An expanded view of bacterial DNA replication

Abstract

A protein-interaction network centered on the replication machinery of Bacillus subtilis was generated by genome-wide two-hybrid screens and systematic specificity assays. The network consists of 91 specific interactions linking 69 proteins. Over one fourth of the interactions take place between homologues of proteins known to interact in other organisms, indicating the high biological significance of the other interactions we report. These interactions provide insights on the relations of DNA replication with recombination and repair, membrane-bound protein complexes, and signaling pathways. They also lead to the biological role of unknown proteins, as illustrated for the highly conserved YabA, which is shown here to act in initiation control. Thus, our interaction map provides a valuable tool for the discovery of aspects of bacterial DNA replication.

The replication of the bacterial chromosome is carried out by a large multiprotein machine, the replisome, in which the activities of individual polypeptides are highly coordinated to achieve efficient and faithful DNA replication. The components of bacterial replisomes have been characterized extensively, revealing the molecular mechanisms at work in a DNA-replication apparatus (1, 2). Localization studies indicated that the replication machinery is preferentially at midcell, suggesting a factory model of replication in which the DNA template moves through a rather stationary polymerase (3). In contrast, the origin regions of the chromosomes are moving toward the cell poles during cell-cycle progression (4, 5). However, other aspects of DNA replication still remain unclear. For instance, it is not known how the replication machinery coordinates its action with other cellular processes in a variety of environmental conditions or what the determinants that specify replisome or origin positions within the cell are. Mutants affected in these biological processes have not been reported yet, possibly because they display weak or inconsistent phenotypes caused by redundant functions.

To gain insight into this unexplored area, we used genome-wide yeast two-hybrid screens (6) to identify the proteins that physically associate with known replication proteins from the Gram-positive bacterium Bacillus subtilis. To circumvent one of the main limitations of the approach, the false-positive interactions, we verified experimentally the specificity of every potential interaction identified in the screens. The resulting protein network is composed of 91 specific interactions connecting 69 proteins. Over one fourth of the interactions were described previously in bacteria or eukaryotes, showing that our approach yields biologically significant interactions. The remaining interactions are previously uncharacterized, and in combination with data from the literature, many of their biological roles can be hypothesized. They link DNA replication with DNA recombination and repair, potential origin- and replisome-anchoring membrane complexes, signaling pathways, and numerous proteins of unknown function. Moreover, we show that the unknown protein YabA, similar only to unknown proteins, acts in initiation control, as predicted from protein interactions.

Materials and Methods

Strains, Plasmids, and Media.

B. subtilis ORFs (7) were PCR-amplified from strain 168 genomic DNA and cloned into a pGBDU bait vector (URA3), and the hybrid plasmids were isolated from Escherichia coli and transformed into PJ69–4a yeast strain (8). DNA sequences of all cloned fragments were verified. The yeast strain PJ69–4α was a gift from P. James. For trihybrid experiments, a p3H vector (TRP1) was generated by deleting the GAL4 DNA binding domain (BD) from pGBD vectors (8) and replacing it with the nuclear localization signal from simian virus 40 T antigen and an S-tag sequence (Novagen). The media for growth of E. coli, B. subtilis, and Saccharomyces cerevisiae were as described (9–11).

Library Construction.

Three B. subtilis genomic libraries (BSLs), BSL-C1, BSL-C2, and BSL-C3, were constructed in E. coli (DH10B, Life Technologies, Cergy Pontoise, France) strain as described previously (8). Restrictions of the 4.2-megabase B. subtilis chromosome produced ≈1.6 × 10⁵ DNA ends that could be ligated into the pGAD prey vectors. Each library contained at least 2.5 × 10⁶ clones, thus providing a 15-fold redundancy. The PJ69–4α yeast strain was transformed by each library DNA as described (12), and at least 1.5 × 10⁷ prey-containing colonies were harvested and pooled. Aliquots containing ≈1 × 10⁸ transformed yeast cells were stored at −80°C.

Library Screening.

A mating strategy derived from that described in ref. 13 was used. For each screen, one BSL-C1, BSL-C2, and BSL-C3 vial was thawed, and cells were mixed with PJ69–4a cells containing the bait plasmid. The mixture then was plated on rich medium and incubated for 5 h at 30°C. Cells were collected, washed, spread on 120 SC-LUH (synthetic complete medium lacking leucine, uracil, and histidine) plates containing an appropriate concentration of 3-aminotriazole, and incubated 10–12 days at 30°C. Mating efficiencies were such that 15–30% of the library-containing cells became diploids, allowing us to test over 3 × 10⁷ possible interactions in one screen. His⁺ colonies were transferred on SC-LUA (synthetic complete medium lacking leucine, uracil, and adenine) plates and incubated 3–5 days at 30°C. All the His⁺ Ade⁺ colonies were organized in 96-well plates and subjected to PCR amplification. Inserts junctions with the GAL4 activation domain (AD) were sequenced and compared with the B. subtilis genome by using the BLAST program and the MICADO database.

Interaction Specificity.

To screen out false-positive interactions, protein-encoding prey plasmids were rescued from His⁺ Ade⁺ colonies in E. coli KC8 strain, purified, and reintroduced in PJ69–4α strain by transformation. For each prey, several transformants mated with bait-containing PJ69–4a strains as described in ref. 14. Cells were transferred to SC-LU medium with a replicating tool and incubated 2–3 days at 30°C to select for diploids, which then were transferred to SC-LUA/SC-LUH supplemented with 0.5 mM 3-aminotriazole and SC-LU plates. Interaction phenotypes were scored after 10–12 days of growth at 30°C.

For tri-hybrid experiments, PJ69–4a strain was cotransformed by different combinations of bait and 3H vectors. Ura⁺ Trp⁺ colonies were mated with PJ69–4α strains containing various prey vectors, and diploids were selected on SC-LUW (synthetic complete medium lacking leucine, uracil, and tryptophan). Interaction phenotypes were scored by replica-plating the diploids onto selective plates. SC-LUWH (SC-LUW lacking histidine and containing 0.5 mM 3-aminotriazole) and SC-LUWA (SC-LUW lacking adenine).

4′,6-Diamidino-2-phenylindole Staining and Nucleoid DNA Content Analysis.

Cells were grown in LB medium to midexponential phase (OD₆₀₀ 0.3–0.5) and fixed with glutaraldehyde as described previously (15). Fixed samples were examined by phase-contrast and fluorescence microscopy on a Zeiss Axioplan equipped with a Plan NEOFLUAR ×100/1.30 objective and a Princeton Instruments Micromax 1300Y/HS charge-coupled device camera (Roper Scientific, Trenton, NJ). For DNA-content determination, an internal standard was added to each sample and used to determine relative DNA content in each field of cells essentially as described in ref. 16. The digital images were quantitated by using METAMORPH 4.6 software, and the data were processed with Microsoft EXCEL. The relative DNA contents given represent an underestimation of the actual DNA content for reasons outlined in ref. 17 and should be multiplied by a factor of ≈1.3 to arrive at actual DNA contents in chromosome equivalents.

OriC Region Copy-Number Determination.

To visualize the location of the replication origin, the oriC-labeling system based on the lacO array (18) or on the Spo0J-green fluorescent protein (GFP) fusion (19) were introduced into the wild-type B. subtilis strain (168) and the yabA null mutant (168ΔyabA∷phleo). The lacO array was amplified and the strains were grown for microscopy essentially as described in ref. 18. The cells carrying GFP fusions were viewed on agarose-coated slides essentially as described in ref. 20. The LacI-GFP strains were grown in minimal medium (TSM) at 30°C, and the Spo0J-GFP strains were grown on nutrient agar overnight at 30°C.

Results

Construction of the Protein Interaction Map.

A BSL was constructed in a GAL4 AD (prey) vector in E. coli. It contained over 10⁷ clones with an average of one fusion endpoint every 52 bp along the chromosome. BSL was transferred into yeast and screened with BD hybrid proteins (baits). The screening was performed by mating, generating large numbers of diploids that were plated directly on selective medium. The preys from all the colonies exhibiting the interaction phenotypes were identified by sequencing. A systematic verification of the specificity of each interaction was carried out to eliminate false positives. For this purpose, prey plasmids were isolated from candidate colonies and reintroduced in haploid yeast. These prey cells were mated with cells containing (i) the initial bait used to screen the library, (ii) an empty bait vector, and (iii) at least six unrelated baits, and the resulting diploids were tested for interaction phenotypes. Interactions were considered specific if reproduced twice independently with the initial bait and not associated with self-activation (tested with the empty bait vector) or stickiness (tested with unrelated baits) of the prey protein.

The 13 proteins chosen as initial baits (Table 1) are key components of the replication apparatus (1). Generally, the full-length proteins were fused to the BD. For the large DNA polymerases (Pols), PolC and DnaE, fusions with two overlapping protein fragments were constructed. For DnaC, only the C-terminal fragment of the DNA helicase was used as a bait because of the self-activation of the reporter genes induced by the N-terminal part of the protein.

Table 1.

Features of two-hybrid screens

Screen	Bait	Bait identification	Diploids (×107)	His+ Ade+ colonies	Pairwise interactions
First cycle	DnaB	Primosome subunit	11.6	152	4
	DnaC_Cter	Replicative DNA helicase [151–454]	7.9	451	7
	DnaC_Nter	Replicative DNA helicase [1–170]	s.a.	—	0
	DnaD	Primosome subunit	6.8	82	1
	DnaE	DNA polymerase [1–1,115]	7.8	48	4
	DnaE_Nter	DNA polymerase [1–826]	5.1	52	1
	DnaE_Cter	DNA polymerase [286–1,115]	14.5	8	1
	DnaG	DNA primase	13.3	51	8
	DnaI	Primosome subunit	8.8	277	2
	DnaN	DNA sliding clamp (β)	4.2	209	6
	DnaX	Replisome subunit (τ)	6.4	11	3
	HolB	Replisome subunit (δ′)	16.7	20	1
	PcrA	DNA helicase	7.6	30	4
	PolC_Nter	DNA polymerase III [1–993]	14.0	11	1
	PolC_Cter	DNA polymerase III [435–1,438]	12.1	5	1
	PriA	Primosome subunit	9.1	8	0
	HolA	Replisome subunit (δ)	9.8	312	6
Second cycle	YabA	Prey of DnaN	12.3	174	7
	YhaM	Prey of DnaC	14.8	204	3
	YhaN	Prey of DnaG	8.5	61	1
	YjcK	Prey of DnaE	10.4	38	1
	YtjP	Prey of DnaE	12.2	24	1
	YxaL	Prey of PcrA	8.8	450	8
	YxeE	Prey of DnaG	s.a.	—	0
Third cycle	DnaA	Prey of YabA	12.6	468	4
	TlpA	Prey of YabA	9.1	440	16

Baits were GAL4 DNA-binding fusion with full-length proteins except for PolC, DnaC, and DnaE, where fusions with two overlapping protein fragments were constructed. The coordinates of the fragments in amino acids are indicated in brackets. Greek letters indicate polymerase subunit designations in bacteria (25). The lack of correlation between the number of His⁺ Ade⁺ colonies and the number of specific pairwise interactions is caused in part by the preys isolated as multiple overlapping fragments and to individual fragments obtained several times because of the library's high redundancy. Also, the specificity assays revealed additional pairwise interactions that were not detected in the screens. s.a., self-activating bait. 

The first cycle of library screenings yielded 38 specific protein–protein interactions. In the second cycle the ORFs corresponding to seven preys were used as baits, and in the third cycle two ORFs were used as baits, adding to a total of 32 new interactions. In addition to the screenings, most of the bait proteins also were expressed as preys and used in specificity assays. These assays revealed 21 interactions that were not detected in the screens, possibly because of nonsaturated screens or the scarcity of full-length ORFs in the prey library. Interestingly, we found that all the preys (46/46) isolated from the screens as independent overlapping fragments interacted specifically with their bait, which is in line with previous observations (21, 22), whereas only one prey out of six (24/148) isolated as a unique fragment did so. Although this proportion depends on our experimental system, it confirms that the level of false positives in genome-wide two-hybrid screens can be very high (23) and indicates that false interactions have to be eliminated to assess correctly the biological significance of the information derived from such screens.

The resulting network is composed of 91 specific interactions involving 69 proteins (Fig. 1). These proteins can be grouped in eight functional categories (Table 2), thus placing the DNA-replication process into a larger biological context. The biological significance of these interactions is discussed below.

Figure 1.

Protein–protein interaction map. Proteins are symbolized by ovals, and protein names are as described in ref. 7 except for HolA (YqeN). Primary baits are indicated by a double blue line except PriA and DnaD, which were omitted. The dashed line symbolizes a bacterial replisome. The arrows are oriented from the bait to the prey. The complete data set including the interacting domains of proteins can be viewed at http://www-mig.versailles.inra.fr/bdsi/SPiD. Functional categories are represented with colors: orange, DNA replication/recombination/repair; dark blue, mobility and chemotaxis; light blue, signal transduction; light green, transcription; pink, protein synthesis; dark green, metabolism of carbohydrates and amino acids; purple, cell division; and yellow, unknown.

Table 2.

Summary of experimental results

Description	Total
Yeast two-hybrid screens
B. subtilis proteins used as baits	22
Two-hybrid screens performed	25
Baits yielding specific interactions	20
Specific interactions identified	91
Discrete interacting protein pairs	86
Average interactions per bait	4
Functional classification*
Interacting proteins	69
DNA replication/repair/recombination	25
Mobility and chemotaxis	5
Signal transduction	3
Transcription	2
Protein synthesis, elongation	1
Metabolism of carbohydrates, amino-acids, and nucleotides	7
Cell division	1
Unknown	25

^*As defined in ref. 7. 

Protein-Interacting Domains.

Our experiments yielded information on the protein domains required for the interaction with the baits. Isolation of multiple independent overlapping fragments of a prey allowed delimitation of the interaction domain. The complete data are available in a dedicated database at http://www-mig.versailles.inra.fr/bdsi/SPiD (24). Users can search and visualize the interactions and the interacting domains, as well as access functional and bibliography information through links with specialized databases.

Deletion of Unknown Genes.

Within the network, we have identified 25 proteins of unknown function (Table 2) having either no significant sequence similarity to other known proteins (8) or similarity to one or more hypothetical proteins (17). We deleted 10 genes by individually replacing them with a phleomycin resistance marker (yabA, yhaM, yjcK, ykaA, ykcC, ykjA, yncB, ytjP, ywhK, and yxeE), and disruption mutants were described for 12 other genes (yacL, yclM, ydeL, yerB, yhcQ, yisN, yklV, ymaF, ytdP, yufO, yxaD, and yxaL) in the two B. subtilis functional databases MICADO (http://locus.jouy.inra.fr/micado) and JAFAN (http://bacillus.genome.ad.jp). A further three genes (yomI, yoqX, and yqaH) belonged to dispensable integrated prophages. Thus, all these genes are individually dispensable for cell survival. In contrast, we found that yqeN encoded a protein essential for B. subtilis. The YqeN protein showed sequence similarity to the Streptococcus pyogenes clamp-loader subunit HolA (25) and interacted with another clamp-loader subunit, DnaX, and the β clamp, DnaN (Fig. 1), which is consistent with HolA function. Thus, YqeN was renamed HolA.

Biologically Relevant Multiprotein Complexes.

To test the relevance of the interactions detected, we examined in more detail the properties of the previously unknown protein YabA, which is well conserved in Gram-positive bacteria but has no similarity with proteins of known function (data not shown). YabA appeared to interact with both the initiator protein DnaA and the sliding-clamp DnaN (Fig. 1), potentially linking the initiation and elongation steps of chromosomal replication. In E. coli, a mechanism of initiation control, mediated by the Hda protein, acts by inhibiting the DnaA initiator activity through interaction with DnaN (26, 27). YabA could accomplish a similar type of initiation control by bringing together DnaA and DnaN in a multiprotein complex. We showed that an interaction between DnaA and DnaN was detected, provided that YabA was coexpressed in the cell (Fig. 2). This result indicates that YabA can act as a bridge between DnaA and DnaN.

Figure 2.

YabA acts as a protein bridge between DnaA and DnaN. DnaA was expressed as bait (BD-DnaA), and DnaN was expressed as prey (AD-DnaN). YabA was expressed as the trihybrid protein (3H-YabA). Negative controls include BD, 3H, and AD expressed from empty vectors. The positive control is AD-YabA. The different combinations of bait and 3H vectors in PJ69–4a are indicated at the left of each row. The preys expressed in each strain are indicated at the top of each column. Interaction phenotypes were scored by replica-plating the diploids onto selective plates SC-LUWA.

We constructed a nonpolar yabA deletion mutation, which was viable but showed a greatly reduced growth rate especially in minimal media (data not shown). Microscopic examination of the mutant (Fig. 3B) revealed an abnormal nucleoid distribution, with increased cell length, which is probably an indirect consequence of the nucleoid effect. Several lines of evidence suggested that this phenotype was associated with increased initiation of DNA replication. First, we measured the relative DNA content of nucleoids of the wild-type and mutant strains (Fig. 3C), which revealed that the mutant had many nucleoids of much higher relative DNA content (>2.5) than the wild type, giving a significantly higher average DNA content of 2.3 versus 1.7 chromosome equivalents per nucleoid. We then directly examined the copy number of the oriC region in wild-type and yabA mutant cells by using two marker systems: a lacO/LacI-GFP tagging system (18) or Spo0J-GFP, which is associated naturally with the oriC region (28, 29). Previous work has shown that each focus represents one copy of the oriC region of the cell or two recently replicated copies. In wild-type cells containing either GFP protein, discrete foci were located mainly at regularly spaced intervals in the cells and cell chains (Fig. 3 D and F). In both cases, the general appearance was similar to that reported previously, with typically 2–4 foci per cell. In contrast, in the yabA mutant, the general number of foci per cell was clearly increased, even taking into account the increased average cell length (Fig. 3 E and G). This increase was particularly apparent with the Spo0J-GFP (Fig. 3G), possibly because the Spo0J protein associates with sites covering a broader region of the chromosome than with the LacO array (29). For the lacI-lacO system, which more specifically tags oriC, we determined the average number of foci per unit cell length; this was 3.46 for the wild type but almost 40% higher, 5.04, for the mutant cells (93 and 143 foci counted, respectively). Furthermore, an underestimation of foci number in the yabA mutant is likely because of the greater probability of two foci coinciding and being counted as a single focus and because some GFP foci would be out of focus in the longer and more twisted cells. Taken together, these results indicate that YabA acts as a negative regulator of initiation at oriC.

Figure 3.

Effects of a yabA mutation on cellular DNA content and initiation of DNA replication in vivo. (A and B) Combined phase-contrast and 4′,6-diamidino-2-phenylindole (DNA) images of wild type (A, WT) and yabA mutant (B) strains grown in LB medium. (C) Analysis of the relative DNA content of cells grown as described for A and B. (D and E) Analysis of oriC copy number in wild-type (D) and yabA mutant (E) cells expressing a LacI-GFP fusion targeted to a lacO cassette located near oriC. (F and G) Analysis of oriC copy number in wild-type (F) and yabA mutant (G) cells expressing an Spo0J-GFP fusion.

Discussion

Specific Interactions Are Biologically Relevant.

To assess the biological relevance of our screens, we compared our data with the known interactions between replisome proteins. We found that the DnaC helicase (the E. coli DnaB homolog) interacts with the DnaG primase and DnaX, the τ subunit of PolIII holoenzyme. DnaX interacts with HolB and HolA (YqeN) to form the β clamp loader (25). The replicative polymerase PolC interacts with DnaX and DnaN, the β sliding clamp. Self-interactions were observed with DnaN, DnaX, and the DnaC helicase. All the corresponding interactions in E. coli have been shown to be required for replication fork function (2). Also, we found that DnaI interacts with the DnaC helicase, as do the DnaI homologues from integrated phages, XkdC and YqaM. Recently it was proposed, based on the DnaI–DnaC interaction, that DnaI acts as a helicase loader (30). Interestingly, the PolC–DnaX interaction was shown to be relatively weak in Gram-positive bacteria (25). Its detection with the B. subtilis proteins shows that our screens are highly sensitive. An expected limitation of our approach, inherent to the use of DNA-replication proteins as baits, is that some interactions might depend on binding to DNA. One such example is that of the primosomal proteins DnaB and DnaD, which we found to interact with themselves, consistent with the multimerization observed with the purified proteins (31, 32). However, we did not detect the PriA–DnaD–DnaB interactions, which depend on the presence of a specific DNA substrate (32). Overall, we detected 16 of 20 interactions involving replisome and initiation proteins, which were reported in the literature, either in B. subtilis or other model organisms. This proportion suggests that the number of false negatives is low in our experiments. Thus, with the limitation that all possible interactions could not be detected, many biologically relevant interactions involved in DNA replication have been revealed.

Other known proteins seem to be associated with the DNA-replication machinery (Fig. 1). Besides PolC, DnaN was found to interact with YqjH, a UmuC/DinB homolog, and MutL, a mismatch repair protein. E. coli DnaN was shown to interact in vitro with the translesional Pols UmuD′₂C and DinB and increase their processivity to various extents (33, 34). The role of the DnaN–MutL interaction is suggested by the function in DNA mismatch repair of the interaction between the yeast sliding clamp, PCNA, and the MutL homolog MLH1 (35). Interestingly, the sliding clamp was found to interact with MutS or a MutS homolog in E. coli and yeast (35, 36), suggesting that interaction with mismatch repair is universally conserved, albeit through different protein complexes in the different organisms. Remarkably, B. subtilis MutL and MutS proteins were shown to form foci that can assemble at the replication machinery (37). PolI interacted with HolB, a subunit of the β clamp loader, suggesting that as in E. coli, PolI might function with the β clamp (36). Strikingly, PolI interacted with PolIII (PolC) and DnaE, another class C Pol. DnaE, similar to PolC, is required for the elongation step of chromosome replication, and both enzymes exhibit similar localization patterns (38), which suggests that PolI might play a structural role within the replisome in addition to a role in the processing of Okazaki fragments. The DnaG primase interacted with YirY and YhaN, which are homologs of E. coli SbcC. Moreover, YhaN interacted with YhaO, a SbcD homolog. The E. coli SbcCD nuclease that belongs to the structural maintenance of chromosomes (SMC) protein family, is involved in DNA recombination and exhibits an ATP-dependent degradation of DNA hairpin structures (39). The interaction of SbcCD-like enzymes with the DnaG primase supports the hypothesis that these enzymes act on the lagging strand at the replication fork (39). The finding of direct interactions between replication and recombination/repair proteins provides insights into how these two processes are linked. DNA repair and recombination functions are of critical importance for the restoration of replication forks, which often are arrested during chromosomal replication (for reviews see refs. 40 and 41). In conclusion, our approach allowed a meaningful reconstruction of the B. subtilis replication machinery, strongly indicating that specific interactions have biological relevance.

Replication Is Linked with the Membrane and Signaling Pathways.

Interestingly, a complex mesh of interactions linking DNA replication with membrane-associated protein complexes and signaling pathways were detected. The DnaC helicase interacted with BfmBAB, and the DnaG primase interacted with AcoC, BfmBAB, and PdhC. These proteins are components of three very large (5–10-megadalton) multienzyme complexes that catalyze the decarboxylation of acetoin and 2-oxo acids and transfer the corresponding acyl groups to CoA (42). PdhC, a subunit of the pyruvate dehydrogenase complex, was identified recently as a membrane-associated regulator of B. subtilis DNA replication in vitro (43). Thus, PdhC might be involved in DNA replication through its direct interaction with DNA primase. Noteworthy is the possibility that the 2-oxo acid dehydrogenases might transfer acyl groups to the DnaC and DnaG proteins. Considering the lack of knowledge about the role of protein acylation in prokaryotes, this hypothesis cannot be ruled out.

Several methyl-accepting chemotaxis proteins (MCPs), which are transmembrane receptors, seem to be linked to DNA replication. The cytosolic part of TlpA interacts with 17 different protein partners. Interactions with TlpA, McpA, the MCP homologs YhfV and YvaQ, and the chemoreceptor-controlled kinase CheA were expected based on E. coli data (44). They support the hypothesis that the different chemoreceptors interact with each other to form composite complexes. The MCPs form large clusters located at the cell poles in E. coli (45) and B. subtilis (46). We found that McpA and TlpA interact with YabA, suggesting that YabA might bring into close proximity the MCPs and the initiator DnaA. Furthermore, the DnaB initiator protein (different from the E. coli DnaB helicase) interacts with McpA and YvaQ. DnaB frequently localizes near the cell poles (30). Thus, our data suggest that the proximity of the polar MCPs with initiation proteins might play a functional role, possibly at a stage of the cell cycle when the origins are also oriented preferentially toward the poles (47). The interactions of replication proteins with massive membrane-associated protein complexes (Pdh and MCPs) might be important in determining some of the striking localization patterns that have been described (3, 30, 47).

MCPs may be involved in yet other aspects of DNA metabolism, because we found that TlpA interacts with GyrA (a subunit of gyrase), RecF and YirY (involved in recombinational repair), and DinG (a damage-inducible helicase). TlpA interacts also with FtsH, an ATP- and Zn²⁺-dependent metalloprotease involved in cell division (48). In dividing cells, FtsH accumulates at the septum (48) where the MCPs are located predivisionally (46). Perhaps, the FtsH–TlpA interaction might fulfill its function at the time of cell division. Interestingly, the link between TlpA and the DegS kinase indicates that in addition to CheA/CheY, the DegS/DegU two-component regulatory system (49) might be related to signal transduction in chemotaxis.

The interaction network reveals additional links between DNA replication and proteins involved in signaling. HolA (YqeN) interacts with DnaX, as expected, but also with CitS, a membrane-bound sensor kinase (49), and LepA, a GTPase of unknown function, universally conserved among bacteria (50). CitS is involved in the regulation of magnesium-citrate transport (51). It appears to interact with HolA through a cytosolic domain flanking the histidine phosphotransferase domain, raising the possibility that the interaction might depend on the phosphorylation state of CitS. LepA is homologous to protein synthesis elongation-factor GTPases (50), suggesting that LepA might act to coordinate DNA replication and protein synthesis. Thus, our data reveal potential signaling pathways that might act to tune DNA replication with the global cell physiology.

Previously Uncharacterized Proteins Involved in DNA Replication.

Strikingly, 36% of the proteins in our interaction network have no known function. We have shown that none of them individually is essential for cell survival. Considering that the major focus of previous research on bacterial DNA replication has been on essential genes (isolated via thermosensitive mutants), this result perhaps is not surprising. The proteins might have redundant functions or play various accessory roles in DNA replication. However, they now can be placed into biological context, and testable hypotheses about their functions can be generated. To illustrate this point, we tested the function of YabA, a protein of unknown function well conserved in Gram-positive bacteria. YabA interacted with both DnaA and DnaN, suggesting that it might be involved in initiation control, based on the Hda-mediated control mechanism described in E. coli (26, 27). We showed that YabA can act as a protein bridge between DnaA and DnaN. In the absence of a functional YabA protein, the average DNA content of nucleoids was much higher, and the number of oriC copies per cell unit was increased significantly compared with wild type. Taken together, these results indicate that YabA acts as a negative regulator of initiation at oriC. In contrast to E. coli Hda, YabA shares no homology with DnaA. Moreover, YabA interacts with the MCPs that are linked to initiation at oriC (see above) and with other proteins that also might have a role in initiation control. These findings suggest that YabA mediates a complex mechanism of initiation control and open the way to the study of initiation control in Gram-positive bacteria.

Abbreviations

AD

activation domain

BD

GAL4 DNA binding domain

Pol

DNA polymerase

BSL

B. subtilis genomic library

GFP

green fluorescent protein

MCP

methyl-accepting chemotaxis protein