Topological entrapment of DNA by cohesin

SMC (structural maintenance of chromosomes) complexes, which include condensin, cohesin, and the SMC5-SMC6 complex, are major components of chromosomes in all living organisms from bacteria to humans. Ring-shaped protein machines, they are powered by ATP hydrolysis, and are responsible for organising and shaping chromosomes, controlling a plethora of chromosomal activities. Most SMC complexes are in dynamic contact with the chromosomes, undergoing repeated cycles of binding and dissociation. Notable exceptions are the cohesin molecules that hold sister chromatids together following DNA replication, which are enduringly stabilised on DNA.

The mechanism by which the SMC complexes trap and release DNA has been of great interest since their discovery some twenty-five years ago. However, the large sizes of the proteins within the multisubunit complexes mean that it has been extremely difficult to produce the pure material required for the biochemical and structural studies needed to elucidate the molecular details.

In 2014, this practical challenge was finally resolved by Yasuto Murayama and Frank Uhlmann (Murayama and Uhlmann, 2014), who succeeded in reconstituting the fission yeast cohesin complex in vitro, in the process settling a long-standing dispute by showing that the cohesin ring topologically encircles the DNA. This work set the stage for a definitive investigation of the molecular sleight of hand by which cohesin, and by extension the other SMC complexes, can move DNA in and out of a closed circular structure.

In this quinquennium, Uhlmann and his group have published a series of groundbreaking papers following on from this work. After a detailed biochemical description of the process (Murayama and Uhlmann, 2015), they went on to describe how a cohesin ring captures a second DNA strand and locks shut around sister chromatids in a cell-cycle dependent manner (Murayama et al, 2018). Most recently, in a collaboration with Alessandro Costa, they have used cryo-EM and biochemical techniques to provide a molecular picture of how ATP-fuelled structural changes drive entry of DNA into the cohesin ring (Higashi et al, 2020).

Cohesin, like all SMC complexes, is characterised by its striking ring shape, made up of long stretches of flexible coiled-coil of the two SMC subunits, which in fission yeast are called Psm1 and Psm3 (Smc1 and Smc3 in budding yeast). The coiled-coil segments are connected at one end by the hinge, a stable dimerisation interface; at the other end lie ATPase domains, known as heads, which dimerise in the presence of ATP, sandwiching two molecules of ATP between them. The heads are held together through this ATP-dependent dimerisation, and by another member of the cohesin complex, a kleisin subunit. While the SMC dimer is largely symmetric, kleisin makes asymmetric contacts with the SMC heads: elements close to the kleisin N terminus form a triple helix with the “neck” of the Psm3 head, and its C terminus associates with the Psm1 head.

Kleisin acts as a docking platform for other important interactions, firstly with the final cohesin subunit Psc3 (Scc3 in budding yeast), which regulates chromatin association. Using its flexible middle domain, kleisin can also bind the auxiliary proteins Pds5 and Mis4. Mis4, with its stable binding partner Ssl3, promotes cohesin loading whereas Pds5, with its substoichiometric binding partner Wapl, dissociates cohesin from chromatin.

In their 2015 paper, Murayama and Uhlmann reconstituted this entire cohesin system, including the auxiliary proteins, in vitro. They proposed and validated a model for DNA entry and exit in which both loading and unloading follow a similar route through a set of double gates at the head ends of Psm1 and Psm3. One gate opens when the kleisin N-terminus loses its grip on the neck of Psm3, aided by the auxiliary unloader protein, Wapl. The other gate, which requires ATP hydrolysis for its operation for both entry and exit, disengages the Psm1 and Psm3 ATPase head domains from each other. Crucially, only one gate at a time can be open.

DNA transit in both directions requires two DNA-sensing lysine residues on the Psm3 head that trigger ATP hydrolysis, necessary for opening the head gate; if these residues are singly or both mutated to glutamines, DNA loading is either partially or wholly obliterated. These lysine residues point inwards into the cohesin ring, positioning them neatly to initiate DNA exit. However, it was less obvious how these same lysines could promote DNA entry, as double-stranded (ds) DNA is too stiff to spontaneously bend sharply enough to reach the lysine DNA sensor from the outside. Murayama and Uhlmann proposed that the flexible cohesin ring could be deformed and flipped to expose the lysine residues, perhaps aided by the Mis4 and Ssl3 cohesin loader complex, which they were able to show makes multiple contacts around the ring circumference that could facilitate a large conformational change. Once incoming DNA engaged the lysine sensors on a folded cohesin ring, the same trajectory through the kleisin N-gate and the head gate would lead to DNA entry.

Simply binding one DNA molecule, whilst decorative, is functionally useless, and Murayama and Uhlmann, now collaborating more distantly after Murayama started his own lab in Mishima, Japan, turned their attention to how cohesin might be able to load two DNA molecules (Murayama et al, 2018). Together, their groups showed that once cohesin has topologically loaded onto one DNA, it can recruit another DNA into the same ring. Capture of the second DNA likely occurs using the same double gates as the initial loading reaction, as it requires ATP, the Mis4/Ssl3 loader and the DNA-sensing lysines on the Psm3 head, but there is one crucial difference: the second piece of DNA must be single-stranded, presumably to give it more flexibility to negotiate around the resident dsDNA molecule. Single-stranded (ss) DNA capture required the presence of an initial dsDNA molecule, and was relatively labile, only being stably tethered within the ring following conversion to dsDNA by DNA synthesis.

The characteristics of capture of this second DNA in vitro have an intriguing parallel with DNA geometry at replication forks. Replicated dsDNA on the leading strand is found next to periodically extended ssDNA on the lagging strand. Cohesin loaded onto the leading strand is therefore perfectly positioned to capture the ssDNA of the lagging strand. Shortly afterwards, Okazaki fragment synthesis could convert cohesin’s fragile ssDNA embrace into stable sister chromatid cohesion. Cohesin can be detected at replication forks, consistent with this model.

How is the cohesin on sister chromatids locked into place? During DNA replication, the two crucial lysines in the head domain of Psm3, required for DNA sensing, are acetylated and inactivated. With the sensor inactivated, the gates are effectively locked, a prerequisite for stable sister chromatid cohesion, although it is not yet clear how acetylation is only triggered by entrapment of the second DNA molecule. At anaphase, the sister chromatids are released by brute-force: kleisin is cleaved by separase, and the complex is destroyed.

This sequential DNA capture appears tailored to trap sister DNAs as they emerge from the replication fork, but it is possible that cohesin and other SMC complexes may operate by a similar mechanism in other situations. There is some evidence for ssDNA structures in other situations where cohesin is used, such as in double-strand break repair and chromosome looping, but further study is necessary to determine if this mechanism is universal.

Uhlmann had always hesitated to take on the challenge of visualising the cohesin complex; the same difficulties inherent in purifying large individual subunits for biochemical investigation also apply to structural biology, where the flexibility of the protein-DNA complex is an added complication. However, the move into the Crick meant that he was able to set up a close collaboration with Alessandro Costa, an electron microscopist whose work focusses on studying the DNA replication machinery. In 2020, the two groups used a combination of cryo-EM, biochemical and biophysical techniques to describe how ATP-fuelled structural changes of the cohesin complex drive the DNA entry reaction into the cohesin ring (Higashi et al, 2020). 

The key to the success of the project was first author Torahiko Higashi’s discovery that a mutant cohesin complex, EQcohesin, whose ATPase was hydrolysisdeficient, could be trapped in a transition state where the heads had engaged but were unable to complete DNA entrapment. Seeking to replicate this state with normal cohesin, Higashi found that substitution of ATP with the non-hydrolysable analogue ADP•BeF₃‾ generated the same frozen complex, which was able to stably bind linear DNA, something cohesin is normally unable to do. This so-called “cohesin gripping state” was amenable to structural analysis.

Cohesin was assembled onto linear dsDNA in the presence of the Mis4-Ssl3 loader and ADP•BeF₃‾, and low- and high-resolution (3.9 Å) structures of the purified gripping state complex were derived by the Costa lab using negative staining and cryo-EM respectively. The resulting atomic model covered the two Psm head domains and adjacent coiled coil, bound to the kleisin N- and C-termini, as well as the Mis4 loader and 32bp of DNA. 

The structure gave unprecedented insight into mechanism. Mis4 was clamping the DNA onto the ATPase head domains, making widespread contacts with both Psm1 and Psm3. To the group’s surprise, the DNA was encircled by Mis4 and Psm3, trapping it in the lumen of a completely distinct protein ring. This gripping state was dependent on ATP head engagement, but not ATP hydrolysis; the subsequent interaction, which does require ATP hydrolysis, topologically embraces the DNA within the main cohesin ring.

The structure enabled accompanying biophysical experiments to draw conclusions about the order in which the kleisin N-gate and head gates opened, and the energy source driving the reactions, and also explained the crucial role of the lysine DNA sensors in the Psm3 head domain.
To enter the cohesin ring, the DNA first passes the kleisin N-gate, which is opened by ATP binding and head engagement, but not hydrolysis. The kleisin N-gate has already shut in the gripping state: if the kleisin N- and C-termini are crosslinked to generate an artificially closed ring, this can be shown to be topologically encircling the gripping state DNA. This crucial insight was obtained using a bifunctional chemical crosslinker, custom-made for the researchers in the Crick’s Peptide Chemistry STP.

Energy for shutting the kleisin N-gate is provided by DNA binding, but entry through the head gate requires ATP hydrolysis. It is here that the lysine DNA sensors come into play. Both lysines are instrumental for ATP hydrolysis, as they sit on a connecting loop that reaches down to contact the γ-phosphate of ATP. However, the cryo-EM structure showed that in the gripping state, one lysine, K106, also engages DNA, and the other, K105, contacts an acidic patch on the Mis4 loader. ATP hydrolysis takes place only when the lysine sensors register that this conjunction of DNA, loader and ATP has occurred, allowing the head gate to open and the DNA to pass into the cohesin ring. In sister chromatid cohesion, acetylation of these lysines removes the ability of the head gate to sense DNA, so it can never open.

While the structure was extremely informative, it only revealed the path of a short segment of DNA through the cohesin complex during one part of the entry process. To track the movement of a longer DNA molecule into the cohesin ring, Higashi drew on a collaboration with the Crick Proteomics STP to invent a new technique, DNA-protein crosslink mass spectrometry (DPC-MS). He designed a DNA probe in which amine-dUTP was incorporated in place of dTTP, and decorated it with a photoactivatable crosslinker. Complexes were assembled with this probe and UV-irradiated to induce DNA-protein crosslinking, and the contact points on the proteins identified by mass spectrometry following peptide fragmentation. These data could then be combined with the structural information to see the footprint of the DNA on the cohesin complex.

Using this approach, the group was able to show that the initial DNA contact was essentially as a linear rod, with the DNA approaching Mis4 and the ATPase heads from the top. New contacts with the N-terminal tail of kleisin were also revealed; a further series of experiments showed that this tail plays a crucial role in guiding DNA through the kleisin gate.

Entry and exit of DNA from the cohesin ring is shown by these papers to be a multistep, rigorous process, as might be expected for something so closely linked to maintaining genome stability. In addition to studying the role of cohesin and other SMC complexes in crucial aspects of chromatin organisation, Uhlmann, with his colleague John Diffley, next aims to biochemically reconstitute the complete sequence of events at a DNA replication fork. Their aim is to fully understand how chromosome replication during S-phase is coupled to the establishment of sister chromatid cohesion, to ensure faithful segregation of the replicated genome to daughter cells during cell division.