University of Miami researchers mined DNA databases and found small viruses whose protein shells resemble encapsulins, the protein compartments that some cells use to hold chemical reactions. They report in Nature Communications that viruses and encapsulins evolved from one another at least three times, not in a single direction.
What the Researchers Found
The study, titled “Small viruses reveal bidirectional evolution between HK97-fold viruses and encapsulins via procapsids,” appears in Nature Communications, volume 17, article 10571. Abelardo Aguilar, a doctoral student in the university’s Department of Biology, is the first author. Antoni Luque Santolaria, an associate professor of biology with secondary appointments in physics and in chemical, environmental and materials engineering, is the senior author.
The work centers on a protein fold named HK97, which forms the shells of many viruses. Encapsulins use the same fold to build compartments inside cells. A common view holds that encapsulins evolved from viruses. The authors report evidence for a two-way exchange instead, and the University of Miami’s announcement says the two groups evolved from one another repeatedly over millions of years. The sources disagree on the publication date. The journal page lists the paper as published Oct. 14, 2026. The announcement, posted on Newswise, carries an Oct. 15 date.
The Numbers Behind the Search
The team searched environmental DNA databases, including IMG/M and MGnify, that hold billions of sequences from soil, seawater, wastewater and the human gut. They looked at what the paper calls twilight-zone viruses, genomes shorter than 11.6 kilobase pairs. The search turned up 4,730 putatively complete genomes in that size range. For comparison, the team examined 2,728 viral contigs between 11.6 and 20 kilobase pairs.
Shell proteins of the HK97 type showed up in 11% of the small genomes, or 536 of them, and in 29% of the comparison group, or 798. The comparison group therefore carried the fold at about 2.6 times the rate, a calculation from the percentages. The authors identified 523 genomes with HK97-fold proteins that were shorter than the shortest isolated virus known to carry the fold of those 523, 93 encoded features that resemble encapsulins, or about 18%.
The pattern depended on genome size. Among what the paper calls mini genomes, 5 to 11.6 kilobase pairs long, 93 of 491 shell proteins, about 19%, carried encapsulin domains. Among micro genomes, under 5 kilobase pairs, none of 45 did. In the comparison group, 20 of 798, or 2.5%, carried them, and in a reference set of known viruses, 17 of 910, or 1.9%, did. The mini genomes therefore carried encapsulin features about 7.6 times as often as the comparison group, a calculation from the paper’s counts.
The evolutionary tree in the paper includes 1,334 environmental HK97-fold proteins, which fall into eight groups and nine singletons among the proteins enriched for encapsulin features. The largest of those groups holds 76 shell proteins from genomes averaging about 8.75 kilobase pairs long. A separate co-occurrence analysis covers 2,321 viral shell proteins and 104 encapsulin proteins of the same fold. When the paper added the comparison group to the count, twilight-zone genomes made up 63% of the viral contigs, which shows that the small genomes are not a rare corner of the data.
How Viruses and Encapsulins Compare in Shape
The authors compared the structures of viral procapsids, which are immature viral shells, with encapsulins. Viral procapsids and encapsulins have inner diameters of about 330 and 322 angstroms, respectively, and both have a projected capacity of about 10.75 kilobase pairs. The paper reports that their amino acid sequences share 14.4% identity over 473 residues. A capacity of 10.75 kilobase pairs is 93% of the 11.6 kilobase pair cutoff the team used to define small genomes, a calculation from the paper’s two figures.
From this evidence, the authors propose that shells built on this fold first appeared as small compartments for chemical reactions, and that the ability to package a genome came as the shells grew and added specialized parts. They also report that procapsids resemble encapsulins more closely than mature viral capsids do, which makes procapsids a plausible bridge between the two.
Aguilar said in the announcement that viral capsids could also help carry out biochemical reactions. The authors suggest researchers may need to pay more attention to non-infective viral particles and their possible other functions.
What the Authors Say Limits the Findings
The paper states several limits. Some metagenomic elements that carry encapsulin-like shell proteins may be misclassified systems. The signatures for encapsulin-like domains come from Pfam profiles. The evolutionary trees built from structure and from sequence have different shapes, and some structural groups have no counterpart in the sequence tree, which the authors attribute to limited resolution of deeper relationships. The comparison between procapsids and encapsulins rests on a small number of available structures, and their sequence identity is low. The paper describes parts of its evolutionary model as qualitative.
Luque said in the announcement that “The majority of viruses cannot be grown in the lab.” The announcement adds that many viruses lack known hosts, which makes the group hard to study and means the findings come from sequence data and computer models, not from cultured viruses.
Methods and Funding
The team combined bioinformatics tools, biophysical models and artificial intelligence tools, including protein language models and AlphaFold, according to the announcement. Luque said the toolset can be applied in other settings, and he named emerging pathogens as one possibility. The Gordon and Betty Moore Foundation and the U.S. National Science Foundation funded the work, the announcement says.
The announcement does not give a count of the viruses identified, and it does not quote a researcher from outside the Miami team. The paper’s author list extends past Aguilar and Luque to colleagues at several institutions. The journal page lists, among others, Anca M. Segall, Forest L. Rohwer and Simon Roux.
