New perspective on how loop extrusion mechanism works
For the 1st question:
It was shown before that C-terminus of CTCF protein interacts with cohesin complexes (probably with condensin too). So, TAD-inward motifs leads to a CTCF-cohesin interaction, if CTCF is bound, of course.
For the 2nd question:
In this paper, they proposed that a unidirectional movement can be achieved if two heads (subcomplexes) have different DNA-binding affinities. The weaker one will move forward upon separation, and the other will follow. However, I couldn’t find any explanation for a possible reverse movement, i.e. in the case of loop dissociation.
For the 3rd, I don’t have an answer too :)