15/n
Also, check out the pre-prints by @fedeteloni.bsky.social from @gerlichlab.bsky.social and by @charlesyeh.bsky.social from @jcornlab.bsky.social with other cool insights about the homology search!
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...

14/n
I also want to thank all the other authors: Daniel Nguyen, Violetta Karwacki-Neisius, Andrew G. Li, Roger Zou, Franklin Aviles-Vazquez and Masato Kanemaki.

And huge thanks to Yang Liu who made vfCRISPR and to
@nucleosomezky.bsky.social and @rezakalhor.bsky.social for discussions!

13/n
Incredibly thankful to my mentor, Taekjip Ha, who supervised and mentored me on this project,

to my co-mentor, Ralph Scully, who designed the mESC exps and mentored me on HR,

and to co-first authors Adam Rybczynski and Namrata Nilavar, who helped make this possible!

12/n
Our model, in a nutshell: cohesin drives homology search via 1D scanning.

During HR, a RAD51 filament locally scans the sister chromatid, but this search could be unproductive (e.g., because the donor is far).

Cohesin loops would then facilitate long-range scanning to help find a donor!

11/n
Does loop-extruding cohesin (Nipbl) regulates HR?

In the short-range HR-GFP reporter, Nipbl depletion mildly reduces HR (45%). But in the long-range reporters, Nipbl depletion substantially reduces HR (75%).

Thus, Nipbl is critical for HR when long searches are required.

10/n
We noticed that the RAD51 domain is constrained by TAD boundaries, suggesting a role for cohesin in mediating the homology search.

Indeed, acute degradation of the cohesin unloader WAPL, known to yield elongated loops, resulted in a broader search.

9/n
So, we generated monoclonal mES cell lines with the donor at +441 kb and +563 kb from the break.

When we induced the DSB…we found a beautiful RAD51 peak at the exact location of the donor!

The RAD51 chromatin domain is thus capturing the homology search.

8/n
We used an HR-GFP reporter system. Upon inducing a DSB, a broken GFP gene can be repaired by HR via use of a 5’-truncated donor, giving a functional GFP.

We reasoned that, upon DSB induction, RAD51 should show a peak at the donor, indicative of a successful homology search.

/n
But why does RAD51 spread to such long distances (~0.5 Mb) from the break?

End resection was constrained to ~ 5 kb, so the broad RAD51 domain does not come from the RAD51-ssDNA filament.

We hypothesized that the RAD51 domain reflects the homology search. We tested this.

6/n
Motivated by this finding, we performed a second time-course experiment, where we looked at the dynamics of RAD51 recruitment, the core HR factor.

RAD51 dynamics mimic break-anchored loop formation, suggesting a causal relationship between loops and HR.

5/n
To answer this, we induced DSBs in cell-cycle-synchronized populations.

Cells in late S/G2 phases, which are HR-proficient, showed break-anchored loops. But cells synchronized in G1, which are HR-deficient, did not show the loops.

Break-anchored loops are HR events!

4/n

Time-course Hi-C, enabled by our light-activated very fast CRISPR, showed that break-anchored loops are late repair events.

Loops form after γH2AX domain, ruling out previous models where break-anchored loops propagate γH2AX at early stages.

What is the role of such loops?

3/n
Using our multi-target CRISPR system – which induces hundreds of DNA double-strand breaks (DSB) on demand – and Hi-C, we found that Cas9 breaks act as anchors for chromatin loops.

Cohesin is recruited to the breaks and is required for break-anchored loop formation.

2/n
The homologous recombination (HR) pathway repairs DNA lesions by copying intact sequence information from a DNA template…

but how does a damaged DNA find a correct DNA template in the 3D genome?

We provide some answers to this Q, known as the homology search problem.