Cytological Characterization of YpsB, a Novel Component of the Bacillus subtilis Divisome

Cell division in bacteria is carried out by an elaborate molecularmachine composed of more than a dozen proteins and known asthe divisome. Here we describe the characterization of a newdivisome protein in Bacillus subtilis called YpsB. Sequencecomparisons and phylogentic analysis demonstrated that YpsBis a paralog of the division site selection protein DivIVA.YpsB is present in several gram-positive bacteria and likelyoriginated from the duplication of a DivIVA-like gene in thelast common ancestor of bacteria of the orders Bacillales andLactobacillales. We used green fluorescent protein microscopyto determine that YpsB localizes to the divisome. Similarlyto that for DivIVA, the recruitment of YpsB to the divisomerequires late division proteins and occurs significantly afterZ-ring formation. In contrast to DivIVA, however, YpsB is notretained at the newly formed cell poles after septation. Deletionanalysis suggests that the N terminus of YpsB is required totarget the protein to the divisome. The high similarity betweenthe N termini of YpsB and DivIVA suggests that the same regionis involved in the targeting of DivIVA. YpsB is not essentialfor septum formation and does not appear to play a role in septumpositioning. However, a ypsB deletion has a synthetic effectwhen combined with a mutation in the cell division gene ftsA.Thus, we conclude that YpsB is a novel B. subtilis cell divisionprotein whose function has diverged from that of its paralogDivIVA.

INTRODUCTION

Top

INTRODUCTION

It is currently recognized that many biochemical processes inbacterial cells are carried out by elaborate macromolecularmachines (36). The case of bacterial division is no exception,as it requires the coordinated action of at least 15 proteinsthat exist as a dynamic assembly known as divisome or septalsome.The divisome is a contractile protein ring that promotes theinvagination of the different layers of the bacterial envelope(typically, cytoplasmic membrane and cell wall in the case ofgram positives and cytoplasmic membrane, cell wall, and outermembrane in the case of gram negatives), culminating with theformation of the division septum (for recent reviews, see references15, 17, and 21).

Divisome proteins can be subdivided into three groups, or subcomplexes,based on their presumed function and how they assemble intothe complex (53). The first of these groups has been calledthe "cytoplasmic protoring" and contains the ubiquitous tubulin-likeprotein FtsZ and several proteins that interact directly withFtsZ (for reviews, see references 31 and 44). Similarly to tubulin,FtsZ is capable of self-assembling into organized polymers.Polymerization of FtsZ into the proper ring structure and associationof the ring with the inner face of the cytoplasmic membranerequires the aid of the FtsZ-binding proteins, notably FtsA(26, 39), ZapA (18), and SepF (YlmF) (20, 25) in the case ofBacillus subtilis. The cytoplasmic protoring is the structuralscaffold of the divisome and determines where division willhappen by recruiting a group of proteins that are known as the"peptidoglycan factory" subcomplex. The peptidoglycan factory,which includes the septum-specific transpeptidase PBP 2B (PBP3 in Escherichia coli) (7), the putative peptidoglycan precursortransporter FtsW, and, presumably, enzymes with transglycosylaseand amidase activities, is responsible for redirecting cellwall synthesis, which occurs longitudinally during growth andbecomes orthogonal to cell length during septation. The connectionbetween the peptidoglycan factory and the cytoplasmic protoringis performed by the third divisome subcomplex. This complex,known as the "periplasmic connector," includes the small bitopicproteins DivIB (FtsQ in E. coli), DivIC (FtsB in E. coli), andFtsL (8, 9, 23, 28). No recognizable enzymatic function hasbeen attributed to any of these proteins. Recent evidence, however,suggests that the periplasmic connector may regulate the timingof divisome maturation through the availability of FtsL, oneof its components (3).

In addition to the components described above, proper cell divisionrequires proteins that control the site of divisome assemblyto ensure that a progenitor cell will give rise to two identicaldaughter cells, each containing a complete chromosome. The best-studiedregulators of division position are the MinC and MinD proteins(for reviews, see references 30 and 45). MinC is an inhibitorof FtsZ polymerization (24) that acts in conjunction with MinDto prevent improper divisome assembly at the cell poles. Theability of the MinCD complex to inhibit polar division stemsfrom the fact that the MinCD proteins are themselves localizedat the cell poles (32, 43). In E. coli, the polar localizationof MinCD results from a peculiar oscillatory behavior of theMinCD complex, which shuttles from one end of the cell to theother with a period of 1 minute (43). Oscillation of MinCD inE. coli is a self-organizing behavior that originates from intrinsicbiochemical properties of MinD and is modulated by the proteinMinE. In the absence of MinE, oscillation does not occur (43).Thus, MinE is known as the topological specificity factor ofE. coli's Min system. In B. subtilis, the MinCD complex is alsopolarly localized (32). However, targeting of MinCD to the polesoccurs by a very different mechanism. First, the MinCD complexdoes not exhibit oscillatory behavior in this species (32).In addition, B. subtilis lacks MinE and instead uses an unrelatedprotein, DivIVA, as its topological specificity factor (4, 11).DivIVA is a 165-amino-acid protein with extensive coiled-coilregions that has been shown to associate with the divisome complexat a late stage in its assembly (32). Thus, DivIVA targets theMinCD complex to the poles by first associating with the divisioncomplex at the nascent septum. Retention of DivIVA and the associatedMinCD at the division site after septation is complete resultsin the accumulation of the three proteins at the newly formedcell poles. In addition to controlling division position, DivIVAplays a second important function in B. subtilis, also relatedto its ability to associate with the cell poles. At the onsetof sporulation, chromosomes must be rearranged into the axialfilament, an extended structure in which their origin regionsbecome fastened to the cell poles. Fastening of the originsto the cell poles appears to be mediated by an interaction betweenDivIVA, at the cell poles, and two DNA binding proteins, RacAand Soj, which serve as adapter proteins between the chromosomeand DivIVA (2, 51, 57).

Here, we characterize YpsB, a new divisome protein of B. subtilis.YpsB is a paralog of DivIVA present in several gram-positivebacteria. Similarly to DivIVA, YpsB is a divisome componentthat associates with the complex late in its assembly. In contrastwith DivIVA, however, YpsB is not retained at cell poles anddoes not appear to be involved in division site selection oraxial filament formation. Instead, we found that YpsB is a proteinrequired for efficient division in cells lacking FtsA. Thus,YpsB is a new B. subtilis cell division protein whose functionhas diverged from that of its paralog DivIVA.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

General methods. All B. subtilis strains were derived from the prototrophic strainPY79 (58) and are listed in Table 1. B. subtilis was grown inLB medium at 37°C unless indicated otherwise. Preparationof competent cells and transformation of B. subtilis were performedaccording to the protocol previously reported by Kunst and Rapoport(27). When necessary, various concentrations of isopropyl-β-D-thiogalactopyranoside(IPTG), xylose, and/or antibiotics were added to the B. subtilisculture. Antibiotics were used at the following concentrations:for chloramphenicol, 5 µg/ml; for erythromycin plus lincomycin,1 µg/ml and 25 µg/ml; for spectinomycin, 100 µg/ml;for kanamycin, 10 µg/ml; and for tetracycline, 10 µg/ml.

View this table:
[in this window]
[in a new window]

TABLE 1. Strains and plasmids

DNA methods. DNA manipulations and cloning were carried out according tostandard methods (46). Phusion DNA polymerase (NEB) was usedin PCRs. Oligonucleotide primers were purchased from IDT (Coralville,IA) and are listed in Table 2. Sequencing was carried out bythe sequencing service of the Departamento de Bioquímica,IQ-USP. Plasmids and their relevant characteristics are listedin Table 1. Plasmid construction is described in detail in thesupplemental material. Sequences and maps of the plasmids areavailable upon request.

View this table:
[in this window]
[in a new window]

TABLE 2. Oligonucleotides

Strain construction. The construction of the ypsB deletion is described below. Forall other strains, see the supplemental material.

The ypsB deletion mutant was created by transformation of competentcells with a PCR-generated fragment containing ypsB flankingsequences and a kan drug resistance cassette. The deletion ofypsB extends from nucleotide –6 to 291 relative to thegene's coding region. This fragment was constructed accordingto established "long flanking homology" PCR protocols (55).Briefly, ypsB upstream and downstream flanking regions wereamplified in a first round of PCR using oligonucleotides OFG158plus OFG159 and OFG160 plus OFG161, respectively. In parallel,a kan resistance cassette was amplified from plasmid pFG4 withprimers OFG112 and OFG113. The three fragments were purifiedand linked in a second PCR in the absence of added primers.An aliquot (usually 1/10 of a microliter) of the linked fragmentsfrom the second PCR was amplified in a third PCR with primersOFG158 and OFG161, purified, and transformed into B. subtiliscompetent cells. The presence of the expected deletion in Kan^rtransformants was confirmed by PCR.

Sequence analysis. Sequences were retrieved from the SSDB (http://www.genome.jp/kegg/ssdb/)or STRING (http://www.bork.embl-heidelberg.de/STRING/) (54)databases. We used the bidirectional best-hit tool to selectYpsB and DiviVA orthologs from SSDB. The alignment was builtusing ClustalW (52) and manually edited and prepared in Jalview(6). The tree was inferred by RAxML (49), using the amino acidsubstitution model WAG with gamma correction, with model parametersestimated by maximum likelihood.

Fluorescence microscopy. Microscopy was performed on a Nikon TE300 inverted microscopeequipped with filters for green fluorescent protein (GFP) (EndowGFP set, 41018; Chroma Technology) and FM4-64 and mCherry (TexasRed Brightline set, TXRED4040-B; Semrock). Images were capturedwith a Roper CoolSnap HQ camera. Exposure times varied from0.5 to 3 s. Images were processed and analyzed with MetaMorphversion 7.1 (Universal Imaging) and ImageJ (http://rsb.info.nih.gov/ij/)software.

To perform microscopy, cells were grown to exponential phaseand mounted on slides covered with a layer of agarose-solidifiedminimal medium or phosphate-buffered saline (PBS). Xylose wasused at 0.25% and IPTG at 50 µM to induce the expressionof GFP or mCherry fusion proteins. Membranes were stained withFM4-64 (Molecular Probes) at a final concentration of 1 µg/ml.Stain was usually added directly to cells in medium 5 min beforemounting and imaging. For some depletion experiments, cellswere grown directly in homemade microscope chambers for visualization.These chambers were assembled the following way: a small volume(1 to 2 µl) of the cells to be visualized was spottedon a square (dimensions, 5 by 5 by 1 mm) of PBS solidified with1.5% agarose and allowed to dry briefly. The agarose squarewas next adhered to a coverslip such that the cells would besandwiched between the agarose and the coverslip. To keep theagarose from moving, 1% low-melting-point agarose in PBS wasused to glue the edges of the square to the glass. The coverslip/cellpad assembly was attached to an aluminum slide with a circularhole (15 mm) by use of silicone grease. The chamber createdthis way was filled with medium containing FM4-64 (0.5 µg/ml)and inducers when appropriate. To avoid evaporation, a secondcoverslip was used to seal the top of the chamber. The chamberswere then incubated at 37°C for the period necessary fordepletion to occur (usually 2 to 3 h). The use of the chambersallowed better visualization of filaments, which are fragileand tend to undergo damage during the manipulations necessaryfor subjecting liquid cultures to microscopy (shaking and pipetting,etc.).

RESULTS AND DISCUSSION

Top

RESULTS AND DISCUSSION

YpsB is a paralog of DivIVA. YpsB is a 98-amino-acid protein annotated in the B. subtilisgenome databases (Subtilist [http://genolist.pasteur.fr/SubtiList/]and BSORF [http://bacillus.genome.jp/]) as "unknown; similarto unknown proteins." The gene for YpsB is located at 200 degreesin the B. subtilis chromosome, apparently in an operon withypsA, the gene for another hypothetical protein of unknown function.Comparison of the gene neighborhoods of ypsB in several genomesrevealed that, besides ypsA, the genes for the Holliday junctionresolvase RecU and the type 1 penicillin binding protein PBP1 are frequently found in the vicinity of ypsB. The potentialsignificance of this gene order conservation is discussed below.

YpsB is predicted to have a central coiled-coil region spanningamino acids 32 to 70 according to the program Paircoil2 (http://groups.csail.mit.edu/cb/paircoil2/paircoil2.html).Under low-stringency conditions (high E-value cutoffs), YpsBcan be detected in BLAST searches using DivIVA as the querysequence, and, indeed, there has been previous mention in theliterature that YpsB could be related to DivIVA (10, 34). Becausecoiled-coil-containing proteins tend to identify other coiled-coilproteins in sequence similarity searches due to the convergent,nonhomologous nature of this signature, and because the similarityscore between DivIVA and YpsB is very low (<25% for B. subtilishomologs), it was unclear whether DivIVA and YpsB are trulyrelated. To investigate this, we carried out detailed sequencecomparisons and phylogenetic analysis of DivIVA and YpsB. Amultiple sequence alignment was constructed using sequencesof several proteins that could be unambiguously identified asDivIVA or YpsB orthologs. Inspection of this alignment (Fig.1A) revealed that the similarity between DivIVA and YpsB wasnot restricted to their coiled-coil regions. In fact, it isthe N-terminal regions of DivIVA and YpsB which exhibit thehighest degree of similarity between the two proteins. Thisregion, which comprises about 30 to 40 amino acids, containsseveral invariant or highly conserved residues which definea signature motif for this protein family. Notably, owing tothis conserved N-terminal domain, we have detected several instancesin which homologs of YpsB appear to be misannotated as DivIVAor DivIVA-like proteins in databases. In contrast, the predictedcoiled-coil segments and the C-terminal portions of both proteinsdisplay lower levels of similarity. The C terminus in particularis quite variable both in sequence and in length among DivIVAproteins. We have noticed, however, that YpsB proteins havea highly conserved block of amino acids at their extreme C termini(Fig. 1A). This sequence block, whose consensus is TNFDILKRLSNLEKHVFG,can be used to clearly distinguish YpsB from DivIVA proteins.Another way to distinguish the two proteins is by their lengths:whereas YpsB proteins are usually less than 120 amino acidslong, DivIVA is always longer than 160 amino acids. Furthermore,in bacteria with genes for both proteins, divIVA was alwayslonger than ypsB.

View larger version (68K):
[in this window]
[in a new window]

FIG. 1. YpsB is a paralog of DivIVA. (A) Multiple sequence alignment of representative YpsB and DivIVA sequences. The conserved C-terminal residues of YpsB are boxed. We have also marked the boundaries used to split YpsB in three segments (N-terminal, coiled-coil, and C-terminal segments) in the deletion analysis shown in Fig. 5. (B) Phylogenetic tree of the same sequences used in panel A. Sequences are identified by organism codes and accession numbers from the SSDB database (http://www.genome.jp/kegg/ssdb/). Organisms are Bacillus anthracis A2012 (baa), Bacillus halodurans (bha), Bacillus subtilis (bsu), Clostridium acetobutylicum (cac), Lactobacillus casei (lca), Listeria monocytogenes EGD-e (lmo), Staphylococcus aureus N315 (sau), Streptococcus pyogenes MGAS10750 (serovar M3) (spi), and Streptococcus pneumoniae TIGR4 (spn).

To infer the evolutionary history of DivIVA and YpsB, we alignedall members of the DivIVA COG (COG3599) in the STRING database(54) and the bidirectional best hits from complete bacterialgenomes from the SSDB database using ClustalW (52), followedby manual editing and selection of sequences (Fig. 1A). A maximumlikelihood phylogenetic tree, depicted in Fig. 1B, was theninferred. We used Clostridium acetobutylicum as an outgroupsince this organism is a member of the Clostridium group, whichis closely related to the orders Lactobacillales and Bacillales(41) and has no DivIVA homologs located close to the recU-ponAlocus or that match the ypsB signature defined above (data notshown). This tree shows that DivIVA and YpsB paralogs correspondto two well-defined groups that most likely originated froma duplication of a DivIVA-like protein in the last common ancestorof the Lactobacillales and Bacillales orders.

Subcellular localization of YpsB. The similarity between DivIVA and YpsB suggested that YpsB couldbe a division protein. To test this, we created B. subtilisstrains expressing N-terminal and C-terminal fusions of GFPto YpsB. The N-terminal fusion was placed under the controlof the xylose-inducible P_xyl promoter and inserted at a nonessentiallocus (amyE) in the B. subtilis chromosome to generate a merodiploidstrain containing the fusion plus wild-type ypsB. The C-terminalfusion was integrated by single crossover at the ypsAB operonitself and therefore is expressed under the control of the nativeypsAB promoter. Fluorescence microscopy experiments with strainsbearing the N-terminal GFP-YpsB fusion revealed strong midcellfluorescent bands in vegetative cells growing in rich media(LB, CH) (Fig. 2A), or in minimal media such as TSS (data notshown) (50). Localization of the GFP-YpsB fusion did not requirethe untagged copy of the protein, since similar results wereobtained in a strain that had the wild-type copy of ypsB deleted(see the first row of Fig. 5 below). This localization pattern,which suggests that YpsB is indeed part of the division complex,was virtually identical for the C-terminal YpsB-GFP fusion (Fig.2B).

View larger version (26K):
[in this window]
[in a new window]

FIG. 2. YpsB is a divisome-associated protein. (A) Localization of GFP-YpsB. Strain JR22 was grown to mid-log phase (optical density at 600 nm = 0.3 to 0.5) and imaged as described in Materials and Methods. Note that besides the ring fluorescence there is significant cytoplasmic fluorescence. Arrows point to examples of cells in which GFP-YpsB or YpsB-GFP (in panel B) are retained at the constricted septa (the new pole). (B) Localization of YpsB-GFP. Strain JR47 was grown to mid-log phase and imaged as described in Materials and Methods. Note here and in Fig. 3 and 4 that YpsB-GFP shows significant peripheral staining besides the ring fluorescence. (C) Localization of DivIVA-GFP. Strain FG96 was grown to mid-log phase and imaged as described in Materials and Methods. Note the close spacing between DivIVA-GFP bands, which indicates that DivIVA is present both at old and new division sites. (D) Localization of GFP-YpsB during sporulation. Strain JR22 was sporulated by the method of Sterlini and Mandelstam (50) and imaged 2 h after resuspension. Arrows point to examples of asymmetric GFP-YpsB rings. Bar = 5 µm.

View larger version (58K):
[in this window]
[in a new window]

FIG. 5. YpsB localization to the divisome is mediated by its N-terminal region. From top to bottom, we show the localization pattern of full-length GFP-YpsB, GFP-N, GFP-N-cc, GFP-C, GFP-cc-C, and GFP-cc, both in the presence of endogenous YpsB (wild-type [WT] column) and in the absence of endogenous YpsB (ypsB mutant column). WT strains are as follows: GFP-YpsB strain, JR22; GFP-N strain, JR133; GFP-N-cc strain, JR134; GFP-C strain, JR136; GFP-cc-C strain, JR135; and GFP-cc strain, JR137. ypsB mutant strains are as follows: GFP-YpsB strain, JR61; GFP-N strain, JR143; GFP-N-cc strain, JR144; GFP-C strain, JR146; GFP-cc-C strain, JR145; and GFP-cc strain, JR147. Microscopy was carried out as described in Materials and Methods with log-phase cells. No membrane stain was added to the samples to avoid potential deleterious effects on localization. That GFP-N-cc is the only deletion that still localizes in the absence of endogenous YpsB suggests that the N terminus of YpsB is the region responsible for targeting to the divisome. For a detailed discussion of these results, see the main text. Bar = 5 µm.

We have noticed in some of our pictures that the GFP fusionsto YpsB exhibited some peripheral membrane staining in additionto the enrichment at the septum. This was clearer for the YpsB-GFPfusion (see Fig. 3 for an example) and was particularly apparentfor some filamentous cells in the dependency experiments shownin Fig. 4. The peripheral staining suggests that YpsB has affinityfor the membrane and, indeed, predictions using the programAmphipaSeeK (47) (http://npsa-pbil.ibcp.fr/) indicate that theN terminus of YpsB is likely an amphipathic alpha helix capableof membrane association. We have tested this prediction by cellfractionation experiments and found that both YpsB-GFP and GFP-YpsBare present in approximately equal amounts in the membrane andsoluble fractions of B. subtilis cells, as one would expectfor a peripherally associated membrane protein (see Fig. S1in the supplemental material). Thus, YpsB is a membrane-associatedprotein. It is unclear why we cannot always see the peripheralpattern of YpsB by fluorescence microscopy. Because the peripheralstaining by YpsB is subtle, we speculate that it is very sensitiveto microscopy conditions (focus precision, the physiologicalstate of the cells on the slides, and the concentration of membranedye, etc.) and thus not as reproducible as the septal signal.Another possibility is that YpsB localization is dynamic, suchthat the protein is associated with the membrane only transiently.

View larger version (13K):
[in this window]
[in a new window]

FIG. 3. YpsB is a late division protein. Colocalization of FtsZ and YpsB. Strain JR132, expressing FtsZ-mCherry and YpsB-GFP fusions, was grown to mid-log phase (optical density at 600 nm = 0.3 to 0.5) and imaged as described in Materials and Methods. Arrows mark Z-rings that are not decorated with YpsB. These presumably represent young division complexes. Asterisks mark YpsB rings that are not decorated with FtsZ. These should correspond to mature divisomes in the final phase of cytokinesis, when FtsZ has already left the complex. Bar = 5 µm.

View larger version (35K):
[in this window]
[in a new window]

FIG. 4. Localization dependence of YpsB-GFP. (A) Assembly hierarchy of the B. subtilis divisome. The question mark next to FtsW indicates that there are no published data to support its placement. Based on the results from panel B, YpsB (in gray) has been included in the appropriate position in the pathway. (B) YpsB-GFP localization in division mutants. The C-terminal YpsB-GFP fusion was introduced into different division gene mutants and localization was assayed in cells grown to mid-log phase. For mutations in essential genes, conditional (depletion or temperature-sensitive) alleles were used. In this case, mutants were grown for at least five generations under restrictive conditions before being imaged. For each mutant, a panel corresponding to the YpsB-GFP image (gray) and a panel with the overlay of the YpsB-GFP channel (green) and the corresponding membrane (FM4-64) channel (red) are shown. JR64, ftsZ depletion strain; JR128, ftsA depletion strain; JR125, divIC temperature-sensitive mutant; JR162, pbpB (PBP 2B) depletion strain; JR31, divIVA minD mutant. Arrow points to minicell. (C) Colocalization of FtsZ-mCherry and YpsB-GFP in divIC(Ts) filaments (strain JR126). Bars = 5 µm.

The localization of YpsB as bands at the division site was foundpreferentially in cells in which a septum (complete or incomplete)was already detectable with the FM4-64 membrane stain (compareGFP and membrane panels in Fig. 2A and B). This was detectedin 70% of 176 cells analyzed and is in contrast with what isseen for proteins such as ZapA, FtsA, and EzrA, which interactdirectly with FtsZ and localize before any sign of septum ingrowth.Thus, YpsB appears to be a late division protein, as is thecase for its paralog, DivIVA. A side-by-side comparison of DivIVAand YpsB, however, revealed that their localizations were notidentical. A feature of DivIVA that is crucial for its functionis its retention at the newly formed cell pole after septationis completed (11, 32). The retention at the cell poles resultsin a localization pattern in which a DivIVA band is presentevery 3.0 to 3.3 µm along a chain of dividing B. subtiliscells (Fig. 2C). It also gives rise to cells that show crescentor dot-like accumulations of DivIVA at their mature poles (heredefined as poles that have already become hemispherical as aconsequence of cell separation) (11, 32). YpsB also seems tobe retained, at least momentarily, at the newly formed poleafter the septum is completed. Examples of this are the markedcells in Fig. 2A and B, in which a bright YpsB dot can stillbe seen in septa that are in the process of becoming maturepoles. The frequency of such cells in the population was 18%(n = 516). Nevertheless, the observations that most old septain a chain have no YpsB bands (thus the wider spacing betweenYpsB bands, i.e., 1 band per 5.5 to 6.0 µm) and that cellswith YpsB accumulation at their mature poles are almost absentfrom the population suggest that YpsB is less efficiently retainedat cell poles than is DivIVA.

Recently, Dervyn et al. identified YpsB in a yeast two-hybridscreen as one of several proteins capable of interacting withScpA, a component of the bacterial condensin complex (10). Thisobservation suggests that ScpA and YpsB should exhibit similarlocalization patterns. However, a comparison of the subcellularlocalizations of YpsB and ScpA, which has been shown to residein a focus over nucleoids (33), indicates that the two proteinsexhibit little if any overlap in their distributions. Althoughthis result per se does not rule out an interaction betweenYpsB and ScpA in vivo, it suggests that if an interaction existsit must involve only a small fraction of the molecules of eachprotein present in a B. subtilis cell.

We also investigated the localization of YpsB during sporulation.A hallmark of sporulation in B. subtilis is the switch in thepositioning of the division machinery from a medial to a polarposition (29). As shown in Fig. 2D, GFP-YpsB localized as asymmetricbands in cells undergoing early sporulation. These cells frequentlycontained a single strong YpsB band close to one of the cellpoles, a pattern indistinguishable from the one reported forseveral division proteins other than FtsZ (7, 14, 18). Therewere also some cells in which there was one strong band at oneof the poles and a weaker band at the other pole. This patterncan also be detected for other division proteins such as ZapAand SpoIIE (F. Gueiros-Filho, unpublished results) and likelyrepresents the transient assembly of a second polar divisome,before its inhibition by the action of proteins responsiblefor the maintenance of the sporangium's asymmetry (40, 12).Thus, YpsB seems to follow the change in divisome positioningthat occurs during sporulation. The presence of YpsB in thesporulation septum represents another difference between thisprotein and DivIVA. Apparently, DivIVA does not localize tothe septum during sporulation (51).

One caveat of the sporulation experiments described above isthat GFP-YpsB was expressed from an inducible promoter. We wereunable to carry out a similar experiment with the C-terminalYpsB-GFP fusion under the control of the native ypsB promoterbecause the strain bearing this fusion sporulates very inefficiently,presumably due to a polar effect of this insertion onto theRNase E RNA gene downstream of ypsB. Thus, it is not certainthat YpsB is normally produced, or part of the divisome, duringsporulation. Indeed, there is evidence from expression profilingand chromatin immunoprecipitation experiments that ypsA, thegene upstream of ypsB, is repressed by Spo0A at the onset ofsporulation (35). Because ypsA and ypsB are likely to be inan operon, these data suggest that there may be a programmedreduction of YpsB levels during sporulation. Nevertheless, thefinding that YpsB, if present, can undergo the switch in localizationthat occurs to division proteins during sporulation reinforcesthe idea that YpsB is a component of the division machinery.

Timing of YpsB localization to the divisome. The data in Fig. 2 suggest that YpsB associates with the divisomesome time after the Z-ring is formed. To confirm this and tomore precisely assess the timing of YpsB association with thedivisome, we colocalized YpsB and FtsZ in actively dividingcells expressing YpsB-GFP and FtsZ-mCherry fusion proteins.Analysis of this strain showed that FtsZ and YpsB colocalizedonly in a fraction of the cells in the population (Fig. 3).In cells in which a clear Z-ring could be detected, we observedcolocalization between the Z-ring and YpsB in about 50% of them.In the other half of the cells, YpsB was still associated withthe old division site, whereas new Z-rings had already formedto start the next round of septation (Fig. 3). In line withthis idea, the cells in which the Z-ring and YpsB did not colocalizetended to be the smallest in the population, as expected forcells in the initial stages of their cycle. Similar frequenciesof colocalization have been reported for DivIVA and FtsZ (reference22 and A. R. Pancetti, G. L. S. Meira, and F. Gueiros-Filho,unpublished results), suggesting that the two proteins associatewith the divisome at a similar stage.

Given that under the conditions of the experiment shown in Fig.3 (rich medium, 37°C, generation time of $~$ 25 min) Z-ringsare present in 90% of the population (data not shown), it canbe estimated that YpsB localizes in 45% (90% times 50%) of thecells in the same culture. Assuming that cell growth in theseexperiments approaches a steady state, this means that it willtake, on average, 45% of a generation for YpsB to get to thedivisome. Similar delays have been estimated for the recruitmentof E. coli late division proteins to the divisome (1). In thecase of E. coli, the delay has been proposed to reflect theneed for the division site to undergo some kind of maturation,probably related to peptidoglycan synthesis, before some ofthe division proteins can assemble into the complex (1). Alternatively,the delayed association of some division proteins with the divisomemay result from their expression being restricted to a periodof the cell cycle. At present we have no evidence to supporteither hypothesis in the case of YpsB.

Recruitment of YpsB to the divisome requires the late division proteins. Studies with E. coli and B. subtilis have found that the assemblyof division proteins into the divisome exhibits a complex patternof interdependencies (for a summary of the B. subtilis work,see reference 13, and for an up-to-date view of E. coli, seereference 16). These interdependencies translate into an "assemblyhierarchy" in which a certain protein depends on the proteinsupstream of it in the pathway, but not on the proteins downstreamof it, to localize to the divisome (Fig. 4A). In the case ofB. subtilis, this hierarchy is composed of four main levels.First, there is FtsZ, which serves as the scaffold of the complexand thus is required for the localization of all other divisionproteins. The next level includes the proteins that bind directlyto FtsZ and whose association with the divisome is likely tobe independent of each other (FtsA, SepF, and ZapA). The thirdlevel includes the proteins with extracellular domains involvedin peptidoglycan synthesis (DivIB, DivIC, FtsL, PBP 2B, andperhaps FtsW). These seem to be mutually interdependent forlocalization and thus are thought to exist as a subcomplex ofthe divisome (13). Lastly, there are DivIVA and the MinC andMinD proteins. DivIVA (and the accompanying MinCD) have beenproposed to go to the cell poles, where they carry out theirfunction by first associating with the divisome (32). Recently,it has been shown that DivIVA can localize to the cell poleseven in the absence of divisome assembly (19, 22). These findingsshow that DivIVA can find the cell poles independently of itsassociation with the divisome. Nevertheless, DivIVA does associatewith the divisome, and this association was shown to dependon PBP 2B (and presumably also on the other components of thePBP 2B subcomplex) (32). Thus, DivIVA/MinCD represent the lastlevel in the divisome assembly hierarchy.

To place YpsB within the divisome hierarchy, we analyzed howthe localization of YpsB was affected in the absence of proteinsfrom each of the four levels (Fig. 4B). In the case of essentialproteins such as FtsZ, FtsA, DivIC, and PBP 2B, depletion ortemperature-sensitive strains were used, whereas for DivIVAa null mutant was used. In cells depleted of FtsZ, YpsB-GFPlost the band-like localization pattern and instead appeareddistributed throughout the cell (Fig. 4B). This finding wasexpected, because no assembly of the division complex occursin the absence of FtsZ, and confirms that YpsB is indeed recruitedto the divisome. The depletion of FtsA, the next protein inthe assembly hierarchy, produced similar results: YpsB-GFP waspredominantly delocalized, although occasional bands were observedin the filaments. Since division can still occur, albeit withlowered frequency, in the absence of FtsA (25, 26), these bandslikely correspond to the rare functional divisomes that manageto form in these cells. The general absence of YpsB bands, however,indicates that YpsB is incorporated into the divisome at somestep after the association of FtsA. We next tested whether thetargeting of YpsB depends on the late division proteins DivICand PBP 2B. By using a DivIC temperature-sensitive mutant [divIC(Ts)]we found that YpsB becomes completely delocalized in divIC(Ts)cells grown at the restrictive temperature (45°C) (Fig.4B). A similar result was obtained in a PBP 2B depletion strain:no YpsB bands could be detected after 3 h of growth in the absenceof PBP 2B expression. Immunoblot analysis of YpsB-GFP in thedifferent division mutant backgrounds showed that the stabilityof the protein is unaffected under filamentation conditions,thus ruling out the degradation of the fusion as the explanationfor the loss of localization detected in our experiments (seeFig. S2 in the supplemental material). To demonstrate that theearly steps of divisome assembly still occur in the absenceof these late proteins, we simultaneously localized FtsZ andYpsB in the divIC(Ts) filaments. As shown in Fig. 4C, the presenceof Z-rings indicates that the assembly of at least part of thedivisome occurs in the absence of DivIC. Because YpsB failsto associate with these "incomplete" divisomes (Fig. 4C), weconclude that YpsB targeting requires the late divisome proteins,perhaps the DivIB-DivIC-FtsL-PBP 2B complex itself, or proteinsrecruited to the divisome by this complex. Finally, we investigatedwhether YpsB required DivIVA for its recruitment. Because DivIVA,similarly to YpsB, depends on PBP 2B for its association withthe divisome (32), we thought that DivIVA could be the linkbetween YpsB and the rest of the divisome. This experiment wascarried out in a divIVA minCD double mutant strain to allowefficient division in the absence of DivIVA (4, 11) and revealedthat YpsB localizes normally in the absence of DivIVA (and MinCD)(Fig. 4B). Typical YpsB bands were found both in regular septaand in septa associated with minicells (Fig. 4B), which arefrequently produced by the divIVA minCD mutant strain. A reciprocalexperiment showed that the localization of DivIVA is normalin a ypsB mutant (data not shown), suggesting that YpsB andDivIVA associate independently with the divisome. From thisanalysis we conclude that the recruitment of YpsB to the divisomeis mediated by interaction either with the late division proteinsor with other, yet-unidentified, division proteins whose recruitmentis dependent on the late proteins. Another formal possibilityis that the recruitment of YpsB is not mediated by protein-proteininteractions with divisome components but instead would be dueto modifications of the membrane and cell wall at the divisionsite promoted by active septation. Unfortunately, we have noway to test this hypothesis at present.

The dependency pattern of YpsB, which is indistinguishable fromthe one exhibited by DivIVA (32), places YpsB among the lastgroup of proteins in the B. subtilis hierarchy (Fig. 4A). Giventhe relatedness of YpsB and DivIVA, it is conceivable that theyinteract with the same protein in the division complex.

In addition to assessing the dependency of YpsB on the divisomeproteins, we also investigated whether the condensin subunitScpA contributes to YpsB localization. Even though the localizationpatterns of ScpA and YpsB are quite different, the proposedinteraction between the two proteins suggests that they mightaffect each other's localization. Microscopic observation ofYpsB-GFP in a scpA mutant, however, revealed that YpsB stilllocalizes to the division complex (data not shown), suggestingthat ScpA plays no role in the targeting of YpsB. Similarly,the localization of ScpA (determined using a ScpA-YFP fusion)was not affected in ypsB mutant cells (data not shown), indicatingthat YpsB is not required for ScpA targeting.

Mapping of YpsB localization determinants. To define the region (or regions) of YpsB required for its associationwith the divisome, we constructed a series of YpsB deletions.All the deletions were fused at their N termini to GFP. Analysisof the YpsB multiple sequence alignment (Fig. 1A) shows thatthe B. subtilis protein has three clearly recognizable regions:an N-terminal region spanning amino acids 1 through 36, a coiled-coilsegment spanning amino acids 37 through 70, and a C-terminalregion from amino acid 71 to 98. Even though these regions areunlikely to represent domains in a structural sense, they roughlycorrespond to the protein's predicted segments of secondarystructure and will be referred to loosely as domains. Usingthese boundaries, we constructed deletions containing the N-terminaldomain alone (GFP-N), the N-terminal domain plus the coiled-coildomain (GFP-N-cc), the C-terminal domain alone (GFP-C), theC-terminal domain plus the coiled-coil domain (GFP-cc-C), andthe coiled-coil domain alone (GFP-cc). Each of these fusionswas introduced into a wild-type (PY79) strain and was assayedfor localization by fluorescence microscopy. Because the presenceof a coiled coil suggests that YpsB can homo-oligomerize, wealso assayed the truncated fusions in a ypsB mutant strain.Comparison of the results in the presence and absence of wild-typeYpsB should allow the distinction between deletions that remaincapable of associating with their target on the divisome anddeletions that associate with the divisome indirectly, throughan interaction with endogenous YpsB.

As shown in Fig. 5, three of the truncated GFP-YpsB fusions,GFP-N-cc, GFP-C, and GFP-cc-C, showed some degree of localizationwhen expressed in a wild-type strain (wild-type column). Thelocalization was weak (as judged by the ratio of fluorescencein the septum versus the cytoplasm) in the deletion containingjust the C-terminal domain fused to GFP (GFP-C) but was significantlystronger in the deletions that contained the coiled-coil segmentin addition to the N- or C-terminal domains. This suggests thatthe coiled-coil segment is capable of enhancing the affinityof YpsB for the divisome, probably by promoting the oligomerizationof YpsB chains. Indeed, Claessen et al. showed recently thatYpsB is capable of binding to itself in a bacterial two-hybridassay (5). It must be noted, however, that the coiled-coil regionalone has a limited capacity to oligomerize, since the GFP-ccconstruct failed to localize in the wild-type strain (Fig. 5,last row). To rule out protein degradation as the reason forthe lack of localization of some mutants, we verified the presenceof the fusion proteins by immunoblotting with anti-GFP antibodies.This control showed that all mutant proteins (in both the wild-typeand ypsB mutant backgrounds) were stable and produced at levelssimilar to that of the full-length protein (see Fig. S3 in thesupplemental material).

Assaying localization in a strain lacking endogenous YpsB producedquite different results. Whereas the GFP-N-cc fusion still exhibitedproper localization, the C terminus-containing fusions becamedelocalized (Fig. 5, ypsB mutant column). Localization of theGFP-N-cc fusion in the ypsB mutant background was significantlyweaker than in the presence of endogenous YpsB, suggesting thatthe full-length protein stabilizes the interaction of GFP-N-ccwith the divisome. Nevertheless, the observation that only theGFP-N-cc fusion still localizes in the absence of full-lengthYpsB suggests that the N terminus is required for the recognitionof the divisome. Because the fusion containing just the N-terminalpart of YpsB (GFP-N) fails to localize, it is likely that thetargeting domain extends beyond the N terminus, probably includingthe coiled-coil portion of the protein. However, at presentwe cannot rule out the possibility that the N terminus containsall the signals necessary for YpsB targeting. In this case,the effect of the coiled-coil segment would be to allow theGFP-N-cc protein to oligomerize, thus increasing the affinityof the N-terminal domain for its target. In support of the conclusionthat the N terminus of YpsB is necessary for divisome targeting,it should be pointed out that the similarity between YpsB andDivIVA is highest in their N-terminal regions (Fig. 1A). Giventhat YpsB and DivIVA associate with the divisome at the samepoint in the assembly hierarchy (Fig. 4), potentially by recognizingthe same divisome protein, it is possible that they will sharetargeting sequences. Furthermore, two previously described mutationsthat affect DivIVA localization (38) map to arginine (R23 inB. subtilis YpsB) and glycine (G24 in B. subtilis YpsB) residuesthat are highly conserved in the members of the DivIVA/YpsBfamily (Fig. 1A). Regarding the role of YpsB's C terminus, wepropose it could be part of the self-oligomerization elementsof YpsB, along with the coiled-coil region. The ability of theC terminus to self-associate would allow it to localize in thewild-type background, where it can bind to the full-length protein,but would not allow it to localize in the ypsB mutant strain.A role for the C terminus of YpsB in self-association is supportedby the observation that the coiled-coil region alone does notseem sufficient for self-association of YpsB (the GFP-cc fusionfails to localize even in the wild-type background).

YpsB is not essential for septum formation. To determine the role of YpsB during division, we constructeda ypsB null mutant by replacing the coding region of the genewith a kanamycin resistance cassette by use of long flankinghomology PCR. Inspection of the mutant by fluorescence microscopyshowed that the ypsB mutant divides as well as its wild-typeparental strain (Fig. 6A). We tested the ypsB mutant also insporulation and found that both polar septum formation and sporetiters were not affected (data not shown). Thus, YpsB is notessential for division during vegetative growth or sporulation.Given the normal sporulation efficiency of the ypsB mutant,we conclude that YpsB is not required for events other thanpolar septation that are necessary for spore development. Thisis in contrast with DivIVA, which, besides its function as thetopological specificity factor of the Min system, plays a rolein the polar anchoring of chromosomes during sporulation (2,51).

View larger version (87K):
[in this window]
[in a new window]

FIG. 6. Phenotypes of the ypsB mutant. (A) Comparison of log-phase cells of wild-type (WT) (PY79) and ypsB mutant (JR46) strains. No difference could be detected in the frequencies or regularities of septa. (B) Synthetic phenotypes of ypsB and ftsA mutants. The wild-type (PY79), the ypsB mutant (strain JR46), a conditional ftsA mutant in which FtsA expression is dependent on xylose (P_xyl-ftsA, strain FG718), and a conditional ftsA ypsB double mutant (strain JR166) were streaked on LB or LB plus 0.5% xylose and grown overnight at 37°C before being photographed. Bar = 5 µm.

Several of the recently characterized division genes such aszapA, noc, and sepF (ylmF) cause no or very mild division defectswhen inactivated. This suggests that these proteins play redundantroles or are involved in nonessential steps necessary for thefine-tuning of the division process. Mutations in these genes,however, often exhibit synergistic or synthetic phenotypes whencombined with mutations in other division genes. Well-documentedexamples are the synthetic division defects of zapA ezrA (18),noc minD (56), sepF (ylmF) ezrA (20), and sepF (ylmF) ftsA (25)double mutants. To test if the same was true of ypsB, we combinedthe ypsB null allele with mutations in several other divisiongenes (ezrA, zapA, noc, minCD, divIVA, and ftsA). No syntheticphenotype could be detected for the ypsB ezrA, ypsB zapA, ypsBnoc, and ypsB minCD double mutants (data not shown). In thecase of ftsA, however, we noticed a synthetic effect. In contrastwith E. coli, in which ftsA is essential and absolutely requiredfor division, the inactivation of ftsA in B. subtilis producesa viable mutant (25, 26) (Pancetti and Gueiros-Filho, unpublished).The ftsA mutant divides at a low frequency but still producescolonies with high efficiency when plated on solid media. Weused an ftsA depletion strain to investigate the interactionbetween ftsA and ypsB and found that depletion of FtsA in anypsB mutant background produced a phenotype much more severethan that produced by depletion of FtsA in a wild-type background.This phenotype could be easily observed on plates, where cellslacking FtsA and YpsB lysed after overnight growth, while cellslacking either YpsB or FtsA alone produced healthy colonies(Fig. 6B). The lysis of the ftsA ypsB double mutant could alsobe inferred from the reduced growth rate of this strain in liquidmedium (see Fig. S4 in the supplemental material). The tendencyof the ftsA ypsB mutant strain to lyse suggested that this mutantwas undergoing extreme filamentation. To test this, we comparedthe cell sizes of the ftsA ypsB mutant strain with those ofthe ftsA mutant strain after different times of FtsA depletion(2, 3, and 5 h). We failed to detect a significant differencein cell length between the two strains in these experiments.This may be due to the difficulty of faithfully measuring thevery long cells generated by FtsA depletion. The longest filamentsoften coil into spaghetti-like swirls or extend beyond the microscopefield of view in ways that compromise their accurate scoring.Very long filaments are also more prone to lysis and thus maybe selectively eliminated from the samples we examined. Alternatively,the increased lysis of the double mutant could be a consequenceof a defect in peptidoglycan synthesis, as discussed below.

Even though the late recruitment of YpsB to the divisome suggestsit does not interact directly with FtsZ, the established roleof FtsA as a promoter of Z-ring formation led us to investigatewhether YpsB could affect FtsZ polymer stability or organization.One way to detect the stabilizing effects of a known proteinon Z-ring formation in vivo is to test whether its overexpressioncan overcome the lethal effect of excess MinD, a Z-ring inhibitor(18). Using this assay, we found that YpsB overexpression wasincapable of suppressing MinD lethality (data not shown), suggestingthat YpsB does not act as a positive modulator of Z-ring formation.Thus, the synergism between the ypsB and ftsA mutations is unlikelyto result from extreme destabilization or disorganization ofthe FtsZ polymers in the cell. More likely, the absence of YpsBaffects a late division event such as septal peptidoglycan synthesis.While the impairment caused by the lack of YpsB may be too mildto be noticed on its own, it may have consequences in strains,such as the ftsA mutant, in which divisome assembly is alreadysignificantly weakened. How would YpsB affect septal peptidoglycansynthesis? We speculate that YpsB may contribute to the recruitmentof PBP 1 (encoded by ponA) to the divisome. PBP 1 is a high-molecular-weightclass A PBP with bifunctional transglycosylase/transpeptidaseactivities (42). Even though PBP 1 has not been traditionallythought of as a divisome protein, it is expected that at leastone PBP with transglycosylase activity must be part of the cellwall-synthesizing complex associated with the division machinery,and localization experiments (37, 48) have shown that PBP 1is indeed enriched at septal locations. The speculation thatPBP 1 is a partner of YpsB is based on the conserved presenceof the ponA gene nearby the gene for ypsB in bacterial genomes.In most genomes in which ypsB and ponA (pbp1) are present, theyare separated by only two or three other genes. Another observationsupporting the link between YpsB and PBP 1 is the finding thatPBP 1 orthologs of some bacteria that lack YpsB (notably, severalenterobacteria) display a potential DivIVA/YpsB signature aminoacid sequence. This signature includes only the most conservedresidues of the N terminus of DivIVA and YpsB (F-x[4,6]-R-G-Y-x-x-x-x-x-x-x-F-L)and thus is too short to be detected in BLAST or PSI-BLAST searches.However, it is detected using programs such as Pattern Search.Assuming that, as suggested by our deletion experiments, theseN-terminal amino acids are indeed the main divisome-targetingdeterminant of YpsB/DivIVA, the addition of this sequence tothe PBP 1 protein of enterobacteria would allow it to associatewith the divisome even in the absence of YpsB. In support ofour speculations, recent work from the Errington lab that waspublished while our manuscript was under revision (5) has shownthat, indeed, YpsB interacts directly with PBP 1 and seems topromote the shuttling of this protein between the lateral andseptal sites of cell wall synthesis. Thus, YpsB is a late divisionprotein that may play an important role in the regulation ofpeptidoglycan synthesis during the division cycle.

Conclusion. We have characterized YpsB, a newly identified B. subtilis divisomeprotein. YpsB is evolutionarily related to DivIVA and, indeed,it displays several properties that are similar to those ofDivIVA. It associates with the divisome at a late stage in thedivision cycle and fits into the same step of the assembly hierarchy(depends on the late proteins DivIC and PBP 2B to be recruitedto the divisome). The molecular role of YpsB, however, seemsto be different from that of DivIVA. YpsB is not essential forseptum formation and does not affect positioning of the divisionseptum. YpsB is not as well retained at cell poles as DivIVAand thus there is no evidence that it serves as a pole-markingprotein. Nevertheless, the synthetic phenotype observed betweenftsA and ypsB mutations suggests that YpsB contributes to properand efficient septum formation, perhaps by regulating septalcell wall synthesis.

The identification of yet another division protein points toan ever-increasing complexity of the divisome. There are now16 proteins which are known to be part of the complex in B.subtilis. The future challenge will be to determine the roleplayed by each of these proteins during cytokinesis.

ACKNOWLEDGMENTS

We thank Jonathan Dworkin, Michael Elowitz, Alan Grossman, andPeter Graumann for plasmids and strains and Peter Graumann andDavid Rudner for suggestions and critical reading of the manuscript.We are also indebted to Walter Colli, Maria Julia Manso Alves,and their team for the lab space and hospitality during partof this project.

This work was supported by a Jovem Pesquisador grant from Fundaçãode Amparo a Pesquisa do Estado de São Paulo (FAPESP).J.R.T. and R.F.D.S. were recipients of doctoral fellowshipsfrom Conselho Nacional de Desenvolvimento Científicoe Tecnológico (CNPq). G.L.S.M. was a recipient of a doctoralfellowship from FAPESP.

FOOTNOTES

* Corresponding author. Mailing address: Departamento de Bioquímica, Instituto de Química, Universidade de São Paulo, Av. Prof. Lineu Prestes 748, 05508-000 São Paulo, SP, Brazil. Phone: 55-11-3091-8520. Fax: 55-11-3091-2186. E-mail: fgueiros{at}iq.usp.br

Published ahead of print on 5 September 2008.

Supplemental material for this article may be found at http://jb.asm.org/.

REFERENCES

Top