The SMC1-SMC3 cohesin heterodimer structures DNA through supercoiling-dependent loop formation

Cohesin plays a critical role in sister chromatid cohesion, double-stranded DNA break repair and regulation of gene expression. However, the mechanism of how cohesin directly interacts with DNA remains unclear. We report single-molecule experiments analyzing the interaction of the budding yeast cohesin Structural Maintenance of Chromosome (SMC)1-SMC3 heterodimer with naked double-helix DNA. The cohesin heterodimer is able to compact DNA molecules against applied forces of 0.45 pN, via a series of extension steps of a well-defined size ≈130 nm. This reaction does not require ATP, but is dependent on DNA supercoiling: DNA with positive torsional stress is compacted more quickly than negatively supercoiled or nicked DNAs. Un-nicked torsionally relaxed DNA is a poor substrate for the compaction reaction. Experiments with mutant proteins indicate that the dimerization hinge region is crucial to the folding reaction. We conclude that the SMC1-SMC3 heterodimer is able to restructure the DNA double helix into a series of loops, with a preference for positive writhe.

The initial extension L i was determined from tether calibration data acquired before addition of protein to the experiment. The lag time was determined to be the time between addition of protein (T 0 , Fig. 2) and the first identifiable step (T 1 ). The final extension L f was determined from the terminal region of the time series; the time T 2 at which the final extension was reached was determined to be the time at which the extension first reached within 50 nm of L f . For extreme cases where there was only one step in the experiment, this amounted to determining T 2 to be the time at which that step occurred. Compaction fraction, lag time, and folding rate showed a marked dependence on DNA topological state. DNA with enough positive torsional stress to reach the corresponding overwinding bucking point (ΔLk = +12 cases of Fig.   S2) generated the most robust reaction with a high compaction fraction (Fig. S2B), shortest lag time (Fig. S2C) and the fastest reaction rate (Fig. S2D). For nicked (torsionally unconstrained) molecules, compaction proceeded similarly to those of positive supercoiled DNA. Remarkably, un-nicked, relaxed molecules (ΔLk = 0) not only had the lowest degree of compaction, but also the longest lag time and the slowest reaction rate. Thus ΔLk = 0 was the least favorable topology for compaction by SMC1/3 among the topological states studied. DNA with enough negative torsional stress to reach its buckling point for plectoneme formation (ΔLk = -15) reached about the same compaction fraction as observed for positively supercoiled DNA (≈60%, Fig. S2B), but with a much longer lag time and a slower reaction rate ( Fig. S2C-D).

Increasing the amount of plectonemically supercoiled DNA does not enhance the SMC1/3 compaction reaction
Given that positively supercoiled DNAs with ΔLk = +12 led to reliable and fast compaction reactions, we examined the effect of varied levels of positive supercoiling. We note that for the 0.45 pN force used, as ΔLk is increased from 0 to +11, the DNA extension only changes slightly; the ΔLk goes largely into DNA twisting, increasing the torque in the molecule.
As ΔLk is increased to values ≥12, plectonemic supercoiling occurs, with almost no change in DNA twisting (constant DNA torque): instead the additional ΔLk increases DNA writhe. Thus by choosing different ΔLk values we could determine whether DNA twist or plectonemic writhe was more important to the compaction reaction.
We carried out experiments for ΔLk = +5, +12, +20 and +30. ΔLk = +5 has about half of the twist of +12, while +20 and +30 have about the same twist as +12 but more plectonemic crossings (writhe). All these levels of supercoiling led to compaction in every experiment (Fig.   S3A). Folding fraction is calculated as a fraction of the extension of plectonemically supercoiled DNA prior to protein addition. Experiments with ΔLk =+12, +20 and +30 had nearly the same compaction fraction, lag time and folding rate ( Fig. S3B-D) indicating that increased plectonemic writhe does not accelerate the compaction. However, experiments with ΔLk = +5 showed a lower compcation fraction, a longer lag time, and a slower reaction rate. This indicates that the compaction by SMC1/3 is driven primarily by DNA torsional stress (which increases steadily before plectonemes start to form at Lk=+11 (1), rather than by plectonemic interwinding (during plectoneme formation DNA torque is nearly constant).

Nucleotide dependence of DNA folding by the SMC1/3/Scc1-C complex
Previous studies have shown that the Smc1/3 heterodimer alone has very low ATP hydrolysis activity (2,3), therefore we did not pursue studies of ATP binding on SMC1/SMC3-DNA interactions. However, addition of the C-terminal domain of Scc1 (Scc1-C, residue 269-566), which binds to SMC1, is known to stimulate SMC1/3 ATPase activity (3,4). We therefore To further examine the role of ATP occupation at the heads, we also carried out experiments using the wild-type SMC1/3 heterodimer bound to a Scc1-C point mutant (L532R) which has suppressed ATP binding by SMC1 and suppressed hydrolysis of ATP bound to SMC3 (3). The results of using SMC1/3/Scc1-C-L532R+ATP (1 mM ATP, rightmost bars of Fig. S11) were reactions which were intermediate in compaction fraction, lag time, and reaction rate for SMC1/3/Scc1-C with no ATP (fastest and most compaction) and SMC1/3/Scc1-C+ATPS (slowest and least compacting). Since for SMC1/3/Scc1-C-L532R+ATP, nucleotide binding occurs at SMC3 but with little or no hydrolysis, while with SMC1/3/Scc1-C+ATPS, binding but no hydrolysis occurs at both SMC1 and SMC3 (3), we conclude that nucleotide binding tends to suppress both the rate and extent of the DNA-compaction reaction by SMC1/SMC3/Scc1-C.

Two-DNA tethers for braiding experiments
Beads tethered by two or more of pFOS-1(9.6 kb) DNA (nicked, with no torsional stiffness) show a rotation-dependent extension distinct from a single supercoiled DNA, characterized by a shape symmetric around Ca=0, a sharp peak at Ca=0 corresponding to the insertion of the first left-or right-handed crossing (Fig. S9A, B), and a spring constant approximately double that a single DNA (crudely, a requirement of a force of approximately 0.2 pN to achieve a 50% extension of the naked braid) (5-7). Constructs satisfying all of these constraints were accepted as two-molecule tethers.