Future work#

Getting assembly on its feet#

Everything else is downstream of this. The immediate queue, in order of how cheaply it discriminates:

1. Direct C-terminal His tag, no linker. The fastest test of the tag-occlusion hypothesis, and the construct that published work reports as assembling.
2. Negative-stain electron microscopy on crude lysate. Bypasses the purification entirely. If cages are present in the cell and being lost during handling, this sees them; SEC and DLS on purified material cannot distinguish "never assembled" from "assembled and lost".
3. Untagged shell with a tagged cargo. Pull on the cargo rather than the shell, so that assembly is not selected against by the affinity step.

Broadening the cargo range#

The λN·boxB handle is not specific to a guide RNA. Any transcript carrying a boxB hairpin becomes a substrate, so a shell evolved for guide capture should transfer directly to mRNA packaging — which is the delivery application that motivated the project. The obvious test is to re-run the selection with the hairpin moved from the sgRNA scaffold onto an unrelated transcript and ask whether the evolved shell still captures it.

Whether that generalises is an empirical question with a real chance of answering "no": evolution under a single selection tends to find the cheapest solution, and the cheapest solution here may be specific to the geometry of the dCas9 complex.

Orthogonal selections#

The OR gate that makes the landscape traversable also makes the output ambiguous. Two follow-ups tighten it.

An AND gate, later. Once shells exist that capture either half, the handles can be split across two selective agents so that both must be sequestered simultaneously. Starting there would have been hopeless; arriving there from a partially evolved population is plausible.

A nuclease challenge. Tetter and co-workers escalated stringency by shrinking the nuclease and lengthening the exposure — benzonase, then RNase A, then RNase T1, from one hour to four. That selects for protection, not merely binding, which is a different and more demanding property than anything our current circuit measures.

Dealing with escape#

The failure mode to plan for is enrichment of cells that restored resistance without improving encapsulation. Two measures are already designed in — holding dCas9 and the sgRNA off the mutagenised plasmid, and re-sequencing the selection plasmid each passage. A third would help: a counter-selection that periodically requires the repressor to be functional, catching populations that have quietly lost it.

Toward delivery#

The end state is a shell that packages a ribonucleoprotein and delivers it to the cytosol of a mammalian cell. The delivery half of that problem has been solved separately for this scaffold, using a pH-sensitive intein to detach cargo in the endosome and a fusogenic peptide to escape it. Nothing in that architecture conflicts with what we are evolving — but it assumes a shell that assembles, which returns to the top of this page.