Rational design of synthetic proteins using a genome-scale CRISPR screen

This study employs a genome-scale CRISPR activation screen to identify native proteins that enhance homology-directed repair, which are then rationally fused to Cas9 to create "TruEditors" that significantly improve precise gene editing efficiency and therapeutic applications in diverse human cell types.

The Gist

Imagine you are a master carpenter trying to fix a very specific, tiny crack in a priceless wooden table. You have a powerful saw (the CRISPR-Cas9 tool) that can cut the wood perfectly, but once you cut it, the wood doesn't know how to glue itself back together the way you want. Usually, it just glues itself back together haphazardly, leaving a messy scar.

For years, scientists have been trying to figure out how to make that "glue" (the cell's DNA repair system) work better. They've tried a few known tricks, like using chemical glue inhibitors, but those often damage the whole table or are too blunt.

This paper is about a new, smarter way to find the perfect glue.

Here is the story of how the researchers did it, broken down into simple steps:

1. The "Who's Who" Search (The Screen)

Instead of guessing which proteins might help fix the DNA, the researchers decided to ask the whole human body for help. They took a library of instructions for 19,000 different human proteins (the "workers" inside our cells) and asked them all to show up at the job site at once.

They set up a massive test:

• The Job: Fix a specific cut in a piece of DNA.
• The Test: If the protein helped fix the cut perfectly, the cell would glow Blue. If it fixed it messily, the cell would stay Green (or go dark).
• The Result: They screened all 19,000 workers and found over 800 "Star Workers" that made the cells glow blue much more often. Some were famous repair experts, but many were unknown workers who had never been seen helping with this job before.

2. Building the "Super-Tool" (TruEditors)

Now that they knew who the star workers were, they didn't just send them to the job site; they strapped them directly to the saw.

They created a new tool called a TruEditor (Targeted Repair fUsion Editor). Think of it like taking a standard power drill and attaching a specialized, high-tech guide attachment to it.

• The Drill: The Cas9 enzyme (the cutter).
• The Attachment: One of those 800 star proteins (the fixer).

By gluing the "fixer" directly to the "cutter," the fixer is right there at the exact moment the cut happens, ready to guide the repair process perfectly.

3. The "Mini-Tool" Discovery

The researchers noticed that some of these star workers were huge, complex machines (thousands of parts long). They wondered: Do we need the whole machine, or just a specific gear?

They chopped the proteins down to their smallest, most essential parts (like taking just the engine out of a car). Surprisingly, these tiny, compact versions worked just as well, or even better, than the full-size versions. This is huge because smaller tools are easier to deliver into cells.

4. How They Work (The Secret Sauce)

The team wanted to know why these tools worked so well. They used a technique called "affinity proteomics," which is like putting a magnet on the tool to see what other proteins stick to it.

They found that the TruEditors act like social butterflies. When they arrive at the cut site, they immediately call over the cell's own repair crew (like a construction foreman calling in the specialized electricians and plumbers). They don't force the cell to do anything; they just organize the existing repair team to do a better job.

5. The Real-World Test (Saving Lives)

Finally, they tested these new tools on things that matter:

• Stem Cells: They fixed genes in human stem cells (the building blocks of the body) with 3 times more success than before. This is important because stem cells are usually very hard to edit.
• Cancer Fighters (T-Cells): They used the tools to engineer T-cells (the body's immune soldiers) to fight cancer. They inserted a "targeting system" (CAR) into the T-cells to hunt down leukemia.
  • The Result: The new tools doubled the number of successful cancer-fighting T-cells created.
  • The Impact: In a test, these new T-cells killed cancer cells much more effectively than the old methods.

The Big Picture

Before this, designing a new gene-editing tool was like trying to invent a new type of glue by guessing the chemical formula.

This paper says: "Let's just test every single ingredient in the kitchen, see which one makes the cake rise best, and then bake that into our recipe."

By testing the entire human "library" of proteins, they found a whole new generation of tools that are:

1. More precise (less accidental damage).
2. More powerful (fixing genes that were previously impossible to fix).
3. Safer (they don't break the cell's other repair systems).

This approach could be the key to curing genetic diseases like sickle cell anemia or muscular dystrophy, where getting the "glue" right is the difference between a cure and a failed attempt.

Technical Summary

1. The Problem

While deep learning has revolutionized protein structure prediction, the rational design of proteins for specific functional tasks remains challenging. In the context of genome editing, improving precise gene editing (Homology-Directed Repair, or HDR) has historically relied on:

• Hypothesis-driven approaches: Fusing known DNA repair proteins to Cas9.
• Global inhibition: Using small molecules to inhibit the competing Non-Homologous End Joining (NHEJ) pathway, which often causes global DNA damage and cytotoxicity.
• Limited scope: Existing strategies are restricted to a handful of well-characterized repair factors.

There is a lack of a systematic, unbiased method to identify novel human proteins that can locally promote HDR without disrupting global DNA repair mechanisms, particularly for difficult-to-edit loci (e.g., low GC content) and in primary cells (e.g., T cells, stem cells).

2. Methodology

The authors adopted a "function-first," massively parallel approach to screen the entire human proteome for proteins that enhance precise DNA repair.

• Genome-Scale CRISPR Activation (CRISPRa) Screen:

  • Cell Line: K562 cells stably expressing dCas9-VP64 and a fluorescent reporter system.
  • Reporter: A "stoplight" system where two precise edits are required to convert EGFP (Green) to EBFP (Blue). This allows for the sorting of cells based on HDR (BFP+) vs. NHEJ/indel (GFP-).
  • Library: A genome-wide library targeting ~19,000 human genes with 6 gRNAs per gene, utilizing an optimized CRISPRa system (dCas9-VP64 + PCP-p65-HSF1) to overexpress endogenous genes.
  • Process: Cells were transduced, nucleofected with Cas9 RNP and a repair template, and sorted into HDR and NHEJ populations. gRNA abundance was quantified via sequencing to identify genes whose overexpression enriched the HDR population.

• Rational Design of TruEditors:

  • Candidate Selection: Top-ranked genes from the screen (including known repair factors like BARD1, BLM and novel candidates like ZNF146, TBX10) were selected.
  • Fusion Construction: Full-length proteins and specific protein domains were fused to the C-terminus of Cas9 using an XTEN linker to create Targeted Repair fUsion Editors (TruEditors).
  • Validation: Editing efficiency was tested across multiple loci (GFP-to-BFP, PRNP, HBB, PARP1, MUT, KRAS, SMARCB1) in diverse cell lines (HEK293, A549, BT16, MDA-MB-231).

• Mechanistic Characterization:

  • Affinity Proteomics (AP-MS): FLAG-tagged TruEditors were expressed in HEK293 cells, immunoprecipitated, and analyzed via mass spectrometry to map protein-protein interactions (PPIs).
  • Structural Modeling: AlphaFold3 was used to predict interaction interfaces (e.g., between BLM domains and TOP3A/RMI1).
  • Mutagenesis: Point mutations and truncations were introduced to validate critical residues required for complex formation and HDR enhancement.

• Therapeutic Application (mRNA Delivery):

  • TruEditor mRNAs (modified with Cap-1 and pseudouridine) were synthesized and delivered via nucleofection to human pluripotent stem cells (hESCs) and primary human CD8+ T cells.
  • CAR-T Engineering: A 2.8 kb CD19 CAR donor was knocked into the TRAC locus to generate allogeneic CAR-T cells.

3. Key Contributions

1. Proteome-Scale Functional Map: Generated the first genome-wide map of ~800 human proteins that positively modulate HDR when overexpressed, identifying both known and novel regulators of DNA repair.
2. TruEditor Platform: Developed a new class of synthetic genome editors (TruEditors) by fusing these identified proteins to Cas9.
3. Minimal Domain Discovery: Demonstrated that full-length proteins are not always necessary; specific N-terminal domains (e.g., BLM residues 2–194) can outperform full-length fusions, offering more compact editors.
4. Mechanistic Insight: Showed that TruEditors function by recruiting endogenous DNA repair complexes (e.g., BTR complex, BRCA1-BARD1) directly to the edit site, rather than globally inhibiting NHEJ.
5. Therapeutic Utility: Proved that mRNA-delivered TruEditors significantly enhance precise editing in clinically relevant primary cells (hESCs and T cells) where standard methods fail.

4. Key Results

• Screen Output: Identified 805 genes enriched in the HDR population. Notable hits included known factors (BARD1, BLM, SLX1A) and novel candidates (ZNF146, TBX10). Many top hits were lowly expressed in K562 cells, highlighting the power of the gain-of-function approach.
• Editing Efficiency:
  • Full-Length TruEditors: 9 out of 20 candidates significantly improved HDR rates (up to ~1.5-fold over Cas9 alone).
  • Minimal TruEditors: The BLM N-terminal domain (residues 2–194) fused to Cas9 improved HDR by ~1.9-fold, outperforming the full-length BLM fusion.
  • Loci Specificity: TruEditors achieved precise editing rates of 25–40% at difficult loci, including the MUT R369H mutation (low GC content) where prime editing failed (<0.1%).
  • Cell Types: Improved editing in lung cancer (A549), brain tumor (BT16), and breast cancer (MDA-MB-231) cells.
• Safety Profile: Unlike global NHEJ inhibitors (e.g., M3814), TruEditors did not impair global DNA repair or increase 53BP1 foci accumulation after DNA damage, suggesting a superior safety profile.
• Mechanism: AP-MS confirmed that TruEditors recruit specific repair complexes (e.g., BLM recruits TOP3A/RMI1; BARD1 recruits BRCA1). Mutations disrupting these interfaces (e.g., N8A in BLM) abolished both the interaction and the HDR enhancement.
• Primary Cell Applications:
  • hESCs: mRNA delivery of TruEditors improved HDR rates by ~3-fold compared to Cas9 alone, while maintaining pluripotency (OCT3/4 expression).
  • CAR-T Cells: TruEditor mRNA delivery doubled the knock-in efficiency of a CD19 CAR into the TRAC locus. These engineered CAR-T cells showed significantly enhanced tumor cell killing compared to Cas9-only controls.

5. Significance

This study establishes a function-guided, genome-scale framework for protein engineering. Instead of relying on deep learning predictions of structure alone, the authors used a functional screen to identify proteins that perform a specific biological task (promoting HDR).

• Overcoming Limitations: TruEditors solve critical bottlenecks in gene therapy, such as low efficiency in primary cells, inability to edit low-GC loci, and the toxicity associated with global DNA repair inhibition.
• Clinical Relevance: The ability to efficiently knock-in large transgenes (like CARs) into primary T cells and stem cells using mRNA-delivered TruEditors represents a major step forward for next-generation cell therapies and the treatment of inborn genetic disorders.
• Future Direction: The study suggests that similar screens can be applied to other editing modalities (base/prime editing) and that the "function-first" approach can guide the design of synthetic proteins for any selectable phenotype. The use of human-derived domains also potentially reduces the immunogenicity risks associated with bacterial fusion proteins.