{"id":"496aaa59-523c-4823-967d-fa87b092f919","arxiv_id":"2411.12515","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"In non-jammed MDCK monolayers, cell migration speed can increase with increasing density and with increasing cell-cell cohesion in specific low-density, low/high-cohesion regimes.","lead":"This experiment maps how MDCK cell monolayers move as cell density and cell-cell stickiness change, finding regimes where more crowding or more sticking makes cells move faster, not slower. The result is a counter-intuitive clue for how cells might move through dense tissues during development or disease.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'speed increases with cohesiveness' branch rests on an uncalibrated DECMA-1-to-cohesion mapping; the authors' own §3 admission leaves the v(ψ,σ) reentrance vulnerable to off-target effects.","rationale":"The paper's main contribution is a qualitative phase-behavior map: at low density, migration speed rises with density, and along the cohesion axis there are two ranges where speed rises with cohesion. The first of these is well supported by the time traces and by the c=0 and c=7 v–σ curves; even a skeptical reader can see the peak. The second is not. The ψ-axis is constructed by linear rescaling of antibody dose, and the manuscript contains no functional assay of adhesion (e.g., dissociation, traction, or single-cell force measurement) to show that DECMA-1 monotonically lowers cell–cell cohesion. The authors explicitly flag this gap in the Conclusion. Since the abstract and title claim results about cell cohesiveness, not merely about antibody dose, the uncalibrated proxy is load-bearing. I do not see an internal inconsistency: the trends in Fig. 4 are visible before fitting. The problem is external validity of the 'cohesiveness' variable. The reader's verdict of CONDITIONAL is exactly right, and the concern is the one they identified. A direct adhesion calibration would settle it. If the calibration shows monotonic reduction and the v-vs-ψ branches persist under the re-mapped axis, the concern is resolved; if not, the cohesion claim should be downgraded to a statement about DECMA-1 dose. Because this is an addressable experimental gap rather than a demonstrated contradiction, the verdict should remain as the reader set it.","tokens_in":12658,"tokens_out":6804,"duration_ms":69519,"concrete_test":"Measure cell–cell adhesion directly as a function of DECMA-1 concentration (0, 1, 2, ... 10 µg/mL) using a calibrated assay such as single-cell force spectroscopy or a dual-micropipette adhesion-frequency assay. Define ψ from the measured adhesion (e.g., normalized detachment force or adhesion probability), refit the v and q landscapes on (σ, ψ), and compare the positions of the 'v increases with ψ' branches. If the reentrant favorable branches disappear or shift outside the measured range under a monotonic adhesion axis, the central cohesion claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim has two prongs: speed increases with density at low density, and speed increases with cohesion in multiple ranges. The density prong is directly supported by raw traces (Fig. 3A) and v–σ curves (Fig. 4D), but the cohesion prong depends entirely on mapping DECMA-1 concentration c to a cohesive index ψ = (cmax−c)/cmax (Fig. 5B). No direct measurement of cell–cell adhesion is reported. The authors concede in Section 3 that 'blocking E-cadherin may produce indirect responses in the monolayer, and that it is preferable to connect DECMA-1 concentration to a quantitative level of cell-cell cohesion through direct measurement.' If DECMA-1 also changes proliferation, signaling, or substrate adhesion, or if its effect on E-cadherin adhesiveness is non-monotonic, then the locations of the 'speed increases with ψ' branches are not what they appear. The double-Gaussian fit propagates this uncertainty, and because the highest-ψ branch is anchored by untreated (c=0) data only, the reentrant shape of v(ψ,σ) could be an artifact of the proxy. This does not invalidate the density finding, but it underdetermines the 'cell cohesiveness' part of the headline claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents time-lapse experiments on non-jammed MDCK monolayer islands in which cell density, migration speed, and cell shape index are measured at ten DECMA-1 concentrations. The authors report three main observations: (i) a strong instantaneous anticorrelation between shape index q and cell density σ (mean extremum −0.88 ± 0.14 over 15 experiments); (ii) at unmodified cohesion, speed rises with density at low densities and falls at high densities; and (iii) a two-Gaussian fit to the speed versus density and cohesion plane, with cohesion encoded as ψ = (cmax−c)/cmax, yields four regimes in which speed increases or decreases with density crossed with speed increases or decreases with ψ. The authors interpret the low-density and high-ψ increasing branches as evidence of cooperative, mechanically stimulated motion, and they note that the transition contour extrapolates to q = 3.81, the vertex-model rigidity point.","tokens_in":12920,"tokens_out":7942,"duration_ms":73434,"significance":"If the cohesion axis were properly calibrated, the central observation—that cell migration speed can increase with packing density and with cell-cell cohesion in non-jammed monolayers—would run counter to the crowding-dominated picture and could be relevant to interpreting transitions in development and disease. The q-σ anticorrelation is robust and extends prior glassy-dynamics results across a broader parameter space. The paper also has concrete strengths: control measurements for PIV and segmentation, raw time traces with replicate means, and an explicit acknowledgment in Section 3 of the DECMA-1 calibration limitation. The main weaknesses are that the 'cohesiveness' axis is an unvalidated linear rescaling of antibody concentration and that the quantitative regime boundaries come from freely fitted surfaces without uncertainty propagation. These weaknesses directly affect the cohesion-based half of the headline claim.","major_comments":[{"comment":"The cohesive index ψ is defined as a linear rescaling of DECMA-1 concentration, ψ = (cmax−c)/cmax, with no direct measurement that DECMA-1 monotonically reduces cell-cell adhesion over 0–10 µg/mL. The authors explicitly acknowledge in Section 3 that 'blocking E-cadherin may produce indirect responses in the monolayer, and that it is preferable to connect DECMA-1 concentration to a quantitative level of cell-cell cohesion through direct measurement.' Because the high-cohesiveness branch (v increasing with ψ near ψ = 1) is anchored solely by untreated c = 0 data, off-target effects of DECMA-1 on proliferation, signaling, or substrate adhesion, or a non-monotonic dose-response, could change the sign of the v-ψ slope and the claimed reentrant topology. Please provide a direct calibration of junctional adhesion or, failing that, restrict the claims to the measured variable 'DECMA-1 concentration' and soften the 'cell cohesiveness' language throughout.","section":"Sections 2.3 and 3; Fig. 5B"},{"comment":"The regime boundaries are derived from the gradient of a sum of two 2D Gaussians with freely varying centers, widths, and amplitudes, but the paper reports only R² values and no parameter uncertainties, cross-validation, or sensitivity of the ridge/valley contours to the fitting basis. With seven of the ten DECMA-1 concentrations represented by a single experiment, the double-Gaussian decomposition is not shown to be unique, and a differently parameterized surface could change the location or even the existence of the valley between the two hills. Please provide bootstrap or leave-one-out analyses that yield confidence intervals on the contours, report the number of fitted parameters, and show residual diagnostics.","section":"Section 2.3; Fig. 5A/C"},{"comment":"Only DECMA-1 concentrations 0, 5, and 7 µg/mL have triplicate experiments; the remaining seven concentrations are single measurements. The reentrant shape of v(ψ) at fixed σ—specifically the local minimum and the second maximum—is determined by these unreplicated points, so a single outlier could alter the apparent topology. Please mark single-replicate points in the scatterplots and test whether the four-regime classification survives omission of any one unreplicated concentration; at minimum, state explicitly which features are supported by replicate data.","section":"Sections 2.2 and 2.3; Fig. 4C and Fig. 5"}],"minor_comments":[{"comment":"The sentence 'A scatterplot of the v datapoints in the σ-c plane shows that the data form two hills centered around the densities and DECMA-1 concentrations where peaks were observed, above' is a fragment ending in 'above'; please rephrase and reference Fig. 5A explicitly.","section":"Section 2.3, first paragraph"},{"comment":"The equation formatting for q(t) and σ(t) is garbled in the manuscript text; please ensure the definitions p_j/A_j and 1/A_j are typeset correctly and that the angle-bracket averaging is defined for all three variables.","section":"Section 2.1"},{"comment":"The reference list contains inconsistent journal capitalization, for example 'Nature reviews Molecular cell biology' and 'Physical Review E—Statistical, Nonlinear, and Soft Matter Physics'; please normalize all journal names to their standard styles.","section":"References"},{"comment":"The claim that a best-fit line to the q-v contour terminates at (v=0, q=3.81) should report the fit range, the number of points, and the confidence interval on the intercept; the current text extrapolates beyond the data and does not give uncertainty.","section":"Fig. S4"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for cond-mat.soft and the density-related finding is credible. My recommendation for major revision is driven entirely by the uncalibrated DECMA-1-to-cohesion mapping and the unvalidated landscape fitting; both are fixable with additional experiments or a substantial narrowing of the claims. If the authors choose the narrowing route, the title should be adjusted to avoid the term 'Cooperative' as a proven mechanism."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe useful thing here is the density result: in non-jammed MDCK monolayers, average migration speed rises with cell density over a range of low densities, before the usual crowding suppression kicks in. That is a clean, reproducible observation, backed by time traces, replicate means, and parametric plots. The q–sigma anticorrelation (about 0.88 across 15 experiments) is also solid and nicely documented. The paper earns credit for careful segmentation validation and for being open about its limitations.\n\nThe soft spot is the cohesion axis. The authors map DECMA-1 concentration to a cohesive index psi with no direct measurement of cell–cell adhesion. They acknowledge in Section 3 that blocking E-cadherin may produce indirect responses and that a quantitative cohesion measurement would be preferable. That is a real hole: the reentrant v(psi,sigma) behavior, including the high-psi branch anchored by untreated cells, could be an artifact of the antibody's off-target effects. So the 'speed increases with cohesion' prong is underdetermined, exactly as the stress-test note says. This does not touch the density finding.\n\nTwo smaller issues. First, the 2D landscape is built from freely fitted double Gaussians with no error propagation; the regime boundaries are derived from that fit. The qualitative shapes are visible in the scatter, so this is more a presentation device than a load-bearing argument, but the boundaries should be treated as approximate. Second, the vertex-model comparison at q = 3.81 relies on an extrapolated line to v = 0; the authors are appropriately cautious, but a referee should push on how much weight that gets.\n\nOverall: the central density observation is genuinely new and worth knowing. The cohesion claim needs either a direct adhesion measurement or a downgrade to 'consistent with' rather than 'shows.' I'd send this to peer review; the referees can ask for the calibration and error propagation. I'd cite it for the density–speed increase, not for the cohesion landscape.","headline":"A valuable experimental map with a solid density finding and a cohesion axis that needs calibration before the headline claim fully lands.","tokens_in":13481,"tokens_out":2230,"would_cite":true,"duration_ms":22749,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper reports that in non-jammed MDCK monolayers, cell migration speed can increase with cell packing density and with cell-cell cohesiveness in multiple regimes, showing that mechanical coupling can promote rather than suppress…","keywords":["collective cell migration","MDCK monolayers","jamming transition","cell-cell cohesion","E-cadherin blocking","shape index","migration speed","cell density"],"falsifier":"Re-plot the speed landscape against directly measured cell-cell adhesion strengths, not antibody concentration, using a technique such as dual-pipette aspiration or adhesion-frequency assays across the same ten concentrations, and check whether the two speed hills and the four regimes persist when adhesion is the horizontal axis.This would settle whether the non-monotonic cohesion dependence is truly an adhesion effect.","tokens_in":12465,"feed_emoji":"🧫","tokens_out":5808,"duration_ms":54568,"temperature":0.7,"pith_summary":"This paper reports that in non-jammed monolayers of MDCK epithelial cells, average migration speed can rise when cells are packed more densely or made stickier, in several well-defined regions of a density–cohesion plane. The authors map average cell speed and cell shape across ten levels of an E-cadherin-blocking antibody and a range of cell densities, and find a hilly landscape with two speed peaks: one at maximal cohesion and low-to-moderate density, and one at reduced cohesion near ψ = 0.27. Four regimes emerge in which speed increases or decreases with density and with cohesion in every combination. The finding matters because it contradicts the usual assumption that packing and sticking cells together always suppresses motion, and it suggests that fluid-like tissues can undergo a second, internal transition before jamming.","feed_headline":"Packing cells tighter can make them move faster","feed_subtitle":"A density–cohesion map of MDCK monolayers reveals cooperative regimes where sticking together speeds cells up.","key_machinery":"The argument is carried by a two-dimensional phenomenological landscape: average migration speed v(ψ,σ) built from particle-image-velocimetry flow fields, with cell density σ and shape index q measured by image segmentation, and cell-cell cohesion parameterized by the DECMA-1 concentration mapped to a cohesive index ψ = (c_max–c)/c_max. The speed data are fit as the sum of two independent 2D Gaussian hills, and the boundaries between regimes are found by computing the gradient of the fitted surface. A second fitted landscape for the shape index q(ψ,σ) is used to test whether shape changes accompany the speed transitions. The nearly universal instantaneous anticorrelation between q and σ (≈ −0.88 across conditions) supplies the geometric baseline against which the speed landscape is read.","core_discovery":"The central claim is that the relationship between collective cell motion and mechanical coupling is non-monotonic: at low densities and at both high and low extremes of cell cohesiveness, migration speed increases with increasing density or cohesion, while at intermediate cohesion or high density it decreases. By fitting the measured speed as a smooth surface v(ψ,σ) over the cohesive index ψ = (c_max–c)/c_max and cell density σ, the authors identify two hills separated by a sigmoidal boundary. On the low-density side of the boundary, speed rises with density; on the high-density side, crowding dominates. Along the cohesion axis, speed falls and then rises again, producing a local minimum at intermediate ψ. The shape-index landscape q(ψ,σ) mirrors the low-cohesion hill, and the boundary in v–q space extrapolates to v=0 at q=3.81, the vertex-model rigidity point, even though the cells remain in a fluid-like state.","pith_inferences":["A direct test of the mechanical-stimulation mechanism would be to measure velocity-velocity correlation functions and traction forces in the high-cohesion, low-density regime; if the speed rise is caused by out-of-phase contractions, the correlation length should grow with density in exactly that regime.","The paper's own caveat about DECMA-1 leaves open the possibility that the apparent reentrant cohesion dependence reflects signaling or proliferation changes rather than adhesion per se; separating those channels would require a different way to reduce adhesion, such as genetic knockdown.","If the landscape generalizes, therapies that modulate cell adhesion (for example, in cancer) could either speed or slow collective invasion depending on where the tissue sits on the density–cohesion plane, so a single directional prediction may be impossible."],"forward_implications":["If the landscape is correct, the usual reading of increased cell density or adhesion as a universal brake on monolayer motion must be revised to a regime-dependent view, with consequences for wound healing and tissue development.","The boundary separating cooperative from crowding-dominated motion shifts with cohesion, so a single critical cell density or aspect ratio cannot fully characterize the transition.","The extrapolation of the fluid-regime boundary to q = 3.81 ties the empirical observations to a geometric rigidity threshold, but places it inside the fluid state, suggesting a pre-jamming transition that models should reproduce.","The near-universal q–σ anticorrelation gives experimenters a single readout that indicates where a monolayer sits on the density axis regardless of cohesion."],"supporting_citations":[{"why":"Establishes that DECMA-1 blocks E-cadherin-mediated cell-cell adhesion, providing the experimental handle for the cohesion axis.","marker":"22,23"},{"why":"Source of the vertex-model rigidity prediction at q = 3.81, the point where the fluid-regime boundary extrapolates to v = 0.","marker":"10"},{"why":"Reports unjamming in airway epithelium driven by cell-cell tugging, used to interpret the speed increase at high cohesion.","marker":"5"},{"why":"Prior observation of glass-like dynamics in MDCK monolayers, supplying the velocity-averaging method and the expected shape-density anticorrelation.","marker":"2"},{"why":"Self-propelled Voronoi model whose glass transition boundary terminates at v = 0 in q-v space, providing the comparison for the newly found boundary.","marker":"6"},{"why":"Image segmentation tool used to measure cell shape and density, with validation errors of about 3% in q and 5% in σ.","marker":"24"},{"why":"Single-cell stretching experiments showing a delayed contractile response, the basis for the hypothesis that cells mechanically stimulate one another.","marker":"21"}],"fun_headline_variants":["Packing cells tighter can speed them up — but not always","When cell crowding boosts motion: a density-cohesion map","Counterintuitive cell speed: tighter packing, faster migration","Cell speed rises with packing in low-density, extreme-cohesion regimes","Cooperative and crowding regimes in collective cell motion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The cohesion axis is built entirely on the assumption that increasing DECMA-1 concentration steadily reduces cell-cell adhesion; the authors state that blocking E-cadherin may produce indirect responses in the monolayer, and they have no direct measurement of adhesion levels.","fun_headline_variants_meta":{"raw":{"variants":["Packing cells tighter can speed them up — but not always","When cell crowding boosts motion: a density-cohesion map","Counterintuitive cell speed: tighter packing, faster migration","Cell speed rises with packing in low-density, extreme-cohesion regimes","Cooperative and crowding regimes in collective cell motion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000354,"raw_usage":{"total_tokens":1896,"prompt_tokens":886,"completion_tokens":1010,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":502,"completion_tokens_details":{"reasoning_tokens":928}},"tokens_in":502,"tokens_out":1010,"duration_ms":8845,"temperature":1.0,"reasoning_tokens":928,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:25:51.756424+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-plot the speed landscape against directly measured cell-cell adhesion strengths, not antibody concentration, using a technique such as dual-pipette aspiration or adhesion-frequency assays across the same ten concentrations, and check whether the two speed hills and the four regimes persist when adhesion is the horizontal axis.This would settle whether the non-monotonic cohesion dependence is truly an adhesion effect.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports unjamming in airway epithelium driven by cell-cell tugging, used to interpret the speed increase at high cohesion."},{"cited_title":"+/+”; (2) increasing v along σ and decreasing v along ψ, denoted as “+/–","cited_arxiv_id":null,"evidence_quote":"Prior observation of glass-like dynamics in MDCK monolayers, supplying the velocity-averaging method and the expected shape-density anticorrelation."},{"cited_title":"& Gilmour, D","cited_arxiv_id":null,"evidence_quote":"Self-propelled Voronoi model whose glass transition boundary terminates at v = 0 in q-v space, providing the comparison for the newly found boundary."}],"review_version":1}