The tiny protein shells that connect viruses with cellular chemistry

A virus and a cell can use the same protein architecture for different kinds of cargo. In viruses, an HK97-fold shell packages genetic material. In bacterial and archaeal cells, related shells called encapsulins enclose the machinery of chemical reactions. The container’s family resemblance survives a striking change of occupation.[2]

A study published on 14 October found 523 candidate viral genomes below 11,600 base pairs that encode HK97-fold proteins. Analyses produced at least three mixed groups of viral and encapsulin proteins, supporting repeated transitions between the two roles.[1]

That raises a richer question than whether two proteins look alike. How much can a biological container change its job while retaining the architecture that makes it a container?

A compartment changes the chemistry inside it

Encapsulins give cells ways to organise reactions without wrapping each compartment in a membrane. Their shells gather enzyme cargo, creating reaction compartments within a cell. Documented functions include iron storage and processes associated with oxidative stress. The surrounding structure helps determine where chemistry happens and what reaches the machinery inside.[2]

A 2021 computational survey identified 6,133 encapsulin-like proteins across 31 bacterial and four archaeal phyla. It organised them into four families and argued that encapsulins probably descended from HK97-fold viruses. The shared fold supplied evidence of relationship; the genes around a shell protein helped distinguish a cellular compartment from a viral component.[2]

Think of that last distinction as asking what else occupies the workshop. An unfamiliar container beside a set of chemical tools suggests a different use from a similar container beside packaging machinery. Shape narrows the possibilities; its neighbours help supply a job description. This is an analogy for interpreting the surrounding genes, rather than an observation of an ancient cell.

AI makes unfamiliar proteins easier to place

Protein language models provide another way into that problem. Trained on protein sequences, they create numerical representations that capture relationships difficult to recover from direct sequence matches. Researchers can train a classifier on those representations to assign broad functional categories. A family resemblance can become detectable even when the amino-acid sequences have diverged.[3]

The classifier behind this approach was described in a 2024 study of viral proteins. Applied to an ocean-virome collection, it added 26,770 annotated protein families to a starting set of 91,156: a 29.4% increase. The authors calibrated class thresholds on a validation set for a maximum 10% false-discovery rate and treated the predictions as a complement to established sequence-based methods.[3]

In the new study, protein-language-model annotation worked alongside conventional sequence profiles, AlphaFold2 structure predictions and evolutionary comparisons. The small genomes came from environmental sequence data; the researchers did not culture these candidate viruses.[1]

That division of labour matters. A useful label helps select something worth examining. Its shape, genomic surroundings and relationship to other proteins then help establish whether the suggested role makes sense. The contribution is a more informative map of material already present in biological databases.

The unfinished shell supplies a possible bridge

The structural connection was strongest with procapsids—the immature shells built before viral capsids reach their mature form. Their resemblance to encapsulins makes that intermediate state a plausible route between biochemical and genome-packaging compartments.[1]

A container is more than its final outline. Assembly, cargo and the changes made during use all matter. Comparing a cellular compartment only with a mature viral shell could therefore miss a resemblance visible at an earlier stage of construction. The procapsid idea makes the history a question about a process as well as an object.

The evolutionary trees do not establish a secure root showing which role came first. The authors’ proposal that simpler biochemical compartments preceded larger genetic containers remains a hypothesis. Their evidence supports a complicated relationship, rather than a reconstruction of life’s first shell.[1]

The enduring idea is that biology can reuse an architecture more freely than our names suggest. A capsid and an encapsulin sound like separate inventions. Looking closely at the shared container opens a history in which its purpose may have changed more than once.

Report an error or suggest a correction