Background#

Protein nanocompartments#

Encapsulins are prokaryotic protein nanocompartments: shell proteins that self-assemble into icosahedral cages roughly 20–45 nm across, with triangulation numbers of T=1 (60 subunits), T=3 (180 subunits) or T=4 (240 subunits).¹ They are the only known class of protein cage that arrives from nature already solving the cargo problem — each shell protein carries an interior binding site that recognises a short cargo-loading peptide (CLP) on its native cargo, so loading is genetically encoded rather than chemically imposed.

That property has made them a popular engineering scaffold. Encapsulins have been used for antigen display, drug delivery, microscopy and enzyme encapsulation, largely because swapping the cargo is, in principle, a matter of appending five residues to whatever protein you want inside.

The permeability result that made this project plausible#

Conventional wisdom held that loading a pre-assembled cage requires taking it apart. Existing approaches relied on shell disassembly under harsh conditions followed by reassembly, on co-expression strategies requiring careful optimisation of expression ratios and timing, or on additional triggering components to initiate assembly.

QtEncapsulin turns out not to need any of that. Despite pores measuring only about 7.2 Å at their narrowest, cargo as large as 482 kDa is internalised simply by mixing it with intact shells. The proposed mechanism is not a global assembly–disassembly equilibrium but local, reversible opening of shell elements that remain attached to the cage — a hinged lid rather than a demolition.

Prior engineering efforts#

Two lines of work bracket this project.

Rational pore engineering. The Giessen lab has widened QtEncapsulin's pores by design, improving substrate access for encapsulated enzymes. This works, and it establishes that the shell tolerates surgery around its symmetry axes — but it optimises a property you can see in a structure.

Directed evolution of the shell itself. Siddiquee and co-workers evolved QtEncapsulin for increased porosity using a chloramphenicol life–death selection, and had to invent a new genetic architecture to do it. Their finding matters for anyone evolving a cage: for multimeric proteins, the impact of a deleterious mutation is amplified by the repeated extensive interactions between adjacent subunits. A variant that would work perfectly well in a mixed shell is lost as a dominant negative when it is the only version present. Their fix was to keep an invariant wild-type copy on a second plasmid, so obligate heteromers could form hybrid assemblies and still be selected.

Evolving an RNA-packaging capsid from scratch. Separately, Tetter and co-workers took a bacterial enzyme that lacks affinity for nucleic acids and converted it, by laboratory evolution under escalating nuclease challenge, into an artificial nucleocapsid that packages and protects its own encoding mRNA. The fraction of particles carrying a full-length genome rose from about 2 % to about 64 %.² Packaging, in other words, is an evolvable trait — provided the selection actually depends on it.

Where the field falls short#

Three gaps, and this project sits at their intersection.

Cargo class. Encapsulins load protein natively and RNA not at all. Lipid nanoparticles do the reverse. Applications that need a protein and its cognate RNA delivered together — ribonucleoprotein genome editors, above all — currently require co-formulating two carriers with different biodistributions.

Measurement. Encapsulation is usually assayed downstream, on purified material, one variant at a time. That is fine for characterising a design and useless as a selection: it cannot be applied to a library, and it does not link a variant's performance to its own survival.

Throughput. Classical directed evolution mutagenises in vitro and transforms the library in, which limits both the depth of mutagenesis and the number of campaigns that can be run in parallel. Continuous systems that hypermutate a designated locus in vivo — PACE, OrthoRep, T7-ORACLE, MutaT7 — remove that ceiling, but they still need a selection worth applying.

We address the third by adopting MutaT7, and the second by building a circuit in which encapsulation is the only cheap route to survival. The first is the point of the exercise.