Genotype specificity and spatial arrangement govern the direction and magnitude of selection in variable environments

Technical Summary: Genotype Specificity and Spatial Arrangement Govern Selection in Variable Environments

Problem Statement
Environmental heterogeneity is a ubiquitous feature of biological systems, ranging from nutrient gradients in microbial biofilms to oxygen and drug gradients in solid tumors. While it is established that spatial structure influences evolutionary outcomes, the theoretical understanding of how environmental variation interacts with population structure to determine the fixation probability of beneficial mutants remains fragmented. Previous studies have reported contradictory results: some models suggest that spatial heterogeneity amplifies selection, while others indicate it suppresses it. Existing frameworks often fail to systematically account for genotype-specific interactions—specifically, whether environmental variation acts symmetrically on competing genotypes or asymmetrically on one. This lack of a unifying principle prevents the reconciliation of disparate theoretical results across different spatial structures and fitness distributions.

Methodology
The authors develop a unified framework for selection on lattice graphs incorporating environmental heterogeneity. The model utilizes a discrete-time Moran birth–death process on undirected graphs (specifically one-dimensional cycles and two-dimensional square lattices).

• Fitness Landscape: The fitness of an individual at node $i$ is defined as $f_{\alpha,i} = r_\alpha + \sigma_\alpha c_i$, where $r_\alpha$ is the inherent fitness of genotype $\alpha$ (mutant $A$ or resident $B$), $\sigma_\alpha$ is the genotype-specific heterogeneity amplitude, and $c_i \in \{+1, -1\}$ represents the local environmental state (rich or poor).
• Interaction Scenarios: The study explicitly distinguishes three classes of genotype-environment interactions:
  1. S1 (Genotype-symmetric): Both genotypes respond equally to environmental variation ($\sigma_A = \sigma_B = \sigma$).
  2. S2 (Mutant-specific): Only the mutant responds to environmental variation ($\sigma_A = \sigma, \sigma_B = 0$).
  3. S3 (Resident-specific): Only the resident responds to environmental variation ($\sigma_A = 0, \sigma_B = \sigma$).
• Spatial Arrangement: The spatial distribution of environmental states is quantified using a spatial correlation index ($M$). This parameter measures the degree of clustering versus intermixing of rich and poor sites, ranging from maximally intermixed (checkerboard, low $M$) to fully segregated (clustered domains, high $M$).
• Analysis: The authors employ a combination of analytical derivations for limiting configurations (checkerboard and segregated) using martingale methods and backward Kolmogorov equations, alongside extensive numerical simulations and stochastic simulations across 946 environmental configurations on 1D and 2D lattices.

Key Contributions
The primary contribution of this work is the identification of two governing principles that determine the direction and magnitude of selection in heterogeneous environments:

1. Genotype Specificity Determines Direction: The authors demonstrate that the direction of the effect (amplification vs. suppression of selection) is primarily dictated by which genotype's fitness is modulated by the environment.

   • Resident-specific (S3): Heterogeneity consistently amplifies selection, increasing the fixation probability of beneficial mutants.
   • Mutant-specific (S2): Heterogeneity generally suppresses selection, decreasing fixation probability, particularly under strong heterogeneity.
   • Genotype-symmetric (S1): Heterogeneity produces a weaker amplification effect compared to S3, often arising from second-order boundary effects.

2. Spatial Arrangement Determines Magnitude: The magnitude of amplification or suppression is modulated by the spatial correlation index ($M$), without reversing the direction established by genotype specificity.

   • In Resident-specific scenarios, highly intermixed (low $M$) environments yield the strongest amplification.
   • In Mutant-specific scenarios, highly clustered (high $M$) environments counteract suppression, leading to higher fixation probabilities than in intermixed environments.
   • In Genotype-symmetric scenarios, the relationship is non-monotonic, with fixation probabilities often minimized at intermediate levels of spatial mixing.

Results

• Analytical Limits: Exact analytical solutions for checkerboard (bipartite) configurations confirm that fixation probabilities depend on the effective geometric mean of fitness ratios, validating the "environmental isothermal theorem" for these specific topologies. Approximations for segregated configurations show that fixation behaves as an average of two homogeneous environments.
• Numerical Validation: Simulations across a continuum of spatial correlations confirm that the analytical limits bound the fixation probabilities observed in random or intermediate configurations.
• Reconciliation of Literature: By mapping previous studies onto this framework, the authors show that seemingly contradictory results in the literature are consistent when genotype specificity and spatial arrangement are explicitly considered. Studies reporting amplification align with resident-specific or weakly heterogeneous segregated models, while those reporting suppression align with mutant-specific or symmetric random models.
• Graph Connectivity: The study notes that while genotype specificity remains predictive across graph types, the influence of spatial arrangement ($M$) diminishes as graph connectivity increases (approaching the complete graph limit), where spatial correlations become irrelevant.

Significance
The paper claims to provide a unified theoretical framework that resolves long-standing inconsistencies in evolutionary graph theory regarding environmental heterogeneity. By distinguishing between genotype-specific and genotype-symmetric modulation, the authors offer predictive criteria for adaptation in heterogeneous landscapes. The findings suggest that environmental heterogeneity does not inherently amplify or suppress selection; rather, the outcome emerges from the interplay between which genotype is affected and how the environment is spatially arranged. This framework is presented as applicable to diverse biological contexts, including microbial communities, somatic evolution, and cancer progression, where understanding the specific interaction between genotype and spatial niche is critical for predicting evolutionary trajectories.