RaDRI: A computational model for radiosensitisation by DNA double strand break repair inhibitors

The authors developed the RaDRI agent-based computational model to investigate radiosensitisation by the DNA-PK inhibitor SN39536, revealing that while radiation-induced cell killing is primarily driven by DSB misjoining, the inhibitor enhances cell death mainly by preventing DSB resolution before mitosis and predicting a ~9-hour exposure duration is required for maximal effect.

The Gist

The Big Picture: A Digital "Cell Simulator"

Imagine you are trying to stop a runaway train (a cancer tumor) by throwing a wrench into its gears. In radiation therapy, the "wrench" is ionizing radiation, which breaks the DNA inside cancer cells. Usually, the cell has a repair crew (DNA repair pathways) that can fix these breaks and keep the train running.

The scientists in this paper built a sophisticated computer simulation called RaDRI (Radiosensitisation by DSB Repair Inhibitors). Think of RaDRI as a highly detailed "flight simulator" for cancer cells. Instead of just guessing how drugs work, they created a virtual world where they can watch thousands of individual cells, see exactly how their DNA breaks, how the repair crew tries to fix it, and what happens when you add a drug that jams the repair crew's tools.

The Problem: The Repair Crew is Too Good

When you zap a cell with radiation, it creates "Double-Strand Breaks" (DSBs). Imagine these as a rope that has been cut completely in half.

• The Repair Crew (NHEJ): Cells have a team called Non-Homologous End-Joining (NHEJ). Their job is to grab the two cut ends of the rope and tie them back together.
• The Risk: Sometimes, the crew gets confused. Instead of tying the two ends of one rope together, they accidentally tie the end of Rope A to the end of Rope B. This is called "Misjoining." It's like tying your shoelaces to someone else's pants; it creates a mess (chromosome aberrations) that usually kills the cell.

The Solution: The "Jammer" Drug (SN39536)

The researchers tested a new drug called SN39536. This drug acts like a super-glue jammer. It stops the repair crew from tying the ropes back together.

• Goal: If the ropes stay broken, the cell dies when it tries to divide.
• The Catch: If you jam the repair crew too early, they might just wait until you leave before fixing the mess. If you jam them too late, they've already finished the job. The big question was: How long do you need to keep the jammer in place to get the best result?

How the Simulator Works (The Metaphors)

1. The Cell Cycle as a Factory Assembly Line

The computer model treats every cell as an individual worker on an assembly line.

• G1, S, G2, M: These are the different stages of the workday.
  • S-Phase: The worker is making a photocopy of the blueprints (DNA replication).
  • M-Phase: The worker tries to split the factory into two new factories (Mitosis).
• The Checkpoint: Before the worker splits the factory, a security guard (the Checkpoint) checks the blueprints. If the blueprints are torn (broken DNA), the guard stops the worker from splitting. This gives the repair crew time to fix the tears.

2. The "Confusion" of the Repair Crew

The model discovered something interesting about how the repair crew works:

• Simple Breaks: If the rope is just cut cleanly, the crew fixes it instantly (Fast NHEJ).
• Complex Breaks: If the rope is frayed and messy, the crew takes a long time to fix it (Slow NHEJ).
• The Drift: While the crew is working on the messy breaks, the broken ends of the rope start to wander around the nucleus (like a lost dog in a park). The longer they wander, the more likely they are to bump into other broken ropes and get tied together incorrectly (Misjoining).
• The Drug's Effect: When you add the jammer drug, the repair crew stops working. The broken ropes keep wandering for hours. By the time the drug wears off, the ropes have wandered so far that they tie themselves into a giant, tangled knot. When the cell tries to divide, it explodes.

3. The "Time Window" Discovery

This is the most important finding of the paper.

• The Myth: You might think you need to keep the drug in the system forever to stop the repair.
• The Reality: The simulation showed that you only need to keep the "jammer" active for about 9 hours.
  • Why? It takes about 9 hours for the broken DNA ends to wander around enough to cause a fatal tangle (misjoin). Once that tangle is formed, the cell is doomed, even if you wash the drug away. Keeping the drug around longer doesn't help much more, but it might hurt healthy cells.

What the Model Predicted vs. Reality

The researchers fed real data from lab experiments (using HCT116 cancer cells) into their computer.

• The Match: The computer predicted exactly how many cells would die at different radiation doses and drug concentrations.
• The Surprise: They found that the main reason the drug works isn't just that it leaves "broken ropes" (unrepaired DNA). It's that it forces the cell to make terrible knots (misjoins).
• The Checkpoint Twist: The model also showed that the cell's "security guard" (the checkpoint) tries to save the cell by delaying division. However, the drug makes the guard work overtime. If the guard is too effective, the cell survives longer, but eventually, the tangled DNA causes a catastrophic failure.

The Bottom Line for Patients

This paper is a blueprint for better cancer treatment.

1. Timing is Everything: You don't need to bombard a patient with the drug for days. A specific window (around 9 hours) is likely the "sweet spot" to maximize tumor killing while minimizing side effects.
2. Combination Therapy: Since the drug relies on the cell's own repair mechanisms failing, it might work even better if combined with drugs that stop the "security guard" (checkpoint inhibitors), forcing the cell to divide before the mess is cleaned up.
3. Virtual Testing: Instead of testing every possible drug dose on real animals or humans, doctors can now use this "flight simulator" (RaDRI) to predict the best dosing schedules for new drugs before they ever reach a patient.

In short: The scientists built a digital microscope to watch cancer cells struggle with broken DNA. They found that by jamming the repair crew for just the right amount of time, you can force the cells to tie their own shoelaces together, causing them to trip and fall (die) when they try to divide.

Technical Summary

1. Problem Statement

Radiation therapy kills cancer cells primarily by inducing DNA double-strand breaks (DSBs). The repair of these breaks, particularly via Non-Homologous End-Joining (NHEJ) mediated by DNA-dependent protein kinase (DNA-PK), is a major mechanism of radioresistance. While DNA-PK inhibitors (DNA-PKi) are effective radiosensitisers, several critical questions remain unanswered regarding their clinical application:

• What are the relative contributions of unrepaired DSBs versus misjoined DSBs (chromosome aberrations) to enhanced cell killing?
• How do cell cycle checkpoints interact with DNA-PK inhibition?
• What is the required duration of DNA-PK inhibition to achieve maximal radiosensitisation?
• Existing models often rely on phenomenological survival curves (e.g., Linear-Quadratic shifts) rather than mechanistic descriptions of DSB repair pathways and cell cycle dynamics, limiting their utility in pharmacokinetic/pharmacodynamic (PK/PD) modeling for tumor microenvironments.

2. Methodology

The authors developed RaDRI (Radiosensitisation by DSB Repair Inhibitors), an agent-based computational model designed to simulate single-cell responses to radiation and DNA-PK inhibition.

• Model Architecture:

  • Agent-Based: Simulates populations of ~10,000 individual cells, each with unique cell cycle progression rates and random variations in cycle duration.
  • Foundation: Builds upon the Medras model (McMahon et al.) for DSB formation and repair but introduces significant extensions.
  • Key Extensions:
    1. Explicit Cell Cycle: Includes S-phase (often the most radioresistant phase) and explicit progression through G1, S, G2, and M.
    2. Checkpoint Dynamics: Incorporates a mechanistic model of ATM and ATR signaling (based on Jaiswal et al.) to simulate G1, S, and G2 checkpoint delays. Crucially, it models the dependence of ATR activation on DNA-PK catalytic activity.
    3. Repair Pathway Assignment: DSBs are assigned to Fast NHEJ, Slow NHEJ, or Homologous Recombination (HR). The probability of HR assignment decreases over time post-replication (due to chromatin maturation) and is suppressed at high radiation doses.
    4. Misjoining Mechanism: Models the probability of DSB misjoining (ligation of ends from different breaks) as increasing over time due to active DSB clustering (mobility), a feature enhanced by DNA-PK inhibition.
    5. Inhibition Logic: SN39536 (a potent DNA-PKi) inhibits DNA-PK catalytic activity, reducing NHEJ repair rates and altering ATR signaling.

• Experimental Validation & Parameterization:

  • Cell Lines: Used HCT116 (pseudodiploid) and isogenic lines (HCT116/ATCC, DLD1 BRCA2-/-).
  • Compounds: SN39536 (DNA-PKi) and ART558 (Pol$\theta$/TMEJ inhibitor).
  • Datasets:
    • CCPD: Cell Cycle Phase Distribution (flow cytometry with EdU/pH3/PI) to fit checkpoint parameters.
    • CSF: Clonogenic Surviving Fraction (colony formation assays) to fit repair kinetics and lethality parameters.
  • Optimization: Used PEST_HP (gradient-based) and CMAES_HP (global stochastic) algorithms to fit ~26 parameters against experimental data.

3. Key Contributions

• Mechanistic Insight into Radiosensitisation: The model dissects the dual role of DNA-PK inhibition. It reveals that while inhibiting NHEJ increases unrepaired DSBs, the primary driver of radiosensitisation in log-phase cells is the enhancement of DSB misjoining (chromosomal aberrations) due to prolonged DSB lifetime and increased mobility.
• Checkpoint-Drug Interaction: The model demonstrates that DNA-PK inhibition compromises ATR activation (which relies on DNA-PK phosphorylation of RPA32). This leads to altered G2 checkpoint dynamics, where ATM signaling becomes the dominant brake on mitotic entry when DNA-PK is inhibited.
• Time-Dependence of Sensitisation: The study quantifies the temporal requirements for effective radiosensitisation, moving beyond static "dose-response" views to dynamic "exposure-time" views.
• Exclusion of TMEJ: Experimental data showed that inhibiting Pol$\theta$ (TMEJ pathway) did not significantly enhance radiosensitisation by SN39536 in HCT116 cells, justifying the exclusion of TMEJ from the current model iteration.

4. Key Results

• Mechanism of Killing:
  • Radiation Alone: Cell killing is driven almost exclusively by misjoining (chromosome aberrations), not by unrepaired DSBs. Unrepaired DSBs contribute minimally because checkpoints allow time for repair.
  • Radiation + SN39536: Unrepaired DSBs (specifically those in post-replication DNA) become a significant driver of cell death. However, the model predicts that checkpoint delays protect cells from this unrepaired damage; thus, the synergy between DNA-PKi and checkpoint inhibitors (like ATRi) is theoretically predicted.
• Drug Exposure Duration:
  • To achieve 90% of maximal radiosensitisation, DNA-PK inhibition must be maintained for approximately 9 hours.
  • 50% of maximal sensitisation is achieved with an exposure time of roughly 2.1 hours.
  • Simulations suggest that a drug with a plasma half-life of ~2 hours is sufficient for 50% efficacy, but longer half-lives (or continuous exposure) are needed for maximal effect.
• Model Performance:
  • RaDRI successfully reproduced the near-linear-quadratic dose dependence of radiation killing and the shift to a more linear survival curve at high DNA-PKi concentrations.
  • The model accurately predicted cell cycle phase distributions (G2 arrest) and clonogenic survival fractions across various radiation doses and drug concentrations.
• Cell Cycle Specificity: The model confirmed that cells irradiated in S-phase are radioresistant due to HR availability, but this resistance is modulated by the time-dependent decay of HR probability (chromatin maturation).

5. Significance

• PK/PD Integration: RaDRI is designed to be computationally tractable for integration into larger-scale agent-based models of tumor microenvironments. This allows researchers to simulate how drug concentration gradients and tumor hypoxia affect radiosensitisation in vivo.
• Clinical Optimization: The finding that ~9 hours of inhibition is required for maximal effect provides a critical guideline for clinical dosing schedules. It suggests that short-acting inhibitors may require frequent dosing or continuous infusion to be effective, while longer-acting analogs could improve therapeutic windows.
• Combination Therapy Rationale: The model provides a mechanistic basis for combining DNA-PKi with ATR inhibitors. Since DNA-PK inhibition weakens ATR signaling (and thus the G2 checkpoint), combining these agents could force cells with unrepaired DSBs into mitosis, leading to catastrophic cell death.
• Future Directions: The authors plan to use RaDRI to model hypoxia-activated prodrugs delivering DNA-PKi to radioresistant tumor regions, addressing the issue of tumor selectivity.

In summary, RaDRI represents a significant advancement in radiobiological modeling by bridging the gap between molecular repair kinetics, cell cycle checkpoint dynamics, and macroscopic cell survival, offering a powerful tool for optimizing DNA-PK inhibitor-based radiotherapy strategies.