Positive feedback can lead to dynamic nanometer-scale clustering on cell membranes

Clustering of molecules on biological membranes is a widely observed phenomenon. In some cases, such as the clustering of Ras proteins on the membranes of mammalian cells, proper cell signaling is critically dependent on the maintenance of these clusters. Yet, the mechanism by which clusters form and are maintained in these systems remains unclear. Recently, it has been discovered that activated Ras promotes further Ras activation. Here we show using particle-based simulation that this positive feedback is sufficient to produce persistent clusters of active Ras molecules at the nanometer scale via a dynamic nucleation mechanism. Furthermore, we find that our cluster statistics are consistent with experimental observations of the Ras system. Interestingly, we show that our model does not support a Turing regime of macroscopic reaction-diffusion patterning, and therefore that the clustering we observe is a purely stochastic effect, arising from the coupling of positive feedback with the discrete nature of individual molecules. These results underscore the importance of stochastic and dynamic properties of reaction diffusion systems for biological behavior.

💡 Research Summary

The paper investigates how the well‑known positive feedback loop in the Ras signaling pathway can generate and maintain nanometer‑scale clusters of active Ras molecules on the plasma membrane. Ras exists in two states: an inactive GDP‑bound form (Ras‑GDP) and an active GTP‑bound form (Ras‑GTP). Activation is catalyzed by the guanine‑exchange factor SOS, which possesses an allosteric site that binds Ras‑GTP and dramatically accelerates SOS’s catalytic activity (≈75‑fold). Consequently, the presence of active Ras locally increases the rate at which neighboring Ras molecules become active—a classic positive feedback loop.

To test whether this feedback alone can produce the experimentally observed clusters (typically 10–30 nm in diameter and containing a few to a few dozen molecules), the authors construct a minimal reaction‑diffusion model that includes only the two Ras species and three reactions: spontaneous activation (k₁), spontaneous deactivation (k₂), and a bimolecular activation step (k₃) that requires a Ras‑GDP molecule to encounter a Ras‑GTP molecule. All other molecular details (SOS, GAPs, cytoskeletal scaffolds, membrane rafts) are coarse‑grained out. Parameter values are taken directly from experimental measurements: particle radius 1.7 nm, surface coverage ≈1 %, diffusion constant of Ras‑GDP κ_D = 0.1 µm² s⁻¹, and a range of diffusion ratios κ_T/κ_D from 0.1 to 1.0 to explore both slower and equal diffusion of Ras‑GTP. Reaction rates are chosen so that the system resides in the diffusion‑limited regime (κ_D/k₃ ≪ 1) and the steady‑state active fraction matches observed values (~10–15 %). The model is simulated in two dimensions using the enhanced Green’s function reaction‑diffusion (eGFRD) algorithm, supplemented by Brownian dynamics when local densities become high.

The authors quantify clustering by computing the pair‑distance distribution P(r) and the corresponding radial distribution function g(r) = P(r)/P_CSR(r), where P_CSR(r) is the analytic distribution for a completely spatially random (CSR) configuration. In simulations without the feedback term (k₃ = 0), g(r) follows the CSR curve, confirming that Ras‑GDP molecules behave as non‑interacting hard spheres. When the feedback is active, g(r) exhibits a pronounced peak at distances around 10 nm, indicating a significantly higher probability of finding two active molecules at short separations than expected by chance. The function also dips below the CSR line at larger distances, a direct consequence of normalization when particles are concentrated into clusters. These signatures persist across all diffusion‑ratio conditions, including the case κ_T = κ_D, demonstrating that differential diffusion is not required for clustering in this system.

To assess whether the observed patterns could be explained by classical Turing instability, the authors perform a linear stability analysis of the corresponding macroscopic reaction‑diffusion equations. The analysis shows that, for the chosen parameter set, the homogeneous steady state is always stable; no Turing bifurcation occurs. Thus, the clusters arise not from deterministic pattern‑forming instabilities but from stochastic fluctuations amplified by the positive feedback. This aligns with recent theoretical work on “fluctuation‑induced patterning,” yet the present study uniquely shows that such patterns can exist even when the macroscopic system lacks any Turing regime.

The robustness of the clustering mechanism is further highlighted by varying the diffusion ratio over an order of magnitude. Even when Ras‑GTP diffuses ten times slower than Ras‑GDP—a scenario supported by single‑molecule tracking experiments—the model still produces clusters with statistics matching experimental observations. Conversely, when both species have identical diffusion constants, clustering remains, underscoring that the feedback loop itself is the dominant driver.

In summary, the paper provides compelling computational evidence that a simple positive feedback loop in Ras activation is sufficient to generate dynamic, nanometer‑scale clusters that are statistically indistinguishable from those observed in living cells. The clusters are stochastic, sustained in steady state, and do not rely on macroscopic Turing mechanisms or on large differences in diffusion coefficients. This work emphasizes the importance of stochastic reaction‑diffusion dynamics in membrane signaling and suggests that similar feedback‑driven clustering could be a general principle for organizing signaling molecules on cellular membranes.