Cell Sorting as Boundary-Energy Minimization

The store↔︎match knob that hides the vacuum, read one rung up. A specialized tissue is one whose internal boundaries have relaxed toward the matched limit Δv → 0; cell-adhesion molecules set that knob, tissue surface tension is its measured face, and Steinberg’s Differential Adhesion Hypothesis is boundary-energy minimization in a Petri dish — the framework’s nesting thesis recurring at the tissue scale. The absolute tension is a deep, descriptive read; but the sorting topology and junction angles are tension ratios, in which the underived cortical dressing cancels — a middle-band, dimensionless test, the biological twin of the quark mass ratio rather than the string tension.

Tissues sort like liquids

Drop two kinds of embryonic cells into a dish, dissociated and mixed at random, and they do not stay mixed. They rearrange — over hours — until one population forms a compact ball wrapped completely inside the other, the same configuration every time regardless of how they started. Malcolm Steinberg’s Differential Adhesion Hypothesis (DAH) reads this as a liquid phenomenon: a tissue behaves like an immiscible fluid, cells rearrange to minimize total interfacial energy, and the population that ends up on the inside is the one whose cells cohere more strongly, exactly as oil and water sort by surface tension.¹ The hierarchy is transitive and the engulfment is mutual and reproducible, the hallmark of a system relaxing a single scalar — interfacial energy — toward its minimum.

The liquid analogy is not a metaphor; it is measured. Foty, Forgacs, and Steinberg compressed embryonic tissue aggregates between parallel plates and read off a genuine surface tension — tens of dyn/cm, scale-invariant and shape-restoring like any liquid’s — and then showed that this tissue surface tension scales, cell type by cell type, with the level of cadherin expression.² Cadherins are the calcium-dependent cell-adhesion molecules (CAMs) — Edelman’s “family of molecules that regulate specialization.” More cadherin, higher tissue surface tension, more cohesion, deeper engulfment. The knob that sorts the tissue is the CAM inventory the cell expresses, and the quantity it sets is an interfacial energy per unit area, measured in the same units as any liquid surface tension.

The same object, one rung up

This is the framework’s boundary-energy object wearing a tissue. The profile chapter writes the energy stored in any boundary that divides the substrate as

E_b \;=\; \tfrac12\,\rho_\text{cr}\,(\Delta v)^2 \cdot A\,\delta,

a kinetic energy density \tfrac12\rho_\text{cr}(\Delta v)^2 held in a thin counter-rotating shear layer of thickness \delta across area A, with \Delta v the velocity jump the boundary has to hold. The profile chapter’s central claim is that this object recurs wherever a boundary divides the substrate, at every scale, with only \rho_\text{cr} and the geometry changing — it is the electron’s mass, the photon it emits, the string that confines a quark. The framework’s nesting thesis says the same grammar should appear at the biological scales too, many nesting layers up, wherever the substrate organizes a boundary.

A cell–cell contact is exactly such a boundary, and tissue surface tension is exactly an energy per unit area held there. The two faces of the profile — store and transmit, governed by the single knob \Delta v — map onto cell sorting term for term:

• The matched limit (\Delta v \to 0). The profile’s low-energy end is the stealth vacuum: a coherence-matched boundary stores nothing and reflects nothing, E_b \to 0. The tissue analogue is a homotypic contact — two cells of the same type, the same cadherins on each face, the same cortical state — across which nothing jumps. Its interfacial tension goes to zero; the cells stay together because there is no stored boundary energy to pay. Like cells minimize the energy of being in contact.
• The mismatched limit (large \Delta v). The profile’s high-energy end stores a wall the coin cannot cross. The tissue analogue is a heterotypic contact — two different cell types, mismatched CAMs, mismatched cortex — across which the flow jumps. It carries a large interfacial tension, and the two populations sort apart to shrink that high-tension area to its minimum, which is exactly a sphere-inside-sphere. High \Delta v → high E_b → the boundary is expensive, and minimizing total boundary energy is what drives the sort.

So Steinberg’s “minimize total interfacial energy” is, line for line, minimize \sum_i E_b^{(i)} over the cell-contact boundaries — the leakage theorem’s optimization run on a tissue. A specialized, correctly organized tissue is one whose internal boundaries have relaxed toward the matched limit: specialization stabilizing is boundaries sliding down the \Delta v knob. That is the framework-native reading of why a sorted tissue is a stable one, and why the cell biologist’s “the cells are happy” is the substrate’s “the boundary stores nothing.”

Why tension goes as the square: DAH → DITH

The boundary-energy form is not just a relabeling — it makes a directional call the bare adhesion picture does not, and the field has already moved in that direction. Naïve DAH treats interfacial energy as the adhesion bond count: tension falls linearly as you add cadherin bonds, because each bond is an energy saved. But the profile says the stored energy is \tfrac12\rho_\text{cr}(\Delta v)^2 — it lives in the shear, the velocity contrast across the boundary, and it scales as the square of that contrast, not as a linear bond tally. The energy is in the counter-rotating layer, not in the count of latches across it.

That is precisely the correction the Differential Interfacial Tension Hypothesis (DITH) and the cortical-tension picture supplied. Brodland, and then Maître, Heisenberg, and others, showed that tissue interfacial tension is dominated by the cortical actomyosin tension at each interface, with cadherins acting less as the energy store themselves and more as the coupling that mechanically links the two cells’ cortices and locally down-regulates cortical tension at the contact.³ In the framework’s terms, the cadherins set the match — how nearly the two cortices move together — and the residual cortical-flow contrast they leave is the \Delta v whose square is the stored interfacial energy. The interfacial tension a cell pays is the boundary energy of the cortical shear the cadherins fail to fully cancel. That is why measured tension tracks cortical actomyosin, why it is modulated by adhesion rather than equal to it, and why the matched homotypic contact (\Delta v \to 0, cortices fully coupled) pays almost nothing. The profile predicts the form the field converged on: a squared-flow-contrast energy that adhesion tunes by matching, not a linear bond ledger.

The cleaner read: a ratio the dressing cannot dim

The squared form does more than predict the field’s correction — it sorts this chapter’s own claims by depth, and it says something the rest of the chapter has not: part of this result is not deep at all. The absolute tension \gamma in dyn/cm is deep — it carries the full factor \tfrac12\rho_\text{cr}\delta, and \rho_\text{cr} at the cell cortex is dressed by every layer of chemistry between it and the substrate, so the framework cannot compute it. But a tissue does not sort on the absolute tension. It sorts on ratios and differences of tensions across its contacts — and in a ratio, the dressing cancels.

Take a single pair of cell types, a and b, in a common medium. There are three interfacial tensions — the two homotypic, \gamma_{aa} and \gamma_{bb}, and the one heterotypic, \gamma_{ab} — and each is the same object \gamma=\tfrac12\rho_\text{cr}(\Delta v)^2\delta read at the same cortex, with the same cortical material density \rho_\text{cr} and the same cortex thickness \delta. What differs across the three contacts is only the match — how nearly the two apposed cortices move together — that is, only \Delta v. So the ratio is pure kinematics,

\frac{\gamma_{ab}}{\gamma_{aa}} \;=\; \left(\frac{\Delta v_{ab}}{\Delta v_{aa}}\right)^{2},

and \rho_\text{cr} and \delta — the deep, underived part — drop out. Whether the tissue mixes, sorts with partial engulfment, or sorts to complete engulfment is fixed by the signs of the spreading coefficients S=\gamma_{ij}-\gamma_{ik}-\gamma_{jk}, and those signs are differences of (\Delta v)^2 terms that share the cancelled prefactor. The sorting outcome — the discrete topological fact of who ends up inside whom — is a ratio read, not an absolute one.

This is the biological twin of the framework’s quark mass ratio: m_\text{down}/m_\text{up}\approx2.2 is a comparatively clean geometric estimate precisely because \rho_\text{cr} and \Delta v are common across the two junction boundaries and cancel, leaving only the area ratio. Here the cancellation runs the other way — \Delta v is what varies and the prefactor \tfrac12\rho_\text{cr}\delta is common — but the structural point is identical: a ratio of two reads of the same object climbs the depth ordering, because the deep, underived dressing is shared and divides out. Steinberg’s transitive hierarchy (A engulfs B, B engulfs C \Rightarrow A engulfs C) is then not a curiosity but a consequence of the form: transitivity is the ordering of a single scalar, and a single scalar per cell type is exactly what one (\Delta v)^2 address provides.

The same cancellation hands over a second, sharper measurable. Where three interfaces meet — two cells and the medium, or three cells at a vertex — the cortical tensions balance like soap films, and the dihedral angles they make are fixed by the tension ratios alone (the tissue Neumann/Young–Dupré balance; equal tensions give the soap-film 120^\circ):

\frac{\sin\theta_a}{\gamma_{bc}} \;=\; \frac{\sin\theta_b}{\gamma_{ac}} \;=\; \frac{\sin\theta_c}{\gamma_{ab}}.

Each angle is a dimensionless read of a ratio of (\Delta v)^2, with the dressing gone — measurable under a microscope at a tri-cellular junction, and exactly the kind of middle-band, “chemistry-sets-one-boundary-condition” observable the depth table places above the deep back-outs. The framework cannot tell you \gamma_{ab} in dyn/cm, but it can tell you that the angle at a junction is the arccosine of a ratio of squared cortical-flow contrasts, and that the ordering of those contrasts fixes the entire sorting topology.

The cancellation is cleanest within a pair, where the three contacts share one set of cortices; across unrelated cell types \rho_\text{cr} and \delta need not be identical, so the cross-type ratio inherits a residual dressing and sits a little deeper than the within-pair angles. But even granting that, the conclusion holds: the honest depth placement of this chapter is two-tiered. The absolute tissue surface tension is deep and owed, as the rest of the chapter says — but the sorting topology and the junction angles are a rung higher, ratio reads where the substrate’s underived dressing largely cancels, the biological analogue of the quark mass ratio rather than of the string tension. That upper tier is the part of this result worth pushing on experimentally, because it is the part the framework can be cleanly wrong about — in a dimensionless number, with no \rho_\text{cr} to hide behind.

A worked case in real numbers: one scalar, and the knob reverses it

The ratio tier is not hypothetical — the measurements exist, and they behave exactly as a single (\Delta v)^2 address per cell type requires. Three datasets make the point: one for the ordering, one for the sign, and one that measures the dimensionless ratio itself and finds the bond ledger empty.

The ordering is one scalar (Foty’s five tissues). Foty, Pfleger, Forgacs, and Steinberg measured the tissue surface tension of five chick embryonic tissues by parallel-plate compression and read off, in dyn/cm, a single column that the whole sorting hierarchy obeys:⁴

Tissue	\gamma (dyn/cm)	sorts
limb-bud mesoderm	20.1	innermost
pigmented epithelium	12.6	↓
heart	8.5	↓
liver	4.6	↓
neural retina	1.6	outermost

Every one of the ten pairwise engulfments follows that single column: the higher-\gamma tissue always ends up inside, with no inconsistency anywhere in the chain. That is the framework’s structural claim made quantitative. If sorting depended on anything richer than one scalar per tissue — two independent adhesion channels that could cross, say — the hierarchy could be intransitive: A inside B, B inside C, yet C inside A. It never is. A single (\Delta v)^2 address per tissue is exactly what forces a total order, and a total order is exactly what ten consistent engulfments report. Transitivity is not a lucky bonus; it is the signature that the sorting variable is one number — which is what \tfrac12\rho_\text{cr}(\Delta v)^2\delta provides and a two-channel adhesion ledger would not.

The sign is the knob (the zebrafish phase reversal). The cleaner test is to turn the address and watch the layout follow. Schötz, Heisenberg, Foty, and colleagues measured zebrafish germ-layer progenitor aggregates the same way:⁵

• ectoderm: \gamma \approx 0.75–0.80 dyn/cm → internal
• mesendoderm: \gamma \approx 0.43 dyn/cm → external

Ectoderm’s address sits above mesendoderm’s, so ectoderm sorts inside — the sign of \gamma_\text{ecto}-\gamma_\text{meso} is positive. Now turn the CAM knob: deplete E-cadherin from the ectoderm and its tension drops to \gamma\approx0.33 dyn/cm — below mesendoderm. The difference changes sign, and the sorting reverses: the depleted ectoderm now wraps to the outside. One molecular knob, moved on one cell type, slid its (\Delta v)^2 across its neighbour’s and flipped the engulfment — the spreading-coefficient sign claim, run as an experiment rather than asserted.

And it carries the cleanest dressing-cancelled number in the chapter. Control and depleted ectoderm are the same cell type with only the adhesion knob changed — and, by the DITH picture the chapter already adopts, E-cadherin couples cortices without itself being the cortical tension, so depleting it changes the match while leaving the actomyosin cortex (\rho_\text{cr}, \delta) essentially intact. The two measurements therefore share a prefactor that divides out exactly:

\frac{\gamma_\text{ecto, E-cad depleted}}{\gamma_\text{ecto, control}} \;=\; \frac{0.33}{0.77} \;\approx\; 0.43 \;=\; \left(\frac{\Delta v_\text{depleted}}{\Delta v_\text{control}}\right)^{2},

a pure (\Delta v)^2 factor of \approx0.43 (taking 0.77 as the ectoderm midpoint; the range 0.75–0.80 gives 0.41–0.44) — the effective cortical-flow contrast that \gamma measures falling by \sqrt{0.43}\approx0.66 when the cadherin match is removed, read with the deep prefactor gone because the same cortex sits on both sides of the ratio. This is the within-type version of the cancellation, and it is the cleanest the biology offers: not a tension in absolute units the framework cannot derive, but a ratio of one cell type to itself under a single perturbation, which the framework reads as kinematics alone. (Which boundary’s \Delta v that effective contrast belongs to at the cortex stays in the deep, owed tier; the value of the ratio does not depend on settling it.)

The ratio itself is measured, and the bond ledger is empty (Maître’s doublets). The sharpest confirmation comes from the experiment built to separate the two tensions directly. Maître, Paluch, Heisenberg, and colleagues measured zebrafish progenitor doublets and triplets and extracted, for each germ layer, the dimensionless ratio of cortical tension at the cell–cell contact to that at the cell–medium interface, \gamma_{cc}/\gamma_{cm}, alongside the adhesion-bond contribution \omega/2\gamma_{cm}:⁶

Germ layer	\gamma_{cc}/\gamma_{cm}	\omega/2\gamma_{cm}	\Delta v_{cc}/\Delta v_{cm}=\sqrt{\gamma_{cc}/\gamma_{cm}}	sorts
ectoderm	0.50\pm0.22	-0.07\pm0.22	0.71	innermost
mesoderm	0.65\pm0.31	-0.06\pm0.50	0.81	middle
endoderm	0.69\pm0.38	-0.01\pm0.55	0.83	outermost

Two readings fall straight out. First, \gamma_{cc}/\gamma_{cm} is a ratio of two tensions on the same cortex, so the deep prefactor \tfrac12\rho_\text{cr}\delta cancels exactly and the framework reads the fourth column with no owed quantity: \gamma_{cc}/\gamma_{cm}=(\Delta v_{cc}/\Delta v_{cm})^2, so the cadherin match pulls the cortical-flow contrast at the homotypic contact down to \Delta v_{cc}/\Delta v_{cm}=\sqrt{\gamma_{cc}/\gamma_{cm}} — 0.71 for ectoderm, 0.83 for endoderm. The cell type that sorts innermost (ectoderm) is the one whose homotypic contact is most matched, its \Delta v_{cc} pulled hardest toward zero — exactly the chapter’s “the matched limit is the cohesive, internal contact,” now a measured ordering. This is the junction-angle/ratio test flagged as the one to run first, already run: a dimensionless cortical-flow ratio, dressing cancelled, ordered as the sorting demands.

Second — and this is the chapter’s squared-shear prediction confirmed by the authors’ own headline — the adhesion-bond term \omega/2\gamma_{cm} is statistically indistinguishable from zero for all three layers (-0.07, -0.06, -0.01, each smaller than its own error bar). The energy that sorts these cells is not in the count of cadherin trans-bonds; it is in the cortical shear the cadherins fail to fully cancel. Maître et al. title the result “adhesion functions in cell sorting by mechanically coupling the cortices of adhering cells” — cadherin as the coupling that sets the match, the actomyosin cortex as where the tension lives. That is the framework’s “\Delta v is the cortical-flow contrast the cadherins tune, and the stored energy is its square, not a linear bond ledger” — written in the cell biologists’ own title and read off their own table.

What the framework does not claim is novelty of the numbers — DAH and DITH organize them already, and the tensions are theirs. What it adds is the reading: that the sorting-controlling quantity is a ratio/sign in which the underived dressing cancels (so it is middle-band, not deep), that the per-type scalar is a squared cortical-flow contrast bounded by the matched and mismatched limits, and that the CAM is the match knob whose turning the phase-reversal experiment makes visible. The ordering (Foty’s transitive five), the sign flip under the knob (the zebrafish phase reversal), the dressing-cancelled ratio (\gamma_{cc}/\gamma_{cm} ordered as the sort), and the empty bond ledger (\omega\approx0) are the framework’s reading made quantitative — and all four are already in the measurements.

Sharp boundaries and the dynamical layer between two states

There is a second, sharper signature, and it is the tissue version of a result this paper already developed in water. Where two cell populations meet at a maximal mismatch — a developmental compartment boundary, like the Drosophila anteroposterior or dorsoventral boundary, or a vertebrate rhombomere boundary — the interface does not stay ragged. It is forced sharp and straight, and a supracellular actomyosin cable assembles right along it: a contractile rope spanning many cells that holds the boundary taut and prevents the two populations from intermixing.⁷ This is the substrate’s most literal prediction made cellular. The cold-wall chapter states the rule — “wherever organized rotational energy meets a different organized regime in the substrate, the boundary is forced sharp” — and reads the sharpness as a counter-rotating shear layer the elastic medium holds against the broadening that simple diffusion would predict. The supracellular actomyosin cable is that shear layer, built in muscle protein: a dynamical contractile interface between two organized cell states, narrower than passive cell-mixing would leave it, exactly where \Delta v is largest.

And this is the same motif the chapter on water found in the liquid itself. The two-state reading of supercooled water — a denser disordered structure and a less-dense ordered one, with the most complex, looped interconversion pathway appearing exactly at the phase boundary where the two states compete most — is the molecular sibling of the same picture. Two organized states, a dynamical layer between them, the boundary forced sharp and most active right at the seam. Water shows it at 10^{-10} m, the cell shows it at 10^{-5} m, the Gulf Stream shows it at 10^{5} m. Cell sorting adds the tissue rung: the actomyosin cable is the dynamical layer the substrate insists on wherever two coherent regimes are asked to share a boundary, and the energy it holds is \tfrac12\rho_\text{cr}(\Delta v)^2 at the cell cortex.

Specialization sets the knob

What sets which cadherins a cell expresses — what fixes its place on the \Delta v knob — is its differentiation state, and the framework has already said what that is. The epigenetics chapter reads cell type as the latch: the non-volatile, heavily-wrapped epigenetic memory that holds “which loci are open” across divisions, the cell’s flash memory. The CAM inventory is one of the things that latch writes. A cell’s adhesion address — its cadherin levels, its CAM repertoire — is downstream of the methyl-and-chromatin marks that fix its lineage. So the causal chain the framework can now draw end to end is:

\underbrace{\text{epigenetic latch}}_{\text{which loci open}} \;\longrightarrow\; \underbrace{\text{CAM inventory}}_{\text{the }\Delta v\text{ address}} \;\longrightarrow\; \underbrace{\text{interfacial tension}}_{E_b=\frac12\rho_\text{cr}(\Delta v)^2} \;\longrightarrow\; \underbrace{\text{sorting \& tissue layout}}_{\text{minimize}\sum E_b}.

The flash sets the address; the address sets the boundary energy; minimizing the boundary energy lays out the tissue. This is why changing the flash changes the shape: re-specify a cell’s epigenetic state and you re-specify its CAMs, which re-specifies the boundary energies it will pay, which re-sorts it into a different position in the tissue. The “green signal when the organization is coherent” is then exactly the matched limit reached: a tissue whose cells have each found the neighbours their CAM address matches, every internal boundary relaxed toward \Delta v \to 0, the total stored boundary energy at its floor. Coherent organization is boundary-energy minimization completed, and the cell’s readout of “I am in the right place, stop moving” — contact-mediated stabilization, the cessation of sorting — is the substrate’s E_b \to 0 read locally.

Topobiology: the chemistry-side accounting

This is the substrate reading of what Gerald Edelman named topobiology: the thesis that morphology is generated by place-dependent regulation of a small set of morphoregulatory molecules — the CAMs and the substrate-adhesion molecules (SAMs, like fibronectin and the laminins) — whose modulation drives the primary processes of development (cell division, movement, death, adhesion, and differentiation) into shape.⁸ Edelman’s CAM-modulation hypothesis says a few adhesion molecules, switched on and off by place, are enough to organize an embryo. The framework reads that the way the cortical-maps chapter already reads his later work — as biology’s chemistry-side accounting of a substrate process. There, Edelman’s re-entrant signalling is the chemistry-side bookkeeping of canonical-loop substrate-current circulation; here, topobiology is the chemistry-side bookkeeping of boundary-energy minimization. The CAMs are the molecules the cell uses to set \Delta v; the morphology that results is the tissue relaxing \sum E_b to its minimum; “place-dependent CAM modulation” is the cell tuning each boundary’s stored energy by tuning the match across it.

The continuity runs straight up into the brain. The same morphoregulatory machinery that sorts a tissue lays out the topographic maps — the retinotopic, tonotopic, somatotopic sheets the cortical-maps chapter reads as substrate-coherence-cell tilings preserved into the cortex. Edelman’s bridge from Topobiology to The Remembered Present is exactly this: the molecules that organize tissue by differential adhesion are the molecules that organize the cortical maps whose re-entrant integration he called the remembered present. In the substrate reading that bridge is one mechanism seen twice — boundary-energy minimization laying out a coherent topology, first as a sorted tissue, then as a cortical map, then as the global mappings the cortical-maps chapter reads as substrate-coherence-cell assemblies. Following the energy of specialization down the \Delta v knob is following it from the cadherin to the cortical column.

Reading depth: a deep, descriptive read

The following-the-energy discipline requires the honest placement, and it is two-tiered. The absolute tissue surface tension is a deep read, descriptive rather than clean: it sits behind a long stack of boundaries — substrate → cadherin and cortical-actomyosin chemistry → cell → tissue — sixteen-plus nesting layers above the dc1 scale where the string tension reads the same object E_b at zero depth. The string tension is the clean read of the boundary-energy profile; absolute tissue surface tension is the same form evaluated far up the nesting stack, where reading depth guarantees the substrate’s own number arrives scrambled. \rho_\text{cr} at the cell cortex is not the substrate’s bare compression — it is dressed by every layer of chemistry between. So the framework reads \Delta v here qualitatively, as the cortical-flow contrast cadherins tune, and does not pretend to compute \gamma in dyn/cm from n_1, m_1.

The sorting topology and junction angles, by contrast, sit a rung higher — in the middle band of the depth table, not the deep one. As the ratio argument shows, they depend only on ratios and differences of tensions across contacts that share the same cortex, so the deep prefactor \tfrac12\rho_\text{cr}\delta divides out and what survives is a dimensionless ordering of (\Delta v)^2. This is the same depth move the framework makes when it reads the quark mass ratio more cleanly than any absolute mass: a ratio of two reads of one object is shallower than either read alone. So the chapter’s honest verdict is not “deep, full stop” but “deep in the absolute, middle-band in the ratio” — and the ratio is where the testable cleanness lives.

Framed that way — the same store↔︎match knob that hides the vacuum governs how tissues sort — the chapter is solid reach: the boundary-energy grammar recurring exactly where the nesting thesis says it should, one rung up. Framed as “we measured the substrate in a Petri dish,” it would be precisely the overclaim the following-the-energy chapter exists to forbid. The honest statement is the first one. This chapter is a lens, like the boundary-energy profile and reading depth themselves: it shows that one object underlies the electron’s mass and the way an embryo sorts its cells, and that Steinberg’s measured surface tension is that object read deep. It adds a row to the framework’s reach, not to its zero-depth catalogue.

Predictions and falsification

The lens sharpens into testable form, in two tiers. The absolute-tension predictions are deep-band (Tier 3), owed and underivable as the reading-depth section sets out; the ratio-level ones are middle-band, dimensionless, and — as the worked numbers show — already borne out by the literature. The value is the unification — one knob predicts the whole pattern — and the ratio reads are where it can be cleanly falsified.

1. Interfacial tension is a squared-flow-contrast energy, not a linear bond count. The profile says \gamma \propto (\Delta v)^2 in the cortical-flow mismatch, with cadherins setting the match. The prediction is that measured tissue interfacial tension tracks cortical actomyosin activity (and its square-law dependence on the residual flow contrast) more tightly than it tracks raw adhesion-bond number — the DAH → DITH shift, which the framework retrodicts as the necessary form. Maître’s doublet measurements already deliver the decisive form: the dimensionless cortical-tension ratio \gamma_{cc}/\gamma_{cm} orders the germ layers while the adhesion-bond term \omega measures zero — the energy is the cortical shear, not the bond tally. A system where tension scaled strictly linearly with cadherin number and was independent of cortical tension would falsify the squared-shear reading; the measured \omega\approx0 is that falsification run in reverse.
2. The matched homotypic contact stores near-zero interfacial energy. Tension across a contact between two identical, identically-conditioned cells should fall toward a floor as their CAM and cortical states are matched — the \Delta v \to 0 limit — and rise monotonically with any imposed mismatch (heterotypic pairing, cadherin-level difference, cortical perturbation). Sorting strength should track that rise.
3. Maximal-mismatch boundaries build a dynamical shear layer and are forced sharp. Wherever \Delta v is largest — compartment and rhombomere boundaries — a supracellular actomyosin cable should assemble and the boundary should be narrower and straighter than passive cell-mixing predicts, the tissue twin of the cold wall and the supercooled-water two-state seam. Abolishing the cable (myosin-II inhibition) should let the boundary broaden and the populations intermix — observed.
4. The latch sets the knob. Re-specifying a cell’s epigenetic state should re-specify its CAM address and therefore its sorting position; the sorting hierarchy of a set of cell types should be predictable from their differentiation-state-set CAM inventories. A lineage change that altered CAMs but left sorting position unchanged, or a sorting change with no CAM/cortical correlate, would break the latch → knob → layout chain. The zebrafish phase reversal — E-cadherin depletion carrying ectoderm’s tension below mesendoderm’s and inverting which engulfs which — is this prediction already observed: turn the knob across a neighbour’s value and the layout flips.
5. The sorting topology and junction angles are the clean, ratio-level test — and the one that climbs the depth ladder. Because tension ratios cancel the underived cortical dressing \tfrac12\rho_\text{cr}\delta, the discrete sorting outcome (mix / partial / complete engulfment) and the dihedral angles at tri-cellular junctions are dimensionless predictions set by the ordering of (\Delta v)^2 across contacts alone — the biological twin of the quark mass ratio, a rung shallower than any absolute tension. Concretely: measure the three cortical/contact tensions of a cell-type pair and predict its junction angles from the Neumann balance and its engulfment from the spreading-coefficient signs, with no fitted prefactor. Both halves already have data. Foty’s five-tissue hierarchy supplies the ordering — ten consistent, transitive engulfments from one scalar per tissue, exactly what a single (\Delta v)^2 address forces; Maître’s \gamma_{cc}/\gamma_{cm} ratios supply the dressing-cancelled angle/ratio read, ordered as the germ layers sort. A junction-angle set, or a sorting hierarchy, that did not follow a single per-type scalar ordering — an intransitive triple, A>B>C>A — would falsify the squared-shear reading more sharply than any dyn/cm value can. This was the prediction to run first because it is the only one with no \rho_\text{cr} to hide behind; it has now been run, and it holds.
6. The knob run pathologically: EMT and metastasis are a cell climbing the \Delta v knob. The same sign-flip the zebrafish depletion demonstrated by hand runs as a disease. The epithelial–mesenchymal transition (EMT) down-regulates E-cadherin as a programmed differentiation switch, and the framework reads that as a tumour cell sliding up the \Delta v knob: its homotypic contacts lose the match, their interfacial tension rises off the matched floor, the spreading coefficient against the surrounding tissue changes sign, and the cell un-sorts — leaves the compartment it was minimizing boundary energy inside. Invasion is the engulfment hierarchy broken from within; metastasis is a cell whose CAM address no longer sorts it home. This keeps the chapter’s tier discipline: the prediction is a ratio/sign claim — invasiveness should track the sign and magnitude of the cadherin-set tension difference to the neighbouring tissue, the spreading-coefficient crossing, not any absolute \gamma in dyn/cm — and it is prediction 4 at its sharpest, because EMT is an epigenetic-latch rewrite of the CAM address: the flash flips, the address re-prices every boundary, the cell relocates — here catastrophically. Two falsifiers follow. Re-imposing the match (forced E-cadherin re-expression, the mesenchymal→epithelial reversion seen at metastatic colonization) should drive the cell back down the knob and re-sort it inward; and an invasive cell that left its tissue with its cadherin match and cortical contrast to neighbours measurably intact — no climb up \Delta v — would break the latch → knob → layout chain the chapter draws. The clinical face of “turn the CAM knob and the layout follows” is that turning it the wrong way is how a tissue comes apart.

Honest assessment

What is solid is the organizing claim: that cell sorting is boundary-energy minimization, that Steinberg’s measured tissue surface tension is the boundary-energy object \tfrac12\rho_\text{cr}(\Delta v)^2 read at the cell cortex, that the matched limit \Delta v \to 0 is the homotypic “green-signal” contact, and that the (\Delta v)^2 form predicts the field’s own DAH → DITH correction. These are faithful re-descriptions that connect measured cell biology to the framework’s central object, and the compartment-boundary actomyosin cable is a clean, already-observed instance of the substrate’s forced-sharp-boundary rule one rung above the cold wall. And the ratio tier now has real numbers behind it, in three independent datasets the field already holds: Foty’s transitive five-tissue hierarchy is the single-scalar prediction met; the zebrafish phase reversal — a within-cell-type tension ratio of 0.33/0.77\approx0.43 read with the dressing cancelled — is the sign-flip-under-the-knob prediction met; and Maître’s doublet measurements are the cleanest of all, giving the dimensionless cortical-tension ratio \gamma_{cc}/\gamma_{cm} (ordered 0.50<0.65<0.69 as the germ layers sort) with the adhesion-bond term \omega measured at zero — the squared-cortical-shear reading and the empty-bond-ledger reading confirmed in one table.

What is not solid, and is flagged as such, is any absolute number: the chapter computes no tissue tension in dyn/cm, because that read is deep and \rho_\text{cr} at the cortex is dressed by all the intervening chemistry. The honest status of the absolute tension matches the boundary-energy profile’s own and reading-depth’s: the form is universal, the value is owed, and at this depth it is owed in a way no clean read can presently pay.

What is solid beyond the re-description, and is this chapter’s one genuine increment, is the ratio tier: because the deep prefactor cancels in any tension ratio, the sorting topology and the tri-cellular junction angles are dimensionless predictions set by the ordering of (\Delta v)^2 alone — the biological twin of the quark mass ratio, a rung shallower than the absolute tension and the part of the result that can be cleanly falsified. The chapter earns its place two ways, then: by unifying — showing that the knob that hides the vacuum and confines the quark also sorts the embryo and lays out the cortical map — and by locating the one clean test inside a deep result, a ratio read where the substrate’s underived dressing divides out. It still adds no sharp number to the scorecard today, but it names exactly which measurement would.

Putting the section in context

The boundary-energy section named one object — the energy a counter-rotating layer stores, \tfrac12\rho_\text{cr}(\Delta v)^2 A\delta, with the single knob \Delta v that makes a boundary either store-nothing-and-pass-all (the matched vacuum) or store-a-wall-and-reflect-all (the nuclear seam). This chapter reads that knob in a tissue. A specialized cell wears a CAM address its epigenetic latch wrote; that address sets the cortical-flow contrast across each of its contacts; the boundary energy of that contrast is its interfacial tension; and a tissue lays itself out by minimizing the sum of those energies — sliding every internal boundary toward the matched limit, which is what a coherent, correctly-sorted, stable tissue is. Steinberg measured the tension and tied it to the cadherins; Edelman named the place-dependent CAM regulation that tunes it topobiology and carried it up into the cortical maps of the remembered present. The framework reads both as the chemistry-side accounting of one substrate process — boundary-energy minimization laying out a coherent topology — and places the read honestly and in two tiers: the absolute tension is deep in the nesting stack, descriptive not derived, the same object as the string tension seen sixteen orders of magnitude up; but the sorting topology and junction angles are tension ratios in which that deep dressing cancels, a middle-band dimensionless test, the twin of the quark mass ratio rather than the string tension. Following the energy of specialization is following \Delta v down to zero, from the cadherin to the cortical column — and the one place the descent leaves a clean, falsifiable footprint is the ratio, where the substrate’s hidden number divides itself out.

Footnotes

1. Steinberg, M.S., “Reconstruction of tissues by dissociated cells,” Science 141, 401–408 (1963); Steinberg, M.S., “Differential adhesion in morphogenesis: a modern view,” Curr. Opin. Genet. Dev. 17, 281–286 (2007). The hypothesis predicts a transitive sorting hierarchy — if A engulfs B and B engulfs C, then A engulfs C — which is observed.

2. Foty, R.A., Forgacs, G., Pfleger, C.M., Steinberg, M.S., “Liquid properties of embryonic tissues: measurement of interfacial tensions,” Phys. Rev. Lett. 72, 2298–2301 (1994); Foty, R.A. and Steinberg, M.S., “The differential adhesion hypothesis: a direct evaluation,” Dev. Biol. 278, 255–263 (2005), which shows tissue surface tension rising linearly with the number of cadherin molecules expressed; Foty, R.A. et al., “Surface tensions of embryonic tissues predict their mutual envelopment behavior,” Development 122, 1611–1620 (1996).

3. Brodland, G.W., “The Differential Interfacial Tension Hypothesis (DITH): a comprehensive theory for the self-rearrangement of embryonic cells and tissues,” J. Biomech. Eng. 124, 188–197 (2002); Maître, J.-L. et al., “Adhesion functions in cell sorting by mechanically coupling the cortices of adhering cells,” Science 338, 253–256 (2012); Heisenberg, C.-P. and Bellaïche, Y., “Forces in tissue morphogenesis and patterning,” Cell 153, 948–962 (2013); Amack, J.D. and Manning, M.L., “Knowing the boundaries: extending the differential adhesion hypothesis in embryonic cell sorting,” Science 338, 212–215 (2012).

4. Foty, R.A., Pfleger, C.M., Forgacs, G., Steinberg, M.S., “Surface tensions of embryonic tissues predict their mutual envelopment behavior,” Development 122, 1611–1620 (1996). The five-tissue hierarchy has been reproduced across labs.

5. Schötz, E.-M., Burdine, R.D., Jülicher, F., Steinberg, M.S., Heisenberg, C.-P., Foty, R.A., “Quantitative differences in tissue surface tension influence zebrafish germ layer positioning,” HFSP J. 2, 42–56 (2008); Krieg, M. et al., “Tensile forces govern germ-layer organization in zebrafish,” Nat. Cell Biol. 10, 429–436 (2008), which traces the germ-layer tension differences to Nodal/TGFβ signalling — the differentiation input that, in the framework’s chain, writes the CAM address.

6. Maître, J.-L., Berthoumieux, H., Krens, S.F.G., Salbreux, G., Jülicher, F., Paluch, E., Heisenberg, C.-P., “Adhesion functions in cell sorting by mechanically coupling the cortices of adhering cells,” Science 338, 253–256 (2012); values from Table S3 of the supplement, with the contact force balance 2\gamma_{cm}\cos\theta = 2\gamma_{cc}-\omega (their Eq. 6) — the cell-doublet Young–Dupré/Neumann relation, dimensionless in the angle \theta and the tension ratio.

7. Major, R.J. and Irvine, K.D., “Localization and requirement for Myosin II at the dorsal-ventral compartment boundary of the Drosophila wing,” Dev. Dyn. 236, 3051–3058 (2007); Landsberg, K.P. et al., “Increased cell bond tension governs cell sorting at the Drosophila anteroposterior compartment boundary,” Curr. Biol. 19, 1950–1955 (2009); Monier, B. et al., “An actomyosin-based barrier inhibits cell mixing at compartmental boundaries in Drosophila embryos,” Nat. Cell Biol. 12, 60–65 (2010); Calzolari, S. et al., “Cell segregation in the vertebrate hindbrain relies on actomyosin cables located at the interhombomeric boundaries,” EMBO J. 33, 686–701 (2014). Eph/ephrin signalling supplies the mismatch cue at many of these boundaries.

8. Edelman, G.M., Topobiology: An Introduction to Molecular Embryology (Basic Books, 1988); Edelman, G.M., “Cell adhesion molecules,” Science 219, 450–457 (1983); Edelman, G.M. and Crossin, K.L., “Cell adhesion molecules: implications for a molecular histology,” Annu. Rev. Biochem. 60, 155–190 (1991). Edelman’s “CAM modulation hypothesis” holds that a few adhesion molecules, regulated in time and place, coordinate the primary processes of development.