A new family of bacterial condensins


E-mail valya@ou.edu; Tel. (+1) 405 325 1677; Fax (+1) 405 325 6111.


Condensins play a central role in global chromatin organization. In bacteria, two families of condensins have been identified, the MukBEF and SMC–ScpAB complexes. Only one of the two complexes is usually found in a given species, giving rise to a paradigm that a single condensin organizes bacterial chromosomes. Using sequence analysis, we identified a third family of condensins, MksBEF (MukBEF-like SMC proteins), which is broadly present in diverse bacteria. The proteins appear distantly related to MukBEF, have a similar operon organization and similar predicted secondary structures albeit with notably shorter coiled-coils. All three subunits of MksBEF exhibit significant sequence variation and can be divided into a series of overlapping subfamilies. MksBEF often coexists with the SMC–ScpAB, MukBEF and, sometimes, other MksBEFs. In Pseudomonas aeruginosa, both SMC and MksB contribute to faithful chromosome partitioning, with their inactivation leading to increased frequencies of anucleate cells. Moreover, MksBEF can complement anucleate cell formation in SMC-deficient cells. Purified PaMksB showed activities typical for condensins including ATP-modulated DNA binding and condensation. Notably, DNA binding by MksB is negatively regulated by ATP, which sets it apart from other known SMC proteins. Thus, several specialized condensins might be involved in organization of bacterial chromosomes.


SMC (structural maintenance of chromosome) proteins are highly conserved V-shaped ATPases that establish global folding of the chromosomes in organisms ranging from bacteria to humans (Swedlow and Hirano, 2003; Cobbe and Heck, 2004; Nasmyth and Haering, 2005; Graumann and Knust, 2009; Rybenkov, 2009). By bringing distant DNA segments together (Strick et al., 2004; Cui et al., 2008; Petrushenko et al., 2010), the proteins are thought to produce the giant-loop chromosome architecture (Saitoh et al., 1995; Swedlow and Hirano, 2003; She et al., 2007) and, as such, are one of the key factors that ensure faithful segregation of genetic information (Alexandrov et al., 1999; Higgins et al., 2005). SMCs act in complex with several other, non-SMC subunits which regulate their activity (Hirano and Hirano, 2004; Petrushenko et al., 2006a) and, at least in some cases, promote oligomerization of the proteins (Matoba et al., 2005; Petrushenko et al., 2006a; Woo et al., 2009). Purified condensins displayed a range of activities including the ability to bind and condense DNA, often in ATP-modulated manner (Hirano and Hirano, 2004; Strick et al., 2004; Cui et al., 2008; Petrushenko et al., 2010). The mechanism of chromatin organization by SMCs and their integration into cell division machinery remain poorly understood.

Eukaryotic cells carry several specialized SMC complexes, each responsible for its own aspect of higher-order chromatin dynamics (Swedlow and Hirano, 2003; Nasmyth and Haering, 2005). In contrast, bacteria are widely believed to encode only one SMC complex that organizes their chromosomes. Two distinct SMC complexes were found in bacteria. The first complex, MukBEF (Niki et al., 1991; Petrushenko et al., 2006a; Danilova et al., 2007), is found in enterobacteria and certain other γ-subdivision proteobacteria (Hiraga et al., 2000). All subunits of the protein are encoded in the same operon in the order mukF–mukE–mukB (Yamanaka et al., 1996). The second complex, SMC–ScpAB (Mascarenhas et al., 2002; Soppa et al., 2002), has been found in many other bacteria and archaebacteria and bears high degree of similarity to eukaryotic condensins and cohesins (Cobbe and Heck, 2004). SMC is encoded apart from its regulatory subunits ScpA and ScpB, often upstream from cell division proteins FtsY or ZipA. Despite significant sequence divergence, MukBEF and SMC–ScpAB share the same structure and appear to play the same role inside the cell, since their inactivation leads to similarly severe defects in chromosome organization and segregation (Niki et al., 1991; Britton et al., 1998; Jensen and Shapiro, 1999; Weitao et al., 1999; Wang et al., 2006). The reported phenotypes of inactivation of condensins include anucleate cell formation, chromosome decondensation and cutting and temperature-sensitive growth. So far, only one of the two complexes was found in any given species.

The defining feature of SMCs is their structure (Fig. 1). They consist of two globular domains connected by two long coiled-coils with a hinge in between (Melby et al., 1998; Anderson et al., 2002; Matoba et al., 2005). The globular N- and C-terminal domains contain, respectively, the Walker A and B motifs and fold to produce a single globular head domain with ABC-type ATPase site on its surface (Hopfner et al., 2000; Lammens et al., 2004; Woo et al., 2009). ATP-mediated dimerization of SMC heads is believed to be central to the activity of the proteins. The hinge domain provides the second dimerization interface for SMCs and is responsible for their characteristic V-shaped appearance on electron micrographs (Haering et al., 2002; Li et al., 2010a). The bulky hinge domain is typical for condensins and cohesins but not for Rad50-like proteins (Hopfner et al., 2002).

Figure 1.

Structural organization of MukBEF. MukB, the SMC subunit of the complex, consists of three globular domains separated by two long α-helices with high propensity of coiled-coil formation. The N- and C-terminal domains fold into a single head domain with ABC-type ATP binding site on its surface. In solution, MukB forms a V-shaped dimer joint at the hinge with the two heads capable of ATP-mediated (ATP is shown as blue box) dimerization. MukEF (stoichiometry FE2) binds the outward side of the heads and is prone to form dimers of its own thereby linking distant MukBs. The inward-facing side of MukB heads contains many basic amino acids, several of which were implicated in DNA binding (Woo et al., 2009).

To gain insight into possible evolutionary origins of MukBEF, we carried out a series of homology searches to each subunit of the protein. We found high similarity between phylogenetic trees built for MukB, MukF and MukE indicating that all three subunits co-evolved together. Intriguingly, we also found distant relatives of MukBEF beyond its previously recognized set of host species. Although sequence homology was barely detectable, the new protein had the same operon organization and predicted secondary structure as MukBEF, including its many telltale features. Divergently, however, the new protein contains a markedly shorter coiled-coil region than that in MukB. In further distinction from MukBEF, ATP negatively regulated MksB–DNA interactions whereas reconstituted MksBEF was able to bind DNA. To acknowledge these similarities and differences, we named the new family MksBEF (Muk-like SMC). Further homology searches revealed that MksBEF is broadly present in diverse bacteria, is highly divergent on the sequence level and often coexists with SMC–ScpAB or MukBEF and, sometimes, another MksBEF. Thus, several condensins might be involved in chromosome organization in bacteria.


Phylogenetic analysis of MukBEF reveals distant relatives

blast (basic local alignment search tool) search using the N-terminal globular domain of the Escherichia coli MukB (amino acids 1–220) as a query sequence revealed 212 homologous bacterial proteins (Fig. 2A). Similar searches using the full-length E. coli MukF and MukE identified 164 and 166 proteins respectively (Fig. 2B and C). The resulting phylogenetic trees were highly similar to each other supporting the view that all three proteins evolved together. In agreement with previous reports (Hiraga et al., 2000), most of the retrieved sequences were found in enterobacteria and several other genera of γ-proteobacteria. Curiously, however, all three phylogenetic trees included Sorangium cellulosum, a δ-proteobacterium, suggesting that MukBEF or its relatives could exist beyond the previously recognized set of host species.

Figure 2.

Phylogenetic analysis of MukBEF reveals its distant relatives. A–C. Phylogenetic trees for sequences retrieved after blast search to the N-terminal domain of the E. coli MukB (A), MukF (B) or MukE (C). Prior to building the trees, the retrieved sequences were further filtered according to their ability to produce a multiple alignment (see Experimental procedures). D. Phylogenetic tree for blast-retrieved sequences using the Sorangium cellulosum MukE as a query.

In support of this view, blast search to the S. cellulosum MukE retrieved several uncharacterized proteins from the Pseudomonas species (Fig. 2D). This finding is significant because Pseudomonades encode the SMC–ScpAB condensin and were not expected to harbour other SMC proteins. In contrast to MukE, no new proteins emerged from blast searches to the S. cellulosum MukF and blast to MukB failed to pick up Pseudomonas proteins (data not shown), pointing to significant sequence divergence between MukBEF and the new family of proteins.

Structural organization of MukBEF-like SMCs in Pseudomonades

The newly identified MukE-like proteins were encoded in three- and four-gene operons, which sometimes contained additional, seemingly unrelated genes (Fig. 3A). We could not detect any pattern in these additional genes and focused instead on the three-gene core of all found operons. These three genes encoded soluble proteins and were organized in a manner that was highly reminiscent of MukBEF operons (Fig. 3A).

Figure 3.

Operon organization and predicted structure of MksBEF. A. Operon organization of MukBEF and MksBEF in several bacteria. Shown are the accession numbers for the SMC protein, B, and, in parentheses, the length of the genes in the operon (in kb). B. Coiled-coil formation in MukB and MksB. The probability of coiled-coil formation was computed using Paircoils2 (McDonnell et al., 2006). Shown are the p-scores of coiled-coil formation, which are inversely related to the probability of coiled-coils. Black arrows mark predicted coiled-coils in the protein. Analysed were MukBs from E. coli and Haemophilus ducreyi (NP_415444 and NP_873974 respectively; top panel), MksBs from Pseudomonas aeruginosa strain PAO1 and Pseudomonas stutzeri (NP_253375 and YP_00117322; middle panel), and the Bacillus cereus MksB and the second copy of the P. stutzeri MksB, MksB2 (YP_002528739 and YP_001171951; bottom). Sequence alignment of the head domains of MukB and MksB, together with the secondary structure predictions, are shown in Fig. S2. Note the bulky hinge domain in MukB and MksB and the longer coiled-coils found in some MksBs. C. Sequence alignment of the ATP-binding motifs in MukB and MksB. In addition to proteins analysed in (B), the alignment includes the Herpetosiphon aurantiacus MksB (YP_001547158) and the Pyrococcus furiosus SMC (NP_579572). Grey boxes mark conserved amino acids.

The first two proteins were about the same size as MukF and MukE and had similar predicted secondary structures (Fig. S1). The protein encoded by the third gene was predicted to contain two long coiled-coils of similar sizes that join together three globular domains – a telltale arrangement typical for the SMC proteins (Fig. 3B). Predicted secondary structures of its N- and C-globular domains were virtually identical to those of MukB even though homology at the sequence level was slim (Fig. S2). Moreover, the N- and C-terminal domains of MukB homologues contained the Walker A and B and signature motifs in their expected places (Fig. 3C) indicating that these proteins are functional ABC-type ATPases. Finally, inspection of the sequence of MukB homologues suggested that the proteins are likely to bind DNA. Indeed, previous structure-based mutagenesis studies implicated several lysines and arginines on the head domain of MukB in DNA binding (Woo et al., 2009). In the Pseudomonas aeruginosa MukB homologue, six out of seven of these amino acids, although not completely conserved, retained their basic character (Fig. S2) as would be expected for amino acids involved in DNA binding. Notably, only three of these amino acids were either lysines or arginines, indicating that the P. aeruginosa protein might have lower affinity to DNA, which might be stimulated by increased protonation at low pH.

Based on these vast structural similarities, we conclude that the newly found protein is a structural and perhaps functional homologue or MukBEF. Previous DNA microarray studies of P. aeruginosa revealed that these proteins are expressed in a growth conditions-dependent manner (Waite et al., 2006), and, therefore, contribute to metabolism of this bacterium. In particular, both MksBEF and SMC were upregulated during exponential phase but were ultimately assigned to different clusters, suggesting that both proteins are needed in growing cells but are differentially regulated. Thus, Pseudomonades appear to encode a second copy of condensins in addition to their previously identified SMC–ScpAB complex.

Certain structural features, however, set these proteins apart from MukBEF. The most prominent among them is the length of the coiled-coil regions within the SMC subunit. We found two groups of MukB-like proteins with the coiled-coils of approximately 200 and 300 amino acids, which falls short of 500 amino acids typical for MukB and is on the low end of lengths found in other SMC proteins (Fig. 3B). This difference, taken together with significant sequence divergence and uncertainty about intracellular function of the proteins, prompted us to give them a new name, MksBEF (MukBEF-like SMCs).

The last protein in many four-gene mksBEF operons is widely conserved in bacteria but its function is unknown. Its N-terminal half is a conserved domain DUF3322 found in numerous hypothetical bacterial proteins. The C-terminal half contains TOPRIM domain, which is found in DNA topoisomerases and primases, OLD family nucleases as well as RecR and Spo11 families of DNA repair proteins. This protein, MksG, was a part of mks operon in numerous diverse bacteria, suggesting that it acts in complex with MksBEF.

MksBEF proteins are widely spread

blast search of referenced bacterial genomes identified 35 proteins homologous to the head domain of the Pseudomonas stutzeri MksB (Fig. 4A), 30 proteins homologous to MksF (data not shown) and 84 proteins homologous to MksE (Fig. 4B). The greater number of leaves in the MksE tree reflects partial sequence conservation between MksE and MukE, which blurs the borders between the two families. This result mirrors our previous finding of Mks proteins in the MukE-, but not MukF-derived trees (Fig. 2).

Figure 4.

MksBEFs are broadly spread across proteobacteria. A and B. Phylogenetic trees for blast-retrieved proteins homologous to the Pseudomonas stutzeri MksB (first 220 amino acids as a query sequence) and MksE. Species that are excluded from the tree by the cobalt step are marked with red. C. Phylogenetic tree for MksB proteins identified in the blast search using the Herpetosiphon aurantiacus MksB (first 200 amino acids) and then filtered using the cobalt-based multi-alignment step. MksFs and MksEs in the found operons were divided into subfamilies (F2 to F4 for MksF and E2 to E7 for MksE) according to their ability to form a single multiple alignment. The branches of the tree were then labelled (grey boxes) according to the type of MksEFs. The full list and classification of the identified proteins are shown in Table S1.

Visual inspection of the corresponding genomes revealed that all identified Mks proteins are encoded together as a part of operons, as described in the previous section (Fig. 3A). Pseudomonades comprised the core of all three protein families with several select members of β-, γ- and δ-proteobacteria found in all three trees (Fig. 4A and B). Thus, MksBEFs indeed represent a distinct protein family distantly related to MukBEF.

We found 90 more MksBEFs, from 13 bacterial subdivisions, when we carried out blast search using the head domain of MksB from Herpetosiphon aurantiacus, A GNS bacterium, as a query sequence (Fig. 4C). As was the case with the Pseudomonas' MksBEFs, the newly found proteins were organized into three- and four-gene operons (Fig. 3A). Notably, much fewer proteins emerged from blast searches using the H. aurantiacus MksE and MksF. This observation further underscores low sequence conservation within the MksBEF family.

Given this difficulty, we organized the found MksEs and MksFs into subfamilies based on their ability to produce a multiple alignment. In this procedure, we conducted a blast search to a given protein and then generated, using cobalt (constraint-based alignment tool), multiple alignment of the retrieved sequences. Thereby generated sets of sequences formed the core of the subfamilies. Several sequences were excluded from the initial multiple alignments but produced significant pairwise alignment with multiple members of an existing subfamily. Such sequences were joined to the said subfamily if we could find at least three its members with high (E-value less than 1) similarity to the query sequence.

Using this approach, we identified two distinct subfamilies of MksF and five subfamilies of MksE in the H. aurantiacus subfamily of MksBEFs (Table S1). In addition, the proteins from Acidovorax delafieldii could not be assigned to any existing group but formed a root of their own subfamilies. Separation of MksFs and MksEs into groups correlated with the clustering of MksB on its phylogenetic tree (Fig. 4C) arguing for co-evolution of all three subunits of MksBEF.

By no means do MksBEFs shown in Fig. 4 comprise a complete list. Homology search to outlying members of the generated trees revealed new subfamilies of the protein (data not shown). Thus, MksBEF proteins are broadly present in diverse bacteria.

Bacteria can encode multiple condensins

Table S2 presents a list of bacteria with several copies of MksBEF. In some cases, these bacteria were identified in the aforementioned homology searches (Fig. 4). In others, the second copy emerged after a series of successive blast s. We could not find a simple rule that would predict which bacteria contain one or two MksBEFs. For example, two MksBEFs are encoded in P. aeruginosa strain UCBPP-PA14 (Table S2) and only one in P. aeruginosa PAO1. Likewise, we did not find any pattern in how similar the two MksBEFs are. For example, MksBEFs in Faecalibacterium prausnitzii were of the same type (Table S1), whereas all subunits of MksBEFs in Pseudomonas syringae belong to different subfamilies (Tables S1 and S2).

Finally, we found that MksBEFs can coexist with MukBEF, SMC–ScpAB or even both condensins together (Table S2), which raises the possibility that several specialized condensins might contribute to chromosome organization in bacteria.

P. aeruginosa MksB contributes to chromosome partitioning

PAO1 strain of P. aeruginosa encodes the conventional SMC–ScpAB condensin (proteins PA1527, PA3197 and PA3198) and MksBEF complex (proteins PA4684, PA4685 and PA4686). To gain insight into the physiological function of MksBEF, we took advantage of the transposon disruption library available at the University of Washington Genome Center (Jacobs et al., 2003). Strain PW8890 from this collection harbours transposon ISphoA/hah insertion at the hinge region of MksB (mksB::ISphoA/hah), which presumably inactivates it.

Examination of PW8890 using fluorescence microscopy revealed increased frequency of anucleate cells (1.2%; 11 out of 937) in cultures grown in M9 at 37°C (Fig. 5A) but not in LB and not in strains harbouring plasmid-encoded MksB (Fig. 5C). These data indicate that MksB functions in chromosome maintenance and that the demand for its activity varies depending on growth conditions. In contrast, Δsmc mutants of PAO1 (Fig. 5A) produced 2.5% anucleate cells both in LB and in M9 (15 out of 626 cells in LB and 28 out of 1189 in M9; Fig. 5C). For both mutants, production of anucleate cells could be suppressed by a plasmid-borne mksB or smc (Fig. 5C), as appropriate, but not by the vector alone (data not shown).

Figure 5.

Chromosome packing defects in condensin-deficient P. aeruginosa. A. Fluorescence micrographs of fixed mksB-deficient PW8890 cells, Δsmc::Gm mutants and the parental PAO1-Lac (PAO1) cells. Anucleate cells (arrows) were found among PW8890 and Δsmc but not PAO1 cells. Size bar, 1 µm. B. Arabinose-induced overproduction of MksB (mksB++) induces chromosome condensation, whereas no chromosome condensation is observed when cell growth and protein synthesis are arrested by treatment with 100 µg ml−1 chloramphenicol (CAM). C. Anucleate cell formation (± SEM) in LB and M9 medium by PAO1-Lac, mksB and Δsmc cells that harbour, when indicated, plasmids pUCP_MksB (pMks) or pUCP_SMC (pSMC). D. Anucleate and guillotine cell formation by the PT7-mksFEBΔsmc OP106 cells (Δsmc mksI) exponentially grown in LB for 16 h in the presence or absence of 5 mM IPTG. E. Anucleate cell formation (OD600 of 0.6) for cells grown in LB for 6 h or, where indicated, for 16 h either in the presence or in absence of 5 mM IPTG. Tested were the parental PAO1-Lac, the Δsmc OP103, the PT7-mksFEB OP105 (mksI) and the PT7-mksFEBΔsmc OP106 cells.

Unlike with the E. coli MukBEF, condensin-deficient P. aeruginosa cells did not increase in size or formed filaments, which could be masking possible chromosome compaction defects. However, overproduction of MksB resulted in marked chromosome condensation (Fig. 5B). No such condensation was detected in cells arrested with chloramphenicol (Fig. 5B) or those harbouring the vector alone (data not shown). Similar condensation of the E. coli chromosome was observed earlier upon overproduction of MukB (Wang et al., 2006) and H-NS (Spurio et al., 1992) but not of MukEF or topoisomerase I (Wang et al., 2006). Thus, the found chromosome condensation reflects the activity of the overproduced protein. This result further supports the notion that MksB might be controlling the chromosome size by bringing distant DNA segments together.

The observed frequencies of anucleate cells are lower than those found for condensin mutants of E. coli and Bacillus subtilis (Niki et al., 1991; Britton et al., 1998; Wang et al., 2006), but exceed the 0.1% reported for the condensin-deficient Caulobacter crescentus (Jensen and Shapiro, 1999). In further discord with other condensins, neither the mksB nor Δsmc strains were deficient in colony formation at 23°C or 37°C, and the MksEF-deficient MPAO1 strain failed to produce anucleate cells either in LB or in M9 (data not shown). Such discrepancies could be explained by high frequency of compensatory mutations that could be masking effects of missing condensins. In this view, MPAO1 cells harbour a mutation that alleviates the absence of MksEF. Alternatively, MksBEF and SMC could be playing partially redundant functions. In tentative support of the latter notion, we found that the frequency of conjugal transfer of Δsmc::GmR into PW8890 was more than three orders of magnitude lower than into the wild-type PAO1 strain (data not shown). In contrast, conjugation-mediated replacement of PMks promoter using pEX_LacI_Mks plasmid occurred with similar frequencies in PAO1 and PW8890 strains (3 × 10−7 and 2 × 10−6 respectively). This result, therefore, suggests synthetic phenotype of mutations in MksB and SMC rather than defects of PW8890 cells in homologous recombination.

We next constructed a conditionally condensin-deficient strain using gene-replacement procedures that minimize accumulation of compensatory mutations (see Experimental procedures). The resulting OP106 strain (LacIq-PT7-mksFEBΔsmc::Gm) lacks smc gene and expresses MksBEF from a tightly controlled IPTG-inducible promoter PT7(A1/04/03) (Lanzer and Bujard, 1988). When overnight OP106 cells, grown in LB in the presence of IPTG, were transferred into the fresh medium without IPTG, formation of anucleate cells could be observed (Fig. 5D and E). As before, the frequency of anucleate cells was relatively low but could be increased if cells were diluted again into the fresh medium and allowed to grow further. After 16 h of continuous growth (compared with 6 h in the ‘regular’ single-inoculation experiment), 5.7% cells displayed severe chromosome partitioning defects (Fig. 5D). Such defects, however, failed to develop in the presence of IPTG (Fig. 5D and E). In contrast, IPTG had only minimal effect on anucleate cell formation in the Δsmc::Gm OP103 cells, which produce MksBEF from its endogenous promoter (Fig. 5E). We conclude therefore that MksBEF, similar to SMC, functions in chromosome partitioning. Curiously, induction of MksBEF in the presence of SMC resulted in a small increase in the frequency of anucleate cells (Fig. 5E), indicating that the functions of the two condensins overlap only partially and that their balanced production is essential for faithful chromosome segregation.

Biochemical activities of PaMksB

We next expressed the P. aeruginosa MksB in E. coli and purified it using C-terminal eight-histidine tag (Fig. 6A). The protein was functionally active since it suppressed the phenotype of MksB-deficient cells (Fig. 5C). The measured sedimentation coefficient of MksB was 6.3 ± 0.3 S. The Stokes radius, measured using gel filtration chromatography, was estimated as 7.4 ± 0.3 nm. Based on these values, the molecular weight and the form factor of MksB can be calculated as 200 kDa and 1.6, respectively, which indicates that the protein, similar to other SMCs, forms a stable, extended in shape dimer in solution.

Figure 6.

Biochemical analysis of the P. aeruginosa MksBEF. A. SDS-PAGE analysis of purified PaMksB and PaMksEF. B. ATPase activity of PaMksB (expressed as ATPs hydrolysed per minute per dimer of MksB) at various buffer and salt conditions. When indicated, the reaction mixtures were supplemented with supercoiled pBR322 DNA (dsDNA) or single-stranded phage φX174 DNA (ssDNA). For DNase treatment, 15 U of DNase I, which was either heat-killed by 10 min boiling (100C) or not (+), was included into reactions 10 min prior to the addition of MgATP and MksB. C. DNA gel-shift analysis of MksB binding to supercoiled (SC), linear (L), nicked circular (NC) pBR322 DNA and single-stranded phage φX174 (ss) DNA. The indicated protein concentrations assume that MksB is a dimer. D. ATP destabilizes MksB–DNA complex. DNA binding reactions (using 10 ng of SC pBR322) were supplemented with MgATP and 25 mU apyrase, as indicated above the gels. To ensure complete ATP hydrolysis, apyrase treatment was carried out for 15 min at 30°C prior to the addition of MksB. E. DNA aggregation by MksB. Twenty nanograms of DNA was incubated with indicated amounts of MksB in the presence or absence of 5 mM MgATP and 50 mU apyrase that was either heat-killed (Ap*) or not (Ap), separated by centrifugation into the top (T) and bottom (B) fractions, deproteinized and analysed by gel electrophoresis. Separation of MksB into the top and bottom fractions was analysed by SDS-PAGE. F. DNA aggregation by the E. coli MukB. The experiment was performed as described in (E) for MksB. G. MksB and MksEF form a 1:1 complex. MksB and MksEF were mixed in indicated proportions and, following reconstitution, analysed by gel filtration through Sephacryl S300 column. Arrows mark positions of the excluded volume of the column (V0) and the peak fractions for MksB, MksEF and MksBEF. H. MksBEF forms a complex with DNA. Following reconstitution, 27 pmol of MksBEF was incubated with 600 ng of pBR322 DNA and resolved by gel filtration through 1.1 ml Sephacryl S300 column.

Purified MksB displayed a potent ATPase activity (Fig. 6B), which co-purified with the protein on the heparin column (data not shown). In agreement with our earlier prediction (Fig. S2), ATPase activity (Fig. 6B) and DNA binding activity (data not shown) of MksB were stimulated by a decrease in pH to 6.5. ATPase activity of MksB markedly declined in the presence of single-stranded but not double-stranded DNA, and this effect was virtually abolished by DNase I treatment (Fig. 6B). This finding indicates that DNA binding and ATP hydrolysis stabilize different conformations of MksB.

DNA gel-shift analysis revealed that MksB can form a stable complex with linear and circular DNA and has about threefold greater affinity to single-stranded than double-stranded DNA (Fig. 6C). Preferred binding to single-stranded and cruciform DNA has been reported earlier for several other SMC proteins (Akhmedov et al., 1998; Hirano and Hirano, 1998; Petrushenko et al., 2006b) and might explain why only single-stranded DNA was inhibiting ATP hydrolysis by MksB (Fig. 6B). Notably, ATP destabilized MksB–DNA complex (Fig. 6D), lending further support to the notion that its function is to displace the protein from DNA.

To evaluate potential DNA condensation activity of MksB, we employed the microfuge DNA aggregation assay (Krasnow and Cozzarelli, 1982). In this assay, multi-molecular DNA aggregates are separated from soluble DNA by short centrifugation in a table-top centrifuge, and the extent of DNA condensation is evaluated by gel electrophoretic analysis of DNA content in the top and bottom halves of the centrifuge tube. During this short centrifugation, large nucleoprotein aggregates sediment to the bottom fraction whereas single DNAs and unbound protein remain evenly distributed throughout the tube. Using this assay, we found that MksB can condense linear, circular and single-stranded DNA and that this activity is inhibited by ATP (Fig. 6E). In contrast, ATP had no notable effects on DNA aggregation activity of the E. coli MukB (Fig. 6F). Thus, MksB can indeed bring distant DNA fragments together, as would be expected from its apparent role in chromosome organization.

PaMksB and PaMksEF form a complex

The P. aeruginosa MksE and MksF were cloned together as a fragment of an operon and purified from overproducing E. coli cells to 95% homogeneity using conventional chromatography (Fig. 6A). The two proteins co-migrated on a gel-filtration column (Fig. 6G; second panel), indicating that they form a complex, MksEF. When mixed at 1:1 ratio, MksB and MksEF formed a complex that migrated faster during gel filtration than MukB or MksEF separately (Fig. 6G). At 1:2 MksB-to-MksEF ratio, only 50% of MksEF co-migrated with MksB (Fig. 6G; bottom panel), indicating that the stoichiometry of the complex is MksB2–(MksEF)2.

Notably, the DNA binding activity of MksB was not inhibited in the presence of MksEF. When preformed MksBEF was incubated with DNA and loaded onto a gel filtration column, both the protein and DNA eluted close to the excluded volume of the column (Fig. 6H), well ahead of the DNA-free MksBEF (Fig. 6G). This result clearly reveals the existence of a ternary MksB–MksEF–DNA complex. Thus, the non-SMC component of MksBEF is dispensable for DNA binding.


A new family of condensins

We describe here a new family of bacterial condensins which is distantly related to MukBEF. The protein has the same operon organization and predicted secondary structure as MukBEF. MksBEF, however, has a markedly shorter coiled-coil region than MukBEF or any other SMC protein. The second distinguishing feature of MksBEF is its low sequence conservation. Based on their sequence, MksBEFs can be organized into a set of overlapping subfamilies where sequence homology between subfamilies is virtually undetectable. This is in stark contrast to SMC proteins, which show significant sequence similarity across kingdoms of life. These features, together with significant biochemical divergence, indicate that MksBEF should be regarded as a new family of proteins rather than an extension of MukBEF superfamily.

MksBEF is broadly present in many diverse proteobacteria and can be found in many more species than we could annotate in this study. Curiously, the protein subfamilies appear to form a network rather than a tree (data not shown) pointing – together with the sequence divergence – to ancient origins of the protein. Such distribution could have also arisen via horizontal gene transfer. MksBEFs could be, for example, an obligatory part of as yet undescribed mobile genetic elements. In either case, further studies of MksBEFs might shed light onto evolution of condensins and other SMC-like proteins.

The P. aeruginosa MksBEF

Biochemical and functional assessment revealed that the P. aeruginosa MksB acts as a condensin. PaMksB demonstrated the key activities expected of condensins including ATP-modulated DNA binding and condensation in vitro (Fig. 6) and in vivo (Fig. 5B). Taken together with the increased frequency of anucleate cells in the MksB-deficient strain (Fig. 5), the synergistic effects of mutations in smc and mksB and the ability of MksBEF to complement anucleate cell formation in SMC-deficient cells (Fig. 5D and E), these data strongly argue that the protein, similar to other condensins, plays a role in chromosome organization and segregation.

There were, however, conspicuous differences. PaMksB proved to be a potent ATPase that hydrolyses ATP two orders of magnitude faster than the E. coli MukB (Petrushenko et al., 2006b). Moreover, the ATPase activity of MksB is inhibited by DNA (Fig. 6B) and, in turn, destabilizes MksB–DNA complex (Fig. 6D). So far, this is the first report that ATP and DNA have opposing effects on an SMC protein. Instead, DNA binding and reshaping activities of other condensins were shown to be stimulated by ATP (Strick et al., 2004; Petrushenko et al., 2010). Finally, the DNA binding activity of MksB was not inhibited by its non-SMC counterpart, which further likens this protein to the B. subtilis SMC.

Ironically, the finding of inhibitory effects of ATP strengthens rather than disproves the emerging mechanistic model of SMCs (Fig. 7). Indeed, ATP has long been argued to mediate dissociation–association of the SMC heads. At the same time, recent structural studies revealed an extensive positively charged patch on the hinge-proximal surface of the SMC heads, and mutational analysis confirmed that many amino acid residues from this patch participate in DNA binding (Woo et al., 2009). It follows then that ATP hydrolysis should displace DNA from the protein if it forces apart the two heads (Fig. 7), whereas the opposite would be expected if disruption of the DNA binding interface occurs at a later stage of the ATPase reaction cycle. This variation in the coupling between the cycles of ATP hydrolysis and DNA binding might be responsible for the differences between SMCs that are stimulated and inhibited by ATP.

Figure 7.

The explosive bolt model of the coupling between ATP hydrolysis and DNA binding. ATP-sandwiched dimerization of the head domains creates an extensive positively charged patch capable of high-affinity binding to DNA. ATP hydrolysis forces conformational transition in MksB that displaces DNA from its binding site. Having lost half of its binding site, DNA is favoured to dissociate from the protein, leaving it competent for the next cycle of ATP binding and hydrolysis. Because ATP and DNA are both postulated to support dimerization of the SMC heads, the rate of ATP hydrolysis is expected to decline in the presence of DNA.

Possible functions of Mks proteins

Our finding of a new family of condensins raises interesting questions about the function of these proteins. It is tempting to propose that, similar to their eukaryotic counterparts, MksBEF, MukBEF and SMC–ScpAB play distinct roles in global packing of bacterial chromosome. Perhaps one of these families acts as a prokaryotic counterpart of eukaryotic cohesins and establishes bridges between sister chromosomes. Alternatively, each of these systems could be optimized for its own set of physiological conditions and, therefore, having all three proteins would benefit environmentally growing and pathogenic bacteria but not necessarily laboratory strains. The latter idea is tentatively supported by the finding that chromosome packing defects in mksB and smc mutants of P. aeruginosa differ depending on growth conditions (Fig. 5C). This interpretation naturally explains why MksBEF subfamilies cut across bacterial subdivisions and appear to correlate with the occupied niche of a bacterium rather than its phylum (Fig. 4) or why the archetypal strains of E. coli and B. subtilis do not encode any MksBEFs. MksBEFs could conceivably be spread from one bacterium to another via horizontal gene transfer or, perhaps, retained because additional adaptive advantage to a given niche. Alternatively, of course, each of these systems could have evolved independently, and their occurrence in specific bacteria reflects little but random selection.

The finding of MksG in many MksBEF operons raises another possibility that MksBEF supports recombination and repair rather than chromosome packing and segregation. Indeed, the presence of TOPRIM domain in MksG likens this protein to the Saccharomyces cerevisiae Spo11 protein, which acts in conjunction with Rad50 in initiation of meiotic recombination (Keeney et al., 1997). MksBEF and MksG could conceivably support an analogous function in bacteria. In fact, two SMC-like proteins, SbcCD (Connelly et al., 2003) and RecN (Meddows et al., 2005), that function in DNA repair have already been identified in bacteria. The putative role MksBEFs in recombination and repair would be in accord with the earlier finding of greater sequence similarity between MukB and Rad50 than between MukB and other SMCs (Cobbe and Heck, 2004). This view, however, does not readily explain the observed chromosome partitioning defects of MksB-deficient P. aeruginosa or the ability of the protein to complement chromosome partitioning defects of SMC-deficient cells (Fig. 5). It seems more tempting to draw parallels with MukB, which was recently found to form a complex with ParC, a subunit of topoisomerase IV (Hayama and Marians, 2010; Li et al., 2010b).

A distinct possibility to consider is that the found MksBEFs represent two or more functionally distinct families. Indeed, many mksBEF operons lack the fourth gene, mksG (Fig. 3A), whereas others contain genes with no relation to mksG (Table S2). It is not inconceivable that the fourth gene- or its absence- defines functional specificity of MksBEFs.

This study reveals that the role of condensins in chromosome maintenance is even more versatile than we previously thought. Other families of condensins, perhaps with novel functions, might emerge in future explorations of genomes. Unravelling the roles of these proteins will likely offer clues to the biogenesis and evolutionary origins of the chromatin structure and will undoubtfully enhance our ability to manipulate chromosomes on the global scale.

Experimental procedures

Sequence analysis

blast search was carried out using default parameters on the NCBI website (McGinnis and Madden, 2004), with the Expect Threshold value set at 1 and limiting search to referenced NCBI sequences (the RefSeq database). To stay on the conservative side, the retrieved protein sequences were aligned together, when indicated, using the cobalt multiple alignment procedure (Papadopoulos and Agarwala, 2007). This approach helped eliminate unrelated sequences with short regions of homology to the bait while retaining most of similar sequences (data not shown). For MukE and MukF, the resulting phylogenetic trees were largely the same whether or not the cobalt multiple alignment procedure was used. Omitting the cobalt step for MukB, however, resulted in trees that included not only MksB but also such distantly related proteins as RecF and putative ABC transporters. The remaining sequences were organized into phylogenetic trees using Fast Minimal Evolution method (Desper and Gascuel, 2004) using blast Tree View. The topology of the resulting trees was essentially the same as for those constructed using Neighbor Joining algorithm (data not shown). The advantage of this approach is that it is integrated into the workflow on the NCBI website and allows simultaneous alignment and classification of all retrieved sequences.

Secondary structure predictions were done using PSIpred (Jones, 1999; Bryson et al., 2005) available via Expasy proteomics server. Coiled-coil predictions were carried out using Paircoil2 program (McDonnell et al., 2006). Protein sequences were aligned using clustalw2 program (Larkin et al., 2007).

Plasmids and strains

The strains and plasmids used in this study are summarized in Table S3. The P. aeruginosa strains MPAO1 and PW8890 (PA4686::ISphoA/hah) were obtained from the University of Washington Genome Center (Jacobs et al., 2003) and single colony purified by plating on LB. The strain UCBPP-PA14 (Lee et al., 2006) was a gift of Dr Ausubel. MPAO1 contains a 1 kb deletion that spans PA4684 and PA4685 (Dotsch et al., 2009) and was used as a ΔmksEF strain. PAO1-Lac [lacIq+Δ(lacZ)M15+tetA+tetR+] was obtained from ATCC and was used as a wild-type PAO1. Plasmids pUCP22, pSP856 and pEX18Ap were a gift of Dr Schweizer. Plasmid pYM101 was a gift of Dr Morita (Morita et al., 2010). Phage F116 was a gift of Dr Lomovskaya. Cells were grown in LB or M9 plus 0.4% glucose at 37°C supplemented, when appropriate, with 20 µg ml−1 tetracycline, 20 µg ml−1 kanamycin or 100 µg ml−1 gentamicin. For all experiments, overnight cell culture was diluted into the fresh medium, and cells were further grown for several doubling times. DNA transformation was done as described earlier (Choi et al., 2006).

pUCP_MksB, which encodes PaMksB-His8 under the control of an arabinose-inducible promoter, was constructed by subcloning XbaI–SphI fragment from pNPA_MksB into the pUCP22 shuttle plasmid (West et al., 1994). The gene encoding PaSMC (PA1527) was amplified using PCR from the genome of UCBPP-PA14 and inserted between HindIII and XbaI sites of pUCP22 yielding plasmid pUCP_SMC. pUCP_MksB and pUCP_SMC were able to complement the phenotype of appropriate condensin-deficient strains in the absence of any inducer (Fig. 5C).

pEX_ΔSMC was constructed by subcloning gentamicin-resistance gene flanked by two FRT sites (Flp recognition sequences) from plasmid pSP856 (Hoang et al., 1998) between two 0.5 kb DNA fragments that flank the smc gene and then inserting the cassette between KpnI and HindIII sites of the conjugal vector pEX18Ap (Hoang et al., 1998). pEX_ΔSMC was then transformed into the E. coli strain SM10, and the Δsmc::Gm fragment transferred, via conjugation followed by sucrose counterselection, into PAO1-Lac and MPAO1 as previously described (Hoang et al., 1998). The success of the replacement was confirmed by PCR analysis of the genome of the resulting OP101 and OP102 strains.

pEX_LacI_Mks was constructed by subcloning the following four fragments between KpnI and HindIII sites of plasmid pEX18Ap: (i) the 0.5 kb fragment upstream from the mks promoter (nucleotides −634 to −138 relative to the start of mksF), (ii) the FRT-flanked gentamicin resistance cassette, (iii) the 1.3 kb fragment of pYM101 plasmid that contains the lacIq gene and the T7 early promoter PT7(A1/04/03), and (iv) the first 503 bp of mksF. PAO1-Lac cells were then transformed with the pEX_LacI_Mks plasmid as previously described (Choi et al., 2006) and plated on LB containing 30 µg ml−1 gentamicin and 5 mM IPTG. Because the plasmid does not carry a P. aeruginosa origin of replication, only cells that integrated gentamicin-resistance gene via homologous recombination are expected to form colonies. PCR analysis of the five resulting colonies identified one clone that underwent a double-cross-over recombination event at the mks promoter, yielding OP104 strain, which encodes IPTG-controlled MksBEF.

The gentamicin-resistance gene was removed from OP104 cells by transforming the cells with pFLP2 plasmid, which encodes Flp recombinase, followed by sucrose counterselection to remove the plasmid (Hoang et al., 1998). Δsmc::GmR locus was then transferred into the resulting unmarked OP105 strain from OP102 via phage F116 transduction, yielding a conditional condensin-deficient double mutant OP106. For all strains, success of the gene replacement procedures was confirmed by PCR analysis of the affected regions.

Fluorescence microscopy

Cells (300 µl) at OD600 of 0.8 were fixed in 70% ice-cold ethanol, incubated on ice for 10 min, rinsed with PBS (phosphate-buffered saline), deposited onto polylysine-coated microscope slides, stained with 100 nM DAPI and 1× SyproOrange and observed by fluorescence microscopy. As described earlier (Wang et al., 2006), this procedure ensures even staining of protein (SyproOrange) and DNA (DAPI) and facilitates thereby reliable detection of anucleate cells. To visualize MksB-overproducing cells, PAO1 cells harbouring pUCP_MksB plasmid were grown in M9 containing 0.4% glycerol up to OD600 of 0.2, supplemented with 0.5% l-arabinose and further incubated for 6 h. Chloramphenicol-arrested PAO1 cells were collected 1 h after the addition of the drug to 100 µg ml−1 (approximately 3× MIC) to the culture (OD600 of 0.4) that was grown at 37°C in M9 medium supplemented with 0.4% glucose.

Protein expression and purification

The mksB-His8 gene, which encodes the P. aeruginosa MksB with C-terminal eight-histidine tag, was amplified, using PCR, from genomic DNA extracted from strain UCBPP-PA14 and subcloned into a pBAD-based plasmid under the control of arabinose-inducible promoter. MksB encoded in UCBPP-PA14 (PA14_61990) is 99.8% identical (100% similar) to MksB of PAO1 (PA4686). The resulting pNPA_MksB01 plasmid was transformed into the E. coli DH5α cells, and the protein was expressed by the addition of 0.2% arabinose. The protein was then purified using nickel-chelate and heparin chromatography, essentially as described earlier for MukB (Petrushenko et al., 2006b), and dialysed against 20 mM HEPES, pH 7.7, 200 mM NaCl, 2 mM EDTA, 1 mM DTT, 50% glycerol for storage. Unlike MukB, MksB eluted as a single sharp peak both from the nickel and from heparin columns and did not contain other co-purifying proteins. As judged by SDS-PAGE analysis, the protein was 99% pure (Fig. 6A).

For MksEF expression, the mksF–mksE fragment of the mksF–mksE–mksB operon was amplified by PCR from the strain UCBPP14-PA14 genomic DNA and subcloned into pET21d(+) vector with C-terminal eight-histidine tag on MksE. Protein expression was carried out overnight at 12°C in BL21(DE3) cells. The protein proved to be highly prone to degradation, which precluded the use of nickel-chelate chromatography. The following protocol yielded more than 95% pure protein with minimal degradation (Fig. 6A).

Following induction, the cells were suspended in ice-cold Cell Disruption Buffer [20 mM Tris-Cl, pH 7.9, 75 mM (NH4)2SO4, 5% glycerol, 2 mM EDTA, 1 mM DTT, 1 mM PMSF, 5 mM benzamidine, 0.7 µg ml−1 pepstatin, 2 µg ml−1 aprotinin (Sigma-Aldrich)] and lysed using a French press. The cell extract was supplemented with 0.2% Triton X-100 and clarified by centrifugation for 5 min at 4 kr.p.m., 4°C. MksEF was then pelleted by 15 min centrifugation at 12 000 r.p.m., resuspended in Cell Disruption Buffer supplemented with 0.2% Triton X-100, and collected again by 20 min centrifugation at 20 000 r.p.m., 4°C. The pellet was then resuspended in the same buffer, sonicated twice for 15 s, clarified by 20 min centrifugation at 20 kr.p.m. and extracted two more times using the same procedure. The combined supernatant was loaded onto a HiTrap QFF column, washed with Triton X-100-containing Cell Disruption Buffer, then with the Chromatography Buffer (20 mM Tris-Cl, pH 8.9, 50 mM NaCl, 5% glycerol, 2 mM EDTA, 1 mM DTT, 1 mM PMSF, 5 mM benzamidine, 0.7 µg ml−1 pepstatin, 2 µg ml−1 aprotinin), and then eluted with 50 mM to 800 mM NaCl gradient. The peak fractions (350 mM to 430 mM NaCl) were pooled, adjusted to 1 M NaCl, loaded onto a HiTrap Butyl-FF column (GE Healthcare) and eluted with 1 M to 0 mM NaCl gradient. The peak fractions were pooled, dialysed against 20 mM HEPES, pH 7.7, 200 mM NaCl, 2 mM EDTA, 1 mM DTT and 50% glycerol and stored at −20°C.

Enzyme assays

ATPase rate– 0.1 µg (0.45 pmol) MksB was added to 10 µl of Reaction Buffer (20 mM HEPES, pH 6.5, 40 mM NaCl, 2 mM MgCl2, 5% glycerol, 1 mM DTT) supplemented with 1 mg ml−1 bovine serum albumin, 0.1 mM MgATP and 0.1 µCi of [γ-32P]-ATP, and reactions were incubated at 37°C. Where indicated, plasmid pBR322 DNA or phage φX174 was included into the reaction or a different buffer was used. Aliquots (1 µl) were removed every 5 min, mixed with 10 µl of Stop Buffer (50 mM Tris-HCl, pH 7.8, 0.5% SDS, 20 mM EDTA, 200 mM NaCl, 0.5 mg ml−1 proteinase K), incubated for 40 min at 55°C and analysed by thin-layer chromatography using polyethyleneimine cellulose plates (J. T. Baker)

DNA gel shift analysis was performed essentially as described earlier (Petrushenko et al., 2006b) with few modifications. Ten nanograms of DNA was incubated in Reaction Buffer supplemented, where indicated, with 5 mM MgATP and 25 mU apyrase (New England Biolabs) for 15 min at 30°C. The heat-killed apyrase was prepared by boiling it for 10 min. MksB was then added, and reactions continued for 5 min at 37°C. Samples were then chilled on ice for 3 min and resolved by electrophoresis (3 V cm−1, 3.5 h, 4°C) through 0.8% agarose gels in TAE7.0 buffer (40 mM Tris-acetate, 1 mM EDTA, pH 7.0).

DNA aggregation assay was adapted from an earlier study (Krasnow and Cozzarelli, 1982). MksB was incubated with 20 ng of DNA in 20 µl of Reaction Buffer supplemented, as indicated, with 5 mM MgATP or 50 mU apyrase for 10 min at 37°C. The reactions were then centrifuged for 15 min at 13 200 r.p.m., 23°C, and the top 10 µl were then immediately transferred to a fresh tube. The top and bottom fractions were either analysed by SDS-PAGE to visualize protein or supplemented with 3.3 µl of 4× Stop Buffer, incubated for 30 min at 55°C, and analysed by gel electrophoresis. Apyrase treatment, when indicated, was performed as described for the gel-shift analysis.

The molecular weight of MksB was determined using the combination of sucrose density gradient centrifugation and gel filtration chromatography. We used the following size markers: thyroglobulin (19.2 S, 8.6 nm), ferritin (17.6 S, 6.2 nm), catalase (11.2 S, 5.3 nm), alcohol dehydrogenase (7.3 S, 4.6 nm), bovine serum albumin (4.3 S, 3.6 nm), carbonic anhydrase (2.8 S, 2 nm) and ribonuclease A (1.9 S, 1.64 nm). Thirty micrograms of MksB was mixed with 30 µg of each marker in the buffer containing 20 mM HEPES, pH 7.7, 200 mM NaCl, 2 mM EDTA, 1 mM DTT, 5% glycerol and analysed by gel filtration through a Sephacryl S400 column and centrifugation through 10–40% sucrose gradient as described earlier (Petrushenko et al., 2006b). MksB migrated as a single peak both during centrifugation and during gel filtration.


This work was supported by Grants 1049755 from the National Science Foundation, AI094124 from the National Institutes of Health and an award from Oklahoma Center for Advancement of Science and Technology.