the replication origins in the ΔseqA mutants is interpreted to reflect replication overinitiation (von Freiesleben et al., 1994; Lu et al., 1994), the reason for this overinitiation being the increased accessibility of oriC to binding by the DnaA initiator protein (Fig. 8A, the ΔseqA shunt). If overinitiation is indeed the problem of seqA mutants, the current understanding (Fig. 1) predicts two possible types of suppression mechanisms: one acting via the oriC loop (Fig. 8A, right) (for example, by providing alternative means of sequestering the origin), the other acting via the DnaA loop (Fig. 8A, left) (for example, via decreasing the ADP-DnaA→ATP-DnaA exchange). We sought to distinguish between the two ideas genetically, by combining the rfaQGP mutants with defects in the replication initiation.
First, employing a mutant DnaA, called DnaA(Cs) [or DnaA(cos)], which does not need the membranes to exchange ADP for ATP [Fig. 8A, the DnaA(Cs) shunt] because of the lower affinity to either nucleotide (Katayama, 1994), we tested the idea that the rfa suppression works through the oriC loop. As a result of its defect, DnaA(Cs) protein binds oriC more aggressively and is thought to overinitiate chromosomal replication in otherwise wild-type strains to such an extent that the cells are inhibited [since the dnaA(Cos) defect was isolated as a suppressor of a dnaA(Ts) defect, dnaA(Cs) cells are inhibited only at lower temperatures] (Kellenberger-Gujer et al., 1978; Simmons et al., 2004). The dnaA(Cs) mutant inhibition at low temperatures is suppressed by a defect in the positive initiation factor, the DNA adenine methylase Dam (Katayama et al., 1997; Nordman et al., 2007). We have confirmed this observation (Fig. 8B) and also showed that the same is true for the defect in DiaA (Ishida et al., 2004) (Fig. 8B), the only other positive factor to the right of the DnaA(Cs) shunt (Fig. 8A). On the other hand, we found that inactivation of Hda has no effect on the DnaA(Cs)-overproducing strain (Fig. 8B), perhaps because the Hda-promoted ATP hydrolysis cannot inhibit the DnaA(Cs) protein, which exchanges nucleotide cofactors freely. Somewhat unexpectedly, inactivation of the other negative replication initiation regulator, SeqA (Fig. 8A), has no effect on the DnaA(Cs)-overproducing strain either (Fig. 8B). Even more surprisingly, we found that the rfaP mutant somewhat exaggerated the inhibition of the DnaA(Cs)-producing strain, instead of alleviating it, as dam and diaA defects did (Fig. 8B). This result suggests the rfa suppression works outside of the oriC cycle – one possibility was that it works in the DnaA cycle, by negatively affecting a step to the left of the DnaA(Cs) shortcut (Fig. 8A).
To test the idea that the rfaQGP suppression works through the DnaA loop, for example by affecting DnaA interaction with membranes, we tested if the rfaQGP defect would exacerbate the temperature sensitivity of the dnaA(Ts) mutants. However, we found no significant interaction between the rfa mutants and four different dnaA(Ts) strains (dnaA5, 46, 177 and 508, Fig. S1A and data not shown) – perhaps because the mutant DnaA proteins were not affected in the nucleotide exchange. Therefore, we introduced the rfaP defect into the dnaA T174P and dnaA A345S mutants that spontaneously hydrolyse ATP and thus underinitiate (Gon et al., 2006); likewise, no interaction between these dnaA defects and the rfaP defect was apparent, with the exception of the peculiar suppression of mucoidy at low temperature (Fig. S1B). We interpret the results of this epistatic analysis to mean that: (i) LPS phosphorylation acts outside of the two known cycles of regulation of replication initiation (Figs 1 and 8A) and (ii) there is no direct interaction between DnaA and LPS of the type proposed for acidic phospholipids.
To test whether LPS phosphorylation affects replication initiation at all, we introduced the rfaP mutant into the double pgsA lpxB (formerly pgsB) mutant that synthesizes reduced amounts of PG and CL at low temperatures and stops their synthesis altogether at 42°C (Nishijima and Raetz, 1979; Nishijima et al., 1981). Since, as mentioned above, these two acidic phospholipids promote the nucleotide exchange in DnaA in vitro (Sekimizu and Kornberg, 1988; Crooke et al., 1992), the inability of pgsA lpxB mutant to form colonies at 42°C may be in part due to a defect in replication initiation, especially since a similar inviability of a different pgsA null mutation was reported to be suppressed by providing alternative means for initiation of chromosomal replication through the rnhA defect (Xia and Dowhan, 1995) [although we were unable to confirm this observation (Figs. S2 and S3)]. We found that we can build the triple mutant pgsA lpxB rfaP at 28°C, but the mutant was severely inhibited at 34°C (mostly due to the pgsA lpxB defect) and essentially non-viable at 23°C [due to the rfaP defect, which is worse at lower temperatures (Parker et al., 1992)] (Fig. 8C). We confirmed that the triple mutant's growth problems were due to the replication initiation defect, by inserting an IPTG-induced plasmid replication origin near oriC– the strains with the plasmid origin grew much better both in the presence and in the absence of IPTG (Fig. 8D, rows 3 and 6). On the basis of our results we propose that rfaQGP inactivation suppresses the seqA recA colethality by lowering the ori/ter ratio to manageable levels via decreasing replication initiation, working outside of the two known regulation cycles and without direct interactions with DnaA itself.
To get insights into the mechanisms of chromosomal fragmentation in ΔseqA mutants we designed a genetic system for isolation of suppressors of the seqA recA lethality. We found that seqA mutants have increased LPS phosphorylation, whereas a major class of ΔseqA recA suppressors, rfaQGP, has reduced LPS phosphorylation, suggesting an initial cause of both lethality and its suppression, but not a mechanism. As expected, the rfaQGP suppressors did reduce the elevated ori/ter ratio of the ΔseqA mutants but, unexpectedly, the suppressed strains still exhibited the high levels of chromosomal fragmentation, characteristic of the non-suppressed seqA mutants. We also found that the rfaP defect is colethal with the defect in the production of acidic phospholipids, and the colethality is suppressed by alternative initiation of chromosomal replication, suggesting a role of LPS phosphorylation in replication initiation. We failed to find genetic evidence for direct interactions between LPS and the DnaA initiator protein, though.
The finding that the rfaQGP suppressors improve growth of the seqA mutants despite the same level of chromosomal fragmentation is surprising. In other words, chromosomal fragmentation does not seem to be as lethal as the increased ori/ter ratio. Other suppressors of the seqA recA lethality restore viability because they concurrently reduce both the ori/ter ratio and chromosomal fragmentation (Fig. 7B and C, the ybfE control; E. Rotman and A. Kuzminov, unpublished). A face-value explanation of the unexpectedly high levels of chromosomal fragmentation in the rfa-suppressed seqA mutants is that the increase in ori/ter ratio and the chromosomal fragmentation are not always causally related, as is currently assumed, but can affect cell viability independently of each other. We are not aware of any additional evidence to corroborate this point, however. Another explanation is that formation of subchromosomal fragments in the seqA mutants is not lethal per se, and an unknown signalling pathway is required to stop the cell growth by the accumulating subchromosomal fragments, while the rfaQGP defects interfere with this signalling independently of lowering the ori/ter ratio. For example, the cell cycle in a lower eukaryote, the budding yeast, can be permanently halted by one double-strand DNA break in a nonessential DNA molecule, while the rad9 defect in the DNA damage-signalling pathway alleviates this lethal block (Bennett et al., 1993). If the RfaQGP proteins are indeed involved in signalling between DNA damage and the bacterial cell cycle, this will be the first example of such function. Yet another way to explain the high chromosomal fragmentation in the rfaQGP-suppressed seqA mutants is to propose two different mechanisms of this fragmentation, with one mechanism producing the bulk of fragmentation, which is not lethal, while the other mechanism being responsible for a small fraction of lethal fragmentation events. According to this explanation, it is the second (lethal) mechanism that the inactivation of rfaQGP blocks.
Another interesting aspect of the rfa suppression of the replication overinitiation defect is that LPS forms the outer leaflet of the outer membrane, whereas DNA metabolism is clearly cytoplasmic. Remarkably, both SeqA and origin DNA have been isolated from outer membrane extracts (Ogden et al., 1988; Chakraborti et al., 1992; Slater et al., 1995; d'Alençon et al., 1999), although in some cases, the origin is associated with a particular intermediate (neither inner nor outer membrane) fraction (Chakraborti et al., 1992). Since it is known that: (i) the DnaA protein is activated for initiation by acidic phospholipids (Sekimizu and Kornberg, 1988; Crooke et al., 1992; Xia and Dowhan, 1995), but this activation is not dependent on a particular lipid chemistry (Castuma et al., 1993; Garner and Crooke, 1996); (ii) as just mentioned, the origin DNA is found in complex with the outer membrane (LPS), rather than with the inner one (phospholipids); and (iii) the rfaQGP mutants reduce phosphorylation of LPS – we imagined that LPS could interact with DnaA protein directly, while its reduced phosphorylation (= acidity) would translate into a reduced activation of DnaA, explaining the mechanism of the rfaQGP suppression of the overinitiation in seqA mutants. Since we found: (i) no evidence for direct interaction between DnaA and LPS, (ii) increased ori/ter ratio in mutants with increased LPS phosphorylation (seqA) and (iii) decreased ori/ter ratio in mutants with decreased LPS phosphorylation (rfaQGP), LPS phosphorylation could stimulate replication initiation outside of the two known regulation cycles (Figs 1 and 8A) via yet-to-be-identified factors.
From the perspective of the cell envelope structure, our rfa suppressors offer a genetic confirmation for the previous reports of the direct contacts between DNA and the outer membrane. For such patently cytoplasmic entities as SeqA and the replication origin DNA to have contacts with the component of the outer surface of the cell, there must be places in the cell, where the outer membrane comes in contact with the cytoplasm. Such places are most likely junctions connecting the outer and the inner membranes. Two types of such junctions are generally acknowledged, although their existence is still debated (Nikaido, 1996; Oliver, 1996). The first type of junctions are periseptal annuli; these are circumferential zones of adhesion of the two membranes positioned either at the poles or in the middle of the cell (Macalister et al., 1983; Cook et al., 1986). The second type is represented by much smaller and dispersed junctions, numbered 200–400 per cell, which have been seen only in growing cells and are known as ‘Bayer patches’ (Bayer, 1968). They are proposed to be sites of fusion of the two membranes, where LPS, capsules and proteins of the cell wall are synthesized (Bayer, 1979) (Fig. 9). Additionally, chromosomal DNA has been photographed interacting with these junctions (Bayer, 1979). It is interesting to note that the rfaQGP suppression of the seqA recA lethality provides genetic evidence to support the idea that parts of the outer membrane, most likely continuous with the inner membrane, are in direct contact with the cytoplasm in general and with the chromosomal DNA in particular (Fig. 9).
Cells were grown in LB (10 g tryptone, 5 g yeast extract, 5 g NaCl, 250 μl 4 M NaOH per 1 l) or on LB agar (LB supplemented with 15 g of agar per 1 l). M9 minimal plates (Miller, 1972) contained 1× M9 salts, 2 mM MgSO4, 0.1 mM CaCl2 and were supplemented per 1 l with 10 mg thiamine (B1), 15 g agar and 2 g galactose or glucose. MOPS minimal phosphate medium was as described (Neidhardt et al., 1974). Ampicillin (100 μg ml−1), kanamycin (50 μg ml−1), spectinomycin (100 μg ml−1) and chloramphenicol (10 μg ml−1) were added when the strains carried the corresponding antibiotic resistance.
Bacterial strains and plasmid
Escherichia coli strains used in these experiments were all K-12, and most of them were derivatives of AB1157 (Table 1). L-216 has an IPTG-dependent ColE1-derived replication origin, inserted in the chromosome at position 4054 Mbp (E. Kouzminova, pers. comm.; Kouzminova and Kuzminov, 2008). Precise deletion-replacement alleles of selected genes were created and confirmed by PCR by the method of Datsenko and Wanner (2000). Alleles were moved between strains by P1 transduction (Miller, 1972). The rec mutants were confirmed by their characteristic UV sensitivities. The dnaA T174P and A345S alleles were confirmed by sequencing. The plasmids are described in Table 2.
Growth at the ‘non-permissive’ temperature was first assayed by diluting an overnight culture 100-fold, growing it to 5 × 108 cells ml−1, diluting 0.2 μl in 5 ml of 1% NaCl and spotting by 10 μl. Since preliminary results indicated that the viability of saturated cultures was similar to that of rapidly growing cultures, in subsequent assays we spotted 10 μl of a 10−6 dilution of a saturated culture. Plates were incubated at either 42°C [permissive temperature for the recA629(Cs) allele] or 30°C (the non-permissive temperature). Colonies were given approximately 24 h to grow at 42°C and 48 h at 30°C.
Mutants in MG1655 background were grown on M9 galactose minimal medium for 2 days (wild type) or for 4 days (pgm mutant) until colonies were approximately 1 mm in diameter. The plates were overlaid with 3 ml top M9 iodine agar containing 1× M9 salts, 7.5 g agar, 10 g potassium iodide and 1 g iodine per 1 l. Within seconds, pgm mutants formed a dark blue border, which then disappeared within 20 min (Adhya and Schwartz, 1971).
pRL27 is a plasmid containing a hyperactive Tn5 transposase gene under the tetA promoter and a separate insertional cassette consisting of kanamycin resistance and the pir-dependent origin of replication oriR6K, bracketed by Tn5 inverted repeats (Larsen et al., 2002). Electroporation of this plasmid in a pir– background causes the mini-transposon to insert randomly into the chromosome. The ΔseqA21 recA629 cells were electroporated with 10 ng of pRL27 and outgrown for 1 h and 20 min at 42°C before being plated on LB kan (10 μg ml−1) at 30°C. Colonies were streaked after 2 days of growth onto LB supplemented with 50 μg ml−1 kanamycin next to the original ΔseqA21 recA629 double mutant and the single recA629 mutant (with a kan-insert at an irrelevant gene to provide kanamycin resistance) as a positive control. If the strain was able to form colonies, a P1 lysate was made at 42°C and used to transduce the parental ΔseqA21 recA629 double mutant, also at 42°C. Ten microlitres of a 10−6 dilution of the transductants was used to test the ability to grow at the non-permissive temperature of 30°C. The confirmed suppressors linked to the kanamycin-resistance insertion were identified as before (Kouzminova et al., 2004; Ting et al., 2008), after digesting chromosomal DNA with MluI (which has no sites in the insertion cassette), circularizing the digested fragments by ligation and transforming into DH5αpir+. The plasmid with interrupted gene was selected for by plating with kanamycin, and the gene was identified using primers (Bradshaw and Kuzminov, 2003) facing outwards from the insertion element's ends.
Pulsed-field gel electrophoresis
Overnight LB cultures were diluted to OD600 = 0.02 in LB and grown in the presence of 2.5–10 μCi of 32P-orthophosphoric acid for 1 h at 22° and 3 h at 37°. All cultures were then brought to OD600 = 0.35. Cells from 0.5 to 1 ml aliquots were spun down, washed in 1 ml of TE and resuspended in 60 μl of TE. To this cell suspension, 2.5 μl of proteinase K (5 mg ml−1) and 65 μl of 1.2% agarose in Lysis buffer (see below) were added, and mixed by pipetting. One hundred and ten microlitres of the mixture was then pipetted into the plug mold and allowed to solidify. The plugs were incubated overnight at 60°C in Lysis buffer (1% sarcosine, 50 mM Tris-HCl and 25 mM EDTA). Samples were loaded into a 1.0% agarose gel in 0.5× TBE buffer and run at 6.5 V Cm−1 with a pulse time of 90 s for 7 h, 105 s for 8 h and 125 s for 8 h in Gene Navigator (Pharmacia). The gel was vacuum-dried onto a piece of chromatography paper (Fisher) for 2 h at 80°C and then exposed to a PhosphorImager screen until signals from the wells reached between 300 000 and 900 000 counts. If unlabelled, the gel was stained for 30 min in 0.5 μg ml−1 ethidium bromide and de-stained for 30 min in deionized water before pictures were taken.
To determine the level of SOS induction in the cell, the ΔseqA20::kan mutation was P1-transduced into AK43 (Kouzminova et al., 2004), a strain containing a Mu ΔX cat derived construct with the lacZ gene fused under the sfiA promoter (Ossanna and Mount, 1989). To generate ΔseqA21 PsfiA→lacZ, we P1-transduced the Mu ΔX cat into ΔseqA21. Other mutants were later P1-transduced into this seqA PsfiA→lacZ reporter background. When the cells are under SOS-induced stress, either by the mutation or external DNA damage, the promoter is expressed, and the level of β-galactosidase can be quantitatively measured by the modified protocol of Miller (1972), using 200 μl of culture (Kouzminova et al., 2004). As a positive control, wild-type cells containing the PsfiA→lacZ fusion were treated with 100 ng ml−1 Mitomycin C, a cross-linking agent. At these Mitomycin C concentrations, cells continue slower growth.
Dot-hybridization to determine ori/ter ratio
Total DNA was extracted from saturated and exponentially growing (OD600 = 0.6) cultures by phenol/chloroform method (Kouzminova and Kuzminov, 2006). Two micrograms of this DNA was denatured in 400 μl of 0.1 M NaOH for 15 min at 37°C and spotted in duplicate on a positively charged Nylon membrane (Amersham) using a vacuum manifold. After cross-linking the DNA to the membrane with UV, the membrane was divided in two with one half hybridizing to an origin-proximal probe and the other half to a terminus-proximal probe (Kouzminova and Kuzminov, 2006).
Total phospholipids were extracted by the method of Bligh and Dyer (1959). 32P-labelled cells were pelleted, resuspended in the remaining liquid and mixed with 4 ml methanol and 2 ml chloroform. After an hour tumble, 2 ml chloroform and 2 ml deionized water were added, and the suspension was mixed and centrifuged to separate the phases. The aqueous phase was removed, and 4 ml of 2 M KCl was used for another extraction, followed by yet another extraction with 3 ml of water. The remaining 3 ml of organic phase was dried under a stream of nitrogen and dissolved in 100 μl 2:1 chloroform:methanol (or 1 ml was dried and dissolved in 33 μl of the chloroform/methanol mixture). Running lanes were generated by scraping on a Silica-G TLC plate (Analtech), which was then baked at 100°C for 30–60 min, and 8–12 μl of sample was loaded per lane. The samples were run in a buffer containing 97.5 ml chloroform, 32.5 ml methanol and 12 ml glacial acetic acid until the front reached the top of the plate. The TLC plate was exposed to a PhosphorImager screen until signals from the PG spots exceeded 500 000 counts.
Detection of LPS phosphorylation
Bacterial cultures were grown in MOPS-minimal phosphate-limiting medium to an OD of 0.35–0.4 and pulse-abelled with ∼10 μCi 32P orthophosphate for 5 min before centrifugation, and then processed according to the ‘total DNA isolation protocol’ (Kouzminova and Kuzminov, 2006). Cells were resuspended in 50 μl of 30% sucrose in TE buffer and lysed for 5 min at 70°C after addition of 350 μl of 2% SDS in TE and careful mixing. The lysate was extracted with 400 μl of phenol, followed by 400 μl of phenol : chloroform (1:1), and 400 μl of chlorofom. The aqueous phase was ethanol-precipitated twice and dissolved in 40 μl of TE buffer. Twenty microlitres of this sample was combined with 20 μl of 2× alkaline agarose loading dye, and 20 μl was separated on a 1.1% alkaline agarose gel (Maniatis et al., 1982) for 620 min at 20 V (1.4 V cm−1).
This work was supported by Grant # RSG-05-135-01-GMC from the American Cancer Society and by Grant # GM 073115 from the National Institutes of Health.