AI-Guided CRISPR Screen Accelerates Discovery of New Drug Targets

This study integrates an AI-driven CRISPR screen with multi-omics validation to identify ALOX5 and OXTR as novel drug targets for psoriasis, demonstrating that their inhibition via existing topical agents effectively suppresses inflammation and epidermal hyperproliferation in preclinical models.

The Gist

Imagine your skin is a bustling city. In a healthy city, the buildings (skin cells) are well-organized, the streets are clean, and the security guards (immune cells) only wake up if there's a real emergency.

Psoriasis is like a city in a state of permanent, chaotic riot. The buildings are growing too fast and piling up on top of each other (thick, scaly skin), and the security guards are screaming "Fire!" even though there is no fire. This causes redness, pain, and itching.

For years, the only way to stop this riot was to send in a massive, expensive, all-hands-on-deck military unit (biologic drugs) that travels through the whole body. While effective, these "military units" are hard to get, very costly, and can sometimes cause side effects because they shut down the whole security system, not just the specific troublemakers.

This paper is about a new, smarter way to stop the riot: finding the specific "alarm buttons" inside the skin cells themselves and pressing them to turn the noise down.

Here is how they did it, broken down into simple steps:

1. The "Wanted" Poster (The CRISPR Screen)

The scientists wanted to find out: What makes the skin cells in psoriasis so loud and reactive?

They used a high-tech tool called CRISPR (think of it as a molecular "find-and-replace" function for DNA). They took a library of instructions for almost every gene in the human body (about 19,000 genes) and "deleted" them one by one in skin cells.

They were looking for a specific reaction: The IL-17 Receptor. You can think of this receptor as the "microphone" on the skin cell that amplifies the inflammatory scream. If they deleted a gene and the microphone suddenly went silent, that gene was a suspect.

2. The AI Detective (VirtualCRISPR)

After deleting 19,000 genes, they had a list of thousands of suspects. Some were obvious (like genes we already knew were involved in skin inflammation). But they wanted the new suspects—the ones nobody had thought of before.

This is where AI came in. They used a smart computer model called VirtualCRISPR.

• The Analogy: Imagine a detective who has read every book ever written about crime. If you show the detective a suspect, the detective says, "I've seen this guy before; he's a known criminal."
• The Twist: The scientists told the AI, "We don't want the known criminals. We want the guys who look guilty based on the evidence, but who never show up in the history books."

The AI scanned the list and pointed to two surprising suspects that no one had ever linked to psoriasis before:

1. ALOX5: A gene involved in making fatty acids (lipids).
2. OXTR: A gene that usually responds to oxytocin (the "love hormone").

3. The "Aha!" Moment (How they work)

The scientists dug deeper to understand why these two genes were making the skin scream.

• ALOX5 is like a "Sticky Note" dispenser. In psoriasis, this gene produces a sticky substance that keeps the "microphone" (IL-17 receptor) glued to the surface of the cell, so it can't be taken away. It keeps the alarm blaring.
• OXTR is like a "Volume Knob" connected to calcium. When this gene is active, it turns up the volume on the alarm by changing how the cell uses energy and calcium.

4. The Solution (Old Drugs, New Use)

The best part? The scientists didn't have to invent new drugs from scratch. They found existing, FDA-approved drugs that already block these two genes:

• Zileuton: A drug already used for asthma (blocks ALOX5).
• Cligosiban: A drug used to stop premature labor (blocks OXTR).

They turned these into creams and applied them directly to the skin of mice with psoriasis.

5. The Result

The creams worked just as well as the expensive, systemic "military unit" drugs, but with a huge advantage:

• Targeted: They only worked on the skin, not the whole body.
• Mechanism: They didn't just silence the alarm; they actually fixed the cell's behavior.
  • The Zileuton cream stopped the "sticky notes," allowing the cell to clean up its own receptors.
  • The Cligosiban cream turned down the "volume knob," calming the cell's energy usage and stopping the overgrowth.

Why This Matters

This paper is a blueprint for the future of medicine. It shows that by combining genetic screening (finding the suspects) with AI (filtering for the unknown) and drug repurposing (using old keys for new locks), we can find cheap, easy-to-make, topical cures for complex diseases like psoriasis.

Instead of sending a tank to the neighborhood, we just found the specific fuse box in the house and turned it off.

Technical Summary

1. Problem Statement

• Clinical Gap: Psoriasis affects over 125 million people globally. While systemic biologics targeting the IL-17/IL-17RA axis (e.g., secukinumab) are effective, they are costly, require systemic administration, and carry risks of immunogenicity and side effects. Topical corticosteroids suffer from tachyphylaxis and skin atrophy. There is an urgent need for novel, topical small-molecule therapeutics.
• Biological Gap: The mechanistic understanding of how keratinocytes (the primary effector cells in psoriasis) regulate IL-17 receptor A (IL17RA) expression is incomplete.
• Technical Barrier: Genome-wide CRISPR screens in primary human adult epidermal keratinocytes (HEKa) have been technically challenging due to low transduction efficiency and limited proliferative capacity, leading to a lack of comprehensive data on keratinocyte-specific regulatory networks.
• Discovery Bottleneck: Traditional CRISPR screens generate thousands of hits, making it difficult to distinguish between well-characterized biology and novel, druggable targets without extensive manual curation.

2. Methodology

The authors developed an integrated platform combining primary cell functional genomics, AI-driven target prioritization, and multi-omics validation.

• Genome-Wide CRISPR Screen:

  • Cell Model: Primary human adult epidermal keratinocytes (HEKa) from two healthy donors.
  • Library: Brunello genome-wide CRISPR knockout library (~77,000 sgRNAs targeting ~19,000 genes).
  • Optimization: Overcame HEKa transduction challenges using spinoculation without polybrene (which inhibits proliferation) and optimized viral incubation (MOI ~0.35).
  • Phenotypic Sorting: Used Fluorescence-Activated Cell Sorting (FACS) to isolate the top 5% (IL17RA-high) and bottom 5% (IL17RA-low) surface expressing cells.
  • Analysis: Deep sequencing of sgRNAs followed by MAGeCK analysis to identify genes whose knockout altered IL17RA surface expression.

• AI-Guided Prioritization (VirtualCRISPR):

  • Model: Utilized VirtualCRISPR, a Large Language Model (LLM) trained on the BioGRID-ORCS database of functional genomics screens.
  • Strategy: Instead of ranking hits by experimental score alone, the model was used as a novelty filter.
    • High Experimental Enrichment + High AI Probability: Known biology (internal validation).
    • High Experimental Enrichment + Low AI Probability: High-novelty candidates (mechanisms not previously linked to the phenotype in the training data).
  • Selection: Identified ALOX5 (Arachidonate 5-lipoxygenase) and OXTR (Oxytocin receptor) as high-confidence novel hits with minimal prior association with psoriasis.

• Multi-Omics Validation:

  • In Vitro: Targeted CRISPR knockout in primary keratinocytes to confirm IL17RA downregulation.
  • 3D Organotypic Skin: Human skin raft cultures treated with inhibitors (Zileuton for ALOX5, Cligosiban for OXTR) followed by quantitative proteomics (LC-MS/MS) to assess cell-autonomous effects.
  • Single-Cell RNA Sequencing (scRNA-seq): Profiling of peripheral blood mononuclear cells (PBMCs) from psoriasis patients vs. controls to assess systemic relevance.
  • In Vivo: Imiquimod (IMQ)-induced psoriasis mouse model treated with topical gels of Zileuton and Cligosiban. Outcomes measured via PASI scoring, histology, flow cytometry, and proteomics.

3. Key Contributions

1. First Genome-Wide CRISPR Screen in Primary Keratinocytes: Successfully executed a screen in difficult-to-transduce primary human keratinocytes to map IL17RA regulators.
2. AI-Enhanced Target Discovery: Demonstrated a novel workflow where LLMs act as a "novelty filter" to prioritize uncharacterized targets from massive screening datasets, significantly accelerating the path from hit identification to drug repurposing.
3. Identification of Novel Targets: Uncovered ALOX5 (lipid metabolism) and OXTR (neurohormonal signaling) as critical upstream regulators of IL17RA in keratinocytes, distinct from the canonical IL-17 signaling axis.
4. Mechanistic Elucidation:
   • ALOX5: Regulates IL17RA surface stability via leukotriene signaling (specifically LTB4), preventing receptor internalization. Inhibition promotes dynamin-dependent endocytosis of IL17RA.
   • OXTR: Regulates IL17RA via a calcium-calcineurin-NFAT signaling axis.
5. Therapeutic Proof-of-Concept: Showed that repurposing FDA-approved drugs (Zileuton and Cligosiban) as topical agents achieves efficacy comparable to systemic anti-IL17RA antibodies.

4. Key Results

• Screen Performance: The screen identified known regulators (e.g., IL17RA, KAT5) validating the assay, while the AI filter successfully isolated ALOX5 and OXTR as high-novelty hits.
• Validation: CRISPR knockout of ALOX5 and OXTR significantly reduced IL17RA surface expression in primary keratinocytes.
• Mechanism of Action (In Vitro):
  • Zileuton (ALOX5 inhibitor): Induced metabolic reprogramming in keratinocytes (shift from anabolic lipid synthesis to catabolic fatty acid oxidation), suppressed inflammatory cytokines (IL-6, IL-1β), and reduced IL17RA surface levels by promoting receptor internalization.
  • Cligosiban (OXTR antagonist): Triggered broad metabolic reprogramming (suppression of mevalonate/cholesterol biosynthesis and sphingolipid pathways) and attenuated calcium-dependent inflammatory signaling.
• In Vivo Efficacy (Mouse Model):
  • Topical application of Zileuton and Cligosiban gels significantly reduced PASI scores, epidermal thickness (acanthosis), and inflammation by Day 7, matching the efficacy of systemic anti-IL17RA antibodies.
  • Immunomodulation: Both treatments suppressed pathogenic Th17/Tc17 cells. Notably, they uniquely polarized macrophages toward an anti-inflammatory M2 phenotype (CD206+, Arg1+), an effect not fully recapitulated by the antibody alone.
  • Proteomics: Both treatments converged on suppressing neutrophil effectors (elastase, myeloperoxidase), alarmins (S100A8/A9), and the IL-17/IL-17RA inflammatory loop, while restoring epidermal differentiation markers (filaggrin, involucrin).
• Systemic Impact: Unlike systemic biologics, the topical treatments did not induce splenomegaly, suggesting a localized therapeutic effect with reduced systemic immune burden.

5. Significance

• Paradigm Shift in Drug Discovery: This study provides a blueprint for integrating AI-driven novelty filtering with functional genomics to accelerate the discovery of non-obvious therapeutic targets in complex inflammatory diseases.
• New Therapeutic Avenue: It establishes that targeting upstream regulators of IL-17RA (specifically lipid and neurohormonal pathways) via topical small molecules is a viable strategy to treat psoriasis, potentially offering a safer, more accessible alternative to systemic biologics.
• Drug Repurposing: The rapid translation of existing FDA-approved drugs (Zileuton for asthma, Cligosiban for preterm labor) to dermatology demonstrates the power of this integrated platform to shorten the timeline from target discovery to preclinical validation.
• Mechanistic Insight: The work reveals that keratinocyte-intrinsic metabolic and signaling pathways (arachidonic acid metabolism and calcium signaling) are central to maintaining the inflammatory state in psoriasis, offering new nodes for intervention beyond direct cytokine blockade.