To identify novel proteins that interact with S. aureus FtsZ or FtsZ-associated proteins in vivo we utilized a previously described GFP affinity-purification strategy (Cristea et al., 2005) using a S. aureus strain that expresses ftsZ-gfp from plasmid pLOW (spa^- pLOW-ftsZ-gfp pGL485; SA103). Expression of ftsZ-gfp was induced with 0.05 mM IPTG as this level of induction has been previously shown to have no observable effect on cellular morphology, cell division or Z-ring formation (Liew et al., 2011). Protein complexes were isolated (see Experimental Procedures for details), separated by SDS-PAGE and bands that were clearly visible by Coomassie staining, as well as remaining gel fragments, were excised and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS-MS; Figure 1). A S. aureus strain expressing only GFP from pLOW (spa^- pLOW-gfp pGL485; SA112) was used as a control to identify proteins not associated specifically with FtsZ, and these were excluded from analysis. Four known division proteins were identified: FtsZ, EzrA, FtsA and SepF. The identification of the self-polymerizing FtsZ protein was expected. EzrA is a known FtsZ-interacting division protein in B. subtilis (Ishikawa et al., 2006) and in S. aureus interacts with FtsZ in a bacterial two-hybrid assay (Steele et al., 2011). SepF and FtsA have also been previously shown to associate with other Staphylococcal cell division proteins using bacterial two-hybrid assays (Steele et al., 2011). In B. subtilis, SepF and FtsA co-purify with FtsZ from in vivo cell extracts (Jensen et al., 2005; Ishikawa et al., 2006). Thus, our results demonstrates the successful isolation of S. aureus division components using GFP-based affinity purification, and provides further evidence that EzrA, SepF and FtsA proteins are genuinely part of an FtsZ-containing complex in S. aureus in vivo.

DnaK was also detected using the FtsZ-GFP pulldown approach. This was of interest as it has been previously observed to be associated with FtsZ in E. coli (Calloni et al., 2012). DnaK is highly conserved in bacteria where it serves as a molecular chaperone during heat stress (Bukau and Horwich, 1998). This protein has not previously been described as being associated with the divisome in S. aureus and localization of DnaK-msfGFP (monomeric superfolder-GFP: Landgraf et al., 2012) using a plasmid-based system (pLOW-dnaK-msfgfp pGL485; SA307) showed a uniform signal throughout the cell cytoplasm with no septal-specific localization pattern (Supplementary Figure 1). This cytoplasmic localization pattern likely reflects interaction of DnaK with various cytoplasmic proteins (Calloni et al., 2012), not just those associated with the divisome.

Although DnaK did not show specific localization to midcell, the potential interaction between DnaK and other divisome components was investigated using the bacterial two-hybrid system (Karimova et al., 1998) anticipating that this might reveal more divisome interactions. Consistent with the affinity-purification results (Figure 1), a direct interaction was detected between DnaK/EzrA, and DnaK/FtsZ (Supplementary Figures 2A,C). Furthermore a novel direct interaction between DnaK/DivIVA was identified. DivIVA has been shown to localize to the division site in S. aureus but its role is unclear (Pinho and Errington, 2004). We therefore further investigated the possible involvement of DivIVA in S. aureus cell division.

Detection of an interaction between DnaK and DivIVA led us to investigate the interaction profile of DivIVA with S. aureus divisome components using the bacterial two-hybrid assay. We detected several pair-wise interactions with DivIVA including FtsZ, EzrA, FtsA, DivIC, DivIB, PBP1 and PBP2 (Supplementary Figures 2B,D).

Using a functional DivIVA-GFP fusion (Supplementary Figure 3), we confirmed midcell localization for this protein (Figure 2) as has been previously shown (Pinho and Errington, 2004). Further analysis of DivIVA-GFP co-localization with the septal membrane (stained with FM4-64) revealed that 60% of cells displayed additional non-septal localization (Figure 2), including foci or arcs that are perpendicular to the septal rings and a small percentage of cells (∼2%) which showed exclusive localization of DivIVA-GFP to the cell periphery (Figure 2).

A previous study to investigate the physiological role of DivIVA in S. aureus using a markerless deletion of divIVA showed no obvious phenotype with regards to growth rates and gross chromosome morphology (Pinho and Errington, 2004). We confirmed these results and also found no observable phenotype using a markerless divIVA deletion in SH1000 (spa^- ΔdivIVA; SA365; data not shown – see Experimental Procedures for more detail).

As DivIVA and DnaK interact (Supplementary Figure 2), the likely relationship between the two proteins was investigated. We found that deletion of both genes led to an exacerbation of the anucleate phenotype exhibited by the dnaK single mutant. We detail our phenotypic analysis of both single and double mutants below.

An S. aureus dnaK::kan insertion mutant (SA210) showed a significant decrease in growth rate (Table 1) at 37°C compared to S. aureus RN4220, consistent with previous observations obtained with a S. aureus COL dnaK::kan insertion mutant (Singh et al., 2007). However, contrary to the previous study, we found that the dnaK::kan insertion mutant had a significantly slower growth rate when also grown at 30°C (29 min for RN4220 vs. 47 min in the dnaK::kan insertion mutant). This discrepancy is possibly due to differences in media types and strain backgrounds in each study. Examination of dnaK::kan cells by microscopic analysis when grown in BHI at 37°C (Figure 3A and Table 1) showed that there was a 22% increase in average cell diameter compared to wild-type (1.1 ± 0.03 μm compared to 0.9 ± 0.02 μm) suggestive of a cell division defect in the dnaK::kan cell population. In addition to this, a significant proportion of cells in the population (10% vs. 0% for RN4220; Table 1) were anucleate. A similar frequency of anucleate cells was also observed in the dnaK mutant when grown at 30°C (Table 1). Importantly, both these morphological changes were not due to a polar effect of the dnaK::kan insertion because ectopic expression of dnaK from pLOW led to a reduction in both cell size and the frequency of anucleate cells back to levels resembling those of wild-type cells (Supplementary Figure 4). Thus, absence of dnaK results in anucleate cells in 10% of the cell population under non-heat stressed conditions.

We found no significant difference in growth rate or cell size in the ΔdivIVA dnaK::kan double mutant (SA213) compared to the dnakK::kan mutant (SA210) in rich broth (BHI; Table 1) or minimal (chemically defined; CDM) media at 37°C (data not shown) indicating that any observable phenotype of the ΔdivIVA dnaK::kan double mutant is not due to changes in growth or a delay in cell division in comparison to the dnaK::kan single mutant. Interestingly, a high proportion of the double mutant cells were anucleate when stained with DAPI and observed under the microscope (Figure 3B). This was similar in both rich and nutrient limiting media: 21% in BHI and 29% in CDM. Importantly, a higher frequency of anucleate cells were observed for the ΔdivIVA dnaK::kan double mutant compared to dnaK::kan alone (two-fold higher; 21% vs. 10%, respectively). Further, a 10-fold and 1000-fold reduction in viability for the ΔdivIVA dnaK::kan double mutant was also observed when stationary phase cells were plated on rich or nutrient-limiting solid media, respectively, compared to the dnaK::kan strain (Figure 3C). This reduction in viability was also observed in mid-exponential phase cultures grown in CDM media but not when cells were grown in BHI (Supplementary Figure 5) suggesting this reduction in viability is more apparent at reduced growth rates. Taken together, the higher frequency of anucleate cells, which will result in non-viable cells and a reduction in cell viability of the ΔdivIVA dnaK::kan double mutant compared to the dnaK single mutant, supports the notion that DivIVA functions with DnaK in maintaining accurate chromosome segregation.

Since DnaK has a well characterized role as a molecular chaperone (Bukau and Horwich, 1998), the direct interaction we observed between DnaK and DivIVA may suggest a role for DnaK in maintaining DivIVA stability. To test this idea, we determined the cellular levels of DivIVA, as well as FtsZ and EzrA (which also showed an association with DnaK; Figure 1) in a dnaK::kan insertion mutant of S. aureus using western blot analysis. Cellular levels of both FtsZ and EzrA-GFP were similar in wild-type S. aureus (LH607 [spa^-] for detection of FtsZ; SA353 [spa^-ezrA::ezrA-gfp pGL485] for detection of EzrA-GFP) and the isogenic dnaK::kan mutant (SA359 [spa^- dnaK::kan] for detection of FtsZ; SA361 [spa^-dnaK::kan ezrA::ezrA-gfp pGL485] for detection of EzrA-GFP; data not shown), demonstrating that DnaK is not required for the stability or septal localization of EzrA or FtsZ in S. aureus at 37°C.

Unlike EzrA and FtsZ, quantitative immunoblotting showed that cellular levels of DivIVA-GFP in the S. aureus dnaK::kan mutant (spa^- dnaK::kan P_{divIV A} divIVA-gfp::P_spac divIVA pGL485; SA363) were reduced by 30 – 50% compared to its parent S. aureus strain expressing divIVA-gfp (spa^- P_{divIV A} divIVA-gfp::P_spac divIVA pGL485; SA356) (Figure 4A). Consistent with this result, fluorescence microscopy showed a marked decrease in the DivIVA-GFP signal intensity in the absence of dnaK (Figure 4B). It is important to note that the reduction in the levels of DivIVA-GFP in the absence of dnaK is unlikely to be caused by the presence of the GFP tag as EzrA-GFP levels remain unaffected by the loss of dnaK (data not shown). Therefore, these results strongly suggest that DnaK plays a specific role in maintaining the stability of DivIVA in S. aureus.

The anucleate phenotype of the dnaK::kan mutant of S. aureus cannot be solely due to the instability of DivIVA since the complete absence of divIVA does not yield a discernible phenotype. The dnaK::kan mutant phenotype is therefore likely to be additionally due to the instability of one of more additional proteins that depend on DnaK for their full stability.

DivIVA is highly conserved in Gram-positive bacteria (Oliva et al., 2010) and DnaK homologs are present in all bacteria (Bukau and Horwich, 1998). We therefore examined whether the requirement for DnaK in maintaining cellular DivIVA levels is conserved in B. subtilis. Indeed, DivIVA-GFP levels in the B. subtilis dnaK::cat mutant (P_{divIV A} divIVA-gfp:: P_{divIV A} divIVA dnaK::cat; SU798) were reduced by 35 – 50% when compared to the B. subtilis strain expressing divIVA-gfp (P_{divIV A} divIVA-gfp:: P_{divIV A} divIVA;SU761) (Supplementary Figures 6A,C). This reduction was also observed with untagged DivIVA (Supplementary Figure 6B). Together, these results demonstrate a requirement for DnaK in maintaining DivIVA stability in both organisms at 37°C.

A role for DivIVA in chromosome segregation has been described previously in some bacterial species. For example, a divIVA::kan insertion mutant in E. faecalis resulted in an approximate100-fold decrease in viability and the majority of cells displayed anucleate cell phenotypes (Ramirez-Arcos et al., 2005) Previous studies in B. subtilis demonstrated that DivIVA functions in chromosome partitioning during sporulation by acting as a tether for RacA to anchor the chromosome origin region at the cell poles (Ben-Yehuda et al., 2003). Since our mutant data points to a role for S. aureus DivIVA in chromosome segregation, we examined whether DivIVA functions with known segregation proteins in S. aureus. As a first step to investigate this, we performed a bacterial two-hybrid analysis to determine if DivIVA interacts with six known S. aureus chromosome segregation proteins: SpoIIIE, FtsK, SMC, SpoOJ, ParC, ParE and the nucleoid occlusion protein, Noc, which displays amino acid sequence similarity to SpoOJ. Interestingly, we found a previously unknown interaction between S. aureus DivIVA and SMC (Supplementary Figure 7). No other pair-wise interactions with DivIVA were observed.

The loss of smc in S. aureus is not lethal but results in the generation of anucleate cells at a frequency of 7% in the SA113 (ATCC 35556) background (Yu et al., 2010). If a smc divIVA double mutant shows an increase in the number of anucleate cells compared to the smc single mutant, it would support our proposal for the involvement of divIVA in chromosome segregation. Accordingly, the divIVA smc double mutant was constructed by introducing the smc::Tn917 mutation (Yu et al., 2010) into RN4220 ΔdivIVA (SA167) creating the smc::Tn917 ΔdivIVA double mutant (SA269). In rich broth, the number of anucleate cells in the double mutant was about three-fold higher (Fisher’s exact test p-value < 0.05) than the smc::Tn917 single mutant (6% ± 2% compared to 2% ± 1%) (Figure 5), suggesting that DivIVA and SMC act together to maintain accurate chromosome segregation.

Bacillus subtilis and Staphylococcus aureus strains used in this study are listed in Supplementary Table 1, plasmids are listed in Supplementary Table 2 and oligonucleotide sequences used (purchased from IDT) are shown in Supplementary Table 3.

Escherichia coli strains were grown in Luria-Bertani broth (LB; Oxoid). S. aureus strains were grown in brain heart infusion broth (BHI; Oxoid) or CDM (chemically defined media) as described previously. B. subtilis strains were grown in PAB medium (Difco Antibiotic Medium 3; BD). Unless otherwise stated, all strains were grown at 37°C. When required, antibiotics were added to a final concentration of 100 μg ml^-1 ampicillin, 10 μg ml^-1 tetracycline, 5 μg ml^-1 erythromycin, 25 μg ml^-1 lincomycin, 50 μg ml^-1 kanamycin, 10 μg ml^-1 spectinomycin, 10 μg ml^-1 chloramphenicol (5 μg ml^-1 for selection in B. subtilis).

Growth rate experiments were performed in triplicate in 96-well flat bottom plates. 150 μl fresh medium (supplemented with 0.05 M IPTG when necessary) was inoculated to an OD₆₀₀ of 0.02 and growth was monitored every 30 min with moderate shaking at the required temperature using a PowerWave HT Microplate Spectrophotometer (Bio-Tek). An unpaired student t-test was used to compare the generation time of RN4220, SA210 (dnaK::kan) and SA213 (ΔdivIVA dnaK::kan).

Gross viability assays were performed as follows: mid-exponential or stationary phase cultures were normalized to an OD₆₀₀ value of 1, serially diluted (10^-3 to 10^-7) in PBS and 10 μl was drop-plated onto BHI or CDM agar. Plates were then air dried and incubated for 24 h at 37°C.

Transformations of S. aureus RN4220 were performed using electroporation as described previously (Kraemer and Iandolo, 1990). Phage transductions into non-RN4220 S. aureus backgrounds using φ11 were performed as described previously (Novick, 1967). B. subtilis SU5 dnaK::cat mutants (SU797 and SU798) were constructed by transferring the dnaK::cat construct from strain BT02 (Schulz et al., 1995) into the SU5 background via natural transformation and homologous recombination using purified genomic DNA from BT02. DNA manipulations and E. coli transformations were performed using standard molecular biology techniques.

Samples were prepared for live cell microscopy as previously described (Pereira et al., 2010) and visualized on a 2% agarose pad. Cells were observed using a Zeiss Axioplan 2 fluorescence microscope equipped with a Plan ApoChromat (100x NA 1.4; Zeiss) objective lens and an AxioCam MRm cooled charged-coupled-device (CCD) camera. To visualize DNA and cell membranes, 1 μg ml^-1 DAPI (Molecular Probes) and 0.25 μg ml^-1 FM 4-64 (Life Technologies) were incubated with S. aureus for 5 and 30 min respectively.

Cell length values were measured directly from digital micrographs using AxioVision software, version 4.6 (Zeiss) with default measurement settings. The Fisher’s exact test was used to compare the number of anucleate cells in the dnaK mutants and was performed with Graphpad Prism software.

To construct plasmid pLOW-dnaK, primers 1280 and 1281, containing SalI and BamHI restriction sites respectively, were used to PCR amplify full-length dnaK from S. aureus RN4220 gDNA and then cloned into plasmid pLOW.

pLOW-msfGFP was constructed by amplification of msfGFP from pDHL1029 (Landgraf et al., 2012) using primer pair 1161 and 1162, and ligation into pLOW using complementary XmaI and EcoRI sites. pLOW-dnaK-msfGFP was subsequently constructed as for pLOW-dnaK.

Plasmid pBCB1-GE IVA700 was used to construct strain SA247 (RN4220 P_{divIV A} divIVA-gfp::P_spac divIVA, pGL485) which expresses divIVA-gfp from its native promoter while placing full length divIVA under the control of the P_spac promoter. Full length S. aureus divIVA was PCR amplified from RN4220 gDNA using primers 1438 and 1452 containing flanking KpnI restriction sites. The PCR product was digested and inserted into pBCB1-GE, to create plasmid pBCB1-GE IVA700. The orientation and sequence of the cloned insert was verified before transformation into RN4220.

pBCB1-GE P_{divIV A} divIVA-gfp was used to construct strain SA289 (RN4220 pBCB1-GE P_{divIV A} divIVA-gfp) which expresses both divIVA-gfp and full length divIVA from the native divIVA promoter. First, full length S. aureus divIVA and the 5’ 200 bps region upstream of divIVA (containing its native promoter) was PCR amplified from RN4220 gDNA using primers 1471 and 1472 containing KpnI restriction sites. The PCR product, along with plasmid pBCB1-GE, was digested with KpnI and ligated, creating plasmid pBCB1-GE P_{divIV A} divIVA-gfp. The orientation and sequence of the cloned insert was verified before transformation into RN4220.

pBCB1-GE 3’dnaK was used to construct strain SA220 (RN4220 dnaK::dnaK-gfp, pGL485) which expresses a dnaK C-terminal GFP fusion from the native dnaK promoter. To do this, a 3′ 800 bps dnaK fragment was PCR amplified from RN4220 gDNA with primer pair 1314 and 1315 (containing KpnI restriction sites) and cloned in-frame with gfp downstream of the P_spac promoter on plasmid pBCB1-GE. The orientation and sequence of the cloned insert was verified and transferred to RN4220.

The thermosensitive plasmid pMAD (Arnaud et al., 2004) was used to make a divIVA null mutant in the laboratory strain SH1000 background (LC102; SH1000 spa::tet). Briefly, PCR fragments containing ∼1.5 kb flanking regions of divIVA were amplified from SH1000 gDNA using primers 1565/1566 and 1567/1568. The PCR products were fused together by Gibson Assembly following manufacturer’s instructions (NEB), followed by gel extraction of the DNA fragment of correct size. This DNA fragment was digested with BamHI and BglII and cloned into pMAD. The plasmid was electroporated into RN4220 at 30°C and subsequently transduced into LC102 at 30°C. Integration of the plasmid into the chromosome was achieved by inoculating BHI with single colonies and incubating at 42°C for 6 h before spreading on BHI agar in the absence of antibiotics and incubation overnight at 42°C. To then allow double cross-over events to occur, resulting in the deletion of divIVA, colonies were initially grown in BHI medium at 37°C overnight, and then subcultured at 25°C for two 24 h passages. The culture was plated in the absence of antibiotics and incubated at 37°C (Kato and Sugai, 2011). Colonies that showed erythromycin sensitivity and were not blue on X-gal were screened by PCR using primers 1563/1564 to confirm the deletion of divIVA to create strain SA365 (SH1000 spa^- ΔdivIVA). No difference in growth rate, viability, cell size or anucleate cells were observed in SA365 compared to LC102. This is in agreement with results seen for RN4220ΔdivIVA (this study; Pinho and Errington, 2004).

To screen for interactions of DnaK and DivIVA with various proteins involved in cell division or chromosome segregation, the coding sequences of dnaK, divIVA, spoIIIE, ftsK, smc and spoOJ were amplified by PCR using appropriate primer pairs (Supplementary Table 3) and cloned into pUT18 (Karimova et al., 1998) and p25-N (Claessen et al., 2008) using restriction sites BamHI and SacI (spoIIIE, ftsK, spoOJ, dnaK), BamHI and KpnI (smc) or BamHI and EcoRI (divIVA) to create C-terminal fusions of the proteins to the T18 or T25 domain of adenylate cyclase respectively. N-terminal fusions of DnaK were also created by PCR amplifying the dnaK orf using primer pairs 1285/1286 and 1289/1290 which introduces flanking BamHI-SacI and BamHI-KpnI sites, respectively, to the PCR products. Inserts were cloned into plasmid pUT18C and pKT25 (Karimova et al., 1998) using the same sites.

To test for pair-wise protein-protein interactions in the BACTH assay, plasmids were co-transformed into BTH101 and β-galactosidase activity produced from the E. coli strains was qualitatively assessed by observing the blue coloration of the cleaved X-Gal substrate as described previously (Steele et al., 2011). Plates are representative from at two independent experiments.

Staphylococcus aureus LH607 (NCTC 8325-4 spa::tet) was used for the immunoblots to prevent non-specific binding of antibodies to Protein A (Pinho and Errington, 2003). Strains were grown to mid-exponential phase and normalized to an OD₆₀₀ value of 1 before collection of cells (∼5 ml) by centrifugation and resuspension in WL buffer ([0.3 mg ml^-1 lysostaphin for S. aureus; 0.3 mg ml^-1 lysozyme for B. subtilis], 1x protease inhibitor (Roche), 25 mM Tris, 0.3 mg ml^-1 PMSF). Cell lysis was performed at 37°C for 30 min and protein concentration was determined by Bradford assay. Samples were separated on 4–12% NuPAGE Novex bis-Tris electrophoresis gels (Invitrogen) and transferred to a PVDF membrane by either 7-min semi-dry western transfer method (iBlot, Invitrogen) or standard wet transfer. Blots were probed with B. subtilis FtsZ antiserum (a kind gift from Moriya S.), anti-GFP antibodies (Roche) or B. subtilis DivIVA antiserum (Marston et al., 1998) at 1:10000, 1:1000 or 1:5000 dilutions, respectively. Immunodetection was performed using the ECL kit (GE) according to the manufacturer’s recommendations. Densitometry analysis was performed using the Quantity One software (Bio-Rad). Percentages indicate the densitometry values of the mutant when compared to its isogenic parent strain.

Protein complexes were isolated from cryogenically lysed cells following the method of Cristea et al. (2005) using 5 mg anti-GFP conjugated M-270 Epoxy Dynabeads. The eluted fraction snap frozen in liquid nitrogen and dried in a vacuum concentrator (Eppendorf) for 4.5 h. The resulting pellet was resuspended in SDS-PAGE sample buffer and separated on a 4–12% NuPAGE Novex bis-Tris electrophoresis gel (Invitrogen) prior to mass spectrometry.

Sample lanes were excised from the gel and cut into smaller gel pieces which were then reduced, alkylated and digested overnight with trypsin. Digested peptides were separated by nano-LC using an Ultimate 3000 HPLC and autosampler system (Dionex). Samples (2.5 μl) were concentrated and desalted onto a micro C18 precolumn (500 μm × 2 mm, Michrom Bioresources, Auburn, CA, United States) with H₂O:CH₃CN (98:2, 0.05% HFBA) at 20 μl/min. After a 4 min wash the pre-column was switched into line with a fritless nano column (75 μ ×∼10 cm) containing C18 media (5 μ, 200 Å Magic, Michrom) manufactured according to specifications described previously (Gatlin et al., 1998). Peptides were eluted using a linear gradient of H₂O:CH₃CN [98:2, 0.1% (v/v) formic acid] to H₂O:CH₃CN [64:36, 0.1% (v/v) formic acid] at 350 nl/min over 30 min. A high voltage of 1800 V was applied to low volume tee (Upchurch Scientific) and the column tip positioned ∼0.5 cm from the heated capillary (T = 250°C) of a LTQ FT Ultra (Thermo Electron, Bremen, Germany) mass spectrometer. Positive ions were generated by electrospray and the LTQ FT Ultra operated in data dependent-acquisition mode (DDA).

A survey scan m/z 350-1750 was acquired in the FT ICR cell (Resolution = 100,000 at m/z 400, with an initial accumulation target value of 1,000,000 ions in the linear ion trap). Up to the 6 most abundant ions (>3,000 counts) with charge states of +2, +3, or +4 were sequentially isolated and fragmented within the linear ion trap using collisionally induced dissociation with an activation q = 0.25 and activation time of 30 ms at a target value of 30,000 ions. M/z ratios selected for MS/ MS were dynamically excluded for 30 s. Peak lists were generated using Mascot Daemon/extract_msn (Matrix Science, London, England, Thermo) using the default parameters, and submitted to the database search program Mascot (version 2.1, Matrix Science). Search parameters were: Precursor tolerance 4 ppm and product ion tolerances ± 0.4 Da; Met(O) specified as variable modification, enzyme specificity was trypsin, 1 missed cleavage was possible and the non-redundant IPI_human protein database (May 2009) searched.

Proteins that are identified as “significant hits” contain at least 2 peptides with a Mascot score of >100 from 2 independent experiments (Cooper et al., 2009). Proteins that were present in both the FtsZ-GFP sample and the GFP control sample were considered to be non-specific interacting proteins and were excluded from further analysis.