Hit the brakes – a new perspective on the loop extrusion mechanism of cohesin and other SMC complexes Free

Structural maintenance of chromosome (SMC) complexes are an evolutionarily conserved family of protein complexes that control the spatial organization of chromatin (Fig. 1). SMC complexes have a common architecture in all domains of life (Aragon, 2018; Hassler et al., 2018; Makela and Sherratt, 2020; Onn et al., 2008; Palecek, 2018; Yatskevich et al., 2019). The core of the complex is composed of a dimer of SMC proteins. These are elongated, rod-like proteins characterized by two globular domains, the head and hinge, which are connected by a long coiled-coil domain of about 50 nm. The head domain contains two halves of an ATP-binding cassette (ABC)-type ATPase domain. The dimerization of the SMC proteins is mediated through the hinges and co-alignment of the coiled coils that result in the spatial proximity of the heads. The binding of two ATP molecules to the Walker box motifs of one head and the ABC signature motifs of the other head induces their engagement (Aragon, 2018; Hassler et al., 2018; Makela and Sherratt, 2020; Onn et al., 2008; Palecek, 2018; Yatskevich et al., 2019). Recent studies have shown that ATP binding and hydrolysis also induce major conformational changes in the SMC proteins (Hassler et al., 2019; Kamada et al., 2017; Muir et al., 2020; Orgil et al., 2016; Rajasekar et al., 2019; Sedeno Cacciatore and Rowland, 2019). The SMC heads are connected by a kleisin subunit that interacts asymmetrically with the heads, docking at the emerging coiled coil on one of the SMC proteins and the head base on the other one. The kleisin serves as a binding hub for regulatory factors, which are characterized by the presence of HEAT repeats, a structural motif associated with protein–protein interactions. In addition to the core subunits, other factors regulate the SMC complex either by interacting or post-translationally modifying its core subunits. The structure and regulatory factors of SMCs have been the subject of many studies and extensively reviewed (see, for example, Aragon, 2018; Hassler et al., 2018; Makela and Sherratt, 2020; Onn et al., 2008; Palecek, 2018; Yatskevich et al., 2019).

Eukaryotes encode six different SMC proteins (Smc1 to Smc6), which are organized into three distinct heterodimers, while most bacteria encode a sole SMC protein that homodimerizes. The three eukaryotic SMC complexes are cohesin, which contains Smc1 and Smc3, condensin, including the Smc2–Smc4 heterodimer, and a complex of Smc5 and Smc6 (denoted hereafter as Smc5/6) (Fig. 1) (Aragon, 2018; Hassler et al., 2018; Onn et al., 2008; Palecek, 2018; Yatskevich et al., 2019). These complexes are involved in many and various cellular functions. Cohesin is a multifunctional complex involved in sister chromatid cohesion, DNA repair and gene expression regulation (Carretero et al., 2010; Moronta-Gines et al., 2019; Nishiyama, 2019), while condensin is a critical factor in the compaction of interphase chromatin into mitotic chromosomes (Kalitsis et al., 2017; Paul et al., 2019). Lower eukaryotes have the canonical condensin, whereas vertebrates contain two variants of condensin, condensin I and condensin II (Onn et al., 2007, 2004, 2003). These complexes share the core Smc2–Smc4 heterodimer, but have a different set of kleisin and HEAT-repeat subunits, and they play distinct roles in the condensation process. Recently, it has been suggested that condensins, in addition to their role in mitotic chromosome organization, regulate transcription and senescence by mediating the local compaction of chromatin during interphase (Hassan et al., 2020; Iwasaki et al., 2019; Marshall et al., 2020). Smc5/6 plays an essential role in sculpting chromatin during DNA replication and repair (Aragon, 2018; Palecek, 2018). However, its exact functions are not fully understood, and this complex has been less characterized than the other SMCs.

Several families of SMC complexes that participate in chromosome organization and partitioning have been found in bacteria (Fig. 1) (Badrinarayanan et al., 2015; Hirano, 2016; Makela and Sherratt, 2020). The most common is the ScpAB family, which shares a high degree of similarity with the eukaryotic condensin and is also found in archaea. The second SMC family, known as MukBEF, is found in enterobacteria and several other related orders of γ-proteobacteria (Badrinarayanan et al., 2015; Hirano, 2016).

Prokaryotic SMC complexes are considered to be ancestors of the eukaryotic condensins (Makela and Sherratt, 2020), although some evidence suggests that they are more related to the Smc5/6 complex (Palecek and Gruber, 2015). Their function has been linked to a compaction of the bacterial chromosome around an axial core (Badrinarayanan et al., 2012a,b; Rajasekar et al., 2019; Wang et al., 2017, 2018).

The cellular functions of SMCs relate to their activity in organizing the three-dimensional structure of chromatin, mainly through the formation of chromatin loops (Hassler et al., 2018; Makela and Sherratt, 2020; Rowley and Corces, 2018; Sedeno Cacciatore and Rowland, 2019; Yatskevich et al., 2019). The similar overall structure of all SMC complexes implies a common mechanism of action. However, the biological roles with regard to their effects on chromosome structure, stability and dynamics significantly differ between the different SMC complexes and even in different functions of the same complex. In this Hypothesis, we examine the two main mechanisms that have been associated with cohesin and propose a unified model for their co-existence in a single complex. Finally, we discuss how this mechanism can be extended and interpreted for both condensin and Smc5/6. As bacterial SMCs are considered to be ancestors of eukaryotic condensin, they will not be further scrutinized here.

Apart from the Smc1–Smc3 heterodimer, cohesin contains the kleisin family protein Scc1 (RAD21 in humans) and Scc3 (Stag1/Sa1 and Stag2/Sa2 in humans) that together define the cohesin core (Fig. 1) (Michaelis et al., 1997; Nasmyth and Haering, 2009; Yatskevich et al., 2019). Key regulatory proteins are the loading dimer, Scc2–Scc4 (NIPBL–MAU2 in humans), the cohesin release factors WAPL and Pds5 (Pds5A and Pds5B in humans), a multitasking protein involved in both cohesin release during G1/S phases of the cell cycle and cohesin maintenance during G2 (Hartman et al., 2000; Losada et al., 2005; Panizza et al., 2000), and the acetyltransferase Eco1 (Esco1 and Esco2 in humans) (Nasmyth and Haering, 2009; Onn et al., 2008; Yatskevich et al., 2019). These proteins regulate cohesin activity by either binding to the core complex or modifying it. Cohesin was initially found to have an essential role in sister chromatid cohesion (Guacci et al., 1997; Michaelis et al., 1997). Here, newly replicated DNA molecules, known as sister chromatids, are tethered once they are formed during S-phase, which ensures their bipolar attachment to the spindle and their accurate segregation to the two daughter cells during mitosis.

Cohesin is loaded onto chromatin in an Scc2–Scc4-dependent manner during mitotic exit (vertebrate cells) or in G1 phase (yeast) (Ciosk et al., 2000; Hinshaw et al., 2015; Kuleszewicz et al., 2013; Shwartz et al., 2016; Watrin et al., 2006). Scc2 has also been identified as an activator of ATPase activity of cohesin, which is mediated by the SMC head domain (Petela et al., 2018). WAPL and Pds5 are loading antagonists that release cohesin from the chromatin (Bernard et al., 2008; Gandhi et al., 2006; Kueng et al., 2006; Kulemzina et al., 2012; Rowland et al., 2009; Sutani et al., 2009; Tanaka et al., 2001). During the S-phase of the cell cycle, in coordination with DNA replication and the formation of the sister chromatid, Eco1 acetylates the Smc3 subunit of cohesin (Rolef Ben-Shahar et al., 2008; Rowland et al., 2009; Unal et al., 2008). This, in turn, counteracts the effect of WAPL and Pds5 (Vaur et al., 2012) and blocks the ATPase activity (Camdere et al., 2015; Elbatsh et al., 2016), and the chromatids become entrapped by cohesin. Interestingly, during the G2 phase of the cell cycle, Pds5 adopts a new function and protects the cohesin from premature dissociation from the chromatin until the onset of prophase (Hartman et al., 2000; Losada et al., 2005; Panizza et al., 2000), when cohesin is removed from chromatin to allow the segregation of the sister chromatids.

The molecular mechanism of tethering has been at the center of a long-standing dispute, and several models have been proposed throughout the years. Three main models have been described: (1) the ring model, in which the sister chromatids are topologically entrapped in the central ring formed by the Smc1–Smc3–kleisin subunits (Fig. 2A) (Haering et al., 2008; Ivanov and Nasmyth, 2005, 2007); (2) the modified ring model, in which one chromatid is topologically entrapped in the cohesin ring, while the second chromatid is locked between the kleisin and the HEAT-repeats subunits (Fig. 2A) (Chapard et al., 2019; Xu and Yanagida, 2019); and (3) a model that is based on the entrapment of a single chromatid in the central lumen of cohesin, while cohesion is achieved through the dimerization of cohesins (Fig. 2A) (Huang et al., 2005; Zhang et al., 2008). More recent structural and spectroscopic studies provide strong support that tethering involves a single copy of the cohesin complex with two distinct chromatin-binding sites (model 2) (Chapard et al., 2019; Liu et al., 2020; Xu and Yanagida, 2019). However, evidence pointing to cohesin dimerization (Eng et al., 2014, 2015) needs to be reconciled with this model.

Elucidating the molecular details of tethering and discussing which one of these models is correct has been the source of many disputes, as recently discussed elsewhere (Skibbens, 2019). Nonetheless, despite the fundamental differences between these models, a common concept is that sister chromatid cohesion is mediated by physical and stable tethering of the chromatids by cohesin in order to prevent their separation during the cell cycle by diffusion or premature pulling by the spindle (Fig. 2A).

For many years, cohesin was not been considered to have a role in interphase chromatin organization. However, this dramatically changed with the identification of cohesin as a regulator of gene expression (Dorsett et al., 2005; Rollins et al., 1999) and the development of genomics techniques that show that cohesin is involved in the organization of interphase chromatin into chromatin loops together with the CCCTC-binding factor (CTCF), a sequence-specific DNA-binding protein known to act as a chromatin insulator (Dixon et al., 2012; Hansen et al., 2017; Lieberman-Aiden et al., 2009; Nora et al., 2012). The colocalization of cohesin and CTCF at chromatin loop boundaries is important for their stability. The CTCF–cohesin loops represent a subset of chromatin domains called topologically associating domains (TADs). These mega-base-pair regions are characterized by extensive interactions between chromatin entrapped inside the domain but only show rare inter-domain interactions. (Belton et al., 2012; Dekker et al., 2013; Dixon et al., 2012; Rowley and Corces, 2018; Yu and Ren, 2017). TADs are a distinctive feature of eukaryotic chromatin during interphase, and they are believed to play roles in the regulation of gene expression and, presumably, also DNA replication and repair (Luppino and Joyce, 2020; Mamberti and Cardoso, 2020; Matthews and Waxman, 2018; Noordermeer and Feil, 2020). TAD size varies in different organisms, and similar chromatin structures have been found in yeast (Eser et al., 2017). However, as yeast does not have a CTCF homolog, the stabilization of their loops is different, as discussed below.

Loop extrusion is the currently accepted mechanism to explain chromatin loop formation, while cohesin has been identified as the main extruder of this process (Alipour and Marko, 2012; Nasmyth, 2001; Wutz et al., 2020). Cohesin acts as a motor protein through which the chromatin is progressively threaded, resulting in the formation of extended loops (Fig. 2B). This model shifted the role of cohesin from a passive tethering factor in mitosis and stabilizer of TADs, as we describe above, into being the principal extruder (Alipour and Marko, 2012; Nasmyth, 2001). Structural studies have shown that chromatin is anchored between the SMC heads and the kleisin, while the ATP-induced conformational change provides the mechanical threading of chromatin. Notably, several models for the exact nature of the conformations have been proposed (Gligoris et al., 2014; Haering et al., 2008; Higashi et al., 2020; Petela et al., 2018; Srinivasan et al., 2018).

The loop extrusion activity of cohesin has been confirmed experimentally in recent single-molecule spectroscopy in vitro studies using recombinant cohesin (Davidson et al., 2019; Kim et al., 2019). Notably, these studies show that under these experimental conditions, extrusion is highly processive and depends on both the ATP-hydrolysis activity of cohesin and its interaction with the Scc2 subunit. As described above, the Scc2–Scc4 dimer is required for cohesin loading onto the chromatin in cells, but not in vitro (Murayama and Uhlmann, 2017; Onn and Koshland, 2011). The structural basis for the interaction between Scc2 and cohesin has only recently been determined (Shi et al., 2020b), and this interaction has been shown to activate the ATPase activity of cohesin (Petela et al., 2018).

Comparing the mechanisms of cohesin function during mitosis and loop extrusion reveals fundamental differences between them. Sister chromatids are tethered by cohesin from the time of their formation until cohesin removal in anaphase initiation (Nasmyth and Haering, 2009; Onn et al., 2008). Therefore, this cohesion function requires physical and stable entrapment of the chromatids. By contrast, loop extrusion is a dynamic process that requires cycles of attachment and dissociation of cohesin from the chromatin, until the loop reaches its final size of tens of kilobases at a rate of ∼0.5 kilobases per second (Davidson et al., 2019; Kim et al., 2019). Thereafter, cohesin engages in passive tethering that ensures loop stability by preventing the diffusion of the chromatin fibers (Haarhuis et al., 2017) (Fig. 2). Hence, a full understanding of the molecular basis of cohesin function requires elucidating how the dynamic loop extrusion activity is turned on and off to mediate physical passive tethering, as during TAD stabilization and in mitosis.

SMC complexes are evolutionary related to ABC transporters, a large superfamily of transmembrane protein complexes (Hopfner and Tainer, 2003; Nasmyth and Haering, 2005). ABC transporters are located in the cell membrane; they are activated by the presence of their substrate and use ATP to induce a conformational change that is associated with the active passage of molecules from the cytoplasm through the transporter cavity into the cell exterior. In many mechanistic aspects, loop extrusion of chromatin by cohesin is similar to the shuttling of molecules through ABC transporters. In both cases, the molecular mechanism depends on cycles of ATP-induced conformational changes that, in turn, push the substrate molecule through the complex (Fig. 2B). With this similarity in mind, we suggest that loop extrusion, rather than passive tethering, is the fundamental biochemical activity of cohesin. Indeed, natural shifting between conformational changes has been described in purified SMC dimers in vitro (Eeftens et al., 2016). The interactions between SMC proteins, kleisin and the regulatory subunits restrain and coordinate these spontaneous conformational changes. Notably, the nature of this conformational change is not fully characterized. Different models that include compartmentalization (Collier et al., 2020; Elbatsh et al., 2016; Srinivasan et al., 2018) or head–hinge engagement (Rowland et al., 2009; Shi et al., 2020b) have been suggested to describe this conformational change. Extrusion is therefore initiated when cohesin interacts with Scc2 and chromatin, which stimulates its ATPase activity (Fig. 3) (Hassler et al., 2018; Hassler et al., 2019; Petela et al., 2018; Shi et al., 2020b). Therefore, according to this scenario, we expect that the dissociation of Scc2 from cohesin after loading of the DNA will halt the loop extrusion activity.

As mentioned above, in vitro, Scc2 is not essential for the association of cohesin with chromatin or its tethering activity (Murayama and Uhlmann, 2017; Onn et al., 2008). However, without Scc2, cohesin is unable to extrude DNA (Davidson et al., 2019; Kim et al., 2019). In cells, Scc2 is essential for the association of cohesin with chromatin and for sister chromatid cohesion (Michaelis et al., 1997), presumably by targeting cohesin to chromatin (Hinshaw et al., 2015). In yeast, Scc2 dissociates from cohesin post-loading, and, in turn, cohesin is pushed away from the loading site and travels along the chromosome, moved along by the transcription machinery (Fig. 3) (Busslinger et al., 2017; Davidson et al., 2016; Hu et al., 2011; Lengronne et al., 2004; Stigler et al., 2016; Wang et al., 2017). This passive movement of cohesin can be explained by the suppression of the loop extrusion activity of cohesin after dissociation of Scc2 and the shift from binding to its passive-topological mode (Fig. 3). However, Scc2 dissociation also reduces the stability of the cohesin–chromatin complex; this results in frequent cohesin dissociations that can be observed in cells by microscopy (Austin et al., 2009; Bernard et al., 2008; Gause et al., 2010; Gerlich et al., 2006).

Interestingly, it has been suggested that in human cells, Scc2 exists in substoichiometric amounts relative to cohesin and ‘hops’ between different chromatin-bound cohesin complexes (Rhodes et al., 2017). This could fine-tune the extrusion activity by turning loop extrusion on and off at certain locations, while allowing cohesin to reside on the chromatin without inducing extrusion at specific loci, such as centromeres, where it is enriched. Therefore, we suggest that dissociation of Scc2 is a key molecular event that suppresses the loop extrusion activity of cohesin but also reduces its binding affinity to chromatin. The remaining issue is, therefore, how cohesin is stabilized on chromatin in a tethering mode for either TAD stabilization and or mitotic sister chromatid cohesion.

Acetylation of the cohesin core subunit Smc3 by Eco1 on two adjacent lysine residues located at the head domain in proximity to the ATP binding site has been identified as a key event in the cohesin activity cycle (Rolef Ben-Shahar et al., 2008; Unal et al., 2008). During the cell cycle, Smc3 is acetylated shortly after the passage of the replication fork and the capturing of the newly formed second chromatid by cohesin (Dauban et al., 2020; Rolef Ben-Shahar et al., 2008; Unal et al., 2008). Mechanistically, the acetylation of Smc3 inhibits both the ATPase activity of cohesin and conformational changes (Camdere et al., 2018). Recently, it has been suggested that acetylation also induces cohesin dimerization (Shi et al., 2020a). Either way, we suggest that, as a result of the acetylation, cohesin loop extrusion is fully turned off, likely by modifying the ATPase activity at the SMC head. As a result, the cohesin–DNA complex is stabilized in a tether-promoting conformation (Fig. 3), such as the recently described conformations that enable the passive capture of the chromatids (Collier et al., 2020; Higashi et al., 2020; Sedeno Cacciatore and Rowland, 2019; Skibbens, 2015; Srinivasan et al., 2018). We imagine that even after the dissociation of Scc2, conformational changes in cohesin occur at a slow rate as they may be essential to capturing the newly formed sister chromatid. As soon as this goal is achieved, the acetylation of Smc3 ensures that loop extrusion is fully halted and cannot be reactivated by the reassociation of Scc2 with cohesin.

In contrast to yeast, which contain a sole Eco1, vertebrates have two Eco1 orthologs, called Esco1 and Esco2 (Hou and Zou, 2005), whose specific activities were reported recently (Alomer et al., 2017). Similar to canonical yeast Eco1, Esco2 plays a key role in the establishment of mitotic cohesion, whereas Esco1 acetylates cohesin throughout interphase, but has little effect on mitotic cohesion and is non-essential for mitosis (Alomer et al., 2017). It is important to note that yeast cells are significantly different from vertebrate cells in several aspects. First, yeast lacks the CTCF homolog. Second, while cohesin is essential for mitotic cohesion, it plays a minor role in the regulation of gene expression in yeast (Bose et al., 2012; Skibbens et al., 2010). Cohesin is important for yeast chromatin organization into TAD-like structures in interphase. However, cohesin is dispensable for maintaining their stability, which instead depends on the forkhead family proteins Fkh1 and Fhd2 (Alipour and Marko, 2012). Therefore, we suggest that the acetylation of cohesin by Esco1 evolved as a safety mechanism in vertebrates to maintain CTCF-dependent extrusion suppression of cohesin in TADs. This activity is dispensable in yeast cells as we assume that the evolutionary divergence of Esco1 and Esco2 is associated with the acquired function of cohesin in TAD stabilization.

In higher eukaryotes, the composition of cohesin complexes can differ depending on the identity of the HEAT repeat subunit, either Stag1 or Stag2 (homologs of yeast Scc3), and the presence of the regulators Pds5A or Pds5B (homologs of yeast Pds5) (Losada et al., 2005, 2000; Sumara et al., 2000). Each one of these protein pairs is mutually exclusive, meaning that cells contain subpopulations of cohesin complexes that differ in their STAG and Pds5 composition. The functional importance of these alternative complexes has been associated with differential localization and functions (Canudas and Smith, 2009; Carretero et al., 2013; Casa et al., 2020; Kojic et al., 2018; Morales et al., 2020; Viny et al., 2019; Zhang et al., 2013). We suggest that an overlooked property of these alternative subunits may affect cohesin loop extrusion processivity, which defines the residency time on chromatin. This property will eventually determine the final size of the extruded loop by allowing differential association times of cohesin with chromatin. To this end, the effect of alternative cohesin subunits on cohesin processivity has not been addressed in detail; it stands as a future challenge to provide biochemical systems that can be used to test loop extrusion activity for complexes with alternative subunit composition to either confirm or reject this idea (Davidson et al., 2019; Kim et al., 2019). Until further experimental data is collected, we hypothesize that high processivity enables the continuous extrusion that is required for TAD formation in interphase until its suppression by CTCF, whereas low processivity results in frequent dissociation of cohesin from the chromatin, as has been observed in mitotic cells (Austin et al., 2009; Bernard et al., 2008; Gause et al., 2010; Gerlich et al., 2006). In support of this model, a recent study revealed differences in the binding of cohesin–Stag1 and cohesin–Stag2 to DNA, with the former lasting hours versus only minutes in case of the latter (Wutz et al., 2020). Furthermore, a distinct role has been reported for Stag1 in the formation of the TAD stem, whereas Stag2 is involved in inter-TAD interactions (Cuadrado et al., 2019). Further studies are needed to unveil the differential functions of the cohesin alternative subunits in tethering and TAD formation.

Loop extrusion is also considered a component of the mitotic chromosome assembly mechanism by the cohesin-related SMC complex condensin (Banigan et al., 2020; Ganji et al., 2018; Kim et al., 2020; Terakawa et al., 2017). Several computational simulations have suggested that condensation is initiated by individual condensins that are distributed along the chromatin. As chromatin condensation continues, the individual condensins aggregate to form the chromosome scaffold (Fig. 4) (Goloborodko et al., 2016). Notably, similar accumulation of bacterial SMC, which is considered to be evolutionary ancestor of condensin, has been reported (Rajasekar et al., 2019; Wang et al., 2017, 2018). We propose that inter-condensin contacts at the chromosome scaffold suppress loop extrusion. This suppression can be mediated by interactions between the HEAT repeat subunits, which is similar to the suppression of loop extrusion by cohesin that is mediated by the CTCF–WAPL interaction (Haarhuis et al., 2017; Makrantoni and Marston, 2018). However, we cannot exclude the possibility that other contacts involving the hinge or the coiled coil domains either mediate or contribute to the suppression. Alternatively, condensin aggregation may contribute to the dynamic crowding of the nuclear nanoenvironment and chromatin viscoelasticity that hinders loop extrusion activity by slowing down dynamic conformational changes (Kschonsak et al., 2017; Shim et al., 2020; Vivante et al., 2020). Notably, condensin activity depends on the phosphorylation of its HEAT repeat subunits by Cdc2 (known as Cdk1 in mammals) (Abe et al., 2011; Kimura et al., 1998; Takemoto et al., 2004), and their dephosphorylation may, therefore, serve as an off switch, corresponding to acetylation of Smc3 in cohesin.

The third eukaryotic SMC complex is Smc5/6, which is involved in DNA replication and repair (Aragon, 2018; Palecek, 2018). However, its molecular activity is not fully understood, and loop extrusion activity has not yet been shown in this complex. Another fundamental difference between Smc5/6 and the other SMC proteins is the presence of a docking site for the SUMO ligase Nse2 in the middle of the Smc5 coiled-coil region (Fig. 1). Nse2 sumoylates Smc5 and the kleisin subunit Nse4, which may affect the structure of the complex (Varejao et al., 2018). Finally, we recently suggested that the organization of Smc5 is different from that of other SMC proteins as it contains non-canonical breaks in the coiled coil domain (Matityahu and Onn, 2018). Taken together, the sequence and anticipated sumoylation-induced rigidity of Smc5/6 indicate that this SMC complex might be less flexible than cohesin and condensin. Accordingly, Smc5/6 may not have the structural flexibility required for the conformational changes that underlie loop extrusion (Higashi et al., 2020; Sedeno Cacciatore and Rowland, 2019). Thus, Smc5/6 is unlikely to have the ability to mediate loop extrusion, and only passively tethers chromatin.

Knowledge about the mechanism of cohesin activity has been accumulating rapidly, but along two separate paths – organization of interphase chromatin by loop extrusion and mitotic tethering of sister chromatids. Herein, we have provided a rational explanation of how these fundamentally different mechanisms could coexist within a single cohesin complex.

Previous work in the field identified the need to develop a model that will explain the co-existence of chromatid cohesion and loop extrusion in a single complex (Yatskevich et al., 2019). We suggest that loop extrusion activity represents a fundamental biochemical activity of SMC complexes that originated in their evolutionary ancestors. Various off-switches dependent on the intra- or inter-molecular interactions evolved, with the capability of controlling SMC complexes under different biological conditions, namely, cohesin–CTCF interactions in TAD boundaries and acetylation of Smc3 by Eco1. Extrapolating this mechanism onto other eukaryotic SMC complexes, condensin and Smc5/6, we have advanced two hypotheses. First, we argue that condensing-mediated regulation of loop extrusion is similar to that of cohesin, but here, the off-switch is condensin crowding at the scaffold of the mitotic chromosome. Second, we suspect that Smc5/6 might have either lost or repressed its extrusion activity and has thus become a passive tethering member of the SMC family.

A unified mechanistic model for cohesion and loop extrusion should also be considered in light of the different cohesin tethering models we present in Fig. 2. Although loop extrusion is mediated by a monomer, in which the chromatin is thread through two binding sites, a simple topological entrapment of the chromatin in the SMC ring is unlikely, as validated in recent studies that show compartmentalization and complex interactions of cohesin with DNA (Collier et al., 2020; Higashi et al., 2020; Liu et al., 2020; Srinivasan et al., 2018). However, this model does not exclude dimerization as a mechanism to suppress extrusion activity. In doing so, the interaction between monomers is expected to be weak. Thus, identification of dimers by methods that do not involve crosslinking will be difficult and might explain why the genetic evidence of dimerization has yet to be validated with molecular data.

Fluorescence resonance energy transfer (FRET) and similar in vivo spectroscopic approaches will help to identify cohesin dimerization in vivo. Furthermore, FRET could be used to explore the conformation dynamics of SMC complexes in live cells. This could be accomplished through dual fluorescent labeling of subunits in the same complex, thereby allowing dynamic extruders to be distinguished from silent tethers. Under this experimental setup, an oscillating on–off fluorescent emission is expected from extruding complexes, while a constant emission would suggest stable conformation complexes involved in passive tethering. While such experiments may well offer exciting results, their success will depend upon the technical ability to differentiate between extruding and tethering complexes within a noisy cellular environment.

Parts of the model we present here can be challenged experimentally using in vitro loop extrusion assays for condensin and cohesin (Banigan et al., 2020; Davidson et al., 2019; Eeftens et al., 2017; Ganji et al., 2018; Kim et al., 2019; Kong et al., 2020; Terakawa et al., 2017). Current data is limited to complexes with a defined subunit composition. Measuring the biochemical properties of complexes with different subunit composition, as well as the effect of Smc3 modifications such as acetylation and modification-mimicking mutations [e.g. Smc3 K112Q K113Q (Unal et al., 2008)] will determine the effects of these factors on loop extrusion activity. These experiments will also reveal whether there are indeed any differences in processivity that depend upon subunit composition, as we suggest here. The mechanism by which loop extrusion is suppressed can be challenged by isolating mutants that induce hyper-condensed mitotic chromosomes, such as those reported in two recent studies. The mutation of a conserved leucine residue to a valine residue in Smc4 ATPase results in hyper-condensation (Elbatsh et al., 2019). We suggest that the phenotype of the mutated condensin can be explained by resistance to the turn-off signal. If correct, the loop size would be expected to be increased, and indeed, about a twofold increase is found in cells carrying this mutant condensin. Previously, we reported on chromosome hyper-condensation in cells infected by the hepatitis C virus (Perez et al., 2019), where viral proteins may dysregulate the condensin off-switch. However, whether this hypercondensation phenotype is related to misregulation of condensin extrusion is yet to be explored. Finally, we expect that Smc5/6 will not extrude DNA in in vitro assays.

Previous experiments on cells in which cohesin has been depleted of STAG1/STAG2, as well as auxiliary factors, such as Pds5 and Esco1/Esco2, did not assess the associated changes in the size of TADs and other structural properties within them. Further Hi-C studies of cells, an experimental technique in which the contact frequency of chromosomal regions is being determined on the genome scale, for systems where these subunits are depleted, could elucidate the differences between complexes with alternative subunits. The development of new advanced imaging technologies to monitor chromatin organization might allow these changes to be assessed within live cells and for their kinetics to be directly measured. Such experiments are anticipated to enhance our understanding of the mechanism that governs SMC complexes and the way they shape the 3D structure of the genome.

I.O. thanks Shay Covo for critical reading and commentary on the manuscript, and the members of the Onn laboratory for their discussions, as well as Steve Spencer for his editorial review and corrections. The authors apologize to colleagues whose work was excluded due to space limitations and the rapid accumulation of new data in the SMC field.