Researchers engineer programmable gene activation cascades

Each proGuide carries not only its own spacer sequence but also a 23-bp Cas9 target site (CTS) embedded within its DNA. This CTS serves as the recognition sequence for the previous guide in the cascade. When that earlier guide becomes active, it binds the CTS and directs Cas9-VPR to cut at that location, excising the Pol III terminator and thereby converting the downstream proGuide into a functional matureGuide. In other words, the CTS provides the address label that determines which proGuide gets activated next, allowing the system to link individual guide RNAs into a programmed, stepwise chain.

Early versions of the design suffered from polymerase read-through, but the authors resolved this by extending the poly-T tract and introducing a second terminator, separated by a short linear segment, which suppressed background activity to almost undetectable levels. They then refined how Cas9 excises the terminator by arranging cut sites as inverted repeats, producing clean deletions and consistently functional matureGuides.

To ensure that each cascade step would fire reliably, the team computationally generated millions of potential Cas9 target sequences and experimentally winnowed them to a panel with strong, low-leak performance. Test cascades showed the expected gradual loss of efficiency with increasing step number but no catastrophic failure, indicating that the optimised parts could support deep sequential logic.

»Traditionally, you build a tissue culture environment, and then cells enter it and will eventually conform to that environment over time. Then, in order to get them to the next stage, you need to build a different environment,« explains Brad Merrill. He is a cofounder and Head of Innovation at Syntax Bio, and a co-senior author of the Science Advances paper.

Because those environments involve 'infinite variables', he argues, the field has struggled to achieve predictable outcomes. The proGuide system attempts something more direct, Brad Merrill points out:

»What we do is different. We’re giving them the instructions from the inside, and then they read that first set of instructions, conform, go to the next step, conform.«

That perspective reflects the work's founding motivation. Merrill says the team began with the view that, despite years of effort, there may be an underlying weakness in the approach to stem-cell differentiation, and that programmable intracellular instruction sets might bypass that weakness to improve the reliability of producing specific cell types.

Although questions remain about how to align cascade pacing with cellular rhythms or extend activity over longer developmental windows, the study demonstrates a practical means of encoding temporal logic directly into cells. By allowing CRISPR to follow developmental syntax rather than forcing abrupt leaps, the system provides a controlled framework for interrogating – and potentially improving – the transitions that underlie engineered cell fates.

Co-first authors of the study were Anupama Puppala and Andrew Nielsen, while Ryan Clarke and Bradley Merrill were co-senior authors – all at Syntax Bio Inc. in Chicago, Illinois. The research was published in Science Advances on 5 December 2025.