Experiments in micro-patterned model membranes support the narrow escape theory

The study addresses how Brownian particles achieve timely encounters with small targets in confined domains, a central question for processes such as receptor activation and intracellular transport. The narrow escape theory (NET) provides a framework to quantify mean first passage times (MFPT) to small boundary targets in domains with reflecting boundaries except for an absorbing window. The authors focus on the analytical solution for a 2D circular domain, where τ ≈ (r²/D)[ln(4r/a) + ln 2 + O(ε)], with ε = a/C ≪ 1, and test this prediction experimentally. Measuring τ in biological systems is challenging due to limited control over parameters and fluorophore photophysics; prior in vitro attempts were not systematic. This work aims to establish a controlled membrane-based model system to systematically vary the escape opening and to devise a practical correction for lifetime-limited trajectories, enabling experimental verification of NET in a disc geometry.

NET has been developed and extended to multiple geometries (discs, spheres, cusps, sheets) with applications in cellular microdomains, calcium dynamics, vesicle trafficking, synaptic clefts, and viral nuclear entry. Foundational works by Holcman, Schuss, Singer, and others provide asymptotic MFPT formulas and full exit time distributions. Prior experimental efforts include nanoparticle diffusion in 3D cavities escaping through nanopores and lipid-anchored quantum dots on vesicles escaping to nanotubes, but a systematic verification across parameters remained lacking. Practical constraints include photobleaching/blinking of fluorescent tags, size-induced perturbations by gold beads or nanodiamonds, and the need for label-free methods like iSCAT in vitro. The concept of mortal walkers has been theoretically incorporated into NET, notably by Grebenkov and Rupprecht, yet an experimentally straightforward correction for lifetime-limited trajectories had been elusive.

Model system fabrication: Micro-structured substrates with circular glass wells in a continuous gold film were prepared via colloidal templating. Polystyrene microspheres (r = 2.5 µm) were randomly deposited, then 2 nm Ti and 100 nm Au thermally evaporated; spheres were removed to yield circular wells surrounded by gold. Gold was passivated with 1-dodecanethiol to prevent membrane formation on Au, ensuring SLBs formed only on the exposed glass discs. Small unilamellar vesicles (SOPC; with 0.005 mol% DOPE-Atto647N for fluorescence) were fused to form circular supported lipid bilayers (SLBs) within wells. The mean membrane radius determined from single-molecule localization patterns across 63 domains was 2.44 ± 0.13 µm.

Imaging and tracking: Single-molecule fluorescence microscopy (100 Hz, 1 kW·cm⁻² illumination, EMCCD) captured up to 10,000 frames (95×95 px, 160 nm pixel). Single emitters were localized by 2D Gaussian fits with ~20 nm precision, consistent with Thompson’s localization precision model; tracking used a probabilistic maximum-likelihood algorithm assuming normal diffusion.

System validation: Membrane integrity assessed via defect area fraction (mean 3.6%, median 2.6%). Diffusive behavior validated by MSD linearity (msd = 4Dt), yielding D = 2.29 ± 0.05 µm²·s⁻¹ (R² = 0.98). Step-length distributions (1D projections) were Gaussian in both center and rim regions (R ≈ 1.00 center, 0.99 rim), with site-specific diffusion coefficients D_center = 2.21 ± 0.01 µm²·s⁻¹ and D_rim = 1.13 ± 0.02 µm²·s⁻¹, indicating reduced diffusion near the rim (possible weak interactions with gold boundaries). Directionality bias at boundaries was excluded by uniform angle distributions between consecutive steps in both center and rim regions.

Escape time analysis: Virtual escape windows of fractional circumference ε ranging from 1/88 to 1/8 were defined along the boundary; the opening was rotated to sample the entire boundary without bias. The window was enlarged by localization precision. Trajectories starting at least a distance r from the window (approximate center) were considered; both pre- and post-hit segments were included if endpoints were also ≥ r away. Escape times (ET) were compiled and τ_exp estimated by fitting the exponential tail of the ET histogram, P(t) ~ exp(−t/τ), with tail region determined by maximizing R².

Lifetime correction: Due to photobleaching/blinking, trajectory lengths were exponentially distributed with a mean of ~0.60 s, much shorter than τ_th predicted by NET (8–14 s over ε ∈ [1/8, 1/88]). The absorbing probability p(ε) was estimated as the fraction of escaped trajectories among all observed trajectories for each ε, modeled by a Weibull distribution across datasets. Using independence between absorption and killing, and in the regime τ_k ≪ τ_a, the theoretical MFPT relates to observed by τ_th ≈ τ_exp / p. Thus τ_exp was rescaled: τ_exp,corr = τ_exp / P_exp(ε). Empirically p ranged from ~3% to 8% across ε.

Simulations: Random walk simulations in a disc (reflecting boundary with an absorbing window) matched experimental parameters (D ~ N(2.29, 0.05) µm²·s⁻¹; r ~ N(2.44, 0.13) µm; localization uncertainty included). Step directions were uniform; step-length statistics followed diffusion with time step dt chosen to ensure l/α ≤ 0.7 at small ε, reducing bias. Mortal-walker simulations used exponentially distributed trace lengths to match experimental p; immortal-walker simulations allowed escape for all particles. Escape times were collected and τ_sim estimated from the tail fit; τ_sim,p was corrected by p to give τ_sim,corr. For very small ε where σ contributed ≥10% of α (ε ∈ [1/72, 1/88]), simulations were also run with σ = 0 to avoid artificial flattening of τ(ε).

Statistical treatment: Theoretical τ_th computed with experimental D and r and propagated uncertainties. Errors in τ_exp,corr combined relative errors of τ_exp and p. Horizontal errors in ε included uncertainties in r and localization precision.

• The model membrane system satisfied NET prerequisites: intact SLBs with low defect fractions (~3.6% mean), normal diffusion (MSD linearity, R² = 0.98), and isotropic step-angle distributions at rim and center.
• Diffusion coefficients: global D = 2.29 ± 0.05 µm²·s⁻¹; D_center = 2.21 ± 0.01 µm²·s⁻¹; D_rim = 1.13 ± 0.02 µm²·s⁻¹ (reduced near rim, indicating minor boundary interactions).
• Domain size: mean radius r = 2.44 ± 0.13 µm (consistent with 2.5 µm template).
• Without correction, τ_exp from ET distributions was much shorter than theory due to lifetime limits (mean trace length ≈ 0.60 s). Theoretical MFPT from NET for ε ∈ [1/88, 1/8] was τ_th ≈ 14–8 s.
• Absorbing probability p(ε) empirically ranged from ~3% to 8%. Example: for ε = 1/88, T_exp = 0.43 s and p = 0.03 yielded T_exp,corr = 14.18 s, matching T_th = 14.17 s.
• Lifetime-corrected experimental MFPTs τ_exp,corr versus ε showed the expected logarithmic dependence ∝ ln(1/ε) and quantitatively matched NET predictions for small openings (ε ≲ 1/20; particularly ε < 1/24), with divergence at larger ε where the asymptotic theory is invalid.
• Simulations corroborated the correction factor: τ_sim,corr (mortal walkers) matched τ_sim (immortal walkers) and aligned with theory at small ε and with experiments across the measured range. At the smallest ε, incorporating σ = 0 avoided artificial leveling and recovered theoretical behavior.
• The reduced rim diffusion had negligible impact on τ for the domain sizes studied.

The findings experimentally validate the narrow escape theory in a controlled 2D circular geometry, demonstrating that MFPTs depend logarithmically on the inverse escape opening size and match theoretical absolute values when corrected for finite trajectory lifetimes. The introduced rescaling by the absorbing probability p provides a practical and theoretically justified method to recover true MFPTs from incomplete single-molecule trajectories dominated by photokinetics. Agreement between in vitro measurements and in silico simulations (with both mortal and immortal walkers) strengthens the conclusion and general applicability of the correction. While a reduced diffusion coefficient near the rim was observed, its effect on MFPTs was negligible in sufficiently large domains; however, edge passivation may be needed for smaller domains. Deviation from theory at larger ε is consistent with the asymptotic nature of the analytical expression (valid for ε ≪ 1). The approach enables MFPT estimation in complex biological contexts where full observation of rare events is impractical and analytical solutions are unavailable.

The study presents a scalable, high-quality model system to test NET in a disc and introduces an experimentally accessible correction to estimate MFPTs from lifetime-limited trajectories. It confirms the logarithmic dependence of τ on the relative escape opening ε and quantitative agreement with theory and simulations for small ε. The platform can be extended to study τ’s dependence on domain area A (by bead size) and diffusion coefficient D (by temperature), with practical limits from edge effects and fluorophore performance. Brighter, more photostable labels could expand accessible parameter ranges. For complex, heterogeneous biological geometries where analytical τ is not tractable, the proposed rescaling method offers a straightforward path to experimentally determine MFPTs.

• The NET analytical expression used is asymptotic and valid for ε ≪ 1; deviations occur at larger openings (ε ≳ 1/20).
• Fluorophore lifetime and tracking intermittency necessitate rescaling via p; accurate estimation of p(ε) is required and may introduce uncertainty.
• Reduced diffusion at the rim suggests weak boundary interactions; for smaller domains, edge effects could influence MFPTs unless additional passivation is applied.
• Localization uncertainty can bias simulations and measurements at very small ε; special handling (e.g., σ → 0 in simulations) was needed to avoid artificial flattening.
• Temperature and phase behavior of lipids, as well as camera/frame rate limits, constrain the range of D and A that can be explored experimentally.