Gradients for Retinotectal Mapping

Geoffrey J. GoodhillNeurIPS 1997

Introduction

In the developing visual system, axons extending from the retina must navigate to specific locations on the optic tectum to create a precise topographic map—essentially a "point-to-point" projection of the visual world onto the brain. Since the 1960s, scientists have hypothesized that this intricate wiring is guided by "chemoaffinity" gradients: molecular coordinates expressed as varying concentrations of proteins across both the retina and the tectum.

While the chemoaffinity hypothesis is decades old, the actual molecules responsible—Eph receptors and ephrin ligands—were only identified in the mid-1990s. This paper bridges the gap between these molecular biology discoveries and classical theoretical models. Specifically, it addresses a critical mathematical question: while biologists have confirmed the existence of these gradients, do their physical shapes (e.g., linear, exponential) mathematically support the creation of a functional, precise map?

Schematic of retinotectal mapping and hypothetical linear gradients.

Theoretical Framework

To evaluate the viability of different gradient shapes, the author employs a theoretical framework based on the law of mass action to model the interaction between receptors ($R$) on the incoming axon and ligands ($L$) on the tectum.

The "topographic signal" ($G$) that instructs an axon to halt its growth is defined by the equation: $$G(x, y) = kR(x)L(y)$$ where $x$ is the position in the retina and $y$ is the position in the tectum.

The paper analyzes two potential biological "stopping rules" that axons might use to interpret this signal:

  1. The Set Point Rule: The axon stops growing when the signal reaches a specific, constant threshold value ($G = c$).
  2. The Local Optimum Rule: The axon stops when it detects a local maximum or minimum in the signal ($\partial G/\partial y = 0$).

Assuming a standard linear mapping (where position $x$ in the retina maps directly to position $y$ in the tectum), the author uses calculus to derive the necessary mathematical shapes for the ligand gradients ($L$) given known receptor distributions ($R$).

Mathematical Insights

By crunching the numbers on these stopping rules, the analysis reveals several key insights into how biological gradients must be structured:

  • The "Infinity" Problem: If the system relies on only one receptor gradient and one ligand gradient, and the receptor gradient is linear starting at zero, the mathematics dictate that the ligand concentration must approach infinity at one end of the tectum to maintain a linear map. This is biologically impossible.
  • The Role of Uniform Receptors: To solve the infinity problem, the math shows that the presence of a uniformly expressed receptor (such as EphA4) is not merely background noise. It is mathematically essential to prevent the need for infinite ligand concentrations, thereby allowing for a physically plausible biological system.
  • Quadratic Gradients: For both the "set point" and "local optimum" rules, the derivations suggest that if the primary receptor gradient is linear, the secondary ligand gradient should follow a quadratic shape ($y^2$). Interestingly, this aligns with some experimental observations of the ephrin-A5 ligand.
  • The "Compass" Advantage: The "local optimum" rule proves to be more biologically robust because it provides a directional gradient. Under this rule, an axon doesn't just know when it has arrived at its destination; it knows which way to turn if it is currently in the wrong location.

The Repulsion Paradox and Future Directions

The most striking result of this theoretical analysis is a fundamental conflict between the mathematical models and experimental reality. The mathematical models analyzed require at least one gradient to be attractive to successfully guide axons to their targets. However, experimental data regarding Eph/ephrin interactions show that they are almost exclusively repulsive.

This "repulsion paradox" highlights a significant limitation in our current understanding: the standard mass action models struggle to explain how a system built entirely on "keep away" signals can create a precise, stable map. Either the mass action model is too simplistic, or there are undiscovered attractive gradients yet to be found.

To resolve this conflict, the author suggests future work must focus on:

  1. Quantitative Mapping: Moving beyond qualitative descriptions to measure the exact, in vivo concentration curves of these proteins.
  2. Testing Knockouts: Analyzing data from Eph/ephrin knockout mice to observe how the topographic map deforms, providing the empirical data needed to refine mathematical models.
  3. Alternative Models: Exploring new theoretical models that analyze how gradients maximize the distance an axon can be guided, rather than focusing solely on where it stops.

Summary updated

BibTeX
@inproceedings{goodhill1997gradients,
  author = {Geoffrey J. Goodhill},
  title = {{Gradients for Retinotectal Mapping}},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {1997},
}

Loading related content...

Figure