Structure-Led Exploration of the Metagenome Yields Novel RNA-Guided Nucleases with Broad PAM Diversity

By leveraging advances in protein structure prediction to explore the metagenome, researchers discovered novel compact RNA-guided nucleases that combine small size with broad PAM diversity, offering superior candidates for in vivo genome editing compared to existing systems.

The Gist

Imagine you are a master locksmith trying to open a million different doors (our genes) to fix broken locks (genetic diseases). You have a giant, heavy keychain (the Cas9 enzyme) that works well, but it's too big to fit through the tiny mail slot (the AAV virus) used to deliver the keys into the patient's body. You need smaller, more agile keys.

The problem is, most of the tiny keys you've found so far only fit one specific type of door lock (a specific DNA sequence called a PAM). If the door you need to fix doesn't have that exact lock, your tiny key is useless.

This paper is about a team of scientists who decided to stop looking for keys by comparing their shapes on paper (sequence homology) and started looking for keys by comparing their 3D blueprints (protein structure). Here is how they did it, explained simply:

1. The Old Way: Looking for Cousins

Usually, scientists find new gene-editing tools by looking for bacteria that look like their cousins. They say, "Hey, this new bacteria looks 80% like the one we already know, so it probably has a similar key."

• The Flaw: This misses the "cousins" that look totally different on the outside but have the exact same internal mechanism. It's like missing a Ferrari because you were only looking for cars that look like Toyotas.

2. The New Way: The "Shape-Shifter" Search

The authors used a super-smart AI (like AlphaFold and ESMFold) to predict what millions of invisible proteins look like in 3D. Instead of asking, "Do these proteins look alike?", they asked, "Do these proteins have the same internal engine?"

They focused on a specific engine part called the RuvC domain. Think of this as the "blade" inside the key that actually cuts the DNA. They scanned a massive library of predicted protein shapes looking for this specific blade shape, regardless of what the rest of the key looked like.

The Result: They found dozens of new, tiny keys that looked nothing like the old ones but had the same cutting blade.

3. The Discovery: A Toolbox of Tiny, Versatile Keys

They found two main types of new keys:

• Rare Subtypes: Tiny versions of known keys (like Cas12f and Cas12n) that are small enough to fit in the mail slot.
• Brand New Subtypes: Completely new designs they call TranC. These are like discovering a whole new brand of keys that no one knew existed.

Why are these special?
Most tiny keys are picky; they only open doors with very specific locks. These new keys are adaptable. They can open a wide variety of door locks (diverse PAMs), almost as well as the giant Cas9 key, but they are small enough to fit in the mail slot.

4. Testing the Keys: From Lab to Life

The team didn't just find them; they tested them.

• The Test: They tried these new keys in human cells (HEK293T cells).
• The Success: Several of the new keys worked surprisingly well. One, called OsTranC3, was so good it could edit human DNA almost as well as the standard tools, but it was much smaller.
• The Real-World Application: They used one of these new keys to cut the "brakes" off of immune cells (T-cells). Imagine a car with the parking brake stuck on; the immune cells can't move fast enough to fight cancer. These new keys cut the brake line, allowing the immune cells to attack solid tumors more effectively.

5. The Big Picture: Evolution's "Convergent" Magic

The paper also tells a cool story about evolution. It turns out that nature invented these tiny keys multiple times, independently, from different ancestors. It's like how birds, bats, and insects all evolved wings independently to fly.

• The Ancestors: These tiny keys evolved from "transposons" (jumping genes that move around the genome).
• The Upgrade: Over time, nature added extra "handles" and "grips" (protein domains) to these keys to make them more precise. The scientists found a version that is still very simple (no extra handles) but works perfectly, proving that you don't need a complex machine to get the job done.

Summary

• The Problem: We need small gene-editing tools to cure diseases, but the small ones we know are too picky about where they can work.
• The Solution: The scientists used AI to search for tools based on their 3D shape rather than their genetic code.
• The Win: They found a treasure trove of tiny, versatile, and precise gene-editing tools that can fit into delivery vehicles (AAVs) and target almost any gene in the human body.
• The Future: This opens the door for "off-the-shelf" gene therapies that can be delivered easily to patients to cure genetic diseases or supercharge immune cells to fight cancer.

In short, they stopped looking for keys by their color and started looking for them by their shape, and they found a whole new set of master keys that fit almost any lock.

Technical Summary

1. Problem Statement

CRISPR-Cas systems are powerful tools for genomic editing, but their therapeutic application is hindered by two main limitations:

• Size Constraints: The widely used Cas9 enzyme is too large to be efficiently packaged into Adeno-Associated Virus (AAV) vectors, which are the standard for in vivo gene therapy delivery.
• PAM Restrictions: Many compact RNA-guided nucleases (RGNs), such as Cas12a (Type V), have restrictive Protospacer Adjacent Motif (PAM) requirements (e.g., 5'-TTN-3'), limiting the number of targetable sites in the genome.
• Discovery Bottleneck: Traditional methods for discovering new CRISPR systems rely heavily on sequence homology. This approach fails to identify highly divergent subtypes or systems that have evolved from different ancestral proteins (convergent evolution) but share similar structural folds. Consequently, many potential compact editors with diverse PAMs remain undiscovered in metagenomic data.

2. Methodology

The authors developed a "Structure-First" discovery pipeline that bypasses sequence homology constraints by leveraging protein structure prediction and comparison.

• Structural Search Strategy:

  • Query: Instead of using full-length protein sequences, the authors extracted the RuvC-like nuclease domain (the catalytic core) from four diverse, known RGNs (SpCas9, FnCas12, Cas12f1, and CasLambda).
  • Database: They utilized the ESMAtlas, a database containing over 600 million protein structures predicted by ESMFold (a protein language model that predicts structure without computationally expensive multiple sequence alignments).
  • Search Tool: They employed Foldseek, a tool designed for rapid structural database searching, to find proteins with RuvC domain folds similar to the queries.
  • Filtering: Hits were filtered to ensure they were located within 10kb of a CRISPR repeat array (indicating a functional CRISPR-Cas system) and possessed catalytic residues (DED motif).

• RNA Component Prediction:

  • Using the hypothesis that structural conservation implies functional conservation, the authors predicted tracrRNA sequences for the discovered systems.
  • They analyzed Minimum Free Energy (MFE) structures and antirepeat-repeat hybridization patterns to identify tracrRNA candidates for Cas12n, Cas12f, and novel TranC (Transposon-CRISPR) clades.

• Experimental Validation:

  • Bacterial Depletion: Active systems were screened in E. coli using an 8-mer PAM library to determine PAM preferences.
  • Eukaryotic Editing: Selected candidates were tested in HEK293T cells for genome editing efficiency and specificity.
  • Specificity Analysis: The GenomePAM assay was used to compare off-target effects and target specificity against known systems (AsCas12a, ISDra2 TnpB).
  • Therapeutic Application: A novel system (OsTranC3) was used to knock out immunological inhibitory receptors (PD-1 and TGFβR2) in human T cells.

3. Key Contributions

• Novel Discovery Method: Demonstrated that structural similarity searches (using Foldseek/ESMFold) are superior to homology-based searches for finding highly divergent CRISPR systems that would otherwise be missed.
• Identification of Novel Subtypes: Discovered and characterized dozens of compact RGNs, including:
  • Rare subtypes: Cas12b, Cas12e, Cas12h.
  • Known compact subtypes: Cas12f, Cas12n.
  • Novel TranC Clades: Identified new clades (3, 9, 11, 20, 24) of Transposon-CRISPR intermediates, some of which are ultra-compact (~414 amino acids).
• PAM Diversity: Showed that these novel compact systems possess PAM diversity comparable to or exceeding Cas9, unlike the highly restrictive Cas12a.
• TracrRNA Prediction Framework: Established a robust method for predicting tracrRNA structures for diverse compact Type V systems based on conserved MFE structures.

4. Key Results

• PAM Diversity & Targetability:
  • The novel compact systems (Cas12f, Cas12n, TranC) exhibited a Median Euclidean Distance (MED) of recognition sequences (0.67) similar to TnpB and Cas9, indicating high diversity.
  • Crucially, unlike ancestral TnpB systems which have diverse but complex TAMs (Transposon-Associated Motifs) with low targetability, the novel CRISPR systems retained high diversity and high targetability (frequent occurrence in genomes).
• Size Efficiency:
  • The discovered systems are significantly smaller than Cas9.
  • CauTranC9 (a novel clade 9) is only 414 amino acids, nearly the size of ancestral TnpB, making it one of the smallest active CRISPR nucleases known.
• Eukaryotic Editing Performance:
  • Six lead candidates (AlCas12f, DeCas12f, Rd2Cas12n, RoCas12n, RuTranC11, OsTranC3) showed detectable editing in HEK293T cells.
  • RoCas12n performed comparably to AsCas12a despite being much smaller (528 aa) and having a purine-rich PAM.
  • OsTranC3 (TranC clade 3) demonstrated robust activity and was used to successfully knock down PD-1 and TGFβR2 in human T cells, proving therapeutic relevance.
• Specificity:
  • The novel systems generally showed higher specificity than ancestral TnpB.
  • Systems with expanded REC and WED domains (Cas12n, TranC11) showed longer target sequence requirements (higher specificity), while ultra-compact systems like OsTranC3 showed intermediate specificity.
• Cleavage Patterns:
  • Distinct deletion profiles were observed: Cas12f systems caused small deletions (median 4bp) outside the target, while TranC systems and Cas12n showed patterns closer to Cas12a or intermediate, confirming functional distinctness.

5. Significance

• Therapeutic Potential: This study provides a new toolkit of compact, high-specificity, and PAM-flexible nucleases that are ideal for AAV-mediated in vivo gene therapy, overcoming the size limitations of Cas9 and the targeting limitations of Cas12a.
• Evolutionary Insight: The findings support the theory of convergent evolution, where independent ancestral TnpB proteins evolved into diverse Type V CRISPR systems. The study highlights that the transition from mobile genetic elements (transposons) to immune defense systems involved the expansion of specific domains (REC/WED) to improve specificity, while retaining compact PAMs for broad pathogen defense.
• Methodological Shift: The paper establishes structural bioinformatics as a critical, scalable approach for mining metagenomic data, suggesting that future discoveries of novel enzymes will increasingly rely on structure rather than sequence similarity.
• Clinical Application: The successful knockout of immunosuppressive checkpoints (PD-1/TGFβR2) in T cells using a novel compact nuclease (OsTranC3) validates the immediate applicability of these tools for engineering "off-the-shelf" CAR-T cells for solid tumor therapy.