Sangamo Biosciences Inc (SGMO)

[jondoeuk]

Seamless will present data at ASGCT this month.

The first set of data is essentially describing how they are industrialising recombinase engineering. They started with massively diverse recombinase libraries from a pool of over 10,000 previously programmed libraries, and looking at useful traits across variants.

On top of this, undertaken alternative evolution with alternating selection. Instead of a single selection pressure, they are using switch selection schemes to refine both activity and specificity. Also, a dual screening strategy (both pooled screening and arrayed screening). The goal is to balance throughput and resolution.

In addition, adding ML in the loop (not just at the start), with protein language models to suggest new variants, which helps guide the next rounds of evolution.

Activity and specificity are further refined through orthogonal protein engineering (e.g., zinc finger fusions).

The key claim is ''target-matched recombinases within months.'' If true, that's huge. Because historically reprogramming recombinases has been slow and bespoke.

The second set of data is on the development of a suite of zinc finger fusions. They are layering domains into recombinases in two distinct ways.

The first is internal ZF fusions. In this, the ZF is inserted inside the recombinase structure. Activity becomes dependent on the recombinase recognition site and ZF binding nearby. Editing happens when both conditions are met.

The second is terminal ZF fusions. In this, the ZF is added to N- or C-terminus in order to help recruit the enzyme to DNA (like bringing the enzyme closer to its target and it works better). Doing so resulted in up to 10× activity increase in tyrosine recombinases and ~4× increase in large serine recombinases.

They didn't just add zinc fingers either. They mapped insertion tolerance with pentapeptide scanning mutagenesis (finding positions that don't break the protein), optimised linker design (length/rigidity/flexibility) and tested across classes, which suggests broad applicability.