Screening metatranscriptomes for ultrastable RNA secondary structures reveals hidden bacteriophages and novel capsid nanomaterials

This study introduces a metatranscriptomic screening method based on ultrastable RNA secondary structures to uncover thousands of previously hidden ssRNA bacteriophages, demonstrating their potential as novel, engineerable capsid nanomaterials for RNA delivery and compiling them into a comprehensive database of over 460,000 RNA molecules and 100,000 coat proteins.

The Gist

Imagine the world of viruses as a massive, bustling library. For decades, scientists have been trying to catalog the books in this library, but they've only been able to read the spines. They've been looking for specific titles (known protein sequences) to find new books. If a book had a completely different spine design, they simply missed it, leaving thousands of volumes hidden on the shelves.

This paper is about a team of researchers who decided to stop looking at the spines and start listening to the rhythm of the books instead.

Here is the story of how they found hidden viruses, built a massive new library, and discovered a way to turn these viruses into tiny delivery trucks.

1. The "Super-Stable" Rhythm

Most things in nature are a bit messy. If you take a random string of letters and try to fold it into a shape, it usually falls apart or forms a weak, wobbly structure.

However, the researchers discovered that single-stranded RNA viruses (tiny parasites that infect bacteria) have a secret superpower: their genetic code folds into shapes that are incredibly, almost unnaturally stable. It's like if you took a piece of paper, crumpled it, and it instantly turned into a diamond-hard gem that could survive being dropped from a skyscraper.

The team realized that while other viruses (like those that infect humans) have "wobbly" folds, these bacterial viruses have "diamond-hard" folds. They decided to use this stability as a new way to find them.

2. The Great Filter: Finding the Hidden Gems

The researchers took massive digital libraries of environmental data (called metatranscriptomes), which are like giant piles of shredded paper from the ocean, soil, and gut bacteria. Most of this data is just noise.

They ran a computer program that acted like a metal detector. Instead of looking for a specific metal (a known protein), the detector looked for the "ring" of stability.

• The Result: They filtered out the wobbly, unstable junk and kept only the "diamond-hard" pieces.
• The Discovery: Suddenly, the pile was full of viruses they had never seen before. They found thousands of new viral species that were hiding in plain sight because their "spines" (proteins) looked too different to be recognized by old methods.

3. The "Coat" Collection: Building a Database

Every virus has a protective shell, or "coat," made of proteins. Think of these coats as the hard plastic casing of a smartphone. The researchers found that these new, hidden viruses had unique casing designs.

They compiled a massive database called SCIB (San Diego State University Coat Information Bank).

• It contains information on over 460,000 unique RNA molecules.
• It lists over 100,000 different types of viral "coats."

This is like finding a catalog of 100,000 different types of smartphone cases, many of which have never been manufactured before.

4. The Factory Test: Do They Actually Work?

Finding the blueprints is one thing; building the product is another. The researchers asked: If we build these viral coats in a lab, do they actually snap together to form a shell?

They created a "factory" inside E. coli bacteria (tiny biological 3D printers). They programmed the bacteria to build 12,000 different viral coats from their new database.

• The Surprise: Almost all of them (93%) successfully snapped together into perfect, round shells.
• The Durability: These shells were tough. When they threw enzymes (nature's scissors) at them, the shells held firm, protecting the RNA inside.

5. The "Eliophage": A New Delivery Truck

To prove these shells could be useful, they picked one specific new virus (which they nicknamed "Eliophage") and studied it closely.

• The Shape: Using a super-powerful microscope (Cryo-EM), they saw it was a perfect icosahedron (a 20-sided die shape), just like a soccer ball.
• The Twist: Unlike the famous MS2 virus (a standard model in science), Eliophage had a different "personality." It was positively charged (like a magnet), while MS2 was negatively charged. This means Eliophage might stick to different types of cells, which is great for targeting specific places in the body.

6. The Ultimate Goal: The "Trojan Horse"

The most exciting part? They showed they could take the shell apart and put it back together with a different cargo.

Imagine taking a FedEx truck, taking it apart, and then reassembling it around a package of your own choosing. The researchers took the Eliophage shell, stripped out its original viral RNA, and reassembled it around a piece of foreign RNA (like a medicine or a gene therapy).

• The Result: The shell held the new cargo tight and protected it from being destroyed.

Why Does This Matter?

This paper changes the game in three ways:

1. Discovery: We can now find viruses that were previously invisible to science, just by listening to their "stable rhythm."
2. Nanotechnology: We now have a massive catalog of 100,000+ tiny, self-assembling containers. These could be used to deliver drugs, vaccines, or genes to specific cells in the human body.
3. Evolution: It helps us understand how life evolves, showing that nature has been building these perfect, stable containers for billions of years, and we are just finally learning how to read the instructions.

In short, the researchers found a new way to spot hidden viruses, proved they can be mass-produced in a lab, and showed that they can be turned into tiny, programmable delivery trucks for the future of medicine.

Technical Summary

1. Problem Statement

While metatranscriptomics has revolutionized the discovery of single-stranded RNA (ssRNA) bacteriophages, current detection methods rely heavily on sequence similarity to known viral proteins, particularly the RNA-dependent RNA polymerase (RdRp). This approach has significant limitations:

• It fails to detect phages with novel RdRp sequences that lack homology to known families.
• It misses partial genomes or defective particles where the RdRp gene is missing or incomplete, even if the capsid genes are intact.
• Consequently, a vast "dark matter" of ssRNA phage diversity remains hidden in environmental datasets.

The authors propose that RNA secondary structure stability is an underutilized, orthogonal signal that can identify these hidden phages, as ssRNA phage genomes are known to adopt exceptionally stable thermodynamic structures essential for replication and packaging.

2. Methodology

The study employs a multi-stage computational and experimental pipeline:

A. Computational Screening (The "Ultrastable" Filter)

• Metric: The authors utilize the Minimum Free Energy (MFE) Z-score. They calculate the MFE of RNA contigs using RNAfold (ViennaRNA) and compare it against the distribution of MFEs from 1,000 random nucleotide shuffles of the same sequence.
• Threshold: Contigs with a Z-score $\le -15$ are retained. This threshold signifies thermodynamic stability significantly higher than random sequences.
• Datasets: Applied to two large, previously published metatranscriptomic datasets (Neri et al. and Hou et al.), totaling over $10^8$ contigs.
• Validation: Retained contigs are screened for the presence of coat protein genes using iterative sequence similarity searches (tBLASTn, MMseqs2, DIAMOND) and structural validation (AlphaFold2, Foldseek) to confirm the canonical ssRNA phage coat-protein fold.

B. Database Construction

• Newly discovered contigs and coat proteins were integrated with known phage data to create the San Diego State University Coat Information Bank (SCIB).
• This database aggregates sequence and structural data for over 467,000 distinct phage contigs and 100,000 unique coat proteins.

C. High-Throughput Experimental Validation

• Library Construction: A plasmid library encoding 12,000 distinct coat proteins (selected from the SCIB database based on length, lack of cysteines, and structural confidence) was created. Each insert included a unique 16-nucleotide barcode.
• Expression & Assembly: The library was expressed in E. coli BL21(DE3). Capsids were purified via nuclease treatment (to degrade non-packaged RNA) and density-gradient centrifugation.
• Quantification: High-throughput sequencing of the packaged RNA (containing the barcodes) versus the input plasmid DNA allowed for the calculation of relative assembly yields for each variant.

D. Biophysical Characterization

• Selected candidates (specifically a novel phage named "eliophage") were characterized using:
  • Negative-stain TEM and Cryo-EM (Single-particle reconstruction) to determine 3D structures (T=3, T=4, and prolate morphologies).
  • Interferometric Scattering Microscopy (iSCAT) for particle mass distribution.
  • Differential Scanning Fluorimetry (DSF) to assess thermal stability across pH ranges.
  • In vitro Reassembly: Disassembly of capsids in acidic conditions followed by reassembly around heterologous RNA (MS2 RNA) to test packaging versatility.

3. Key Results

A. Discovery of Hidden Phages

• Enrichment: Filtering metatranscriptomes for Z-scores $\le -15$ enriched known ssRNA phage sequences by ~60-fold and phage-like viruses by ~35-fold.
• Novelty: The filter identified 94,920 new phage-derived contigs that were previously unclassified by RdRp-based methods. Many of these lacked RdRp genes entirely but encoded properly folded coat proteins, confirming they are genuine phage fragments or novel phages.
• Database: The resulting SCIB database contains structural and sequence data for 467,168 RNA molecules and 100,403 distinct coat proteins.

B. Capsid Assembly Landscape

• High Yield: Of the 12,000 expressed coat proteins, 11,144 (92.8%) successfully assembled into nuclease-resistant capsids.
• Variability: Assembly yields varied by five orders of magnitude. The top 1% of variants produced ~400-fold more capsids than the median.
• Sequence-Structure Relationship: While similar sequences generally yielded similar assembly efficiencies, single amino acid substitutions could cause drastic (100-fold) changes in yield, suggesting assembly is a complex, context-dependent process not easily predicted by simple physicochemical properties.

C. Structural and Functional Characterization of "Eliophage"

• Structure: The novel "eliophage" (SCIB_ID_0008658_0000000) forms T=3, T=4, and rare prolate icosahedral capsids. Its coat protein dimer structure is similar to MS2 but shares only 18% sequence identity.
• Biophysics: Unlike MS2, eliophage capsids possess a positive net surface charge and denature at significantly lower temperatures across all tested pH values.
• Repackaging: Eliophage capsids could be disassembled and reassembled in vitro to successfully package heterologous RNA (MS2 genome), demonstrating their potential as programmable delivery vehicles.

4. Significance and Impact

• New Discovery Paradigm: The study establishes thermodynamic RNA stability as a robust, sequence-independent metric for viral discovery. This "structure-first" approach uncovers viral diversity that sequence homology methods miss, particularly for viruses with highly divergent or missing polymerase genes.
• Nanomaterial Resource: The creation of the SCIB database and the demonstration that >90% of novel coat proteins assemble into stable capsids provides a massive, validated library of nanomaterials. These particles offer diverse surface chemistries (charge, stability) for engineering applications in drug delivery, imaging, and molecular display.
• Platform for RNA Delivery: The successful in vitro reassembly and packaging of foreign RNA by a novel capsid proves that these uncultured phages can be repurposed as RNA delivery vehicles, a critical step for therapeutic applications.
• Future Directions: The authors highlight that the structural data generated from these uncultured phages will be essential for training next-generation AI tools (like AlphaFold for RNA) to predict long RNA structures, closing the loop between viral discovery and structural biology.

In summary, this work bridges computational virology and synthetic biology, transforming "dark matter" in metatranscriptomes into a functional, characterizable library of viral nanomaterials.