Multiple Interactions between the Transmembrane Division Proteins of Bacillus subtilis and the Role of FtsL Instability in Divisome Assembly

Abstract

About 11 essential proteins assemble into a ring structure at the surface of the cell to bring about cytokinesis in bacteria. Several of these proteins have their major domains located outside the membrane, forming an assembly that we call the outer ring (OR). Previous work on division in Bacillus subtilis has shown that four of the OR proteins—FtsL, DivIC, DivIB, and PBP 2B—are interdependent for assembly. This contrasts with the mainly linear pathway for the equivalent proteins in Escherichia coli. Here we show that the interdependent nature of the B. subtilis pathway could be due to effects on FtsL and DivIC stability and that DivIB is an important player in regulating this turnover. Two-hybrid approaches suggest that a multiplicity of protein-protein interactions contribute to the assembly of the OR. DivIC is unusual in interacting strongly only with FtsL. We propose a model for the formation of the OR through the mutual association of the membrane proteins directed by the cytosolic inner-ring proteins.

Cell division in gram-positive bacteria involves the synthesis of a plate of new cell wall material in parallel with invagination of the cytoplasmic membrane at the midpoint of the cell. Genetic experiments have shown that division in Bacillus subtilis requires the correct function of about 11 proteins, most of which are conserved in other bacterial species. Protein localization techniques have shown that these proteins all accumulate at the division site prior to any detectable structural changes in the cell membrane or wall (for reviews, see references 10 and 11). Thus, it seems that complete assembly of the proteins into a complex, referred to as the divisome, is required before the initiation of cell division can occur. Although several factors have been identified as being involved in regulating the initial stages of divisome assembly, it is not known whether the process of assembly, once initiated, is subject to further regulation.

The hierarchy of assembly of the cell division complex has been studied in both Escherichia coli and B. subtilis by examining the effects of ablation of one division protein on recruitment of other proteins (10, 11). In E. coli these experiments have revealed an almost linear pathway of assembly, thus: FtsZ → [FtsA, ZipA] → FtsK → FtsQ → [FtsB, FtsL] → FtsW → FtsI → FtsN. For any given protein in this sequence, deletion of genes encoding proteins preceding it abolishes recruitment of the protein to potential division sites, but the loss of proteins to the right has no effect. Therefore, FtsN, for example, at the end of the sequence, depends on all of the other proteins for targeting. Recently, work from several laboratories has suggested that the actual process of assembly may involve a plethora of protein-protein interactions (reviewed in reference 12).

In general, the proteins preceding FtsQ in the sequence have their major domains inside the cell membrane, whereas those following are integral membrane proteins or have their major domains outside the cell membrane. These two groups of proteins can also be distinguished on the basis of “early” versus “late” arrival at the division site (1, 8). Henceforth, we refer to the two sets of proteins as the inner-ring (IR) and outer-ring (OR) proteins. The timing and localization of IR formation is thought to be orchestrated largely by GTP-dependent polymerization of the FtsZ protein, a tubulin homologue, regulated by interactions with a large number of other proteins, whereas the formation of the OR is dependent on the presence of the IR (for recent reviews, see references 10 and 11).

In the gram-positive bacterium B. subtilis, analysis of the dependence hierarchy for assembly of the divisome is incomplete, but the work that has been done is consistent with a similar IR and OR organization. Curiously, however, the OR proteins that have been well studied—FtsL, DivIB, DivIC, and PBP 2B—are apparently interdependent for assembly (10). These four OR proteins all have a single membrane span with their major domains on the outside of the cell. FtsL and DivIC are small bitopic proteins, each of which has a single transmembrane span, and a C-terminal domain predicted to have a coiled-coil structure, outside the cell (7, 22). These two proteins probably have equivalents in most bacteria with cell walls; the E. coli counterparts are called FtsL (14, 33) and FtsB (3). FtsL in B. subtilis is intrinsically unstable, and the protein rapidly disappears when transcription is shut down (e.g., using a xylose-inducible promoter). Thus, FtsL could represent a key control point in division because division arrests rapidly when the transcription of its gene ceases (7). DivIC appears to be a relatively abundant membrane protein (22) that is dependent on FtsL for its stability (7), but little is otherwise known of its function.

DivIB is related to FtsQ of E. coli but is unusual in that complete loss of the protein can be tolerated and gives rise to a serious division block only at higher temperatures (31), whereas FtsQ seems to be essential. Previous work showed that the main function of DivIB could be to stabilize FtsL because the temperature sensitivity of the divIB-null mutant for division can be overcome by the overexpression of ftsL (5). PBP 2B is one of the few division proteins to have a well-defined activity. Its large extracellular portion has two distinct domains. The C-terminal domain has a transpeptidase activity that is thought to play an essential role in synthesis of peptidoglycan during division and closely resembles many other penicillin-binding proteins (26). The other domain, which lies adjacent to the membrane span (the N-domain), has no precisely defined function, although it is thought to be involved in protein-protein interactions that could help to target it to the division site (25, 29, 35).

We have analyzed here the stability of the FtsL, DivIB, DivIC, and PBP 2B protein in various division-mutant backgrounds. The results confirm previous observations that FtsL is intrinsically unstable in vivo and show that both its stability and that of DivIC are markedly affected by the inhibition of division and by the presence or absence of DivIB. We have also studied the interactions between these proteins by bacterial and yeast two-hybrid assays and by genetic suppression. These results indicate that the proteins interact together in a complex manner, supporting the idea that multiple weak interactions are involved in divisome assembly. DivIC, strikingly, showed strong interactions with FtsL but neither of the other proteins. On the basis of the data, we present a mechanism for the regulated assembly of the OR.

MATERIALS AND METHODS

Media, strains, and general methods.

Bacterial strains and plasmids used are listed in Table 1. B. subtilis strains were grown in Difco antibiotic medium 3 (PAB) supplemented where required with xylose (0.5%) and/or 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside). E. coli strains were grown at 37 or 30°C (for the bacterial two-hybrid plasmid preparation and interaction analysis) in 2×YT (32), supplemented with ampicillin (100 μg/ml) and/or kanamycin (25 μg/ml), as well as glucose (0.4%), as necessary. B. subtilis strains were transformed according to the method of (2) as modified by Jenkinson and Mandelstam (19). Selection for B. subtilis transformants was carried out on nutrient agar (NA; Oxoid) supplemented with chloramphenicol (5 μg/ml), kanamycin (5 μg/ml), and/or spectinomycin (50 μg/ml), along with xylose (0.5%) and IPTG (0.5 mM), as necessary.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant features or genotypea	Construction, source, or reference
Strains
B. subtilis
168	trpC2	Laboratory stock
813	trpC2 Ω amyE846 (cat Pxyl ftsL-HA) Ω ftsL799 (lacI aph-A3 Pspac-pbpB)	5
818	trpC2 ftsL::pRD175 (ftsL-HA cat Pxyl-′ftsL pbpB)	5
826	trpC2 Ω divIB1306 (spc)	5
828	trpC2 Ω divIB1306 (spc) Ω amyE846 (cat Pxyl-ftsL-HA) Ω ftsL799 (lacI aph-A3 Pspac-pbpB)	826 DNA transformed into strain 813
830	trpC2 Ω amyE830 (cat Pxyl-divIB)	pRD202 integration into strain 168
831	trpC2 Ω amyE830 (cat Pxyl-divIB) Ω divIB1306 (spc)	826 DNA transformed into strain 830
832	trpC2 ftsL::pRD209 (ftsL-HA neo ΔlacI Pspac-′ftsL pbpB)	pRD209 integration into strain 168
833	trpC2 Ω amyE830 (cat Pxyl-divIB) ftsL::pRD209 (ftsL-HA neo Δ lacI Pspac-′ftsL pbpB)	pRD202 integration into strain 832
835	trpC2 Ω divIB1306 (spc) Ω amyE830 (cat Pxyl-divIB) ftsL::pRD209 (ftsL-HA neo Δ lacI Pspac-′ftsL pbpB)	826 DNA transformed into strain 833
836	trpC2 Ω divIB1306 (spc) sib-2	This study
837	trpC2 Ω divIB1306 (spc) sib-5	This study
E. coli strains
XL1-Blue	F′::Tn10 proA+B+ lacIq lacZΔ M15/recA endA1 gyrA96 thi hsdR17 supE44 relA1 lac	Stratagene, Ltd.
BTH101	F−cya-99 araD139 galE15 galK16 rpsL1 (Strr) hsdR2 mcrA1 mcrB1	21
Plasmids
pJPR1	bla amyE-3′ cat Pxyl amyE-5′	J. Rawlins
pUK19	bla neo	W. G. Haldenwang
pSG441	bla aph-A3 lacI Pspac	16
pRD200	bla neo pbpB (sib-2)	This study
pRD201	bla neo pbpB (sib-5)	This study
pRD202	bla amyE 3′-cat Pxyl-divIB amyE-5′	This study
pRD207	bla aph-A3 Pspac-′ftsl-HA	This study
pRD209	bla neo Pspac-′ftsl-HA	This study
pRD133	bla cat Pxyl-ftsL-HA	5

^aStr^r, streptomycin resistance.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western blotting was done as described previously (7) with polyclonal antisera and anti-hemagglutinin (HA) monoclonal antibodies (Roche). Cells were observed by microscopy using methods described earlier (7). A Zeiss Axiovert 137 epifluorescence microscope attached to a Cool-Snap HQ camera (Universal Imaging Corp.) was used for the cell imaging. The images obtained were processed by using Metamorph v6.2 and Adobe Photoshop 6.0.

B. subtilis strain constructions.

Strain 828 was constructed by transforming strain 813, where the ftsL gene is under the control of the P_xyl promoter, with chromosomal DNA from strain 826 (divIB-null mutant) and selecting for spectinomycin resistance in the presence of xylose. A clone with both chloramphenicol and spectinomycin resistance and exhibiting xylose dependence for division was then isolated and denoted 828. Plasmid integration or transfer of genetic markers by genomic DNA transformations from one strain to another, as described in Table 1, was used to construct further strains (as described in Table 1). In all cases the strain constructions were confirmed by both phenotypic and genetic tests.

Depletion of PBP 2B and FtsL.

Depletion of the division proteins FtsL and PBP 2B was carried out in essentially the same way. The strains were grown in standard medium supplemented with xylose (0.5%) for ftsL expression and IPTG (0.5 mM) for pbpB. Strains grown overnight in supplemented PAB at 30°C were then diluted 10-fold and incubated at 30 or 37°C for about 1 h before being diluted to an A₆₀₀ of 0.05. When the culture reached an A₆₀₀ of 0.5, it was centrifuged for 3 min; washed with fresh, prewarmed, unsupplemented PAB; and then suspended in prewarmed PAB and split into two aliquots, each of which was then supplemented with xylose (0.5%) or IPTG (0.5 mM). The cultures were then incubated at the appropriate temperature for the experiment, and samples were obtained and processed at the required intervals as described previously (5).

Plasmid constructions.

Plasmid pRD202 was constructed by the insertion of the divIB coding sequence, including the ribosome-binding site, obtained by PCR, into plasmid pJPR1 (Joy Rawlins, unpublished data). This placed the divIB gene under the control of the P_xyl promoter and allowed the genetic construction to be inserted into the amyE locus of the B. subtilis genome by homologous recombination.

Plasmid pRD209 was made in two steps. First, the 212-bp HindIII fragment was liberated from pRD133 and cloned between the HindIII sites of the vector portion of pSG441, to give pRD207. This plasmid contained the P_spac-′ftsL-HA cassette. However, this plasmid was found to be difficult to select for when transformed into B. subtilis, presumably because the resistance marker had been damaged. Plasmid pRD209 was constructed by cloning the cassette liberated from pRD207 by digestion with EcoRI and PstI and cloned into pUK19 digested with the same restriction enzymes. This plasmid could then be integrated into the Bacillus genome to provide a C-terminal epitope tag of the FtsL protein and provided for the constitutive expression of PBP 2B (by the unrepressed P_spac promoter, since the gene encoding the Lac repressor [lacI] had been removed from the construction).

To determine the location of the sib mutations, the pbpB locus was cloned into an integration plasmid (pUK19). A 2.2-kbp PCR product corresponding to the coding sequence of pbpB was amplified from purified genomic DNA of strains 836 and 837 with oligonucleotides designed to have BamHI restriction sites located at both termini. This fragment of DNA was then cloned into pUK19 to yield plasmids pRD200 and pRD201. When the plasmids were transformed into the divIB-null mutant (strain 826) the sib phenotype was obtained in a proportion of the transformants, confirming that the sib mutation was present in the cloned segment of DNA.

Mapping and identification of the sib mutations.

To test whether the suppressor mutations were separable from the original divIB mutation, genomic DNA from the sib-2 and sib-5 strains was transformed into the wild-type strain 168, with selection for the divIB insertion marker, spectinomycin resistance (spc). About 20% of the transformants were able to grow at 49°C, indicating that they had acquired the sib mutation. This frequency of cotransformation suggested that the sib-2 and sib-5 mutations lay close to the divIB gene. From these transformations clones bearing both the sib mutation and the divIB1306 (spc) markers were purified and denoted strains 836 for sib-2 and 837 for sib-5.

Three factor crosses were used to locate the positions of the sib-2 and sib-5 mutations relative to other known genes (4). The linkage results obtained by this method indicated that both sib mutations lay close to spoVD and that divIB was further away from spoVD than were the sib mutations.

To define the positions of the sib mutations more precisely, segments of DNA spanning the region from yllA to spoIIIG (19 kb) were amplified as two overlapping fragments from chromosomal DNA isolated from strains 837 (sib-5) and 836 (sib-2) by long-range PCR. Cotransformation of these DNA fragments with genomic DNA from strain 826, with selection for spectinomycin resistance and screening for growth at 49°C, confirmed the initial mapping results. This method was then used with smaller DNA fragments to define the position of the sib alleles. From the analysis, the most probable location for the sib mutations was within pbpB, potentially toward the 5′ end of the gene. The cloning of the coding sequence of pbpB from the sib-2 and sib-5 strains into plasmid pUK19, producing plasmids pRD200 and pRD201, confirmed this as transformation of these plasmids into strain 826 conferred the Sib phenotype at a frequency of 15 and 18%, respectively. The relatively low frequencies of sib transformants suggested that the mutations responsible for the Sib phenotype were located in the N-terminal quarter of the gene. Sequencing of the entire insert in these plasmids then allowed the identification of the mutations conferring the Sib phenotypes.

Construction of yeast two-hybrid bait and prey plasmids.

The B. subtilis divIB, divIC, ftsL, and pbpB full-size open reading frames were PCR amplified from the wild-type strain 168 genomic DNA and fused in frame with (i) the Gal4 BD into the pGBDU-C2 bait vector and (ii) the Gal4 AD into the pGAD-C2 prey vector. The divIC and ftsL genes were cloned by using the EcoRI and BamHI restriction sites, and divIB was cloned by using the EcoRI and PstII restriction sites. The resulting constructs were isolated from E. coli and transformed into yeast haploid strain PJ69-4a (for the pGBDU derivatives, Ura⁺) and PJ69-4α (for the pGAD derivatives, Leu⁺), using URA3 and LEU2 as selective markers, respectively (18). In addition, the different open reading frames were subcloned into the three-hybrid vector p3HB(Trp⁺) using the EcoRI and SalI restriction sites (28). The constructs were isolated from E. coli and cotransformed with the pGBDU derivatives into PJ69-4a, selecting for Ura+ and Trp+.

Yeast trihybrid assay.

For yeast trihybrid experiments, PJ69-4a strains cotransformed by different combinations of bait and 3HB vectors expressing DivIB, DivIC, FtsL, and PBP2P proteins (Ura+ Trp+) were mated with PJ69-4α strains containing the various prey vectors (Leu+) and the empty pGAD-C1 vector. 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), as described previously (9).

Bacterial two-hybrid plasmid construction.

The method used was essentially that of Karimova et al. (21). The coding sequence of each gene was amplified by PCR using oligonucleotides that introduced XbaI and KpnI sites to facilitate cloning into the plasmids pKT25 and pUT18C in such a way as to fuse the coding sequence of the truncated adenyl cyclase gene in the plasmids to the N terminus of each of the Bacillus genes. These plasmids were then used in cotransformations of BTH101 to test for interaction between the proteins by plating on a minimal medium composed of 1% Agar No. 1 (Oxoid), 10 mM ammonium chloride, 1.2 mM ammonium nitrate, 1 mM magnesium sulfate, 0.75 mM sodium sulfite, 0.5 mM potassium dihydrogen phosphate, 0.1 mM manganese (II) chloride, and 4 μM iron (III) chloride. The pH was adjusted to 7 with sodium hydroxide. Then, 0.8% glucose, 0.4% Casamino Acids, 3 μM thiamine, 100 μg of ampicillin/ml, 25 μg of kanamycin/ml, and 0.004% X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) were added to the medium prior to pouring the plates.

Since the analysis suggested that DivIC and FtsL were likely to form at least a dimer and DivIC only exhibited significant interaction with FtsL, a further pair of plasmids was constructed in which the coding sequence of FtsL with a ribosome-binding site was inserted into the KpnI site of pB19 and pA19. This then provided a plasmid for which both the divIC fusion and ftsL were cotranscribed from the same promoter.

To assay for interactions, the simplest and most robust assay was to spot 10-μl aliquots of the cotransformations of each test pair of plasmids onto a defined medium containing the selective antibiotics and X-Gal. This minimized the occurrence of false positives, since on rich medium the host strain (BTH101) expressed sufficient background β-galactosidase activity to produce blue colonies after 24 h of incubation at 30°C, whereas on minimal medium control transformations with empty vectors remained white for up to 72 h of incubation.

RESULTS

Depletion of PBP 2B destabilizes FtsL and DivIC.

It has previously been shown that depletion of FtsL prevents the localization of the other OR proteins—DivIB, DivIC, and PBP 2B—to potential division sites, even though FtsZ ring formation still occurs. Western blot analysis showed that FtsL depletion was accompanied by the disappearance of DivIC protein, whereas FtsZ, DivIB, and PBP 2B were maintained at approximately normal levels (5, 7). To examine the effects of depleting PBP 2B, we placed pbpB under the control of an IPTG-inducible promoter and analyzed the levels of various division proteins by Western blotting. Figure 1A shows the reduction of PBP 2B levels after the removal of inducer in cells incubated at 37 or 48°C. This was accompanied by an arrest of cell division beginning about 80 min after repression at 37°C (6) and beginning about 40 min after repression at 48°C (data not shown). Figure 1B to D shows that the levels of FtsL, and to a lesser extent DivIC, were reduced once cell division was perturbed, with a more severe effect at 48°C than at 37°C. In contrast, DivIB levels (Fig. 1E) were virtually unaffected by the PBP 2B depletion. It was also evident that the levels of FtsL and DivIC were reduced in the cultures once the culture reached stationary phase (100 min at 37°C and 80 min at 48°C).

FIG. 1.

Stability of FtsL, DivIC, and DivIB when cell division is arrested by depletion of PBP 2B. Strain 832 was grown at 37 or 48°C in the presence of IPTG, and then at time zero the cells were washed to remove IPTG. Samples of the culture were taken at regular intervals, and the total protein was extracted. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis combined with Western blotting was then used to detect the presence of PBP 2B (A), FtsL (B), FtsL at a threefold-longer exposure (C), DivIC (D), and DivIB (E). The lower-molecular-weight signals in panel C probably represent degradation products of the epitope-tagged FtsL derivative. The quantity of sample loaded in each lane was adjusted using the culture's optical density at 600 nm to ensure that the same cellular material was applied. The increase in cell length due to the depletion of PBP 2B was detectable after about 80 (at 37°C) and 40 min (at 48°C). The time (in minutes) of sampling, relative to time zero, is shown above each lane.

Synthetic lethal phenotype associated with C-terminal modifications of DivIC and FtsL.

Because we have not been successful in detecting FtsL with anti-FtsL polyclonal antisera, FtsL was detected in the experiments described above by using a C-terminal HA epitope tag. This construction was reported to be functional, as judged by normal cell division when present as the only copy of ftsL in the cell (5). However, in the course of these experiments, it emerged that this construction could not be combined with the divIC355^ts mutation, which encodes a C-terminally modified DivIC protein (24). The HA fusion construct also gave a mildly enhanced cell division defect when combined with a divIB-null mutation, but no effect was seen with various other cell-division-deficient mutations, including ftsZ ts1 and ftsA279 (data not shown). DivIC and FtsL are predicted to be predominantly coiled-coil proteins and to have a similar transmembrane topology and size (7). The synthetic lethality of the two C-terminally modified proteins supports the idea that they interact through their C-terminal domains and that interaction between these proteins is important for their function.

DivIB is required for the destabilization of DivIC in the absence of FtsL.

Previous work has shown that FtsL is destabilized in the absence of DivIB, particularly at higher growth temperatures (5). To further examine the relationship between DivIB, DivIC, and FtsL stability, the effects of FtsL repression in strains with or without DivIB were compared by Western blot analysis. As expected, at 30°C the relative cellular abundance of both FtsL and DivIC were essentially unaffected by the absence of DivIB (Fig. 2A and D). However, when FtsL was depleted in the presence of DivIB (Fig. 2B), the amount of DivIC decreased rapidly and by 40 min was essentially undetectable (as had been shown previously at 37°C) (7). Surprisingly, however, in the absence of DivIB (Fig. 2C) there was no significant reduction in the DivIC signal, even when FtsL was undetectable. Thus, the rapid turnover of DivIC in the absence of FtsL depends on the presence of DivIB. (We do not know why the levels of both FtsL and DivIC fell in the samples shown in Fig. 2B and D. This result may be related to physiological changes associated with the entry of these cultures into stationary phase.)

FIG. 2.

Effects of DivIB levels on the stability of FtsL and DivIC. (A to D) Isogenic divIB⁺ (strain 813) and ΔdivIB (strain 828) strains were grown in the presence (A and C) and absence (B and D) of the inducer xylose at 30°C. The figure shows Western blots of total cell protein probed with antisera against DivIB and HA (FtsL) or DivIC alone. The values above each lane indicate the time (in minutes) the sample was taken relative to the removal of xylose (inducer of ftsL-HA) from the culture. (E to H) Strains expressing different levels of DivIB were shifted from 30 to 37°C at time zero, and the levels of DivIB and DivIC were monitored by Western blotting. The strains were as follows: E, divIB⁺ (strain 168); F, divIB mutant (strain 826); G, divIB mutant P_xyl divIB (strain 831) without xylose; and H, divIB P_xyl divIB mutant (strain 831) upon addition of xylose.

Overexpression of DivIB destabilizes DivIC but has no effect on FtsL.

The experiments described above suggested that DivIB has some role in controlling the stability of DivIC, and previous experiments have indicated a role in the stabilization of FtsL (5). If so, it was interesting to test the effects of overexpression of divIB on these proteins. For these experiments strains were constructed in which the levels of DivIB could be regulated by transcription from the strong xylose-inducible P_xyl promoter. The results of typical experiments examining the effects on DivIC levels are shown in Fig. 2E to H. (These experiments were carried out using cultures grown at 30°C and then shifted to 37°C at t₀. These conditions provided a clear indication of the changes in DivIC abundance relative to DivIB expression levels. Comparable but less clear results were obtained with cultures incubated at 30°C.) The expected wild-type levels of DivIB and DivIC are shown in Fig. 2E. Figure 2F, G, and H show the effects of no (null mutant) (F), low (uninduced) (G), and high (induced) (H) levels of DivIB on DivIC abundance. It appeared that the levels of DivIC varied inversely to those of DivIB, a finding consistent with DivIB acting as a negative regulator of DivIC accumulation.

In other experiments, the effects of DivIB overexpression on the levels of FtsL were examined. No significant changes in the cellular levels of FtsL were observed at 30, 37, or 48°C (data not shown).

Isolation of sib (suppressor of divIB) mutations as alleles of pbpB.

To shed further light on the nature of the interactions between the OR proteins, we looked for mutations that could suppress the thermosensitive lethal division phenotype of divIB-null mutants. A divIB-null mutant was incubated at 49°C on NA plates, and eight independent spontaneous mutants were isolated. To examine their phenotypes, cultures of each of the mutants were shifted from 30 to 48°C, and samples of the cultures were examined by phase-contrast microscopy. Mutants 1, 3, 4, 6, 7, and 8 exhibited a filamentous cell morphology that was similar to that of the parent divIB-null strain (i.e., strain 826); these strains seemed to survive by being less prone to cell lysis than the parent at 48°C. However, mutants 2 and 5 appeared to be substantially rescued for cell division, with cell lengths similar to those of the wild-type strain. Quantitation of the cell division frequencies of the bypass mutants compared to the wild type and parent divIB-null mutant is shown in Fig. 3. At 30°C all four strains showed typical narrow cell length distributions, with mean cell lengths of about 3 to 4 μm (left panels). After growth at 48°C, the wild-type distribution was unchanged, whereas the divIB parent strain showed a broad spectrum of cell lengths, with many long/filamentous cell, as expected for cells that had grown but were impaired in division. Mutants 2 and 5 clearly divided much more frequently than the divIB parent strain. Mutant 5 in particular, had a length distribution that was almost the same as that of the wild-type strain.

FIG. 3.

Phenotypes of sib mutants. Cultures of the sib mutants and control strains were grown at 30°C to early exponential phase and then shifted to 48°C. The cell length was measured from ethanol-fixed samples. The graphs show the cell length distribution at 30°C and then after 60 min of incubation at 48°C. Phase-contrast images of fields of cells characteristic for each strain after 60 min of incubation at 48°C are shown on the right-hand side of each panel. The strains were as follows: A, ΩdivIB1306 (strain 826); B, ΩdivIB1306 sib-2 (strain 836); C; ΩdivIB1306 sib-5 (strain 837); and D, wild type (strain 168).

Genetic mapping revealed that both sib-2 and sib-5 were alleles of the pbpB gene. To confirm this, the entire pbpB gene from each mutant was cloned, and the resulting plasmids (pRD200 and pRD201, respectively) were transformed into the divIB-null strain (strain 826). Both plasmids gave mixtures of transformants with or without the suppressed phenotype. The frequency with which the suppression phenotype segregated indicated that the mutations were both in the 5′ end of the gene. Sequencing the inserts in the plasmids showed that sib-2 was a T-to-A transversion generating the substitution V165E and that sib-5 was a G-to-A transition resulting in the substitution D213N. No phenotype was observed when the sib mutations were introduced into the divIB⁺ background (not shown). These experiments showed that single amino acid substitutions in the N-terminal domain of PBP 2B (the function of which is currently unknown) are sufficient to bypass the requirement for DivIB in cell division in B. subtilis. This region of the protein is predicted to be located on the outside of the cell, probably close to the cell membrane. Thus, it is possible that this region interacts with the other division proteins during divisome assembly.

Interactions of the OR proteins.

Various attempts to detect interactions between the OR proteins by biochemical methods were unsuccessful. Therefore, we turned to two-hybrid analysis, beginning with the yeast two-hybrid system (18). The full-length DivIB, DivIC, FtsL, and PBP2P proteins were fused to the Gal4 DNA-binding domain (BD) and to the Gal4 activation domain (AD), and fusion proteins were expressed in different combinations in yeast cells. None of these pairwise and “self” combinations activated expression of the interaction reporters (data not shown). This suggested that the binary interactions taking place in the yeast nucleus, if they occurred, were too weak to be detected. The possibility that more stable ternary complexes involving the DivIB, DivIC, FtsL, and PBP 2B proteins could form was then tested by using the yeast trihybrid assay (see Materials and Methods). Indeed, certain ternary interactions were revealed when the interacting proteins were expressed with a specific third protein (3HB) (Fig. 4A). Thus, the BD-DivIC (bait) appeared to interact with AD-DivIB (prey) provided that the FtsL protein was coexpressed in the cell. Similarly, interaction between BD-FtsL and AD-DivIB was observed, but only in the presence of DivIC. No interaction was detected of AD-DivIC, AD-FtsL, or AD-Pbp2B with the available BD constructions, indicating that the observed interactions were specific to the AD-DivIB prey (data not shown). Altogether, these results show that DivIB, DivIC, and FtsL are able to form a ternary protein complex in yeast.

FIG. 4.

Yeast three-hybrid and bacterial two-hybrid interaction assays. (A) Ternary complexes assayed by yeast three-hybrid analysis. DivIB, DivIC, FtsL, and PBP 2B proteins were expressed as bait (BD, left,), as additional partners (3HB, top), and DivIB was expressed as prey (AD-DivIB). Negative controls are the empty bait and 3HB vectors (BD and 3HB). The empty prey vector (AD) is not shown. Pairs of independent colonies containing the various BD/3HB combinations were arrayed and mated with the strains expressing the AD-DivIB fusion protein, and diploid cells were subjected to selection for expression of the interaction reporters as described in Materials and Methods. Replica plating of the diploids onto selective SC-LUWH plates was used to score interaction phenotypes. (B) Bacterial two-hybrid assays. The top left portion of the image shows the E. coli primary transformants representing all possible pairwise combinations of the division proteins PBP 2B, DivIB, DivIC, and FtsL after incubation at 30°C for 32 h. The labels along the top of the image refer to the fusion construction cloned into the pKT25 (cya′-gene fusion) plasmid, and the labels on the left correspond to the fusion construction in pUT18C (′cya-gene fusion). DivIC (FtsL) indicates that these constructs express a nonfused copy of ftsL, as well as the divIC fusion protein.

We then turned to the bacterial two-hybrid system (20) to explore the interactions further. Positive interactions in this system result in the formation of an active adenyl cyclase enzyme, resulting in the relaxation of catabolite repression in the host cell, rather than an active transcriptional activator. This has the advantage that the association of the proteins does not need to also involve association with the DNA. Figure 4B shows a typical result, based on a β-galactosidase readout of colonies directly selected on the plate from cotransformed cells. (In a series of control experiments we found this to be the most reliable and reproducible means of scoring relative interaction levels.) In this case, many positive reciprocal interactions were detected. Controls, in which only one fusion construct was present or with one fusion construct and an empty vector (with the adenyl cyclase fragment only), gave only white colonies under the same conditions (not shown). PBP 2B, DivIB, and FtsL all exhibited self-interaction, whereas DivIC did not, within the limits of detection. FtsL also interacted with all three of the other proteins and gave a particularly strong signal with DivIC. DivIB, PBP 2B, and FtsL all gave positive reciprocal interactions with each other, but they did not detectably interact with DivIC. To assess the specificity of these interactions, we included in the tests other proteins, including various transmembrane proteins not directly associated with division. No interaction with these control proteins was detected (data not shown), confirming that the interactions detected were specific.

Since DivIC did not show significant interaction with DivIB or PBP 2B but did interact strongly with FtsL, it was possible that the association of DivIC with other proteins required FtsL, as was indicated by the yeast trihybrid method. To test this possibility, plasmids were constructed in which unfused ftsL would be coexpressed with the divIC test fusion. In the presence of unfused FtsL, DivIC now showed weak interactions with itself, DivIB, and PBP 2B, as well as a strong interaction with FtsL, and all of these interactions were seen in reciprocal tests (Fig. 4).

To try and understand the mechanism by which the sib mutants suppressed the requirements for DivIB, pairwise interaction tests using the sib-2 and sib-5 mutant forms of PBP 2B were used. These indicated that the mutant proteins behaved like the wild-type PBP 2B, except that the sib-2 mutant appeared to exhibit an increased interaction with DivIB (Fig. 4).

DISCUSSION

FtsL turnover as a key step in assembly of the OR.

We have previously shown that FtsL is intrinsically unstable; that it is further destabilized by mutations in divIC and particularly divIB (but not ftsZ); and that when it is depleted, DivIC is also destabilized (7). We have now shown that both FtsL and DivIC are also destabilized by depletion of PBP 2B. Most importantly, the loss of DivIC is completely dependent on DivIB (summarized in Table 2). These findings raise a couple of interesting issues. First, the effects on turnover change our view of the dependence hierarchy for assembly. It now seems possible that B. subtilis does have a near linear hierarchy of assembly of the OR proteins, as in E. coli, but that this is masked by degradation when assembly is perturbed. According to this view, when one or another of the “downstream” proteins is depleted, the “upstream” OR proteins may still be recruited to the division site, but they are either rapidly degraded or their assembly is transient and so undetected. Thus far, substantial effects on stability do not seem to have been reported for the equivalent E. coli proteins, and the detection of a stepwise assembly hierarchy suggests that turnover is less important in this organism.

TABLE 2.

Summary of genetic effects on the stability of OR proteins

Division defect	Protein stabilitya
	DivIB	DivIC	FtsL	PBP 2B
Wild type	+	+	+	+
FtsZ ts (48°C)b	+	+	+	+
PBP 2B depletionc	+	−	−	(−)
FtsL depletion	+d	−d	(−)	+
DivIB null (49°C)	0	++e	−b	+
FtsL depletion, DivIB nullc	0	+	(−)	+
DivIB overexpressedc	++	+/−	+	+
DivIB null, sib (48°C)c	0	+	+	+

^a+, Normal stability; −, severe instability; (−) partial instability; 0, null mutant for that protein.

^bFirst shown by Daniel and Errington (5).

^cThis work.

^dShown by Daniel et al. (7).

^eIdentified independently by V. Katis and A. Moire (unpublished data).

Second, why is turnover of FtsL and DivIC enhanced by a late block in divisome assembly (e.g., pbpB repression or divIB deletion) but not by an early block in assembly (e.g., ftsZ) (Fig. 1) (6)? One possibility is that an intermediate step in assembly involves increased turnover and that this is regulated such that it only begins at some point after OR formation. Perhaps one or more proteases coassembles with the OR (see below). Turnover of FtsL could be a normal reaction during OR assembly to limit the assembly of the OR to a defined region. When division is perturbed, turnover continues and results in the complete loss of FtsL, preventing the inappropriate assembly of the OR. An alternative or additional function of FtsL degradation might lie in the constriction of the division complex during division. During membrane constriction, the OR must be continuously remodeled by virtue of its association with the constricting IR. This remodeling will result in the displacement of excess subunits from the OR to allow it to follow the constricting IR. Degradation of displaced FtsL could help to drive division by ensuring that the cell wall synthetic apparatus (OR) is restricted to a narrow zone defined by the shrinking IR.

Apparent role of DivIB in enhancing the turnover of DivIC.

The only interactions detected in yeast were tripartite interactions involving DivIB, FtsL, and DivIC. The existence of complexes involving the homologues of these proteins has also been reported in other organisms, i.e., E. coli and Streptococcus pneumoniae, on the basis of both in vivo and in vitro experiments (3, 27). We currently favor a model in which DivIB interacts with a heterodimer of FtsL and DivIC. We do not understand why the DivIC-FtsL heterodimer is not always readily detectable (30), but for the moment we suppose that it does exist and, indeed, this was the strongest interaction detected by the bacterial two-hybrid experiments (Fig. 4). Conceivably, the role of DivIB could be to promote the formation of, or somehow stabilize, the FtsL-DivIC complex. The temperature-sensitive effect of divIB disruption suggests that the complex can form spontaneously but requires DivIB at higher temperatures.

The data described above provide further insights into aspects of this tripartite complex. Previous work had suggested that DivIB has a role in stabilizing FtsL (5). In contrast, the results shown in Fig. 2 suggest that DivIB is involved in DivIC degradation, and elimination of DivIB turned out to prevent the loss of DivIC when FtsL is depleted. Also, the results in Fig. 2E to H show that the levels of DivIC vary in inverse proportion to those of DivIB. This suggests that DivIB may act directly, either as a DivIC protease or by regulating the access of DivIC to a protease. According to this scenario, the formation of an FtsL-DivIC heterodimer would protect DivIC from degradation by the DivIB proteolytic pathway and also stabilize FtsL. This would explain why DivIC is degraded in parallel with FtsL under various conditions. The fact that FtsL is degraded via a DivIB-independent pathway and FtsL is destabilized in the absence of DivIB suggests that the FtsL-DivIC multimers are dynamic and that DivIB interacts with FtsL and DivIC homo- and heterodimers in multiple ways, stabilizing some forms and targeting others for degradation. Thus, DivIB may promote the formation of heterodimers of FtsL and DivIC, thereby indirectly stabilizing FtsL and removing excess DivIC. The result could be a mechanism that regulates the formation of FtsL-DivIC heterodimers (or higher hetero- oligomers) and their entry into the OR (see below).

Whatever the precise function of DivIB, it is not essential for division, because null mutants are able to divide quite well at lower temperatures (15). To shed more light on DivIB function, we isolated suppressor mutants that were able to divide well at a high temperature in the complete absence of DivIB. The sib mutations that we mapped were located in pbpB and presumably enable OR assembly to occur at high temperatures in the absence of DivIB. The mutations affected the N-terminal portion of PBP 2B in a domain that has been suggested to be involved in protein-protein interactions (29, 34). Further work is needed to establish how these mutations rescue division in the divIB mutant background but, presumably, they act at least in part to stabilize FtsL and allow it to accumulate in the OR structure. These results reinforce the notion that DivIB does not have a crucial enzymatic or structural role in the divisome and raise the question of whether FtsQ may be similarly dispensable under certain conditions in E. coli.

Significance of multiple protein-protein interactions for OR assembly.

The bacterial two-hybrid system provided evidence for a much more extensive network of interactions than the yeast system (summarized in Fig. 5). This may be due to the bacterial system being more suited for detecting interactions between integral membrane proteins. The same bacterial two-hybrid system we used also revealed multiple interactions when applied to E. coli division proteins (20). In our experiments with the B. subtilis proteins, all possible pairwise interactions of FtsL, DivIB, and PBP 2B were detected by the system. Unfortunately, these kinds of two-hybrid data give neither quantitative information about the strength of the interaction nor information about the three-dimensional configuration of the interacting surfaces on the proteins. The interactions could be strong and involve very specific surfaces on the interacting parties or be relatively weak and topologically imprecise, resulting in no more than a loose tendency to associate. Considering that we (and others [20]) have detected a multiplicity of interactions between the proteins, the implication would be that multiple weak interactions promote a tendency for the proteins to inhabit the same space in the membrane. Karimova et al. (20) have discussed in detail the significance of this attractive idea and its parallels in other biological systems, such as the neuronal synapse. Within this space, the formation of specific multiprotein complexes such as those detected by the yeast three-hybrid assay could be facilitated, leading to the assembly of a properly ordered OR system capable of coordinating membrane constriction and cell wall synthesis. We propose that this assembly requires an additional driving force, probably provided by association with the FtsZ-orchestrated IR system.

FIG. 5.

Pairwise interactions between the OR proteins. A summary of the positive interactions from the OR proteins of B. subtilis, as detected by bacterial two-hybrid analysis, is presented. Interactions detected between pairs of proteins are indicated by arrows, with thicker lines representing stronger interactions.

Interestingly, it has recently been shown that DnaA, a component of the DNA replication apparatus, directly acts to regulate the expression of FtsL (13). This and other analyses (17, 23) provide further support to the idea that the instability of FtsL acts as a key point of regulation for cell division and provides an interesting link between DNA replication and cell division. The results presented here also imply the existence of specific degradation pathways for division proteins. Characterization of the genes involved may improve our understanding of the regulation of cell division.