Discovery of Selective Nrf2 Activators from Natural Products: AComputational Screening Approach to Minimize Off-Target Effects on PXR and CYP2D6

This study presents a large-scale computational screening of nearly 630,000 natural products using a novel three-tier selectivity strategy to identify 10 ultraselective Nrf2 activators that effectively bind KEAP1 while minimizing off-target interactions with PXR and CYP2D6, thereby offering a promising pathway for developing safer therapies for oxidative stress-related diseases.

The Gist

Imagine your body is a bustling city. Sometimes, due to pollution, stress, or aging, the city gets covered in "rust" (oxidative stress). To clean this up, the city has a special cleanup crew called Nrf2. When Nrf2 is activated, it sends out thousands of workers to fix the damage and protect the city.

Scientists have long wanted to create a "super-activator" drug to wake up Nrf2 and treat diseases like Alzheimer's, heart disease, and cancer. But there's a catch: in the past, the drugs designed to wake up Nrf2 were like clumsy construction workers. While they fixed the rust, they accidentally knocked over other important things:

1. PXR (The Traffic Cop): They triggered the traffic cop to go crazy, causing massive traffic jams (drug-drug interactions) where other medicines couldn't work.
2. CYP2D6 (The Waste Management): They clogged the garbage trucks, meaning the body couldn't process other drugs properly, leading to dangerous toxic buildups.

This paper is about a team of researchers who decided to find a "smart" activator—one that wakes up the cleanup crew (Nrf2) without bothering the traffic cop or the garbage trucks.

The Great Natural Product Hunt

Instead of building new chemicals from scratch in a lab (which is like trying to invent a new shape from a lump of clay), the researchers went on a treasure hunt through nature. Nature has been evolving for millions of years, creating millions of unique chemical shapes (natural products) that are often very good at fitting into specific locks in our bodies.

They looked at a massive digital library called COCONUT, which contains nearly 630,000 different natural compounds. That's like searching through a library with 630,000 books to find just a few perfect sentences.

The "Three-Lock" Test

To find the perfect candidate, they didn't just look for something that fits the Nrf2 lock. They used a computer simulation to test every single one of the 630,000 compounds against three locks at the same time:

1. The Good Lock (KEAP1/Nrf2): They wanted a key that fits perfectly here to turn on the cleanup crew.
2. The Bad Lock 1 (PXR): They wanted a key that doesn't fit here at all, so it doesn't cause traffic jams.
3. The Bad Lock 2 (CYP2D6): They wanted a key that fits just a little bit (not too tight, not too loose) so it doesn't clog the garbage trucks.

The "Funnel" Strategy

Imagine a giant funnel with three layers of mesh screens:

• Layer 1 (The Rough Screen): They threw all 630,000 compounds through. About 46,000 made it through because they were "okay" at the good lock and "okay" at avoiding the bad locks.
• Layer 2 (The Fine Screen): They tightened the mesh. Only 3,700 compounds were good enough to pass. These were the "Quality Candidates."
• Layer 3 (The Super-Fine Screen): They made the holes tiny. Only 10 compounds made it through. These are the "Star Candidates."

The Winners: What Do They Look Like?

The researchers found that the 10 "Star Candidates" had some interesting features:

• They are "Fat" and "Round": Many of them looked like lipids (fats) or steroids. Think of them as having a 3D, bumpy shape (like a golf ball) rather than a flat, pancake-like shape.
• Why does shape matter? The "Bad Locks" (PXR and CYP2D6) are like big, open rooms that accept many different shapes (flat pancakes). The "Good Lock" (Nrf2) is a specific, narrow groove. The researchers found that the "bumpy, 3D" shapes fit perfectly into the narrow groove but were too weird-shaped to fit into the big open rooms. This is why they are so selective!

The Result: A New Blueprint for Safer Drugs

The team didn't just find 10 drugs; they created a new rulebook for drug discovery.

• Safety by Design: Instead of finding a drug and then checking if it's safe, they built safety into the search from the very beginning.
• The "Selectivity Score": They invented a math formula (like a report card) that gives a single number to tell you how good a drug is at being specific. The top candidate got a score that was 12 times better than average at being specific.

The Takeaway

This paper is like finding a needle in a haystack, but instead of just finding the needle, the researchers figured out exactly what the needle looks like so we can find more of them in the future.

They proved that by using computers to screen nature's library and focusing on shape and safety (not just power), we can find medicines that treat diseases without causing dangerous side effects. It's a shift from "throwing darts and hoping to hit the bullseye" to "aiming with a laser."

In short: They used a super-computer to find 10 nature-based "smart keys" that unlock the body's healing power without accidentally locking the doors to other vital systems. This paves the way for safer, more effective medicines for the future.

Technical Summary

1. Problem Statement

The activation of the Nuclear factor erythroid 2-related factor 2 (Nrf2) pathway is a promising therapeutic strategy for oxidative stress-related diseases (e.g., neurodegeneration, cardiovascular disease, cancer). However, the clinical translation of Nrf2 activators is severely hindered by off-target toxicities:

• Pregnane X Receptor (PXR) Activation: Unintended activation of PXR leads to severe drug-drug interactions (DDIs) by upregulating xenobiotic metabolism enzymes.
• CYP2D6 Inhibition: Potent inhibition of Cytochrome P450 2D6 disrupts the metabolism of ~25% of prescribed drugs, causing adverse metabolic complications.

Existing natural product studies often prioritize potency over selectivity. There is a critical lack of systematic efforts to identify natural Nrf2 activators that simultaneously possess high efficacy against KEAP1 (the Nrf2 repressor) while minimizing interactions with PXR and CYP2D6.

2. Methodology

The authors employed a "Safety-by-Design" computational framework involving a large-scale virtual screening campaign.

• Data Source: The study utilized the COCONUT database, screening 628,898 structurally diverse natural products.
• Target Selection: Molecular docking was performed against three targets:
  1. KEAP1 (Kelch domain): The primary target for Nrf2 activation.
  2. PXR (Ligand-binding domain): To assess DDI risk.
  3. CYP2D6 (Active site): To assess metabolic inhibition risk.
• Docking Engine: The KARMA (Kinetic And Rigid-body Molecular docking with Autodock) scoring function was used to generate binding affinity scores.
• Three-Tier Selectivity Screening Strategy:
  • Tier 1 (Candidate Pool): Relaxed criteria (KEAP1 >50th percentile, PXR <50th, CYP2D6 30th–70th).
  • Tier 2 (Quality Candidates): Stringent criteria (KEAP1 >70th, PXR <30th, CYP2D6 30th–70th).
  • Tier 3 (Star Candidates): Ultra-stringent criteria (KEAP1 >90th, PXR <10th, CYP2D6 40th–60th).
  • Note: The CYP2D6 range was intentionally set to "moderate" (40th–60th percentile) to avoid compounds that are either non-metabolized or strong inhibitors.
• Selectivity Indices: Three quantitative metrics were developed to rank compounds:
  • Nrf2 Selectivity Index (NSI): Ratio of KEAP1/PXR scores.
  • Safety Index (SI): Ratio of KEAP1/CYP2D6 scores.
  • Comprehensive Selectivity Score (CSS): A weighted metric integrating both selectivity dimensions.
• ADME & Chemical Analysis: Compounds were evaluated for Lipinski's Rule of Five, QED (Quantitative Estimate of Drug-likeness), TPSA, and chemical class enrichment (using NP Classifier).

3. Key Results

A. Screening Outcomes

• Funnel Efficiency: From 628,898 compounds, the screening identified:
  • Tier 1: 46,443 compounds (7.38%).
  • Tier 2: 3,767 compounds (0.60%).
  • Tier 3: 10 ultraselective "Star" candidates (0.002%).
• Selectivity Performance: The top candidate, CNP0371205.2, achieved an NSI of 12.29, indicating a >12-fold preference for KEAP1 over PXR. The top 10 compounds showed CSS values up to 2.64, significantly higher than the population average (0.66).

B. Structural and Chemical Insights

• Enriched Scaffolds:
  • Lipids/Lipid-like molecules: Highly enriched in Tier 2 (1.49-fold), particularly terpenoids and steroids. Their lipophilic nature likely aids in engaging the hydrophobic KEAP1 pocket while avoiding the promiscuous PXR pocket.
  • Nucleosides: Dramatically enriched (2.85-fold), suggesting unique hydrogen bonding patterns favor KEAP1 selectivity.
• Depleted Scaffolds: Phenylpropanoids, polyketides, and benzenoids were under-represented. Correlation analysis confirmed that high aromatic ring counts negatively correlate with selectivity (promoting promiscuous binding), while higher Fsp³ (fraction of sp³ carbons) correlates positively with selectivity.
• Drug-Likeness: Despite stringent selectivity filters, the top candidates maintained favorable drug-like properties. Most Tier 1 and 2 compounds complied with Lipinski's Rule of Five, and QED scores remained moderate to good (0.37–0.46), indicating that high selectivity does not necessitate "undruggable" structures.

C. Top Candidates

The study highlighted 10 Tier 3 compounds, including:

1. CNP0371205.2: A phenylpropanoid with the highest NSI (12.29) and CSS (2.64).
2. CNP0012224.1: An imidazopyridine derivative.
3. CNP0540007.1: A benzodiazepine derivative.
4. Several terpenoid/steroid derivatives (e.g., CNP0117966.1, CNP0348608.7).

4. Key Contributions

1. First Large-Scale Safety-Integrated Screening: This is the first study to simultaneously screen nearly 630,000 natural products against KEAP1, PXR, and CYP2D6 to prioritize safety alongside efficacy.
2. Novel Selectivity Metrics: The introduction of NSI, SI, and CSS provides a quantitative framework for ranking compounds based on off-target risk, moving beyond simple docking score thresholds.
3. Identification of Novel Scaffolds: The discovery of lipid-based and nucleoside-inspired scaffolds as highly selective Nrf2 activators offers new chemical starting points for drug design.
4. Open Data Resource: The authors released a comprehensive dataset and analysis code (Zenodo), bridging the gap between molecular docking, toxicological filtering, and cheminformatics.

5. Significance

This work establishes a scalable, data-driven pipeline for the rational development of safer therapeutics. By proving that selectivity and drug-likeness are synergistic (positive correlation between QED and CSS), the study challenges the notion that highly selective compounds must be structurally complex or "undruggable."

The "Safety-by-Design" approach demonstrated here can be adapted to other drug discovery campaigns to preemptively filter out off-target liabilities, potentially reducing late-stage clinical failures due to drug-drug interactions or metabolic toxicity. The identified 10 ultraselective candidates provide immediate leads for experimental validation in Nrf2-targeted therapies for oxidative stress-related diseases.