Fully computational design of PAM-relaxed Staphylococcus aureus Cas9 with expanded targeting capability using UniDesign

This study presents KRH, a fully computationally designed Staphylococcus aureus Cas9 variant that expands PAM recognition to NNNRRT and achieves editing efficiencies comparable to or exceeding evolution-derived variants, demonstrating the power of computational design to rationally engineer high-performance genome-editing tools without experimental optimization.

The Gist

Imagine you have a super-smart, microscopic pair of molecular scissors called CRISPR-Cas9. These scissors are famous for being able to cut DNA at very specific locations, which allows scientists to edit genes to cure diseases or improve crops.

However, there's a catch. These scissors are very picky about where they can cut. They only work if they see a specific "password" next to the target spot. This password is called a PAM (Protospacer Adjacent Motif).

The Problem: The Picky Scissors

The version of these scissors used in this study comes from a bacteria called Staphylococcus aureus (the kind that causes skin infections). Let's call this version SaCas9.

SaCas9 is special because it's tiny. Think of it as a compact sports car compared to the larger, bulkier versions of CRISPR. Because it's small, it fits easily into a tiny delivery truck (a virus called AAV) that can carry it into human cells to fix genetic diseases.

But here's the problem: SaCas9 is extremely picky. It only recognizes a password that looks like NNGRRT.

• N = Any letter (A, C, G, or T)
• R = A specific letter (A or G)
• T = A specific letter (T)

Because of this strict rule, there are huge gaps in the human genome where SaCas9 simply cannot go. It's like having a master key that only opens doors with a specific, rare lock, leaving thousands of other doors locked forever.

The Old Way: Trial and Error

For years, scientists tried to make SaCas9 less picky. They used a method called "directed evolution." Imagine trying to fix a lock by randomly changing the shape of the key thousands of times, testing it, and hoping one works. It works eventually, but it's slow, expensive, and you don't really understand why the new key works.

The New Way: The Digital Architect

This paper introduces a new approach using a computer program called UniDesign. Instead of randomly guessing, the scientists used a digital architect.

1. The Blueprint: They fed the computer the 3D blueprint of the SaCas9 scissors and the DNA lock.
2. The Simulation: The computer simulated millions of tiny changes (mutations) to the scissors' "fingers" (amino acids) that touch the password.
3. The Goal: The computer's job was to find a new set of fingers that would still hold the scissors tight but would accept a wider variety of passwords (changing the rule from NNGRRT to NNNRRT). This means the scissors could now accept almost any letter in that third spot.

The Result: The "KRH" Scissors

The computer designed a new version of the scissors called KRH.

• How it works: The computer figured out that to make the scissors less picky, you need to loosen their grip on the specific "password" letters but tighten their grip on the "backbone" of the DNA (the railing the password sits on). It's like loosening a specific screw on a door handle so it fits more keys, while tightening the handle itself so it doesn't fall off.
• The Performance: When the scientists built the KRH scissors in the lab, they worked perfectly.
  • They could cut DNA at spots the old scissors couldn't touch.
  • In some cases, they were 116 times more efficient at these new spots than the original scissors.
  • They were just as good as, or even better than, the best "evolved" scissors scientists had made before.

Why This Matters

This is a huge leap forward for two reasons:

1. More Targets: Because KRH is less picky, scientists can now target many more genes in the human body. It's like having a master key that opens 90% of the doors in a building instead of just 10%.
2. Speed and Precision: This proves that we don't need to wait for nature to evolve better tools. We can design them on a computer in days, predict exactly how they will work, and build them immediately. It's the difference from "guessing and checking" to "engineering and building."

The Bottom Line

The scientists took a tiny, picky pair of molecular scissors and used a super-smart computer program to redesign them. The result is a new tool (KRH) that is just as small and safe but can cut DNA in many more places. This brings us one step closer to curing genetic diseases that were previously out of reach because the "lock" was too specific.

Technical Summary

1. Problem Statement

CRISPR-Cas9 genome editing is constrained by the requirement for a specific Protospacer Adjacent Motif (PAM) sequence adjacent to the target DNA. Staphylococcus aureus Cas9 (SaCas9) is highly valued for therapeutic applications due to its compact size, which allows efficient packaging into adeno-associated virus (AAV) vectors. However, wild-type (WT) SaCas9 recognizes a restrictive PAM sequence (NNGRRT), significantly limiting the number of targetable genomic loci.

Previous efforts to relax this specificity (e.g., the KKH variant) relied on iterative molecular evolution or chimera engineering. These methods often lack mechanistic insight, require extensive wet-lab screening, or rely on computationally expensive simulations (e.g., molecular dynamics) that hinder high-throughput design. The challenge remains to rationally design SaCas9 variants with expanded PAM compatibility (specifically NNNRRT) while maintaining high nuclease activity without relying on experimental evolution.

2. Methodology

The authors employed a fully computational design workflow using an improved version of their protein design framework, UniDesign.

• Workflow Improvements: The standard UniDesign pipeline was enhanced to specifically generate point-mutation variants with a constrained number of mutations.
  • Mutation Count Constraints: A penalty term was added to the scoring function to bias the search toward sequences with a specific number of mutations (e.g., exactly 1, 2, or 3).
  • Redundancy Reduction: A duplicate-design penalty was introduced to ensure that independent Monte Carlo simulations produced unique low-energy sequences, increasing design diversity.
• Iterative Design Process:
  1. Iteration 1 (Relaxing Specificity): The design focused on Arg1015, the residue responsible for recognizing the third guanine (G) in the NNGRRT PAM. The goal was to find substitutions that reduced the Mean Absolute Deviation (MAD) of binding energies across four representative PAMs (TTAGGT, TTCGGT, TTGGGT, TTTGGT). R1015H was identified as the optimal single mutation to reduce positional bias.
  2. Iteration 2 (Restoring Affinity): The R1015H mutation weakened binding to the native PAM. The workflow introduced a second mutation to restore binding energy via nonspecific interactions with the DNA backbone. Positively charged residues (Lys, Arg) were selected to form salt bridges with phosphate groups.
  3. Iteration 3 (Optimization): Promising double mutants were combined into triple mutants. The top candidate, KRH (E782K/N968R/R1015H), was selected based on its ability to achieve binding energies comparable to WT SaCas9 on the native PAM while maintaining low MAD across relaxed PAMs.
• Experimental Validation:
  • Genome Editing: WT, KRH, and the evolution-derived KKH variants were tested in HEK293T, A549, HeLa, and U2OS cells across various PAM contexts.
  • Base Editing: KRH was fused with the ABE8e deaminase (D10A nickase) to create KRH-ABE and compared against WT-ABE and KKH-ABE.
  • Off-Target Analysis: Potential off-target effects were assessed using deep sequencing.

3. Key Contributions

• First Fully Computational SaCas9 Variant: The paper reports KRH, a SaCas9 variant designed entirely in silico without any additional wet-lab optimization or directed evolution.
• Mechanistic Insight: The study provides a structural and energetic explanation for PAM relaxation. It demonstrates that relaxing PAM specificity requires a balance between disrupting sequence-specific interactions (at the PAM base) and enhancing nonspecific interactions (with the DNA backbone) to maintain overall binding affinity.
• Methodological Advancement: The improved UniDesign workflow proves capable of navigating the sequence space to identify minimal mutation sets (3 mutations) that remodel protein-DNA interfaces effectively, offering a scalable alternative to directed evolution.

4. Key Results

• Expanded PAM Recognition: KRH successfully recognizes the expanded NNNRRT PAM motif.
• Editing Efficiency:
  • At non-canonical PAM sites (e.g., NNHRRT), KRH showed 1.4-fold to 116.3-fold higher editing efficiency compared to WT SaCas9.
  • At the RUNX1 locus, efficiency increased from 0.47% (WT) to 54.68% (KRH).
  • KRH maintained high activity at canonical NNGRRT sites, with only a modest decrease compared to WT.
• Comparison with KKH:
  • KRH (E782K/N968R/R1015H) and the evolution-derived KKH (E782K/N968K/R1015H) share two key mutations (E782K, R1015H) but differ at position 968 (Arg vs. Lys).
  • Structurally, N968R in KRH forms two salt bridges with the target strand phosphates, whereas N968K in KKH forms one.
  • Performance: KRH exhibited editing efficiencies comparable to or slightly superior to KKH across multiple cell types and PAM contexts.
• Base Editing: The KRH-ABE construct showed significantly improved base-editing efficiency (up to 65-fold at specific loci) compared to WT-ABE at non-canonical PAMs, while maintaining comparable specificity (low off-target effects).
• Cell-Type Robustness: The performance gains of KRH were consistent across HEK293T, A549, HeLa, and U2OS cells, though absolute efficiency varied by cell type due to chromatin accessibility.

5. Significance

• Therapeutic Potential: By expanding the targeting range of SaCas9 to NNNRRT, KRH significantly increases the number of disease-relevant genomic sites accessible for AAV-delivered gene therapies, potentially eliminating the need for less efficient dual-AAV split-Cas9 strategies.
• Paradigm Shift in Protein Engineering: This work demonstrates that computational protein design can not only recapitulate but in some cases surpass the performance of variants derived from labor-intensive directed evolution. It validates UniDesign as a predictive framework capable of identifying minimal, high-impact mutations.
• Scalability: The approach offers a rapid, scalable strategy for engineering CRISPR nucleases with custom specificities, paving the way for the rational design of "catalogs" of Cas9 variants optimized for specific PAM subsets rather than a single "PAM-free" enzyme that may suffer from kinetic trapping.

In conclusion, the study establishes KRH as a blueprint for rationally engineered, PAM-relaxed nucleases and underscores the power of atomic-level computational modeling to accelerate the development of next-generation genome editing tools.