A Mathematical Model of Axon Guidance by Diffusible Factors

Geoffrey J. GoodhillNeurIPS 1997

Introduction

In a developing embryo, neurons must send out long probes called axons to connect with distant targets. One primary way they navigate is by "sniffing" out gradients of diffusible proteins, such as netrin-1, secreted by their destination. However, experimental data consistently shows a biological "range limit": axons seem to lose their way if the target is more than about 0.5 to 1 mm away.

Researchers have long debated whether this limitation occurs because the signal becomes too faint (low concentration) or because the gradient becomes too flat to detect. This paper provides a mathematical explanation for this phenomenon, demonstrating that the physical limits of a growth cone's sensitivity dictate the maximum range of this biological GPS.

Mathematical Framework

To understand these limits, the author develops a mathematical model based on radially symmetric Fickian diffusion in an infinite three-dimensional volume. The model relies on a specific solution to the diffusion equation for a source releasing a factor at a constant rate ($q$) with a specific diffusion constant ($D$).

The author defines two critical constraints for a growth cone (the sensing apparatus at the tip of the axon) to successfully navigate:

  1. Absolute Concentration ($C$): The signal must be high enough to be detected but low enough not to saturate the receptors.
  2. Fractional Change ($p$): There must be a sufficient percentage difference in concentration across the physical width of the growth cone ($\Delta r$) to overcome stochastic noise.

By pulling empirical parameters from existing biological literature—such as growth cone diameters ($\approx 10\text{--}20\ \mu\text{m}$) and minimum detectable concentration gradients ($\approx 1\text{--}2\%$)-the model establishes a "ground truth" to validate its predictions against observed biological phenomena.

Key Insights and Results

The 1 mm Wall and the Asymptotic Limit

For "large times" (after the gradient has stabilized), the fractional change $p$ simplifies to approximately $\Delta r / r$. This means the maximum guidance range is essentially the width of the growth cone divided by its minimum sensing threshold. Using a $10\ \mu\text{m}$ growth cone and a $1\%$ sensitivity threshold, the math yields a maximum range of exactly $1000\ \mu\text{m}$ (1 mm). This provides a robust, parameter-independent explanation for the $500\text{--}1000\ \mu\text{m}$ limit observed in in vivo and in vitro experiments.

Spatiotemporal Mapping

The paper provides "guidance maps" showing how the window of detectable signal evolves over days.

Guidance map illustrating the spatiotemporal window where concentration (C) and gradient (p) constraints allow for successful axon navigation.

For a typical diffusion constant, the concentration limit dominates for the first few days. After this initial period, the gradient limit ($p$) takes over and rapidly shrinks the maximum guidance range toward the 1 mm asymptote. The author finds that this gradient constraint (the slope of the signal) is generally more restrictive and "robust" than the absolute concentration constraint, as the former is independent of the rate of factor production.

The "Slow Diffusion" Advantage

Counter-intuitively, the model suggests that larger, more slowly diffusing molecules might actually be better for guidance. Slow diffusion maintains a steeper, more detectable gradient for a longer period during the initial phase of factor production. Furthermore, the model predicts that guidance over ranges longer than 1 mm might be possible very early in development, before the gradient has fully spread out and flattened.

Limitations and Future Directions

While the model elegantly explains observed biological limits, it relies on several simplifications:

  • Simplified Environment: The model assumes an infinite, uniform 3D volume. In reality, the extracellular matrix is crowded, and factors can bind to surfaces or be degraded by enzymes, altering the gradient shape.
  • Steady Production: The assumption of a constant rate of release ($q$) may not capture biological systems that use pulsed releases or varying rates to extend guidance range.
  • Noise Modeling: The paper uses a fixed percentage threshold ($p$) rather than a full stochastic simulation of receptor binding to account for statistical noise.

Finally, the model's most provocative prediction—that guidance over ranges $>1\text{ mm}$ is possible at very early time points—remains an exciting avenue for future rigorous laboratory testing.

Summary updated

BibTeX
@inproceedings{goodhill1997mathematical,
  author = {Geoffrey J. Goodhill},
  title = {{A Mathematical Model of Axon Guidance by Diffusible Factors}},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {1997},
}

Loading related content...

Figure