Thread by @Anders_S_Hansen

(1/n) excited to share our new preprint:

We visualize and track a CTCF/cohesin loop live using super-res live-cell imaging.

CTCF/cohesin loop is rare (~3-6.5% of time) and very dynamic (~10-30 min lifetime), ruling out models of TADs as stable loops.

biorxiv.org/content/10.110…

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE (2/n) We focus on: Fbn2 TAD in mESC since simple and verified by @deWitLab to be CTCF-dependent.

Using genome-editing, we fluorescently label the 2 CTCF anchors (C36) and generate pos and neg looping controls (C27 and C65).

Micro-C verifies that tagging does not perturb loop

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (3/n) We can now track loop dynamics in live cells using 3D distance as readout (slightly old movie example for twitter:) )

CTCF anchors separated by 505 kb.

low 3D dist ~ loop state more likely

high 3D dist ~ loop state unlikely

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (4/n) with system we ask: what's the role of CTCF and cohesin?

We AID-tag and deplete cohesin, CTCF, & WAPL.

By super-res live-cell imaging and by Micro-C, our results are consistent with loop extrusion by cohesin, CTCF being the boundary factor, & WAPL the cohesin unloader.

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (5/n) from polymer consideration and steady-state variance of wt (C36) and ΔCTCF to the fully unextruded ΔRAD21 state, we can roughly estimate the extruded fraction:

On average, ~61% of DNA in Fbn2 TAD is inside extruding cohesins, but this drops to ~46% in ΔCTCF.

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (6/n) What's the lifetime & loop fraction?

To call loops in trajectories, we develop Bayesian Inference of Looping Dynamics (BILD).

We validate BILD on polymer sims with loop extrusion and noise.

We define loop as sustained CTCF/cohesin complex state that disappears in ΔCTCF.

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (7/n) Fbn2 loop is dynamic (median ~10-30 min) and rare (~3-6.5%).

During typical ~12 hr mESC cell cycle:
- loop state forms ~1-2 times
- cumulative lasts ~20-45 min
- remaining ~11.5 hours are fully unlooped or partially extruded

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (8/n) But are rare (~2-6% of time) loops consistent with corner peaks in Hi-C/Micro-C?

Yes! We can simultaneously reproduce
1) Micro-C maps
2) ~2-6% looped fraction
3) short loop lifetime
using polymer simulations with loop extrusion.

Often >1 cohesin holds together the loop!

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (9/n) Merging our sims and data, we can fully parameterize the Fbn2 TAD!

The fully looped (~2-3%) and the fully unlooped (~6%) states are very rare & transient.

Instead, TADs overwhelmingly exist in partially extruded state (~92%), which may be functionally important state.

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (10/n) Wonderful tri-lab collab with @zechnerlab and @leonidmirny

Project lead by @labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 who put it together despite lab getting shut-down with pandemic 6 weeks after opening!

Collab with Gina Dailey, @CCattoglio @Stanley_TH

@labronic_mike @hugoresearch @SGrosseHolz @AsmitaJha16 @CCattoglio @Stanley_TH @leonidmirny @zechnerlab @hansen_lab @MITdeptofBE @deWitLab (11/n) Also thanks to @TjianDarzacq for hosting early stages of this project and to @Jeff_m_Alex for sharing RNA-destabilization elements.

We'd very much welcome critical feedback and suggestions for improvement.

More info at biorxiv.org/content/10.110…