A synthetic Escherichia coli system identifies a conserved origin tethering factor in Actinobacteria


E-mail marc.bramkamp@uni-koeln.de; Tel. (+49) 221 4706472; Fax (+49) 221 4705091.


In eukaryotic and prokaryotic cells the establishment and maintenance of cell polarity is essential for numerous biological processes. In some bacterial species, the chromosome origins have been identified as molecular markers of cell polarity and polar chromosome anchoring factors have been identified, for example in Caulobacter crescentus. Although speculated, polar chromosome tethering factors have not been identified for Actinobacteria, to date. Here, using a minimal synthetic Escherichia coli system, biochemical and in vivo experiments, we provide evidence that Corynebacterium glutamicum cells tether the chromosome origins at the cell poles through direct physical interactions between the ParB–parS chromosomal centromere and the apical growth determinant DivIVA. The interaction between ParB and DivIVA proteins was also shown for other members of the Actinobacteria phylum, including Mycobacterium tuberculosis and Streptomyces coelicolor.


Cell polarity in both prokaryotic and eukaryotic cells is essential for countless biological processes (Ahringer, 2003; Dworkin, 2009; Schofield et al., 2010). Proteins or protein complexes are distinctly and selectively localized to specific subcellular regions, spatially organizing the cytoplasmic environment and in turn influencing temporal regulation. In rod-shaped bacteria, the cell poles have emerged as important sites for the intracellular localization of proteins, lipids and nucleic acids, affecting numerous cellular processes including chemotaxis, pole morphogenesis, symmetric and asymmetric cell division and chromosome segregation (Edwards and Errington, 1997; Lam et al., 2006; Goley et al., 2007; Ringgaard et al., 2011).

How cell polarity is generated and subsequently maintained remains poorly understood. Polarity can be generated as a result of cell division. Cytokinesis in rod-shaped bacteria gives rise to offspring with differently organized cell poles (Lawler and Brun, 2007). Specific proteins are inherited at the site of most recent division (new pole) historically distinguishing the two poles, setting up spatial cues in the progeny cells that once again guide the cell through the cell cycle. In some bacteria, the origin of replication (oriC) is another molecular marker of cell polarity, which is firmly localized to one cell pole, the old pole immediately after division (Webb et al., 1997). Following the initiation of DNA replication the newly replicated origin is translocated to the opposite cell pole. This directed movement is often mediated by interaction between components of the ParAB system, comprising of (i) centromere-like elements, parS sites, (ii) a centromere-binding protein, ParB and (iii) a Walker-type ATPase, ParA (Fogel and Waldor, 2006; Gerdes et al., 2010; Ptacin et al., 2010). The replicated origin is subsequently attached to the opposite cell pole.

Although polar localization of the oriC is not common for all bacteria, Caulobacter crescentus and sporulating Bacillus subtilis exhibit polar attachment of the origin. However, this is not the case for vegetatively growing B. subtilis and Escherichia coli. In C. crescentus, a multimeric self-organizing protein (PopZ) was recently identified as a chromosome origin tethering factor (Bowman et al., 2008; Ebersbach et al., 2008). Polar-localized PopZ sets up a polar epicentre that not only anchors the chromosome origins, but also stabilizes bipolar gradients of the cell division inhibitor MipZ (Thanbichler and Shapiro, 2006) and mediates the localization of morphogenic and cell cycle regulating proteins (Bowman et al., 2010).

However, PopZ is not highly conserved and many bacterial species exhibit polar localization of the chromosome origins. Thus, the mechanism of polar origin anchoring must differ in unrelated bacteria. We set out to identify the origin tethering factor in Actinobacteria. Antibiotic producing Streptomyces species, amino acid producing Corynebacterium glutamicum and notorious pathogens such as Mycobacterium tuberculosis, Mycobacterium leprae and Corynebacterium diphtheriae are members of this family, and hence understanding the fundamental cell biological mechanisms are of medical and industrial importance. Unlike the conventional model organisms, many Actinobacteria lack an MreB homologue and all grow apically, inserting new cell wall material at the cell poles. This polar mode of growth is co-ordinated by the DivIVA protein (Flärdh, 2003; Nguyen et al., 2007; Hempel et al., 2008; Kang et al., 2008; Letek et al., 2008; 2009).

Using C. glutamicum as a representative of Actinobacteria, we previously showed that the origins are localized at the cell poles (Donovan et al., 2010). Collectively, the components of the ParAB system relocalize the replicated origin to the opposite cell pole. Deletion mutants of parA or parB have severe chromosome segregation defects (Donovan et al., 2010). However, disruption of parB additionally results in impaired polar cell growth. We speculated that the ParB–oriC complex interacts with a polar-localized protein that is directly or indirectly involved in cell wall growth. One potential candidate was DivIVA. Furthermore, in an unrelated study it was observed that Enterococcus faecalis DivIVA is involved in maintaining proper cell division and chromosome segregation (Ramirez-Arcos et al., 2005). As alterations in DivIVA expression severely influence C. glutamicum cell morphology (Letek et al., 2008), we analysed potential DivIVA–ParB interactions in a synthetic E. coli system. This system is advantageous as E. coli does not encode DivIVA or Par protein homologues, allowing direct in vivo interaction analysis of DivIVA and ParB, independent of other components of the Par system. We observed that ParB, and ParB in complex with parS DNA, is recruited by and interacts directly with DivIVA. Moreover, the ParB–DivIVA interaction could also be shown M. tuberculosis and Streptomyces coelicolor, two model organisms of the Actinobacteria phylum.


DivIVA physically interacts with ParB in vivo

We constructed a C. glutamicum strain in which DivIVA–mCHERRY is chromosomally expressed as a single copy from its native promoter (Fig. 1C). Growth rates and morphology of the strain was identical to wild type, suggesting that the DivIVA–mCHERRY construct is fully functional. Upon analysis of this strain it became apparent that DivIVA forms polar-localized polymers that are partly intruding into the cytoplasm, to some extent resembling the localization of the functionally analogous PopZ protein in C. crescentus.

Figure 1.

C. glutamicum ParB interacts with and is recruited by DivIVA. A. C. glutamicum ParB–CFP localizes as patches over the chromosome (first panel), while DivIVA–GFP localizes to the cell poles and division septa (second panel) when expressed in E. coli. Coexpressed in E. coli, ParB–CFP is recruited by DivIVA–GFP to the cell poles and septa and thus, ParB and DivIVA physically interact (third panel). DivIVA recruitment is also observed if ParB is in complex with parS DNA (lower panel). Expression was induced with 0.1 mM of IPTG. B. Co-elution assays show that ParB and DivIVA interact in vitro. Purified proteins used as bait are indicated on the right. Bait proteins were detected by immunoblotting using α-Penta-His antibodies. Interacting prey proteins [ParB–CFP expressed in C. glutamicum (top) or DivIVA–GFP expressed in E. coli (bottom)], which co-eluted with the bait protein, were detected by immunoblotting using α-GFP antibodies. As a control, prey proteins were incubated with Ni-NTA resin in the absence of bait proteins (-bait). Shown is a montage of immunoblotted proteins. C. Chromosomal expression of DivIVA–mCHERRY from the native promoter in C. glutamicum. D. Fluorescent profile intensities demonstrate that DivIVA localization is similar in C. glutamicum and E. coli. DivIVA exhibits membrane-associated cytoplasmic intruding localization in both C. glutamicum and E. coli. The fluorescence profile of chromosomally expressed DivIVA–mCHERRY in C. glutamicum is shown in red and DivIVA–GFP expressed in E. coli is shown in green. The black line shows the light intensity of the phase contrast. The profiles represent the mean of 10 individual cells. E. Coexpression of ParBR21A–CFP and DivIVA–mCHERRY, in E. coli. Mutation of a conserved arginine (R21) in ParB abolishes interaction with DivIVA. Scale bar 2 µm.

In a previous study, we observed that a parB mutant exhibits severe chromosome segregation defects and also impaired apical growth (Donovan et al., 2010), similar to a DivIVA knock-down mutant (Letek et al., 2008). As the ParB–parS nucleoprotein complex is localized at the cell poles in C. glutamicum (Donovan et al., 2010), we hypothesized that DivIVA anchors the chromosome at this subcellular position through interaction with the ParB–parS nucleoprotein complex. Indeed, a bacterial two-hybrid interaction analysis demonstrated that ParB and DivIVA might interact (Fig. 5).

Since DivIVA is essential in C. glutamicum and alterations in the expression level alter cell morphology, we decided to further test in vivo interactions between ParB and DivIVA by reconstitution in E. coli cells. The heterologous E. coli system is particularly advantageous as E. coli does not encode homologues of divIVA and the parAB chromosome partitioning system. We constructed E. coli strains in which DivIVA–GFP and ParB–CFP are expressed, either individually or in combination. Unless otherwise stated, expression in E. coli was induced with 0.1 mM IPTG, for 1 h at 37°C. Similar to other DivIVA homologues, DivIVA–GFP localized to the curved membranes at the cell poles and sites of septation, in E. coli cells (Fig. 1A). Fluorescence intensity profile measurements demonstrated that DivIVA localization is similar in C. glutamicum and E. coli (Fig. 1D). In both organisms, DivIVA associates to the pole membrane, partially intruding into the cytoplasm, and is also targeted to the division site. Thus, in our heterologous system DivIVA behaves similar as in the native host. Therefore, polar localization of DivIVA in E. coli is not the result of insoluble material accumulating at the cell poles.

We tested interaction between ParB–CFP and DivIVA–GFP in the presence and absence of plasmid containing parS DNA. When DivIVA–GFP and ParB–CFP were coexpressed in E. coli, ParB–CFP was recruited to the cell poles and septa by interaction with DivIVA (Fig. 1A). To exclude unspecific interactions, ParB–CFP was coexpressed with DivIVA lacking a fluorescent tag (for detection purposes DivIVA contains a C-terminal S-tag). Similarly, ParB–CFP was recruited by the native DivIVA (data not shown). Also, in the presence of parS DNA ParB–CFP was recruited by DivIVA to the cell poles and division septa; however, the localization of ParB–CFP sometimes resembled an assembly of ParB–CFP foci (Fig. 1A). Thus, the ParB–DivIVA interaction also occurs in complex with parS DNA. Together, these results demonstrate that ParB and DivIVA physically interact under in vivo conditions, providing evidence that DivIVA anchors the ParB-bound origins at the cell poles in C. glutamicum.

DivIVA interacts with ParB in vitro

The direct and specific interaction between ParB and DivIVA was tested in vitro by co-elution experiments. Therefore, purified His-tagged DivIVA (bait protein) was bound to Ni-NTA resin. After removal of unbound protein, bound His-DivIVA was incubated with the cleared cell lysate of C. glutamicum ParB–CFP expressing cells (prey protein). After extensive washing, interacting protein complexes were eluted and subsequently analysed by immunoblot. DivIVA and ParB were detected with anti-Penta-His and anti-GFP antibodies respectively. Similar to the in vivo situation, ParB specifically co-eluted with DivIVA (Fig. 1B). The reciprocal experiment, in which purified His-tagged ParB was bound to Ni-NTA resin and probed with DivIVA–GFP, reproduced this interaction (Fig. 1B). To exclude unspecific interactions, prey proteins (ParB–CFP and DivIVA–GFP) were incubated with the Ni-NTA resin in the absence of bait protein [Fig. 1B (-Bait)]. Here, we provide additional evidence for the direct physical in vitro interaction of C. glutamicum DivIVA and ParB, and subsequent implications in tethering the origins to the cell poles.

The N-terminal domain of ParB interacts with DivIVA

In several organisms, ParB domains have been attributed specific functions. The N-terminal domain of C. crescentus ParB, as well as ParB of plasmid P1 and SopB of F plasmid interact with ParA/SopA (Radnedge et al., 1998; Figge et al., 2003; Ravin et al., 2003), while the C-terminal domain of C. crescentus and Thermus thermophilus ParB is required for dimerization and subsequent DNA binding (Figge et al., 2003; Leonard et al., 2004). In B. subtilis, the N-terminal domain of Spo0J (ParB) stimulates the ATPase activity of Soj (ParA), driving Soj to an ADP-bound monomer which in turn inhibits DNA replication initiation (Scholefield et al., 2011).

In C. glutamicum, ParB domains have not been attributed to specific function, yet. We took advantage of the concentration-dependent localization of ParB to the E. coli nucleoid and carried out mutational analysis to gain additional insight into the function of the ParB domains and interaction with DivIVA. It has been shown that the C-terminal domain of ParB is the initial dimerization domain, and dimerization is a prerequisite for DNA binding (Figge et al., 2003; Leonard et al., 2004; Murray et al., 2006). Truncation of the C-terminal domain of C. glutamicum ParB (ParBΔc50) exhibited a diffuse cytoplasmic localization when expressed in E. coli (Fig. S1A), indicating that binding of this ParB mutant to the nucleoid is weakened, if not completely abrogated. As this ParB variant still contained the putative central DNA binding domain (HTH domain), it appears that deletion of the C-terminal abroginates dimerization rather than directly inhibiting binding to DNA. Thus, the dimerization potential of the ParB mutant was assayed using size exclusion chromatography. Purified full-length ParB (theoretical molecular weight of 40.8 kDa) eluted as a dimer, while the ParBΔc50 mutant protein eluted as a monomer (data not shown). Thus, it appears likely that C-terminal domain of ParB is necessary for dimerization and subsequent DNA binding. When coexpressed with DivIVA in E. coli, recruitment of ParBΔc50 to the cell poles was not altered (Fig. S1A). Therefore, we turned our attention to the N-terminal domain of ParB. Thus, a mutant ParB protein, lacking 100 amino acids of the N-terminal region of ParB (ParBΔN100–CFP), was expressed in combination with DivIVA, which lacked a fluorescent tag. Interaction of the ParB variant deleted for the N-terminal 100 amino acids and full-length DivIVA was completely abolished (Fig. S1B).

In a previous study, sequence alignments of putative ParB (Spo0J) proteins revealed a high degree of conservation of the N-terminal regions (Leonard et al., 2004). Indeed, alignment of C. glutamicum ParB and B. subtilis Spo0J (ParB) sequences revealed that the N-terminal domains of both sequences contain a conserved stretch of aliphatic residues (Fig. S2). We hypothesized that one or all of these residues might be required for interaction with DivIVA. Hence, the N-terminal domain of ParB was truncated, deleting the first 21 amino acids (ParBΔN21–CFP). This mutant ParB protein lacks the highly conserved arginine (R21) within the stretch of aliphatic residues, which in other organisms is reported to be imperative for ParA interaction (Leonard et al., 2005). Despite localizing identical to full-length ParB in E. coli, DivIVA-dependent recruitment of ParBΔN21–CFP was completely abolished (Fig. S1C). It was reasoned that the conserved arginine is important for interaction with DivIVA. Therefore, a ParB mutant protein still possessing the conserved arginine but lacking the preceding 20 amino acids (ParBΔN20–CFP) was coexpressed with DivIVA–mCHERRY. Interestingly, ParBΔN20–CFP still interacted with DivIVA, showing that the N-terminal arginine does indeed strongly influence interaction with DivIVA (Fig. S1C). Keeping with this line of investigation, the conserved arginine was mutated to an alanine, in the otherwise wild-type protein. Again, this mutant ParB protein (ParBR21A–CFP) localized as patches over the E. coli nucleoid; however, interaction with DivIVA was completely eliminated (Fig. 1E). Thus, the N-terminal motif of ParB (Fig. S2) is essential for interaction with DivIVA.

In C. glutamicum, mutation of a conserved arginine (R21) of ParB significantly alters interaction with DivIVA

Using the synthetic E. coli system, the conserved arginine (R21) of the N-terminal motif (Fig. S2) of ParB was identified as an important residue for interaction with DivIVA (Fig. 1E). To further support this finding, we constructed two C. glutamicum strains that express ParB–eCFP and ParBR21A–eCFP in a ΔparB DivIVA–mCHERRY mutant background. Expression of ParB–eCFP and ParBR21A–eCFP was induced with 0.5 mM IPTG for 90 min before microscopic analysis. The subcellular localization of ParB–eCFP and ParBR21A–eCFP foci were analysed statistically. The foci were categorized as either (i) unipolar or bipolar, (ii) bipolar with one or two foci at midcell or cell quarter positions, (iii) unipolar with two foci at midcell or cell quarter positions and (iv) foci localized at midcell or cell quarter cell positions in the absence of polar-localized foci. In cells expressing ParB–eCFP, ParB foci were always anchored to the cell poles, either uniquely as polar foci or as polar and midcell/cell quarter foci. In contrast, unipolar or bipolar localization of ParBR21A–eCFP foci was greatly reduced (20%) compared to ParB–eCFP expressing cells (60%). Although ParBR21A–eCFP foci were sometimes localized at the cell poles, a significant number of cells (20%) completely lacked polar-localized foci (Fig. 2A and C and Fig. S3).

Figure 2.

Mutation of the invariant arginine (R21) of ParB alters polar localization of ParB foci in C. glutamicum. A. Localization of ParB–eCFP and ParBR21A–eCFP expressed from a plasmid in C. glutamicum cells in a ΔparB DivIVA–mCHERRY background. Cells were induced with 0.5 mM IPTG for 90 min prior to microscopic analysis. ParB–eCFP foci are localized to the cell poles (upper panel, arrows). In cells expressing ParBR21A–eCFP, foci are often observed that are not anchored to the cell poles (lower panel, arrow). B. Quantification of anucleate cell production in absence and presence of ParB–eCFP and ParBR21A–eCFP. Expression of ParB–eCFP almost completely complements the anucleate cell phenotype of a ΔparB mutant (1.69% anucleate cells). However, expression of ParBR21A–eCFP only partially complemented the ΔparB mutant phenotype, producing 10.45% anucleate cells. C. Statistical analysis of ParB–eCFP and ParBR21A–eCFP localization. ParB–eCFP localizes predominantly at the cell poles, exhibiting either exclusive polar localization or polar and midcell/cell quarter localization (black). However, in 20% of cells expressing ParBR21A–eCFP, the poles are completely devoid of ParB foci (grey). Scale bar 2 µm.

Expression of ParB–eCFP almost fully complemented the ΔparB mutant phenotype. In the absence of inducer, 32.8% anucleate cells were produced, which was reduced to 1.7% upon expression of ParB–eCFP (Fig. 2B). However, C. glutamicum cells expressing ParBR21A–eCFP only partially complemented the parB mutant phenotype, producing 10.4% anucleate cells (Fig. 2B). Thus, mutation of the invariant arginine (R21) of ParB significantly reduces the affinity of ParB for DivIVA and impairs polar anchoring of the chromosome origin.

A central region of DivIVA is necessary for ParB interaction

With the exception of the N-terminal domain, the amino acid sequences of functionally different DivIVA homologues display significant variation (Fig. S4). The Actinobacteria DivIVA is larger, containing a central insertion that is absent in the Firmicutes DivIVA (Lenarcic et al., 2009; Oliva et al., 2010). In sporulating B. subtilis, attachment of the origin to the prespore pole is mediated by centromere-bound RacA and polar DivIVA (Ben-Yehuda et al., 2003; 2005; Wu and Errington, 2003). Although a soj (parA) spo0J (parB) racA triple mutant exhibited a higher level of chromosome misorientation than the single racA mutant, no direct interaction between membrane-associated DivIVA and Spo0J has been shown, to date. We used the synthetic E. coli system to test the potential interaction between B. subtilis DivIVA (DivIVAbsu) and Spo0J (Fig. S5). Expressed alone, Spo0J–CFP localized as patches over the E. coli nucleoid. This pattern was not altered when expressed in combination with DivIVAbsu, showing that these proteins presumably do not interact physically (Fig. S5). This result highlights the reliability of the synthetic E. coli system for assaying protein–protein interactions.

We speculated that the central region of C. glutamicum DivIVA would be important for interaction with ParB. To test this idea, we coexpressed C. glutamicum ParB–CFP with the DivIVAbsu in E. coli. Testifying that our postulation was correct, C. glutamicum ParB was not recruited to the cell poles and septa by DivIVAbsu (Fig. S5). Consequently, part of the central region of C. glutamicum DivIVA (amino acids 144–298) was deleted. This mutant DivIVA protein (DivIVAΔ1–GFP) still localized to the cell poles and division septum when expressed in E. coli cells; however, when coexpressed with ParB–CFP, ParB remained bound to the nucleoid (Fig. 3B). To further support this finding, that ParB interacts with the central region of DivIVA, potential interaction between the central region of DivIVA (DivIVA144–298) and full-length ParB was assayed in a LexA bacterial two-hybrid. The bacterial two-hybrid experiment indicates that DivIVA144–298 and ParB can interact; however, in comparison to interaction analysis of full-length ParB and DivIVA, the interaction is slightly reduced (Fig. 5). However, these results indicate that the ParB interaction site lies within this central region of DivIVA.

Figure 3.

ParB interacts with a central region of DivIVA. A. Illustration of the domain architecture of DivIVA and the different DivIVA truncation mutants where central regions were deleted (deleted regions are shown in white together with deleted amino acids). A summary of the interaction potential of the various DivIVA mutants with ParB or ParBΔC50 is indicated on the right. B. Examples of the subcellular localization and ParB interaction potential of the various DivIVA truncation mutants when expressed in E. coli. DivIVAΔ1–GFP, and all other DivIVA truncation mutants, localized to the cell poles and septa, identical to full-length DivIVA (first panel). DivIVAΔ1–mCHERRY does not recruit ParB–CFP to the cell poles (second panel). DivIVAΔ3–mCHERRY shows partial recruitment of ParB–CFP (third panel, arrow heads). ParB remains mostly associated with the chromosome (arrow). A ParB mutant, defective in DNA binding (ParBΔC50–CFP) and free to diffuse in the cell, is not recruited by DivIVAΔ1 (fourth panel) but is recruited by DivIVAΔ3 to the cell poles (lower panel). Scale bar 2 µm.

To more accurately reveal the region of DivIVA that is necessary for ParB interaction, a series of shorter DivIVA truncation mutants were constructed. The various DivIVA truncation mutants are shown in Fig. 3A. All DivIVA truncation mutant proteins localized to the poles and division septa, identical to the full-length protein (Fig. 3B, one example shown), suggesting that these DivIVA variants are functional in respect to membrane binding and oligomerization in E. coli. These DivIVA truncation mutants were coexpressed with ParB–CFP and subsequently analysed for interaction. Similar to DivIVAΔ1–mCHERRY, deletion of amino acids 144–229 of DivIVA (DivIVAΔ2–mCHERRY) inhibited ParB recruitment and interaction (data not shown). As DivIVAΔ1 and DivIVAΔ2 mutants lack the C-terminal end of coiled-coil 1 and the N-terminal end of coiled-coil 2, we speculated that both coiled-coil domains are necessary for ParB recruitment. Coexpression of DivIVA variants lacking the N-terminal end of coiled-coil 2 but still containing the entire coiled-coil 1 domain (DivIVAΔ3–mCHERRY, DivIVAΔ4–mCHERRY) exhibited weak recruitment of ParB–CFP (Fig. 3B). Finally, a DivIVA mutant lacking amino acid 160–200 (DivIVAΔ5–mCHERRY) was expressed with ParB–CFP and exhibited again partial recruitment of ParB–CFP. Hence, both coiled-coil domains of DivIVA are necessary for ParB interaction.

In the synthetic E. coli system, ParB interacts with DivIVA via a diffusion-capture mechanism

In the synthetic E. coli system, ParB exhibits affinity for the chromosome and, when coexpressed, a stronger affinity for DivIVA. The affinity of the ParB–DivIVA interaction can be eliminated or reduced by mutating the N-terminal domain of ParB or truncating the central region of DivIVA. On the other hand, truncation of the ParB C-terminal domain (ParBΔC50–CFP) results in a diffuse cytoplasmic localization.

To unambiguously determine if ParB can interact with the diverse DivIVA truncation mutants, we coexpressed the various DivIVA mutants with ParBΔC50–CFP, which is free to diffuse in the cytoplasm. If the C-terminal domain of ParB is indeed involved in DNA binding, it would be expected that ParBΔC50–CFP would interact with the DivIVA mutants where full-length ParB showed only partial interaction, also supporting the idea of a diffusion-capture mechanism, in E. coli cells. Coexpression of ParBΔC50–CFP with DivIVAΔ1 did not alter the diffuse cytoplasmic localization of ParB (Fig. 3B). However, when expressed in combination with DivIVAΔ3, DivIVAΔ4 or DivIVAΔ5, ParBΔC50–CFP was again recruited to the cell poles via interaction with the DivIVA mutants (Fig. 3B, one example shown). These results support the idea that the central region of DivIVA is required for interaction with ParB. Also, the C-terminal domain of ParB is required for DNA binding. Consequently, in the absence of the other components of the Par system, ParB that is free to diffuse in the cytoplasm interacts with DivIVA via a diffusion-capture mechanism, in the synthetic E. coli system. The fact that no primary active segregation protein, such as ParA needs to be involved in tethering of ParB to DivIVA in synthetic systems, suggests that these two proteins might be used as minimal and sufficient components required for replicon maintenance in synthetic cells.

The interaction between ParB and DivIVA is conserved in Actinobacteria

The conserved role of DivIVA in apical growth tempted us to determine if anchoring the origin at the cell pole is additionally a common feature of DivIVA from Actinobacteria. Therefore, we selected two medically and industrially important members, M. tuberculosis and S. coelicolor, for further analysis.

Similar to C. glutamicum, ParB localizes to the cell poles in M. tuberculosis (Jakimowicz et al., 2007a; Maloney et al., 2009). We expressed fluorescent fusion constructs of the M. tuberculosis DivIVA homologue, Wag31, and ParB in the synthetic E. coli system. Similar to the C. glutamicum DivIVA, Wag31–mCHERRY localizes to the cell poles and division septa, while M. tuberculosis ParB (ParBmtu–CFP) formed patches over the E. coli nucleoid (Fig. 4A). Expression of ParBmtu–CFP in combination with Wag31, without a fluorescent tag, again resulted in Wag31-dependent recruitment of ParBmtu–CFP to the cell poles and division septa. The interaction between ParBmtu and Wag31 could also be shown in a LexA based bacterial two-hybrid (Fig. 5). These results indicate a direct physical interaction between ParBmtu and Wag31 and, thus, anchoring of the chromosome origins to the cell poles (Fig. 4A).

Figure 4.

DivIVA-mediated recruitment of ParB is conserved in Actinobacteria. A. Heterologous expression of M. tuberculosis Wag31 and ParBmtu–CFP. In E. coli, Wag31–mCHERRY localizes to the cell poles and septum (upper panel, arrowheads), while ParBmtu–CFP forms patches over the chromosome (middle panel). Expressed in combination, Wag31 recruits ParBmtu–CFP to the cell poles and septa (lower panel, arrowheads). B. Expression of S. coelicolor DivIVA and ParB in the synthetic E. coli system. DivIVAsco–mCHERRY localizes at the cell poles and at the division septum. DivIVAsco–mCHERRY localization often extends laterally along the membrane (upper panel, arrow). The parBsco gene contains a native internal parS site, and thus ParBsco–CFP forms foci localized mostly near the cell poles (middle panel, arrowhead). Coexpressed, ParBsco–CFP is mostly recruited by DivIVAsco–mCHERRY (lower panel). Scale bar 2 µm.

Figure 5.

Analysis of the interaction between ParB and DivIVA by a LexA-based two-hybrid system. A LexA-based bacterial two-hybrid (Dmitrova et al., 1998) approach was performed to study protein–protein interactions between ParB and DivIVA from Actinobacteria. The experiments were performed in E. coli SU202 (PsulAlacZ, op408/op+). The proteins of interest were fused to either wild-type LexA DNA binding domain (pMS604) or a mutant LexA DNA binding domain (pDP804). Heterodimerization of the proteins of interest bring the wild-type and mutant portions of the LexA DNA binding domains together, repressing transcription of a LexA-regulated PsulA–lacZ fusion, which was measured by β-galactosidase assays. The leucine zipper domains of Fos and Jun fused to the WT and mutant LexA DNA binding domain respectively, serve as positive control (+). Unrepressed levels of β-galactosidase activity was determined in the presence of empty vector (pMS604/pDP804) and the LexA DNA binding domain fusion to the test protein, in both combinations [negative control (−)]. The % repression of PsulA–lacZ of interacting proteins was determined against the negative control. Shown are interaction between C. glutamicum ParB and DivIVA (Cgl) (pMS–ParB/pDP–DivIVA, negative control – 1761 Miller units), C. glutamicum DivIVA144–298 and ParB (Clg*) (pMS–DivIVA144–298/pDP–ParB, negative control – 2011 Miller units), S. coelicolor ParB and DivIVA (Sco) (pMS–DivIVA/pDP–ParB, negative control – 2537 Miller units) and M. tuberculosis ParB and Wag31 (Mtu) (pMS–ParB/pDP–Wag31, negative control – 1977 Miller units). Standard deviations are derived from three independent experiments.

The complex life cycle of Streptomyces species differs drastically from C. glutamicum and M. tuberculosis. In vegetative mycelia, ParBsco–eGFP forms irregular foci that are often closely associated with hyphal tips (Jakimowicz et al., 2005). The foci localized close to the hyphal tips have been suggested by the authors to be anchored at these positions, while the non-apical ParBsco foci have been implicated in regulating replication initiation (Jakimowicz et al., 2005). During the reproductive phase, chromosome condensation, segregation and multiple septation lead to the development of chains of unigenomic spore compartments. At the onset of sporulation, the expression of ParABsco is upregulated. ParAsco, which accumulates at the hyphal tips, spreading along the hyphae as a pair of helical filaments, appears to be necessary for the proper assembly of segregation complexes (Jakimowicz et al., 2007b). ParBsco assembles into regularly spaced foci, usually situated in the middle of the prespore (Jakimowicz et al., 2005); however, some foci can also be seen close to the septum. The requirement of ParBsco (and also ParAsco) in chromosome organization is reflected in the increase of anucleate spores in a parBsco mutant (Kim et al., 1999; Jakimowicz et al., 2005; 2007b). If ParBsco and DivIVAsco might interact, this could support chromosome organization in Streptomyces.

The S. coelicolor parB gene contains an internal parS site, thus when expressed in E. coli, ParBsco was often seen as foci localized near the cell pole, but some ParBsco was also diffused in the cytoplasm (Fig. 4B). DivIVAsco–mCHERRY initially localized as foci at the cell poles, then over time lined the inner membrane of the curved poles and also localized along the lateral axis, beginning around the midcell region then extending laterally. On occasion, some lateral DivIVAsco–mCHERRY foci were also seen. A Z-stack series and subsequent 3-D construction demonstrated that S. coelicolor DivIVA forms a ring-like structure at this midcell position, indicating that DivIVAsco binds to the curved membrane at the invaginating septum as expected (Fig. 4B and Fig. S6). In general, the behaviour of DivIVAsco is different to the C. glutamicum DivIVA when expressed in E. coli, highlighting intrinsic properties of DivIVAsco in regards to subcellular assembly and organization. In light of the complex life cycle of S. coelicolor, multiple additional roles of DivIVAsco might also be taken into account. Coexpression of DivIVAsco–mCHERRY and ParBsco–CFP in E. coli did not reveal complete recruitment of ParBsco to the cell poles and division septa; however, interaction with DivIVAsco was often seen (Fig. 4B). Interestingly, ParBsco does not contain the conserved arginine seen in ParBcgl and ParBmtu; however, the corresponding stretch of amino acids is aliphatic (Fig. S2), and thus the entire motif might be necessary for interaction with DivIVA. As ParBsco can bind the parS site in the native parBsco coding sequence, reduced cytoplasmic diffusion of ParBsco may perturb interaction with DivIVAsco. Therefore, interaction between ParBsco and DivIVAsco was also assayed in a LexA bacterial two-hybrid (Fig. 5). Together, in both M. tuberculosis and S. coelicolor, ParB interacts with and is recruited by DivIVA, providing evidence for a possible role in chromosome organization and/or polar anchoring of the origins.


Here, we present in vivo and in vitro evidence that the apical growth determinant DivIVA anchors the chromosomal origins at the cell poles through direct physical interaction with ParB, in C. glutamicum. A minimal synthetic E. coli system was employed to demonstrate ParB–DivIVA interactions in three members of the Actinobacteria phylum and, subsequently, the C. glutamicum ParB–DivIVA interaction sites were mapped. In this study, the modular nature of C. glutamicum DivIVA is highlighted, emphasizing the importance of DivIVA in organizing the subcellular environment. Other DivIVA homologues, although also involved in anchoring chromosome origins during development processes, require mediator proteins, such as RacA in B. subtilis (Ben-Yehuda et al., 2003).

Although the ParB–DivIVA interaction in E. coli probably relies on a diffusion capture mechanism, the situation in C. glutamicum cells is more complex. In vivo, ParB and DivIVA alone are not sufficient for faithful localization of the origins. Once replicated the origins must be segregated. The requirement of ParA in the segregation of replicated origins is highlighted by the altered localization of ParB foci and increased number of anucleate cells in a parA deletion mutant (Donovan et al., 2010). Using the synthetic E. coli system, the highly conserved N-terminal motif of C. glutamicum ParB was identified as an important motif for interaction with DivIVA. In particular, mutation of arginine R21of ParB to an alanine altered the subcellular localization of ParB foci in C. glutamicum cells. While wild-type ParB foci are mostly localized at the cell poles, 20% of the cells expressing the mutant ParB protein (ParBR21A) completely lacked polar foci. The consequence of this aberrant localization is reflected in the high frequency of anucleate cells (10.45%). Thus, it appears that ParA is necessary for moving the origin to the opposite cell pole, while, once at the cell pole, ParB interacts with DivIVA tethering the origins at this subcelluar position. However, the discrepancy between the ParBR21A–DivIVA interaction in the synthetic E. coli system and the in vivo situation in C. glutamicum cells hints that more complex systems contribute to the polar localization and tethering of the origins.

We also tested the interaction between ParB and DivIVA from M. tuberculosis and S. coelicolor in the synthetic E. coli system. In M. tuberculosis ParB is localized at the cell poles (Jakimowicz et al., 2007a; Maloney et al., 2009). Thus, similar to the C. glutamicum situation, the ParB-bound origins might be tethered at the cell poles through interaction with Wag31 in M. tuberculosis. In comparison to C. glutamicum and M. tuberculosis, the interaction of S. coelicolor ParB and DivIVA appeared to be weaker. When assayed in a LexA-based bacterial two-hybrid the strength of interaction was comparable to C. glutamicum and M. tuberculosis. Also, the conserved arginine of the N-terminal of ParB is not present in the S. coelicolor ParB. Nevertheless, the corresponding stretch of amino acids is aliphatic. Similar to the situation in C. glutamicum, the entire motif may be necessary for interaction with DivIVA. In S. coelicolor DivIVA and ParB do not directly colocalize in vegetative mycelium (Kim et al., 1999; Jakimowicz et al., 2005; 2007b). Also deletion of parB, and even more pronounced deletion of parA, leads to a dramatic increase in anucleate spores (Jakimowicz et al., 2005). Thus, a meaningful interpretation of the interaction observed in the synthetic E. coli system could be that the ParB–DivIVA interaction plays a role during spore development. During sporulation ParB foci are mostly localized in the middle of the prespore; however, some foci are localized to the inward growing septum. Although the results presented here hint at a possible interaction between ParBsco and DivIVAsoc, a more detailed analysis in S. coelicolor cells must be carried out.

Collectively, all the components of the Par system function to aid efficient chromosome segregation. In Mycobacterium and Streptomyces, ParA promotes binding of ParB to parS sites (Jakimowicz et al., 2007a,b). It has been postulated that ParA provides the driving force that pushes or pulls the replicated origins towards the cell poles. This has been nicely shown for C. crescentus and Vibrio cholerae where ParA depolymerization, stimulated by ParB, moves the origin to the opposite cell pole (Fogel and Waldor, 2006; Ptacin et al., 2010). The importance of ParA in chromosome segregation can not be underestimated and a contribution of ParA in origin tethering via interaction with DivIVA can not be ruled out.

From the results presented here we propose the following model for polar anchoring of the origins in C. glutamicum cells (Fig. 6). In C. glutamicum and likely other Actinobacteria, the chromosome origins are orientated towards the cell pole prior to replication initiation (Fig. 6I). After initiation of replication, the new origin is segregated from the old origin and actively moved towards the opposite cell pole, most likely mediated by ParA (Fig. 6II). At the opposite cell pole, the replicated origin is again anchored in position through interactions with polar DivIVA (Fig. 6III). The remainder of the chromosome is segregated and condensed resulting in a DNA free zone. Polymerization of the cell division protein, FtsZ, is initiated, forming the contractile Z ring, on which the cytokinesis apparatus assembles (Fig. 6IV). The DivIVA protein is recruited late to the site of cytokinesis, where after completion of septation, DivIVA remains attached to the newly formed cell poles (Fig. 6V and VI). Interaction of ParB at the growing septum might even be more complex, since a potential interaction between ParB and FtsZ in C. glutamicum has been demonstrated earlier (Donovan et al., 2010).

Figure 6.

Proposed model for origin anchoring during the cell cycle in C. glutamicum. (I) Prior to replication initiation, the chromosome origin is localized close to one cell pole. (II) After initiation of replication, the polar ParB-origin nucleoprotein complex is segregated towards the opposite cell pole. Active mechanisms help segregate the remaining of the chromosome. (III) When the replication origin reaches the opposite cell pole, ParB interacts with polar-localized DivIVA, tethering the origin at the cell pole. We speculate that the anchoring of the origin at the opposite cell pole influences growth of this cell pole, which in turn could aid the segregation process. (IV) Once the chromosomes have been segregated, FtsZ can polymerize in the internucleoid space and mature to a functional Z-ring. (V) C. glutamicum is capable of fast growth and multiple rounds of replication can be initiated before completion of cell division. In such cases, the newly replicated origins are segregated towards the division site where they eventually stop. DivIVA is recruited late to the division septum. Thus, the ParB–parS nucleoprotein complexes are held in position, close to the division septum by interaction with DivIVA. (VI) The division septum matures and constructs, and DivIVA is divided between the two new cell poles, tethering the newly replicated origins at the young cell poles.

Corynebacterium glutamicum cells are capable of rapid growth to high cell densities, initiating multiple rounds of replication prior to cell division completion. Although ParB–parS nucleoprotein complexes are mostly statically localized at the cell poles, additional ParB–parS complexes are often seen near the midcell region (Donovan et al., 2010), where DivIVA binds the inward growing membrane. DivIVA is recruited to the division septum before cytokinesis is completed, where it forms a ring structure. The midcell localized ParB-bound origins are held in position through interaction with DivIVA. As the septum matures and constricts forming the new cell pole, the origins are again tethered at the correct subcellular localization (Fig. 6V and VI).

The loss of polar growth in both a parB mutant and a DivIVA knock-down mutant suggest that ParB–DivIVA interactions have implications on polar growth. Unlike organisms that grow by insertion of new material along the lateral axis, apical growth in C. glutamicum is not always equal at both cell poles. It will be interesting to see if anchoring the origin regions at the cell poles via interaction with DivIVA actively influences growth at a specific pole, thus further assisting segregation of the chromosomes. If this proves true, the DivIVA–ParB binding interaction surface would be an interesting antimicrobial target.

Experimental procedures

Strains, plasmids and oligonucleotides

All plasmids, oligonucleotides and strains used in this study are listed in supplementary material in Tables S1, S2 and S3 respectively. Growth media and conditions are detailed in the supplemental material.

Heterologous overexpression and protein purification

DivIVA purification.  His-DivIVA was heterologously overexpressed in BL21 (DE3). Expression was induced by addition of 0.1 mM IPTG and cells were grown overnight in TB (Terrific Broth) at 18°C. Cells were collected by centrifugation at 5000 g (4°C) for 10 min, resuspended in buffer A [50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 5 mM MgCl2, 10% glycerol] supplemented with DNase I, protease inhibitor (Complete Roche) and 1% Triton X-100 and disrupted by using an Emulsiflex C4 (Avestin). Cell debris was removed by centrifugation at 31 000 g (4°C) for 45 min. The supernatant was applied to a 1 ml His Trap FF column (GE Healthcare), washed with 20 ml of buffer A and the bound protein was eluted with buffer A supplemented with 250 mM imidazole. Fractions were analysed by SDS-PAGE and relevant fractions were pooled.

ParB purification.  His-ParB was heterologously overexpressed in BL21 (DE3). Expression was induced by addition of 0.5 mM IPTG and cells were grown in LB medium at 30°C for 3 h. Cells were collected by centrifugation at 5000 g (4°C) for 10 min, resuspended in buffer B [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 150 mM KCl, 10% glycerol] supplemented with DNase I and protease inhibitor (Complete Roche). Cells were disrupted by using an Emulsiflex C4 (Avestin). Cell debris was removed by centrifugation at 31 000 g (4°C) for 45 min. The cleared cell lysate was applied to a 1 ml His Trap FF column (GE Healthcare), washed with 20 ml of buffer B and the bound protein was eluted with buffer B supplemented with 250 mM imidazole. Fractions were analysed by SDS-PAGE and relevant fractions were pooled. The pooled fractions were applied to a Superdex 200 gel filtration column (GE Healthcare) pre-equilibrated with ParB gel filtration buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 150 mM KCl, 1 mM DTT, 5 mM EDTA, 10% glycerol]. Relevant fractions were analysed by SDS-PAGE. The His-ParBΔC50 mutant was purified in an identical manner.

Co-elution assays

Affinity purified His-DivIVA was dialysed against buffer C [20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 5 mM MgCl2] at 4°C. His-DivIVA was incubated with a Ni-NTA resin (Macherey-Nagel) for 1 h at 4°C. Unbound DivIVA was removed by washing with buffer C. The cleared lysate of strain CDC011 (ΔparB ParB+–CFP) was added. Unbound protein was removed by wash steps with buffer C. Protein complexes were eluted with buffer C supplemented with 250 mM imidazole and analysed by immunoblotting. His-DivIVA was detected with α-Penta-His antibody and ParB–CFP was detected with α-GFP antibody.

Similarly, affinity purified His-ParB was dialysed against buffer B at 4°C. His-ParB (10 µM) was applied to a Ni-NTA resin (Macherey-Nagel) and incubated at 4°C for 1 h. Unbound protein was removed by wash steps with buffer B. The cleared cell lysate of DivIVA–GFP expressing E. coli BL21 (DE3) cells was added and incubated for 1 h at 4°C. Unbound protein was removed by wash steps with buffer B. Interacting protein complexes were eluted with buffer B containing 500 mM imidazole and subsequently analysed by immunoblotting. His-ParB was detected with α-Penta-His antibody and DivIVA–GFP was detected with α-GFP antibody.

Fluorescence microscopy

Escherichia coli cells were routinely transformed fresh and prepared for microscopy by inoculating single colonies into LB medium. Cells were grown at 37°C in LB medium to OD600 0.5–1.0. Protein expression from the pETDuet-1 vector was induced with 0.1 mM IPTG for 1 h before analysis, unless otherwise stated. C. glutamicum cells were grown at 30°C in LB medium to OD600 2.0–3.0 before induction with 0.5 mM IPTG, for 90 min where appropriate. Cells were then mounted on agarose coated slides (1.5% agarose). Images were acquired on an Axio Imager.M1 fluorescence microscope (Carl Zeiss) and analysed with AxioVision version 4.6 software. Final image preparation was done in Adobe Photoshop (Adobe Systems Incorporated).

Bacterial two-hybrid and β-galactosidase assay

Protein–protein interactions were analysed by a LexA-based bacterial two-hybrid analysis (Dmitrova et al., 1998). E. coli strain SU202 was co-transformed with derivatives of pDP804 and pMS604. Heterodimerization of two heterologous proteins was determined by β-galactosidase assays. Therefore, a 5 ml LB culture supplemented with appropriate antibiotics and 1 mM IPTG were inoculated (OD600 0.05) with an overnight culture also containing 1 mM IPTG. After reaching an OD600 of 0.5–0.6, 20 µl of culture was mixed with 80 µl of permeabilization buffer [100 mM Na2HPO4, 20 mM KCl, 2 mM MgSO4, 0.8 mg ml−1 CTAB, 0.4 mg ml−1 deoxycholic acid (sodium salt), 5.4 µl ml−1β-mercaptoethanol] and incubated for 30 min at 30°C. Afterwards 600 µl of pre-warmed substrate solution (30°C) [60 mM Na2HPO4, 40 mM NaH2PO4, 1 mg ml−1 o-nitrophenyl-β-D-galactopyranoside (ONPG), 2 µl ml−1β-mercaptoethanol] was added. When the samples turned yellow, the reaction was stopped by adding 700 µl of stop solution (1 M Na2CO3). Subsequently, the samples were centrifuged and the supernatant was measured photometrically at 420 nm. Water was used as blank. Miller units were calculated with the following equation: 1000 ∗ {OD420/[OD600 ∗ volume (0.02 ml) ∗ reaction time (min)]}.

Supplemental data

Supplemental data include six figures, three tables, supplementary material and supplementary references.


We would like to thank Anja Wittmann for excellent technical assistance, Hildgund Schrempf (University of Osnabrück) for S. coelicolor genomic DNA, David Rudner (Harvard Medical School) for pDR200 and Martin Thanbichler (Philipps University, Marburg) for pCHYC-2.