{"id":"af5ac57c-c55f-4955-a67b-a9e6b8553fe7","arxiv_id":"2501.01868","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"A three-scale mechanical model predicts that axon growth cones generate stiffness-dependent traction with an optimal stiffness, producing positive or negative durotaxis, attraction to optimal-stiffness lines, and reflection or refraction at stiffness boundaries.","lead":"Axons navigate by feeling the stiffness of their surroundings, and this paper builds a three-scale model for that process. The model predicts that stiffness changes can steer or even bounce axons, and suggests that lines of preferred stiffness act as guides.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reflection/refraction phase diagram (Fig. 7) rests on the edge-only, non-interacting traction law in Eq. 31; if real adhesions are distributed over the growth-cone interior, the reflective branch may vanish because the stiffness sampling is smoothed out.","rationale":"The paper's central claims are (1) stiffness-dependent sign of durotaxis, (2) attraction to lines of optimal stiffness, and (3) reflection/refraction at stiffness interfaces. Claims (1) and (2) are protected to leading order: the durotactic force in Eq. (32) is proportional to T'(χ) P∇χ with P symmetric positive-definite, so the sign of the component along ∇χ is fixed by T'; an area-distributed traction law yields a similar positive-definite tensor with different coefficients, so the sign of durotaxis and the potential-well structure near χ* persist. Claim (3), by contrast, is a nonlinear threshold effect computed numerically from the edge-only integral (31). When the growth cone straddles a sharp interface, the transverse force arises from the difference in traction on the two halves of the boundary; if adhesions are spread across the interior, the effective stiffness field is averaged over the cone area, smearing the interface and changing the transverse force-displacement relation. The paper provides no code or parameter sweeps to show that the reflective branch in Fig. 7 is robust to this idealization. A concrete numerical experiment with a distributed-adhesion force law can settle this: if the reflective branch persists after area integration, the concern is resolved; if not, the second main prediction of the abstract is an artifact of the edge-only assumption. We therefore keep the reader's CONDITIONAL verdict; the concern reinforces the need for the conditions to be checked rather than changing the verdict.","tokens_in":26110,"tokens_out":24069,"duration_ms":244583,"concrete_test":"Implement a distributed-adhesion counterpart of Eq. (31): F_area(r,θ)=ρ_A ∫_{0}^{R}∫_{-φ0}^{φ0} T(χ(r+ρ e_φ)) e_φ ρ dφ dρ, with ρ_A chosen so the total traction on a uniform substrate matches the edge-only model. Rerun the straight-interface simulations of §5.1 for α_T = 4/5 and β_a = 0.1 over the full range of incidence angles θ1. If the reflective branch (Fig. 7b) disappears for all θ1, the reported reflection is an artifact of the edge-localized force law; if a reflective branch persists with a modified critical angle, the central qualitative claim survives. Report the critical angle as a function of the ratio of adhesion area to R.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3 derives the growth-cone force F(r,θ) by integrating the single-filament traction T(χ) only over the circular-sector edge: F = ∫_{-φ0}^{φ0} ρ T(χ(r+R e_φ)) e_φ R dφ (Eq. 31). This assumes (i) all adhesion sites live on the lamellipodial rim, and (ii) filaments are mechanically independent. The subsequent reflection/refraction study in §5.1 uses this exact F with a tanh-smoothed interface and reports a reflection branch for α_T = 4/5, β_a = 0.1 (Fig. 7b). Reflection is a threshold phenomenon: it arises because the portion of the edge entering the low-traction (stiff) region produces a transverse force that overcomes the axon's bending resistance. If instead adhesions are spread over the growth-cone area, the traction field is effectively averaged over the cone interior; the 'interface' the cone experiences is smeared over a length scale comparable to R, and the transverse force near the boundary is smaller and develops more gradually. This can shift the critical incidence angle for reflection or eliminate the reflective branch entirely. The same averaging changes the coefficients in the stability expansion (56)-(57), though the sign of the first-order durotactic force is protected by the positive-definiteness of P(φ0). Since no code, data, or parameter sensitivity analysis is provided, there is currently no evidence that the reflective trajectories of Fig. 7 survive a distributed-adhesion version of the model.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a three-scale mechanical model of axonal durotaxis. At the molecular scale, a Bell-type molecular-clutch model of a single actin filament on a deformable substrate yields a traction-stiffness relation T(χ) that can be non-monotonic, with an optimal stiffness χ* if myosin contractility M exceeds a critical value. At the growth-cone scale, the traction is integrated over a circular-sector lamellipodium, and a slow-gradient expansion gives a durotactic force proportional to T′(χ)P(φ0)∇χ, so the sign of durotaxis is set by the sign of T′(χ). At the cellular scale, the axon is modelled as a morphoelastic rod with tip growth, friction, and a growth-cone end load. The model is used to predict reflection and refraction at a straight stiffness interface, attraction to lines of optimal stiffness in a linear gradient, and a qualitative comparison with Xenopus retinal axon guidance. The paper is clearly written and the derivations are mostly explicit, but the headline predictions rest on several strong simplifications, especially the edge-only adhesion assumption in the growth-cone force law.","tokens_in":26558,"tokens_out":12152,"duration_ms":117240,"significance":"If the model's assumptions hold, this is a valuable contribution: it connects molecular clutch mechanics to axon-scale trajectories, gives closed-form expressions for the optimal stiffness and the durotactic sensitivity, and introduces the appealing 'axonal optics' concepts of reflection, refraction, and focalisation. The qualitative reproduction of the Xenopus growth patterns and the non-monotonic dependence of success rate on myosin contractility are suggestive. The paper is honest about its limitations and does not overclaim quantitative biological accuracy. However, the central predictions are not yet backed by robustness tests, and one of the two main physical assumptions (edge-localized, non-interacting adhesion) is load-bearing for the reflection/refraction phenomenology and for the attractive-line result. With additional sensitivity analysis and a finite-width adhesion test, the framework could become a strong foundation for further work.","major_comments":[{"comment":"The growth-cone force is evaluated as an integral of the single-filament traction T(χ) over the lamellipodial edge only: F = ∫ ρ T(χ(r+Reφ)) eφ R dφ. This edge-localized, non-interacting-filament assumption is load-bearing for the interface reflection/refraction diagram in Section 5.1 (Fig. 7) and for the restoring-force expansion in Section 5.2 (Eqs. 56–57). In real growth cones, adhesion and force generation are distributed over the interior of the lamellipodium, so the stiffness field is sampled over an area rather than a one-dimensional rim. Averaging over the cone interior would smooth the tanh interface (Eq. 52) over a length comparable to R and would reduce the transverse force that produces the reflective branch. The manuscript offers no argument, numerical test, or literature estimate showing that Fig. 7(b) survives this averaging. Please provide a distributed-adhesion version of the model, or at least a sensitivity analysis with adhesion patches of finite width, and show how the reflection/refraction phase boundary changes.","section":"Section 3, Eq. (31)"},{"comment":"The biological application asserts a non-monotonic success-rate curve and concludes that a combination of reflection and refraction, rather than optimal-stiffness tracking, is the most plausible guidance mechanism. This conclusion rests on a single parameter set (K=0.1, β=f0K, Ec=6000 Pa, R=0.5L, φ0=π/2, βa=0.1) and on a small number of manually chosen M/f0K values (10, 35, 50, >50). No sensitivity analysis over R/L, φ0, βa, Ec, or the initial-condition distribution is reported, and no code or data are provided. Since the success rate includes a catastrophic 0% point (M/f0K=35), it is important to know whether that point and the 89% point are robust to perturbations of these parameters. Please add parameter sweeps and, if possible, release the simulation code.","section":"Section 6, Fig. 11"},{"comment":"The claim that the optimal-stiffness line D is 'a locally attracting set' is supported only by the sign of the instantaneous transverse force Fn in Eq. (57) together with the simulations in Fig. 10. However, the axon trajectory is governed by the full free-boundary beam problem (49), where clamp advection, bending resistance, the growth law (51), and the feedback of the force on the shape all enter. A negative Fn for r>0 and θ=0 does not by itself guarantee asymptotic convergence under this dynamics, especially since the same expansion shows Fn depends on θ with a coefficient of the same sign as -θ; the manuscript itself notes escape angles. Either provide a linearized stability analysis of the coupled system (49)–(51) about D, or soften the statement to a local, parameter-dependent tendency.","section":"Section 5.2, Eqs. (56)–(57)"}],"minor_comments":[{"comment":"The caption is internally inconsistent: panel (a) is described as 'smaller on the right' with α_T=5/4, and panel (b) as 'larger on the right' with α_T=4/5, while the labels in the figure panels show T1>T2. Please correct the caption and ensure that the α_T values and the T1/T2 labels match the actual simulation parameters.","section":"Fig. 7 caption"},{"comment":"The displayed argument of H appears as qχν(χ+N) or is otherwise ambiguous; from the definition u = q(1+χ^{-1}N)/ν in Eq. (19) and the subsequent equations, T should be f0K H(q(χ+N)/(χν)). Please verify the typesetting and correct the expression.","section":"Eq. (15)"},{"comment":"The symbol β is used for the dimensionless clutch parameter in Eq. (19), for the dimensionless adhesion parameter in Eq. (46), and then β_a appears in Eq. (54). This overloading is confusing; please rename at least one of these parameters.","section":"Eqs. (19), (46), (54)"},{"comment":"No statement of code or data availability is provided. The numerical results in Figs. 7–11 would be considerably easier to assess if the simulation code or a reproducibility description were included.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid modeling contribution with a clear multiscale structure and honest caveats. My main reservation is that the headline phenomena—especially reflection/refraction and attraction to optimal-stiffness lines—are produced by the specific edge-only adhesion assumption and by a small number of hand-picked parameters. If the authors add a finite-width adhesion test and parameter sweeps, and if they clarify the analytical status of the attractive-line claim, I would be supportive of publication. The relation to prior work by Oliveri et al. (2021) is properly acknowledged, and I see no citation concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe short version: this is a serious multiscale modeling paper with real new content—the three-scale coupling of molecular clutch mechanics, growth cone force integration, and morphoelastic rod growth—and it deserves a real referee. But the central predictions are more proposal than proof, and the paper would be stronger if it acknowledged more directly how much hangs on the edge-localized adhesion assumption.\n\nWhat's new: the explicit closed-form optimal stiffness (χ*), the attraction to lines of optimal stiffness, and the reflection/refraction phase diagram for single axons are all new relative to prior work by Sens or Oliveri et al. The derivations are careful and readable. The use of morphoelastic rod theory with tip growth is a nice extension of earlier axon models. I especially liked the observation that the durotactic steering force is governed solely by the sign of T'(χ) and that a nonmonotonic T(χ) naturally produces attractive lines—that's a clean, teachable result.\n\nThe soft spots are real but not fatal. The stress-test note about distributed adhesions is on point: Eq. 31 assumes all traction acts at the lamellipodial edge, and the reflection/refraction phase diagram (Fig. 7) could change if adhesions are spread over the cone interior. That's a modeling assumption, not an error, but it deserves a sensitivity analysis or at least a discussion. The positive/negative durotaxis result is essentially a restatement of the sign of T'(χ) in the mesoscopic expansion—not an independent prediction. The Xenopus comparison is qualitative, and no code or data are provided, which limits reproducibility. The parameter choices are numerous and somewhat arbitrary, though the qualitative behavior seems robust.\n\nWho should read this: anyone modeling axon guidance or cell durotaxis, especially people interested in multiscale mechanotransduction. It's a good paper for a journal that accepts mechanism proposals without demanding quantitative experimental validation. I'd recommend sending it to peer review, with a request for sensitivity analysis around the adhesion distribution assumption and a clearer separation between model outputs and experimentally falsifiable predictions.","headline":"A serious three-scale model of axon durotaxis with genuine new content; worth refereeing, but the predictions hang on assumptions that need sensitivity analysis.","tokens_in":27068,"tokens_out":1927,"would_cite":true,"duration_ms":20189,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C10","74L15","92C17","74K10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper argues that a single traction–stiffness curve, derived from molecular clutches and integrated over the growth cone, determines whether an axon is steered by stiffness and that this mechanism produces attractive stiffness lines…","keywords":["axon guidance","durotaxis","growth cone","molecular clutch","morphoelastic rod","stiffness gradient","mechanotaxis","axonal optics"],"falsifier":"Measure the traction–stiffness curve of isolated growth cones on calibrated substrates spanning both sides of the predicted optimal stiffness $\\chi^*$ and compare the sign of $T'(\\chi)$ with the turning direction in a linear stiffness gradient; if axons on stiffer-than-optimal substrate still turn up-gradient, the central claim fails.","tokens_in":25901,"feed_emoji":"🧠","tokens_out":7185,"duration_ms":67769,"temperature":0.7,"pith_summary":"During development, growing axons are guided not only by chemicals but by the stiffness of the tissue they grow through—durotaxis. This paper builds a three-scale mechanical theory of axon durotaxis starting from molecular clutches that link actin filaments to the substrate, integrating their forces over the growth cone, and then treating the whole axon as a growing, bendable rod. The central claim is that the sign of the derivative of the traction–stiffness curve $T(\\chi)$ decides whether an axon moves up or down a stiffness gradient, and that because this curve can be nonmonotonic, both positive and negative durotaxis can arise from one mechanism. As a consequence, lines of optimal stiffness act as attractors for axon trajectories, and axons hitting a straight stiffness interface behave analogously to light rays—refracting across it or reflecting off it, though without time-reversal symmetry. The model is tested against measurements of the optic tract in Xenopus, where it reproduces the observed deflection of retinal axons and predicts how myosin contractility tunes guidance success.","feed_headline":"One traction curve predicts axon steering and bounce","feed_subtitle":"Nonmonotonic force–stiffness relation yields positive and negative durotaxis, plus reflection at stiffness boundaries.","key_machinery":"The load-bearing object is the traction–stiffness function $T(\\chi)$ obtained from a mean-field molecular-clutch model. A rigid actin filament drags at velocity $V$ over a substrate while cross-bridges attach unstretched and detach according to a Bell law $p_{\\rm off}\\propto e^{f/f_0}$, and all bridges share a single substrate spring of stiffness $\\kappa'$; the balance yields $T(\\chi)=f_0 K H(u)$ where $\\chi=\\kappa'/\\kappa$ is dimensionless substrate stiffness, $H$ is the special function $H(x)=x e^x\\int_0^\\infty y\\,e^{-x e^y}\\,dy$, and $u$ is fixed by a two-equation system. This curve has a global maximum at the optimal stiffness $\\chi^*$, expressible in closed form with the incomplete gamma function, and a curvature $T''(\\chi^*)$ that sets how sharply axons react. Integration over a circular-sector growth cone gives a mesoscopic force whose leading terms are $2R\\rho\\sin\\varphi_0\\,T(\\chi)\\,\\mathbf t + R^2\\rho\\varphi_0\\,T'(\\chi)\\,\\mathbb{P}(\\varphi_0)\\nabla\\chi$, so the sign of $T'$ sets the direction of steering. The axon is then a morphoelastic rod with exponential growth law and a moving clamped tip, reduced to a beam equation for the active distal segment.","core_discovery":"The paper's central discovery is that the direction and strength of axonal durotaxis are controlled by the local slope $T'(\\chi)$ of the traction generated by a single actin filament as a function of substrate stiffness, and that this function is inherently nonmonotonic under Bell-type adhesion kinetics and a deformable substrate. When the local stiffness is below the optimal stiffness $\\chi^*$ where traction is maximal, axons exhibit positive durotaxis; above it, negative durotaxis. Because the traction peaks at $\\chi^*$, the stiffness landscape has attractive lines: an axon near the line of optimal stiffness experiences a restoring force that pulls it back toward the line, while sufficiently oblique crossings escape. At a sharp interface between two uniform stiffness regions, the growth cone's finite size generates either refraction or reflection depending on the ratio of tractions and the incidence angle, with no simple Snell's law and no time-reversibility. In a realistic stiffness field from the developing Xenopus brain, the model shows that reflection and refraction at internal interfaces, rather than attraction to optimal-stiffness lines, best account for guiding axons to the optic tectum.","pith_inferences":["Editorial extension: the same nonmonotonic traction–stiffness mechanism could apply to other cell types that use molecular clutches, so durotaxis in fibroblasts or cancer cells may also switch sign around an optimal matrix stiffness.","Editorial extension: the optical analogy suggests a design principle—by patterning stiffness landscapes, one could steer growing axons or other cells along prescribed paths, but the absence of time-reversal means such 'axonal lenses' are intrinsically directional.","Editorial extension: the prediction of attraction to optimal stiffness lines could be tested directly in vitro by culturing growth cones on stiffness gradients that bracket $\\chi^*$ and measuring the distribution of final trajectories; the paper does not report such an experiment.","Editorial extension: because the tip-growth approximation assumes a short active zone, the model may underestimate long-range mechanical coupling along the axon; testing with full-axon simulations where the friction lengthscale is not small would reveal whether reflection is robust."],"forward_implications":["On substrates softer than the optimal stiffness, axons can only turn toward stiffer regions; on stiffer substrates, only toward softer regions.","If the traction–stiffness curve is monotonic (myosin contraction below a critical value $M_{\\rm crit}$), negative durotaxis is impossible; positive durotaxis is the only mode.","The line of optimal stiffness in a graded field is a local attractor: nearby axons with small inclination converge to it, whereas sufficiently steep crossings escape.","At a straight stiffness interface, axons can refract into the stiffer traction region or reflect back, with reflection angles not equal to incidence angles.","In the Xenopus optic-tract scenario, the success rate of reaching the tectum is a nonmonotonic function of myosin contractility $M$, with the best guidance arising from reflection and refraction at internal interfaces rather than tracking optimal-stiffness contours."],"supporting_citations":[{"why":"Supplies the microscopic friction model of actin sliding against a deformable substrate that the paper adapts.","marker":"(Sens, 2013)"},{"why":"Provides the molecular-clutch traction dynamics on compliant substrates that underlies the cross-bridge description.","marker":"(Chan and Odde, 2008)"},{"why":"Gives the Bell-type detachment law used for cross-bridge unbinding kinetics.","marker":"(Bell, 1978)"},{"why":"Establishes the motor-clutch model and its nonmonotonic force–stiffness relation that motivates the optimal stiffness.","marker":"(Bangasser and Odde, 2013)"},{"why":"Earlier durotactic axon guidance theory and the optical analogy for deflection that this paper generalizes to single axons.","marker":"(Oliveri et al., 2021)"},{"why":"Physical model of axonal elongation with traction-driven growth speed that the rod model builds on.","marker":"(O’Toole et al., 2008)"},{"why":"Provides morphoelastic rod theory used for axon shaft growth and bending.","marker":"(Goriely, 2017)"},{"why":"Xenopus optic nerve experiments showing stiffness-gradient guidance, used as qualitative comparison.","marker":"(Koser et al., 2016)"},{"why":"Supplies the AFM brain rigidity field used as input in the optic-tract simulations.","marker":"(Thompson et al., 2019)"}],"fun_headline_variants":["Axon durotaxis: one curve to steer them all","Stiffness gradients: how axons decide to turn or bounce","The traction slope that flips axon guidance direction","Reflection and refraction in axon durotaxis"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the growth cone's net steering force is just the sum of independent edge forces from actin filaments that do not interact mechanically and act only at the lamellipodium rim.","fun_headline_variants_meta":{"raw":{"variants":["Axon durotaxis: one curve to steer them all","Stiffness gradients: how axons decide to turn or bounce","The traction slope that flips axon guidance direction","Reflection and refraction in axon durotaxis"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000651,"raw_usage":{"total_tokens":3029,"prompt_tokens":1034,"completion_tokens":1995,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":650,"completion_tokens_details":{"reasoning_tokens":1928}},"tokens_in":650,"tokens_out":1995,"duration_ms":20174,"temperature":1.0,"reasoning_tokens":1928,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:18:35.967140+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the traction–stiffness curve of isolated growth cones on calibrated substrates spanning both sides of the predicted optimal stiffness $\\chi^*$ and compare the sign of $T'(\\chi)$ with the turning direction in a linear stiffness gradient; if axons on stiffer-than-optimal substrate still turn up-gradient, the central claim fails.","supporting_citations":[{"cited_title":", author Pillai, E.K","cited_arxiv_id":null,"evidence_quote":"Supplies the AFM brain rigidity field used as input in the optic-tract simulations."},{"cited_title":", year 2013","cited_arxiv_id":null,"evidence_quote":"Supplies the microscopic friction model of actin sliding against a deformable substrate that the paper adapts."},{"cited_title":", year 1978","cited_arxiv_id":null,"evidence_quote":"Gives the Bell-type detachment law used for cross-bridge unbinding kinetics."},{"cited_title":", author Franze, K","cited_arxiv_id":null,"evidence_quote":"Earlier durotactic axon guidance theory and the optical analogy for deflection that this paper generalizes to single axons."},{"cited_title":", author Lamoureux, P","cited_arxiv_id":null,"evidence_quote":"Physical model of axonal elongation with traction-driven growth speed that the rod model builds on."}],"review_version":1}