The bacterial chromosome: architecture and action of bacterial SMC and SMC-like complexes

1Our story starts with the isolation of an Escherichia coli mukB mutant (from the Japanese word ‘Mukaku’, meaning ‘anucleate’) that generated anucleate cells as a consequence of impaired chromosome segregation, by Soto Hiraga, a pioneer of bacterial chromosome biology (Hiraga et al., ). Subsequent analysis characterized the mukB gene and its 177-kDa protein, MukB, which was shown to form V-shaped dimers in which the C- and N-terminal regions came together to form a globular ATP-binding ‘head’, with a rod-like central domain and a distant ‘hinge’ (Fig. 1; Niki et al.,, ). It was proposed that this might be a force-generating motor-like protein, akin to myosin, kinesin or dynein, which functioned directly in chromosome segregation. Later, the E. coli mukB gene was shown to be in an operon in which two other genes, mukE and mukF, encoded proteins necessary for Muk function, whereas the first gene of the operon encodes a presumptive methyl transferease, SmtA, of unknown function (Yamanaka et al., ). Other bacteria that encode mukBEF genes also have them organized into a single operon, whereas most other bacterial SMC proteins are encoded in an operon distinct from those of their cognate accessory proteins, which themselves tend to be cotranslationally expressed from within the same operon.

2In 1993, Strunnikov et al., characterized a yeast gene, SMC1, and its product, Smc1p, identified earlier by a mutant phenotype of impaired minichromosome segregation (Larionov et al., ). This essential protein had the same general organization of a central coiled-coil ‘rod’, bounded by a dimerization domain and an ATPase formed by the N- and C-termini, as in MukB, and the SMC-like Rad50 protein, involved in DNA repair (Alani et al., ). Furthermore, Strunnikov et al. identified putative proteins related to Smc1p from bioinformatic searches in a number of other eubacteria. Melby et al. extended the bioinformatic and phylogenetic analysis and defined MukB as a bona fide, yet distant, SMC family member. In addition, their biochemical and electron microscopy analysis led them to articulate clearly the distinctive and conserved architecture of SMC V-shaped dimers in which a c. 50-nm-long antiparallel intramolecular coiled-coil folded about a flexible dimerization hinge, with the ATPase head domain created by the SMC N- and C-termini at the opposite end to the hinge (discussed in Haering et al., ). Some O-shaped dimers arose by interaction between the two heads of a dimer.

3Later, it became clear that SMC proteins in general do not function in isolation, and like MukB, they interact with essential accessory proteins (Fig. 1; Michaelis et al.,; Yamazoe et al.,; Mascarenhas et al.,; Nasmyth & Haering, ). One of these, termed ‘the kleisin’, bridges the two ATPase heads of an SMC dimer and interacts with a second accessory protein. Together, these play key regulatory roles in the ATP binding–hydrolysis–release cycles. Eukaryotic SMCs invariably form heterodimers, with the N- and C-terminal domains of the kleisin interacting with specific SMC heads (Fig. 1; Haering et al.,; Onn et al., ). In contrast, bacterial SMCs are invariably homodimers. In eukaryotes, SMC complexes have been classified into three groups, cohesins, condensins and Smc5-6 complexes. Classically, cohesins maintain cohesion between replicated sister chromatids by entrapping the sisters within a tripartite SMC complex ring, while condensins act in condensation–organization of mitotic chromosomes and Smc5-6 complexes act in DNA repair (Thadani et al.,; Wu & Yu,; Remeseiro & Losada, ). Furthermore, all of these complexes have been implicated directly or indirectly in other chromosome processing activities. Typically, bacterial SMC complexes have been described as condensins although it remains unclear as to whether their primary role is in chromosome condensation. Certainly, E. coli MukBEF is not a cohesin because the newly replicated sister chromosomes of Muk− cells have increased rather than decreased cohesion (Danilova et al., ). SMC protein complexes are ubiquitous and, as far as we are aware, are present in all organisms except a few bacteria.

4Until recently, a single characteristic SMC complex was thought to be encoded by any given bacterium. For example, the Gm+ bacterium Bacillus subtilis and Gm− alphaproteobacterium Caulobacter crescentus encode a typical SMC protein, a kleisin ScpA (Segregation and condensation protein A), and ScpB, which binds ScpA (Britton et al.,; Melby et al.,; Jensen & Shapiro,; Mascarenhas et al.,; Bürmann et al., ). The E. coli SMC protein, MukB, and its accessory proteins, MukE and MukF (kleisin), have a low primary sequence homology to the more typical SMC complexes. MukBEF complexes are restricted to some Gamma- and Deltaproteobacteria, which additionally contain a number of restricted and characteristic genes that include dam, seqA, mutH, matP and zapB and a number of other genes of unknown function (Hiraga,; Brézellec et al., ). The significance of this cluster of genes remains to be elucidated, but it is intriguing that those characterized all participate in chromosome organization processing, and mutations in at least two of these genes suppress some of the phenotypes of muk mutations (Onogi et al., ). A third bacterial SMC complex was identified recently through bioinformatic analysis and has been termed ‘MksBEF’ for MukBEF-like SMC (Petrushenko et al., ). Unlike MukBEF, these genes are widely distributed in bacterial species and often found in combination with MukBEF, SMC-ScpAB and/or other MksBEF complexes. MksB has a shorter coiled-coil than typical SMCs, and its precise functions remain unclear.

5In addition to true SMC complexes, a family of conserved SMC-like proteins that include the related bacterial SbcC and archaeal and eukaryotic Rad50 has important roles in DNA repair at DSB ends; A c. 50-nm antiparallel coiled-coil is bounded by a zinc-hook dimerization domain, and the typical ATPase head is bridged by the nuclease SbcD/Mre11 (Fig. 1; Lammens et al.,; Lim et al.,; Möckel et al.,; Stracker & Petrini, ). Finally, another SMC-like bacterial repair–recombination protein, RecN, has been proposed also to act in DSB repair, although atypically no accessory proteins appear to be involved in its action, and its coiled-coil is much shorter than the typical c. 50 nm (Pellegrino et al., ).

6Mutations in SMC/MukB (hereafter termed ‘SMC’ for brevity), or in either of the accessory proteins, lead to the same phenotype, showing that all three proteins are required for SMC complex function. Impairment of SMC function generally leads to an increase in the production of anucleate cells, indicating a direct or indirect role in chromosome segregation (Table 1). In C. crescentus, B. subtilis and E. coli, this is accompanied by temperature sensitivity, which is particularly acute under fast growth conditions. Furthermore, global chromosome decondensation has been reported (Niki et al.,; Moriya et al.,; Britton et al.,; Jensen & Shapiro,; Sawitzke & Austin,; Table 1). These defects are frequently enhanced under conditions that lead to decreased negative supercoiling (gyrase inhibitors or appropriate gyrase mutations; Hiraga et al.,; Britton et al.,; Moriya et al.,; Jensen & Shapiro,; Lindow et al.,; Adachi & Hiraga, ), while conditions that increase negative supercoiling can suppress the temperature-sensitive phenotype resulting from Muk/SMC impairment (for example, TopA impairment in E. coli and B. subtilis; Sawitzke & Austin,; Lindow et al., ).

7If SMC complexes have a direct role in chromosome segregation, then in their absence, what other mechanisms mediate or facilitate chromosome segregation? Two other systems that act in chromosome segregation are the ParAB–parS systems that are best characterized for their roles in low copy plasmid segregation and the FtsK/SpoIIIE family of DNA translocases (Kaimer & Graumann,; Mierzejewska & Jagura-Burdzy,; Reyes-Lamothe et al., ). Mutation of either of these systems does not generally lead to a loss of cell viability (an exception being loss of functional ParAB in C. crescentus; Mohl & Gober, ), but can be synthetically lethal or sick in combination with SMC impairment. For example, in Staphylococcus aureus and Streptomyces coelicolor, the double mutants ΔsmcΔspoIIIE and ΔsmcΔparB, respectively, are viable, but have phenotypes considerably stronger than the relatively mild phenotypes of the single parB and spoIIIE mutants (Dedrick et al.,; Yu et al., ). When the Muk−/SMC− phenotype is strong, the phenotypes of some double combinations are synthetically lethal, for example, an E. coli mukB-ftsKC double mutant (Yu et al.,; Sivanathan et al., ) and the double mutants in SMC-SpoIIIE, SMC-SftA (an FtsK orthologue) and SMC-Soj/Spo0J in B. subtilis (Britton et al.,; Britton & Grossman,; Lee & Grossman,; Kaimer et al., ). The overlapping activities of multiple SMC complexes may account for some weak SMC− phenotypes like in Pseudomonas aeruginosa (Petrushenko et al., ).

8Analysis of mutant phenotypes as outlined above has failed to determine conclusively whether the primary role of SMC/MukB complexes is in chromosome organization, which would reasonably require that the SMC complex interacts globally with different regions of the chromosome, or whether the primary role is in chromosome segregation as proposed in the original experiments of Hiraga, Niki and colleagues. In the latter case, one would expect the SMC complex to interact with a specific chromosomal region. A third possibility is that these complexes act independently in both organization and segregation. If this is the case, it should be possible to separate these functions genetically. A limitation of these classical genetic studies results from the strong pleiotropic phenotypes that arise from mutation in Muk/SMC, ParAB-parS or FtsK and its orthologs. Quantitative cell biology studies combined with real-time depletion and repletion experiments have begun to address this (later).

9The SMC ATPase head domain is a characteristic ABC transporter ATPase, characterized by Walker A and Walker B consensus motifs (Walker et al., ), the signature motifs or C-motif unique to the ABC superfamily (Ames & Lecar, ) and a D-loop (Fig. 2). The Walker A motif, which is essential for ATP binding, is encoded by the SMC N-terminus, whereas the Walker B motif, required for ATP hydrolysis, is encoded by the C-terminus, with the N- and C-termini folding into a compact globular domain. Within an SMC dimer, the C-terminal domain of the second monomer carries the C- or signature motif and the D-loop that stabilize the binding of ATP by the first monomer, and is required for ATP hydrolysis. Consequently, two composite ATP-binding sites are created by dimerization of the two head domains (Fig. 2). ATP binding leads to head engagement, resulting in the formation of the ring-shaped SMC dimer in which the kleisin association with both heads in a dimer reinforces the O-shaped structure by forming an SMC–kleisin tripartite ring (or a MukB-MukF tetrapartite ring in the case of MukBEF; Fig. 1; Woo et al.,; Bürmann et al.,; reviewed in Hirano & Hirano,; Lim & Oh, ). ATP binding and head engagement are invariably required for stable association with DNA, while ATP hydrolysis, which can only occur when the heads are engaged, is required for normal function (Hirano & Hirano,; Schwartz & Shapiro,; Badrinarayanan et al.,; Bürmann et al., ). In vitro ATPase assays of MukBEF have shown that robust ATP hydrolysis can occur in the absence of added DNA, but that requires the presence of both MukE and MukF in addition to MukB (Woo et al.,; reviewed in Lim & Oh, ) Similarly, B. subtilis SMC ATPase activity is stimulated by accessory proteins (Kamada et al., ). Note that both ATP binding and kleisin association with the heads act to keep the heads within a dimer ‘closed’.

10Eukaryotic SMC complexes are intrinsically asymmetric because they are formed from SMC heterodimers. Despite bacterial complexes containing SMC homodimers, asymmetric complexes form at least during part of the ATP binding–hydrolysis–release cycles, indicating that asymmetry underlies their conserved biochemical action. An analysis that focussed on the B. subtilis SMC complex showed that a kleisin monomer bridges the two heads of an SMC homodimer, thereby introducing asymmetry into the tripartite ring-shaped complex (Bürmann et al., ). The kleisin C-terminus interacts with the ‘cap’ region of one head, while the N-terminal domain interacts with the SMC ‘neck’, a region of the coiled-coil adjacent to the ATPase (Fig. 1). This asymmetric interaction is facilitated during head engagement resulting from ATP binding, because a steric clash prevents the binding of two kleisin C-terminal domains to the two heads of a single SMC homodimer. The architecture of this asymmetric complex is remarkably similar to that proposed for eukaryotic SMC complexes (Bürmann et al.,; Upton & Sherratt, ). At first sight, the architecture of the MukBEF complex looks dissimilar because the kleisin, MukF, dimerizes through its N-terminal domain, while the C-terminal domains interact with a ‘cap’ in the heads of the MukB dimer in the absence of nucleotide, giving a symmetrical complex (Fig. 1; Woo et al., ). However, ATP binding and consequent head engagement sterically prevent occupation of both heads by kleisin C-terminal domains, thereby generating an asymmetric complex (Fig. 1). This could lead to dimer of dimer complexes in the presence of ATP, as have been observed in vivo (Badrinarayanan et al., ) and in vitro (Petrushenko et al.,; Woo et al., ). Intriguingly, in Rad50-Mre11 complexes, Mre11 dimerization is accompanied by both cap and neck interactions with the SMC-like Rad50 (Fig. 1; Lim et al.,; Lammens et al.,; Möckel et al., ), and we speculate that MukF might in addition have a neck interaction, potentially leading to asymmetric head engagement through the kleisin interaction in the presence of ATP and thereby helping close a tripartite ring. The theme of ATP-dependent head engagement leading to steric clashes is continued with Rad50-Mre11, where head engagement occludes the nuclease within Mre11 (Lammens et al.,; Lim et al.,; Möckel et al., ).

11Structures of the dimerization hinges of several SMC proteins have demonstrated a strong dimerization interface that contains some conserved features (Haering et al.,; Ku et al.,; Li et al.,; Griese & Hopfner, ). Nevertheless, several studies have provided evidence that the dimerization interface must open or change for normal function (Gruber et al.,; Hu et al.,; Thadani et al., ). Indeed, it has been proposed that this interface provides the loading gate for DNA, while the head domains are the exit gate (later; Chan et al.,; Bürmann et al., ), thereby providing an analogy to type II topoisomerases that have separate DNA entry and exit gates.

12A range of studies has assessed SMC binding to DNA in vitro (Volkov et al.,; Hirano & Hirano,,; Petrushenko et al.,,; Cui et al.,; Woo et al.,; Griese & Hopfner,; Borgmann et al., ). Some of these have been in the absence of accessory proteins, and few have used conditions in which DNA-binding requirements mimic in vivo loading onto chromosomes. Nevertheless, structural studies of E. coli MukBEF and an archaeal SMC identified a conserved flat positively charged region on top of the SMC head that interacts with DNA and is likely to play an important role, possibly once DNA is properly loaded into the tripartite ring. Some assays have also assessed DNA compaction, intermolecular DNA bridging or an ability to restrain DNA topology (Petrushenko et al.,,; Chen et al.,; Borgmann et al., ), but the physiological significance of these remains unclear, as does the study in which SMC overexpression led to DNA condensation in vivo (Volkov et al.,; Wang et al.,; Petrushenko et al., ).

13Although the early analyses of Hiraga, Niki and colleagues (Niki et al.,, ) proposed that MukB(EF) might function as a force-generating motor-like myosin, kinesin or dyneins, using DNA as a track rather than actin or tubulin, the analogy may well be mechanistically superficial. Nevertheless, there are at least two similarities. For example, movement of cytoskeletal motors on actin and tubulin tracks bears similarities to how MukBEF has been proposed to move on DNA. The slow turnover of MukBEF dimers of dimers on DNA, as compared to measured ATPase rates in vitro, led to the proposal of a ‘rock climber’ model in which release of DNA from one dimer upon head opening, mediated by ATP hydrolysis, would allow potentially reloading of a second DNA segment, while the second dimer remained bound to DNA (Badrinarayanan et al., ). Another similarity is that cytoskeletal motors carry cargo bound to sites distant from the ATPase heads, analogous to the observed in vitro interaction of topoisomerase IV with the MukB hinge (Hayama & Marians,; Li et al., ). This indicates that TopoIV is a MukBEF cargo that would allow coordination of decatenation and segregation, consistent with the observation that sister cohesion is increased in Muk− cells (Danilova et al., ).

14Do all SMC complexes entrap DNA within a closed tripartite ring as shown for cohesin (Haering et al., ) and supported by experiments with B. subtilis Smc (Bürmann et al., )? The conserved ability to engage heads on ATP binding and open them on ATP hydrolysis, the closure being reinforced by the kleisin interaction, taken together with observations that led to proposals that ring opening at both the heads and the hinge is necessary for in vivo function (Arumugam et al.,; Weitzer et al.,; Gruber et al.,; Chan et al., ), all support the DNA entrapment model, based on the observed topological entrapment of yeast minichromosomes within cohesin rings (Haering et al.,; Farcas et al., ). In the case of cohesin, it is proposed that the two newly replicated sisters are effectively topologically entrapped, because any chromosome ends are distant from a bound SMC complex (Fig. 3). In other situations, for example, where the SMC complex is acting as a condensin, it has been proposed that two duplex segments derived from the same chromosome are contained within a ring. If these were formed from a simple loop of limited size, there would be no topological entrapment, because the loop could be released from (or indeed loaded into) the SMC ring without the ring opening. However, trapping of toroidal links or catenated loops would need ring opening to load or release the DNA (Fig. 3). Despite the strong evidence for tripartite SMC complex rings, in vitro experiments have also provided evidence that SMC complexes may ‘daisy-chain’ into oligomeric or rosette structures, and it has been proposed that these may be physiologically relevant (Mascarenhas et al.,; Matoba et al., ). Such structures could be mediated simply by kleisin–SMC interactions or through higher-order interactions mediated by other accessory proteins; for example, MukE and B. subtilis ScpB have been proposed to be able to mediate such higher-order complexes (Hirano & Hirano,; Gloyd et al., ). Nevertheless, both in vivo and in vitro analysis of B. subtilis SMC complexes has found predominantly simple dimeric SMC complexes compatible with tripartite rings being the major players in SMC action (Fuentes-Perez et al.,; Bürmann et al., ).

15Why is the c. 50-nm intramolecular coiled-coil a conserved part of SMC complex architecture? We consider two nonexclusive scenarios. In one, the length of the coiled-coils is needed to make a ring that can accommodate chromosomal DNA and anything associated with it; in this, it is simply a long linker that should be relatively immune to mutagenesis. In the other, the coiled-coil is a communication device between the head and hinge; for example, enabling the energy derived from ATP binding–hydrolysis–release of phosphate and/or ADP being used to do work at the hinge (discussed in Thadani et al., ). In Rad50-Mre11 complexes, the angle of the coiled-coils leaving the heads changes markedly on head engagement (Lammens et al.,; Lim et al.,; Möckel et al., ). In this context, it is intriguing to note that in dynein, energy obtained from the ATP binding–hydrolysis–release cycle is transmitted through the coiled-coil to do work at the distant microtubule-binding domain (the equivalent of the dimerization hinge in SMCs), by changing the register of the two chains in the coiled-coil (Carter, ). A similar mechanism could operate within SMCs, with kleisin–neck interactions and other changes associated with the ATPase cycle facilitating this transmission from the head to the hinge.

16Analysis of chromosome organization in Muk− E. coli cells showed that the position of genetic loci changes dramatically in the absence of MukBEF, corresponding to a 90° rotation of the whole chromosome with respect to the long axis of the cell, such that newborn cells growing in minimal medium have their origins at the old pole and replication terminus regions at the new pole (Danilova et al., ). In such cells, DNA replication now initiates at a pole-proximal origin, showing that origin position defines replisome position at initiation rather than vice versa (Reyes-Lamothe et al., ). Reorganization of the chromosome after fast Muk repletion and its altered organization on rapid degron-mediated Muk impairment were relatively slow, taking > 30% of a cell generation equivalent for the altered state to be achieved and occurred in the absence of DNA replication (Badrinarayanan et al., ).

17The in vivo visualization of fluorescent SMC complexes (Table 2) and their relationship with genetic locus position and the replication machinery have shown that E. coli MukBEF and other SMC complexes can form a discrete number of fluorescent foci per cell. Niki, Hiraga and colleagues were the first to demonstrate the patterns of MukBEF fluorescent focus positioning in cells (Ohsumi et al., ). Further analyses showed that MukBEF and B. subtilis SMC are ori-associated (Fig. 4; Danilova et al.,; Gruber & Errington,; Sullivan et al.,; Badrinarayanan et al., ) and not associated with functional replisomes. In C. crescentus, between two to five foci were observed depending on the cell cycle stage, and one or two bright polar foci were formed only in a minority of late predivisional cells (Jensen & Shapiro, ).

18These studies using relatively long capture times in epifluorescence microscopy could not reveal the proportion or action of nonfocus molecules at other positions of the chromosome. Nevertheless, the simplest interpretation of the ori association is that at least some of the E. coli MukBEF action is local at ori. In B. subtilis and Streptococcus pneumoniae, ParB (Spo0J) targets SMC to multiple parS sites in the ori region, although intriguingly, B. subtilis ParB mutants have much weaker phenotypes with respect to chromosome segregation than SMC mutants (Gruber & Errington,; Sullivan et al.,; Minnen et al., ), indicating that an alternative mechanism must be available to target the SMC complex to chromosomes or that disruption of ori localization does not lead to an SMC− phenotype. In this respect, it is noteworthy that in general, SMC complexes reside at specific regions of chromosomes, although they themselves do not exhibit sequence-specific DNA binding. In C. crescentus, no evidence was found for an SMC–ParB interaction (Table 2; Schwartz & Shapiro, ). What targets E. coli MukBEF complexes to the ori region is unknown, but the association requires MukEF and the ability to hydrolyse ATP. Functional association of eukaryotic cohesin with chromosomes has the same requirements (Hu et al., ). High-resolution quantitative imaging showed that a single ori-associated E. coli MukBEF focus contains 10 immobile dimer of dimer complexes, on average. These were shown to have a dwell time of c. 50 s, while in B. subtilis, a somewhat slower turnover rate was calculated (Borgmann et al., ). These in vivo turnover rates are slower than one might have expected based on in vitro ATPase activities, leading to the proposal of a ‘rock climber’ model (above; Badrinarayanan et al., ).

19Does the ori region position SMC or vice versa? The observation that oris are mispositioned in Muk− cells points to MukBEF positioning oris. This view is strengthened by repletion experiments, which indicated that an ori-independent MukBEF focus positioning system operates (Badrinarayanan et al., ). If this is indeed the case, it indicates that ori positioning by MukBEF is central to its role in chromosome segregation. Revealing the mechanism that positions MukBEF will be crucial to understanding the roles of SMC in chromosome segregation.

20Mutant MukB proteins that fail to bind ATP or bind ATP, but are defective in head engagement, failed to form Muk fluorescent foci. In contrast, a mutant MukB that binds ATP but is impaired in hydrolysis (MukBEQ) has a classical Muk− phenotype, but forms fluorescent foci that do not turnover (Badrinarayanan et al., ), consistent with the view that ATP binding and consequent head engagement are necessary for stable loading onto DNA. Intriguingly, fluorescent foci containing this mutant protein had an altered chromosome location (Badrinarayanan et al.,; D.J. Sherratt, unpublished data), leading to the conclusion that although ATP binding is sufficient for DNA association, ATP hydrolysis is required for dissociation from DNA and for normal chromosomal location and function. An equivalent mutation in cohesin gave an SMC complex that loaded efficiently at centromeres, but could not establish cohesion or relocate to the normal cohesin binding sites on chromosomes (Hu et al., ).

21The rapidly diffusing E. coli MukBEF molecules, which constitute the majority population, appear to have MukB, MukE and MukF complexed together, because they exhibited the same low diffusion coefficient (Badrinarayanan et al., ). This contrasts with the conclusion from experiments in B. subtilis where the majority diffusing population appeared not to have ScpAB complexed with SMC (Borgmann et al., ). Whether these rapidly diffusing molecules are physiologically active in global chromosome organization is unclear; the results with B. subtilis would suggest not, at least for this organism. If MukBEF and ori-associated bacterial SMC complexes have a direct role in global chromosome organization, then either the non-ori-associated molecules must have a role in this or those associated stably with the ori must transiently interact with (or ‘gather in’) other regions of the chromosome. In our view, this latter scenario seems unlikely. Alternatively, the primary roles of MukBEF and ori-associated SMC complexes may be to position ori regions, thereby facilitating bipolar segregation. A failure to do this could indirectly lead to phenotypes of chromosome disorganization and decondensation.

22The remarkable conservation of the architecture of different SMC complexes leads us to be confident that all such complexes share a common biochemical mechanism during their association with, and action on, chromosomal DNA. In our view, the available data support a model in which functional SMC complexes have DNA within the ring formed by the association of SMC with kleisin proteins (Fig. 3). Loading of DNA requires ATP binding and head engagement, with the possibility that DNA is loaded through the dimerization hinge, which may open on head engagement. Loaded DNA may also interact with specific sequences in the SMC, leading to closure of the entrance gate. Given that SMCs are dimeric and cohesin entraps two DNA duplexes, it seems likely that all SMCs will load two DNA segments each interacting with a specific SMC region. Asymmetry in at least the head-engaged complexes appears to be universal for reasons that need to be determined, although it possibly relates to regulation of DNA release from the head exit gate during the ATP binding–hydrolysis–release cycles. The observation that MukBEF dimer of dimer complexes appear to be physiologically relevant, points to mechanisms in which release of DNA from one dimeric complex allows the other dimer of a complex to remain associated with DNA (Badrinarayanan et al., ). The likely functional association of MukBEF with TopoIV hints at coordination of decatenation with segregation, two likely sequential steps in the segregation process. Indeed, just as cohesin provides the tension between sister chromatids that is vital for sensing their amphitelic attachment to opposing spindles during successful eukaryotic chromosome segregation, catenation between newly replicated bacterial chromosomes may achieve the same cohesion-like role, with separating ori-associated sister MukBEF clusters loaded onto separate sisters fulfilling the role of the eukaryote spindle (Fig. 3d).

23Future experiments aimed at revealing the molecular action of SMC complexes will require the multidisciplinary use of structural biology, biochemistry, genetics and cell biology, with an increasing focus on the quantitative live cell imaging methods that will enable real-time in vivo biochemistry within the paradigm of ‘observe–measure–perturb’.