On repeatability and directionality of collateral effects of drug resistance evolution

This paper proposes a framework integrating pharmacodynamics and population genetics to explain the non-repeatable and unidirectional patterns of collateral effects observed during drug resistance evolution, demonstrating how drug concentration and selection regimes critically influence these evolutionary outcomes.

The Gist

The Big Picture: The "Evolutionary Trade-Off"

Imagine you are a bacteria living in a world full of antibiotics (drugs). To survive, you have to evolve. But evolution isn't magic; it's a series of compromises.

This paper asks a simple question: When a bacteria evolves to survive Drug A, does it become stronger or weaker against Drug B?

• Cross-Resistance: The bacteria gets a "super shield" that protects it from both drugs. (Bad for us, good for the bacteria).
• Collateral Sensitivity: The bacteria evolves a shield against Drug A, but in doing so, it accidentally breaks its armor against Drug B, making it more vulnerable. (Good for us, bad for the bacteria).

The authors want to know: Can we predict this? If we repeat the experiment 100 times, will the bacteria always make the same mistake (Collateral Sensitivity), or will they sometimes get lucky and become super-resistant? And does it matter which drug we hit them with first?

---

The Core Concepts: The "Fitness Landscape" Analogy

To understand the paper, imagine a mountain range where the height of the mountain represents how well the bacteria can survive.

• Peaks are high places where bacteria are very fit (resistant).
• Valleys are low places where bacteria die.

The paper explores how bacteria climb these mountains and what happens to their "other skills" (susceptibility to other drugs) along the way.

1. Repeatability: The "Lucky Dice" vs. The "Fixed Path"

• Repeatable: Imagine a narrow, single-file path up a mountain. No matter how many times you send a group of bacteria up, they all take the exact same path and end up at the same spot. If they end up vulnerable to Drug B every time, the effect is repeatable.
• Non-Repeatable: Imagine a mountain with two different paths to the top. Sometimes, the bacteria take Path 1 and end up vulnerable to Drug B. Other times, they take Path 2 and end up resistant to Drug B. Because the outcome changes every time, it is non-repeatable.

The Paper's Insight: Whether the path is fixed or random depends on how many "steps" (mutations) are available and how crowded the mountain is (population size).

2. Directionality: The "One-Way Street" vs. The "Roundabout"

• Bidirectional: It's a roundabout. If you start at Drug A and go to Drug B, you get a specific result. If you start at Drug B and go to Drug A, you get the same result. The order doesn't matter.
• Unidirectional: It's a one-way street. If you start with Drug A, you end up vulnerable to Drug B. But if you start with Drug B, you don't end up vulnerable to Drug A. The order matters completely.

---

The Secret Sauce: How the Paper Explains It

The authors built a mathematical model (a "recipe") to explain why these patterns happen. They used two main ingredients:

1. The "Cost of Resistance" (The Backpack Analogy)

Evolution isn't free. Gaining a shield against a drug often weighs you down, like wearing a heavy backpack.

• Single Mutation: You put on one heavy backpack (resistance to Drug A). You are strong against A, but the backpack makes you slow against B.
• Double Mutation: You put on two heavy backpacks (resistance to A and B). You are super strong, but you are so slow and clumsy that you might not survive at all unless the drugs are very strong.

The paper shows that drug concentration acts like the terrain.

• Low Drug Concentration: The "single backpack" is enough to survive. The bacteria takes the easy path.
• High Drug Concentration: The single backpack isn't enough. The bacteria must take the double backpack to survive. This changes the outcome!

2. The "Selection Regime" (The Race Analogy)

How do the bacteria compete?

• The "Sweep" (Strong Selection): Imagine a race where the fastest runner wins instantly, and everyone else is eliminated. If there is only one "fastest" mutation, the result is repeatable.
• The "Clonal Interference" (Competition): Imagine a crowded race where many runners are close in speed. They bump into each other, and the winner is decided by luck. If two different mutations can both win, the result is non-repeatable (sometimes one wins, sometimes the other).

---

The "Aha!" Moments from the Paper

The authors discovered that the "rules" of evolution change based on the environment:

1. Concentration is King: If you change the dose of the drug, you can flip the script. A scenario that was "repeatable" (always the same outcome) can become "non-repeatable" (random outcome) just by increasing the drug concentration.
2. Order Matters: Sometimes, treating a patient with Drug A then Drug B works perfectly (because Drug A makes the bacteria weak to Drug B). But if you treat them with Drug B first, that weakness disappears. This is Unidirectionality.
3. Time is a Factor: Evolution takes time. A bacteria might start with a mutation that makes it weak to Drug B, but if you wait long enough, a second mutation might appear that fixes that weakness. The "collateral effect" isn't static; it changes over time.

The Takeaway for Real Life

Why does this matter?

If doctors want to use Collateral Sensitivity (the strategy of using Drug B to kill bacteria that are resistant to Drug A), they need to know:

• Will it work every time? (Repeatability)
• Does the order of drugs matter? (Directionality)

This paper provides a framework to predict the answer. It tells us that we can't just look at the bacteria; we have to look at the drug dose, the speed of evolution, and the genetic options available to the bacteria.

In short: Evolution is like a game of chess. The paper explains that the best move (the drug strategy) depends entirely on the board state (drug concentration) and how many pieces are on the board (population dynamics). If you understand the rules of the board, you can predict where the bacteria will go next.

Technical Summary

1. Problem Statement

The evolution of drug resistance often leads to collateral effects, where adaptation to one drug alters susceptibility to another. These effects manifest as:

• Cross-resistance: Increased resistance to a second drug.
• Collateral Sensitivity (CS): Increased susceptibility to a second drug.

Experimental studies have observed diverse patterns in these effects, specifically regarding:

• Repeatability: Whether repeated evolution from the same parental strain yields the same collateral effect.
• Directionality (Reciprocity): Whether the collateral effect is symmetric (bidirectional) or asymmetric (unidirectional) depending on the order of drug exposure.
• Temporal Dynamics: How these effects change over time during adaptation.

Despite extensive experimental data, the underlying genetic architecture and evolutionary conditions (e.g., drug concentration, selection regimes) that generate these specific patterns remain incompletely characterized. There is a need for a theoretical framework that links pharmacodynamics (drug response) with population genetics to explain why certain patterns emerge and others do not.

2. Methodology

The authors propose a theoretical framework integrating pharmacodynamics and population genetics to model the evolution of collateral effects.

• Pharmacodynamic Model:

  • Uses a dose-response curve to describe the relationship between drug concentration ($c$) and population growth rate ($\psi$).
  • Key parameters include the Minimum Inhibitory Concentration (MIC) and the maximum growth rate ($\psi_{max}$).
  • Mutations are assumed to shift the MIC (increasing for resistance, decreasing for sensitivity) and potentially incur a fitness cost (reducing $\psi_{max}$).
  • The model assumes additive effects for multiple mutations regarding MIC shifts and costs.

• Population Genetic Model:

  • Assumes resistance is encoded by multiple loci, each with two alleles (wild-type and mutant).
  • Defines Mutation-MIC Shifts: A mapping of how specific mutations at specific loci affect the MIC of different drugs (Drug A vs. Drug B).
  • Analyzes Mutant Selection Windows (MSW): The drug concentration ranges where specific mutants outcompete the wild type.
  • Considers two primary selection regimes:
    1. Selective Sweeps: Strong selection, rare mutations (one mutant fixes before another arises).
    2. Clonal Interference: Competition between multiple beneficial mutants arising simultaneously.

• Approach:

  • The authors construct minimal examples using the fewest possible loci required to generate specific patterns of repeatability and directionality.
  • They simulate evolutionary trajectories under varying drug concentrations and selection regimes to determine the stable genotype composition and resulting collateral effects.

3. Key Contributions

The paper provides a mechanistic explanation for the diversity of collateral effect patterns observed in experiments:

1. Bidirectional, Repeatable Cross-Resistance:

   • Mechanism: Can be explained by a single locus where a mutation confers resistance to both Drug A and Drug B.
   • Outcome: Regardless of the starting drug, the same mutant fixes, leading to consistent cross-resistance in both directions.

2. Bidirectional, Repeatable Collateral Sensitivity:

   • Mechanism: Requires at least two loci. One mutation confers resistance to A (sensitivity to B), and a different mutation confers resistance to B (sensitivity to A).
   • Outcome: Selection by either drug forces the population to fix the specific mutation that confers resistance to that drug, inevitably creating sensitivity to the other.

3. Unidirectional, Repeatable Cross-Resistance:

   • Mechanism: Requires two loci with asymmetric effects.
     • Mutation 1: Low resistance to A, Low resistance to B.
     • Mutation 2: High resistance to B, No resistance to A.
   • Outcome:
     • A $\to$ B: Adaptation to A fixes Mutation 1 (low resistance), which confers cross-resistance to B.
     • B $\to$ A: Adaptation to B fixes Mutation 2 (high resistance), which has no effect on A.
   • Condition: This relies on a specific drug concentration where the "low resistance" mutant cannot survive alone (extinction regime) or is outcompeted by the "high resistance" mutant (competition regime).

4. Non-Repeatability:

   • Mechanism: Requires multiple loci (at least three) where different resistant genotypes have distinct collateral effects.
   • Outcome: Depending on stochasticity (drift) or clonal interference, different resistant mutants may fix in different experimental replicates. Some may confer cross-resistance, others may not, leading to non-repeatable outcomes.

5. Temporal Dynamics:

   • The paper demonstrates that collateral effects are not static. If the double mutant (resistant to both) is the fittest genotype, the population may evolve through intermediate states (single mutants) that exhibit different collateral effects (e.g., transient sensitivity followed by cross-resistance) before the double mutant fixes.

4. Results

• Drug Concentration is Critical: The same genetic architecture can produce different patterns (repeatable vs. non-repeatable, unidirectional vs. bidirectional) solely based on the drug concentration.
  • Low/Moderate Concentration: May favor single mutants, leading to specific directional outcomes.
  • High Concentration: May favor double mutants, potentially turning non-repeatable scenarios into repeatable ones (as the double mutant always fixes).
• Selection Regime Matters:
  • Selective Sweeps (rare mutations) tend to produce repeatable outcomes if a single mutant is clearly superior.
  • Clonal Interference (common mutations) increases the likelihood of non-repeatability, as different mutants compete, and the "winner" depends on stochastic arrival times.
• Cost of Resistance: The assumption that resistance carries a fitness cost is crucial. If costs are absent, double mutants might fix via drift, altering the predicted patterns of directionality and repeatability.

5. Significance

• Clinical Optimization: The framework helps predict when collateral sensitivity can be reliably exploited for sequential drug therapies. It highlights that the success of such strategies depends heavily on the specific drug concentration and the evolutionary history of the pathogen.
• Experimental Design: The authors argue that measuring only the final MIC is insufficient. To understand collateral effects, researchers must measure full dose-response curves for resistant genotypes and account for the selection regime (mutation rates, population size) and drug concentration used during adaptation.
• Generalizability: While focused on bacteria and antibiotics, the framework applies to any system involving drug resistance evolution, including cancer chemotherapy and insecticide resistance, providing a unified theoretical basis for interpreting complex evolutionary trajectories.

In summary, the paper establishes that the repeatability and directionality of collateral effects are not intrinsic properties of the drugs alone but are emergent properties resulting from the interaction between genetic architecture (mutation-MIC maps), drug concentration, and population dynamics.