Organizer Molecules and Antagonist Gradients
Why does the fly's gut sit on the same side as our spinal cord — but upside down? This is not a trick question; it has a precise molecular answer. When you look at a cross-section of an insect and a cross-section of a vertebrate, the nerve cord and the gut swap positions relative to each other. For over a century this dorsoventral (D-V) inversion was a curiosity of comparative anatomy. It fueled speculation that arthropods and chordates evolved their body plans independently. The resolution came not from fossils but from molecules. The Spemann-Mangold Organizer was known to dorsalize vertebrate embryos. But for sixty years nobody knew what it secreted. In the 1990s a set of experiments in frogs, and in parallel in flies, revealed something new. The organizer works by exporting antagonist proteins — Chordin and Noggin. These bind and neutralize a diffusible growth factor called BMP (Bone Morphogenetic Protein). The fly homolog of this same antagonist, Short gastrulation (Sog), does the identical job. It works on the opposite side of the embryo. This single discovery unified the D-V patterning systems of protostomes and deuterostomes. It explained the anatomical inversion as a literal flip of the same molecular map. The big picture is this: pattern along an embryonic axis is not built by a single gradient alone. It is built by a gradient of a signal opposed by a gradient of its own antagonist. This diffuses from a discrete source — the organizer. It shapes a field of cells that must interpret concentration as positional identity. This is the mechanistic partner to the concept of the Morphogen Gradient. BMP is the morphogen; Chordin/Noggin/Sog are what sculpt its shape. This article works through the molecular logic of BMP antagonism step by step. It covers the classic embryological and biochemical experiments that nailed it down. It includes two worked quantitative examples and the common conceptual traps. It also covers the scientists whose careers intersected on this problem. Finally, it looks at where the same antagonist logic is exploited in medicine today.
- The Spemann-Mangold organizer works largely by secreting BMP antagonists — Chordin and Noggin — not by secreting a novel "dorsalizing" morphogen from scratch.
- Chordin and Noggin bind BMP4/BMP2 directly, preventing receptor binding; this converts a uniform BMP field into a D-V gradient of free BMP.
- High free BMP → ventral/epidermal fates; low free BMP (organizer side) → dorsal/neural fates — this is an inverse morphogen gradient.
- Drosophila uses the same logic inverted: Sog (Chordin homolog) is made dorsally, Dpp/Scw (BMP homologs) signal ventrally — a 180° flip that explains why insect nerve cords are ventral and vertebrate neural tubes are dorsal.
- The metalloprotease Tolloid/BMP-1 cleaves Chordin/Sog, releasing active BMP — adding a second layer of spatial control beyond simple binding.
- Direct biochemical proof came from Sasai, De Robertis, and colleagues (1994–1995): purified Chordin protein blocks BMP4 signaling and induces neural tissue in ventral ectoderm explants.
Dorsoventral pattern is carved not by one morphogen but by a diffusible morphogen (BMP) sculpted into a gradient by a diffusible antagonist (Chordin/Noggin/Sog) secreted from the organizer — and this system is so ancient that flies and vertebrates use the same two molecules, flipped top-to-bottom.
The core idea
A morphogen is a signaling molecule whose local concentration is read out by cells as positional information, producing different cell fates at different distances from a source — this was established for structures like the vertebrate neural tube and limb bud (see Morphogen Gradients). Bone Morphogenetic Proteins (BMPs), members of the TGF-β superfamily, are the master morphogens of the dorsoventral axis in both vertebrate and invertebrate embryos. But a morphogen gradient needs a source and, critically, a way to prevent it from simply diffusing into a flat, uninformative plateau. This is where antagonists come in. A BMP antagonist is a secreted protein that binds BMP ligand directly, sterically blocking it from engaging its cell-surface receptors. The two central antagonists of the D-V system are Chordin (a large, cysteine-rich secreted protein with four "chordin domains" that each bind BMP) and Noggin (a smaller disulfide-linked dimer that clamps around a BMP dimer, plugging its receptor-binding epitopes). Both are secreted by the Spemann-Mangold organizer — the dorsal blastopore lip in amphibians, and its homologous structures (Hensen's node in birds, the node in mammals) in other vertebrates. Because BMP itself is expressed broadly across the ventral and lateral ectoderm/mesoderm, while the antagonist is expressed only in the small dorsal organizer domain, the result is a smooth gradient of free, receptor-competent BMP running from low (dorsal) to high (ventral) — even though total BMP mRNA/protein might be present much more uniformly. Position along the D-V axis is thus read out not from BMP concentration alone, but from the local ratio of BMP to antagonist. The deep insight, cemented in the mid-1990s, is that this exact same molecular pair operates in Drosophila, with the geometry flipped. The fly antagonist Short gastrulation (Sog) is Chordin's homolog; the fly BMP ligands Decapentaplegic (Dpp) and Screw (Scw) are homologs of vertebrate BMP4/BMP2/BMP7. In the fly, Sog is expressed ventrolaterally and Dpp signals dorsally to specify the amnioserosa and dorsal ectoderm, while the ventral neuroectoderm (which makes the nerve cord) forms where BMP signaling is low. In vertebrates, Chordin is expressed dorsally (organizer) and BMP signals ventrally, with the dorsal neural tube forming where BMP is low. The same molecular circuit, dorsal-high-BMP in flies versus dorsal-low-BMP in vertebrates, produces mirror-image anatomy from a conserved wiring diagram — strong molecular evidence for a common bilaterian ancestor with an inverted body plan relative to one lineage or the other (a hypothesis originally floated on anatomical grounds by Geoffroy Saint-Hilaire in 1822, vindicated genetically 170 years later).
Mechanism
Step 1 — BMP ligand is broadly transcribed. In the Xenopus gastrula, BMP4 mRNA is expressed throughout the ventral and lateral marginal zone and ectoderm — essentially "everywhere except where the organizer actively suppresses it." Left alone, uniform BMP4 signaling would ventralize the entire ectoderm, producing epidermis and blood, and no neural tissue or dorsal mesoderm at all. This is exactly what happens when BMP4 is over-expressed experimentally, or when the organizer is surgically removed.
Step 2 — The organizer secretes antagonists into the extracellular space. The dorsal blastopore lip expresses chordin, noggin, and a third antagonist, follistatin, at high levels. These are freely diffusible secreted proteins. Chordin in particular can diffuse several cell diameters from its source, aided by transport mechanisms (see Step 4) that actively shuttle it away from the organizer.
Step 3 — Antagonists bind BMP with high affinity, blocking receptor engagement. Noggin binds BMP2/BMP4 dimers with sub-nanomolar affinity, physically occluding the epitopes on BMP that would otherwise engage type I and type II BMP receptor serine/threonine kinases. Chordin does the same via its cysteine-rich domains. With BMP sequestered, cells cannot activate their receptors, so the intracellular signal transducers — Smad1/5/8 — are not phosphorylated. Unphosphorylated Smad1/5/8 cannot partner with the common mediator Smad4 and cannot enter the nucleus to switch on ventralizing target genes.
Step 4 — A gradient of free BMP forms across the D-V axis. Because antagonist concentration is highest near the organizer and falls off with distance, while BMP ligand production is roughly uniform, the ratio of free (unbound, receptor-active) BMP to bound (inactive) BMP increases moving away from the organizer. The result is an inverse gradient: BMP signaling (as read out by phospho-Smad1/5/8 immunostaining) is lowest dorsally and highest ventrally in the vertebrate embryo. Remarkably, this gradient is not just passive diffusion — Chordin-BMP complexes are actively transported along the embryo surface toward the ventral side by a second protease system, redistributing BMP away from the dorsal side even further and sharpening the gradient (a phenomenon sometimes called "facilitated diffusion" or shuttling).
Step 5 — A protease, Tolloid (BMP-1 homolog), locally destroys the antagonist. Chordin-BMP complexes are substrates for the zinc metalloprotease Xolloid/Tolloid (BMP-1 in vertebrates, Tolloid in flies). Cleavage of Chordin releases active BMP at specific locations, adding a spatial "release valve" independent of simple diffusion. This step is essential in the fly, where it concentrates peak Dpp/Scw signaling in a narrow dorsal stripe rather than a shallow gradient.
Step 6 — Cells read out the local BMP/Smad signal as a fate decision. High pSmad1/5/8 in ventral ectoderm activates genes like vent1/vent2 and msx1, driving epidermal and ventral mesodermal (blood island) fates. Low pSmad1/5/8 in dorsal ectoderm permits default neural induction: ectodermal cells, freed from BMP-driven repression, express proneural genes (sox2, ngnr) and become neural plate. This is the celebrated "default model" of neural induction — ectoderm becomes neural not because an inducer instructs it to, but because an inhibitor removes a standing repression.
Step 7 — In the mesoderm, the same axis is set in parallel. BMP signaling in the dorsal mesoderm is similarly suppressed by organizer-derived Chordin/Noggin, allowing dorsal mesoderm to become notochord and somite rather than ventral blood/mesenchyme. So one antagonist gradient patterns two adjacent germ layers simultaneously.
The evidence
Nieuwkoop and the "default state" problem (1950s–1970s). Pieter Nieuwkoop's classic recombination experiments in Triturus showed that isolated animal cap ectoderm, cultured alone, forms only epidermis, but adjacent to vegetal endoderm it forms neural tissue and mesoderm — establishing that neuralization requires a signal from a neighboring tissue, not an autonomous ectodermal program. This set up the puzzle: what is that signal, and is it an inducer (instructive, "make neural tissue") or a de-repressor (permissive, "stop making epidermis")?
Spemann and Mangold (1924) had already shown that a graft of the dorsal blastopore lip into a host embryo induces a complete secondary axis, including a second neural tube built largely from host tissue — the organizer transplant experiment that founded the entire field (Spemann-Mangold Organizer). But the molecular identity of "the organizer signal" remained unknown for seventy years, despite exhaustive biochemical hunting (this gap was sometimes called "the Spemann problem").
Sasai, Lin, Steinbeisser, Geissert, Yamada & De Robertis (1994, Cell). Working in Eddy De Robertis's lab, Yoshiki Sasai cloned chordin by differential screening for genes expressed specifically in the Xenopus organizer. Injection of synthetic chordin mRNA into ventral blastomeres — which normally make only ventral epidermis and blood — induced a secondary dorsal axis with organized neural tissue, notochord, and somites, phenocopying an organizer graft. This was the first molecular candidate that, on its own, could recapitulate organizer activity from a single injected mRNA.
Sasai, Lin, De Robertis and colleagues (1995, Nature, and companion papers). The decisive biochemical proof came next: purified, secreted Chordin protein was shown to bind BMP4 directly in solution (co-immunoprecipitation and cross-linking assays) and to block BMP4 from binding its receptors. In ectodermal explant assays ("animal cap assays"), BMP4 protein alone drove explants to epidermal fate; adding Chordin protein alongside BMP4 blocked this and instead induced neural markers, showing Chordin acts by direct ligand sequestration, not by activating a separate receptor.
Smith and Harland (1992) cloned Noggin as a secreted factor from a Xenopus organizer-region expression screen; injected noggin mRNA into ventral cells induced a secondary axis with neural tissue, exactly as Chordin later would. Zimmerman, De Jesús-Escobar & Harland (1996) then showed biochemically that Noggin binds BMP4 with high affinity and blocks its signaling — placing Noggin and Chordin in the same mechanistic class.
The clinching "double knockout" and loss-of-function experiments. If Chordin and Noggin together really constitute the organizer's dorsalizing activity, removing both should largely abolish organizer function. Depleting Chordin and Noggin simultaneously (using antisense/dominant-negative approaches) in Xenopus substantially reduces dorsal development, though residual dorsalizing activity (from Follistatin and other factors) shows the organizer uses redundant antagonists — an important nuance, not a clean single-gene story.
The fly side: Francois, Solloway, O'Neill, Emery & Bier (1994) and Holley, Jackson, Sasai, Lu, De Robertis, Hoffmann & Ferguson (1995, Nature). Genetic analysis in Drosophila had already shown that loss of sog (short gastrulation) expands dpp-dependent dorsal fates ventrally, and loss of dpp expands ventral neuroectoderm dorsally — establishing Sog as a Dpp antagonist genetically. The molecular bombshell came when Sasai and colleagues showed Sog protein directly binds BMP4/Dpp and blocks its signaling in Xenopus assays, and reciprocally that vertebrate Chordin can substitute for Sog function when expressed in flies. This cross-phylum rescue — a vertebrate gene functioning in an insect, and vice versa — is exceptionally strong molecular evidence for deep homology of the D-V system, not mere convergent evolution.
Genetic epistasis nails the direction of the D-V flip. In flies, ectopic Sog expression dorsally converts dorsal ectoderm toward neuroectodermal fate; in frogs, ectopic Chordin ventrally converts ventral ectoderm toward neural fate. Since fly dorsal = low BMP = neural-like outcome under Sog, and frog dorsal = low BMP = neural under Chordin, the molecular logic (antagonist blocks BMP → neural tissue) is identical, while the anatomical mapping onto dorsal/ventral is inverted between phyla — precisely the pattern predicted by Geoffroy Saint-Hilaire's 1822 "inversion hypothesis," now given a mechanistic, testable basis 170 years later. This is widely regarded as one of the most elegant convergences of comparative anatomy and molecular genetics in developmental biology.
Worked examples
Example 1 — Predicting phenotypes from a dose-response logic. Suppose you have four experimental Xenopus embryos, injected ventrally at the 4-cell stage with different combinations of mRNA. Assume: BMP4 signaling above a high threshold gives epidermis; intermediate BMP4 signaling gives ventral mesoderm/blood; and BMP4 signaling below a low threshold gives neural tissue/dorsal mesoderm (a simple three-threshold interpretation of the gradient, consistent with the French-flag logic of a morphogen).
| Injection | Predicted local BMP signaling | Predicted fate | Rationale |
|---|---|---|---|
| Uninjected ventral cell (control) | High (baseline BMP4 present, no antagonist) | Epidermis/blood | Default ventral state |
| + chordin mRNA (high dose) | Very low | Neural tissue, secondary axis | Chordin sequesters local BMP4 below threshold |
| + chordin mRNA + excess BMP4 mRNA | Restored to high/intermediate | Epidermis (rescue of ventral fate) | Excess ligand titrates out fixed antagonist dose — a stoichiometric rescue, the classic test for direct binding |
| + noggin mRNA + tolloid/BMP-1 mRNA | Intermediate-high | Partial rescue toward ventral, weaker axis | Tolloid protease degrades Chordin/antagonist complexes (works on Chordin more efficiently than Noggin, since Noggin is not a Tolloid substrate) — so co-injecting Tolloid rescues Chordin's effect far more than Noggin's |
This table encodes the actual historical logic of the "rescue" experiments: because Chordin's antagonism is stoichiometric (one Chordin binds and inactivates roughly one BMP dimer) and reversible by protease cleavage, titrating BMP4 or Tolloid dosage against a fixed Chordin dose predictably shifts the phenotype back toward ventral — direct evidence the two proteins act through binding rather than through separate, parallel pathways.
Example 2 — Reading a genetic mosaic experiment quantitatively. In Drosophila, embryos from mothers heterozygous for a null allele of sog (sog⁻/sog⁺) are crossed appropriately so that a quarter of embryos are maternally and zygotically sog⁻/sog⁻ (assume a simple recessive zygotic-effect scenario for this exercise, ignoring maternal rescue for simplicity). In wild-type embryos, the neuroectoderm (ventral, low-Dpp region) occupies roughly the ventral 40% of the D-V circumference, and the dorsal Dpp-dependent ectoderm (amnioserosa + dorsal epidermis) occupies the remaining 60%. In sog⁻/sog⁻ embryos, loss of the antagonist allows Dpp/Scw signaling to spread unopposed toward the ventral side. Suppose experimentally the dorsal Dpp-dependent domain now expands to occupy 85% of the circumference (a real order-of-magnitude consistent with published sog mutant phenotypes, which show a dramatic dorsalization and loss of most ventral neuroectoderm).
- Change in dorsal domain: 85% − 60% = 25 percentage points gained
- Corresponding loss in neuroectoderm: 40% − 15% = 25 percentage points lost
- Fractional loss of neuroectoderm: 25/40 = 62.5% reduction in the tissue that should give rise to the CNS
Because roughly two-thirds of the presumptive nervous system is lost when the only ventral antagonist is removed, this quantifies why sog mutants show severe reductions in CNS material and expanded dorsal cuticle structures — and it is the mirror image, in effect size, of injecting excess Chordin ventrally in frogs, which converts ventral epidermis-fated cells to neural tissue at a comparably large scale. Reasoning through the percentages this way makes concrete what "the antagonist sets the boundary position" means: move the antagonist source or dose, and the fate boundary moves by a predictable, roughly stoichiometric amount.
Common misconceptions & traps
| Misconception | Correction |
|---|---|
| "Chordin is the dorsalizing morphogen itself." | Chordin has no receptor of its own and does not signal into cells; it works purely by binding and neutralizing BMP. The morphogen being read out by cells is (free) BMP, not Chordin. |
| "Neural induction requires an instructive dorsal signal to switch ectoderm to neural fate." | The "default model" shows the opposite: ectoderm becomes neural when a repressive BMP signal is removed. Neural fate is the ectodermal default in the absence of BMP signaling; epidermis is the induced/imposed fate. |
| "Fly and vertebrate D-V axes are patterned by unrelated systems since their body plans look so different." | Sog/Chordin and Dpp-Scw/BMP4-BMP7 are direct molecular homologs performing the same binding chemistry; the anatomical mapping is inverted, but the molecular circuit is conserved — deep homology, not convergence. |
| "More BMP protein always means more BMP signaling." | Signaling depends on free (receptor-accessible) BMP, which depends on the local ratio of ligand to antagonist and on protease activity (Tolloid) that liberates BMP from antagonist complexes — total ligand concentration alone is not predictive. |
| "The organizer works through one master gene." | Organizer activity is built from multiple partially redundant antagonists (Chordin, Noggin, Follistatin) plus Wnt antagonists (Frzb, Dickkopf) acting on separate axes — no single gene knockout fully abolishes organizer function. |
The scientist behind it
Origins · 1 minEddy Marcel De Robertis and how this was discovered
Eddy De Robertis trained in Uruguay and Germany before joining Harvard and then UCLA, where his laboratory became the epicenter of the search for the organizer's molecular identity in the late 1980s and 1990s. De Robertis had earlier shown, with colleagues, that homeobox genes are conserved between flies and vertebrates — priming him to suspect that D-V patterning genes might also be conserved despite the inverted anatomy long noted by comparative anatomists. His lab's cloning of chordin, led by postdoctoral fellow Yoshiki Sasai, and the subsequent cross-species rescue experiments with Drosophila Sog, are widely credited as the decisive proof that vertebrate and invertebrate D-V axes share a single ancestral molecular logic. De Robertis has continued to work on BMP/Chordin biology, extending it to Wnt antagonists, cnidarian axis evolution, and planarian regeneration, treating the organizer as a general-purpose module reused throughout the animal kingdom.
Origins · 1 minhow this was discovered
Where this matters today
Regenerative medicine and stem-cell differentiation protocols. Recombinant Noggin is a standard reagent in human pluripotent stem cell culture, used to block BMP signaling and drive efficient neural differentiation — directly exploiting the "default model" discovered in frog embryos to manufacture neurons and neural progenitors for research and cell-therapy pipelines (e.g., dopaminergic neuron production for Parkinson's disease research).
Skeletal and connective tissue disease. BMP signaling drives bone formation, and Noggin mutations in humans cause skeletal disorders including proximal symphalangism and multiple synostoses syndrome, in which joints that should stay separate fuse into single bones — a direct clinical demonstration that antagonist dose sets the boundary between "make bone here" and "don't." This connects organizer biology directly to orthopedic genetics.
Cancer and fibrosis. Chordin-like and Noggin-like antagonists, along with BMP receptor signaling components, are dysregulated in several cancers and in pulmonary/renal fibrosis, motivating BMP-pathway-targeted small molecules and biologics currently in clinical development.
Agricultural and evolutionary developmental biology. Comparative studies of BMP/Chordin orthologs across cnidarians (e.g., Nematostella), echinoderms, and protostomes are used to reconstruct how the bilaterian body plan's D-V axis evolved, informing broader theories of animal body plan origins — a direct extension of the fly/frog unification into deeper phylogeny, and a natural bridge to how organizers establish Secondary Signaling Centers elsewhere in the embryo, such as the limb's apical ectodermal ridge and the vertebrate floor plate.
Diagnostic and teratology relevance. Because BMP antagonist balance is dose-sensitive, environmental or genetic perturbations that shift Chordin/Noggin/BMP stoichiometry are studied as potential contributors to neural tube defects, reinforcing why this pathway remains a live target for prenatal and reproductive toxicology research.
Ask the tutor: How does this same BMP-antagonist logic get reused to build secondary signaling centers, like the zone of polarizing activity in the limb bud?
Is Chordin the only molecule the organizer uses to pattern the D-V axis?
No. The organizer secretes multiple antagonists with partially overlapping function, including Chordin, Noggin, and Follistatin (all blocking BMP), plus separate antagonists of the Wnt pathway such as Frzb and Dickkopf, and antagonists of Nodal-related signaling like Cerberus. This redundancy explains why single-gene knockouts (losing just chordin, say) often produce milder phenotypes than removing the whole organizer, since the remaining antagonists partially compensate. The organizer is better thought of as a signaling hub broadcasting several distinct antagonist gradients simultaneously, each shaping a different axis or germ layer, rather than a single-molecule dorsalizing switch.
Why is BMP considered the "ventral" signal in vertebrates but the "dorsal" signal in flies — doesn't that contradict conservation?
It looks contradictory only if you map molecules onto anatomy before checking the anatomy itself. The molecular circuit — a BMP-family ligand opposed by a Chordin-family antagonist secreted from a discrete source — is conserved. What differs is which side of the body the antagonist source sits on: the vertebrate organizer is dorsal, so BMP signaling is high ventrally; the fly's Sog source is ventrolateral, so Dpp/Scw signaling is high dorsally. This maps onto the classical Geoffroy Saint-Hilaire inversion hypothesis: insects and vertebrates may derive from a common ancestor whose D-V axis became flipped in one lineage, while the molecular wiring stayed the same.
What role does the protease Tolloid/BMP-1 actually play, and why is it needed if antagonist binding already blocks BMP?
Simple binding gives a graded, diffusion-shaped antagonist field, but development often needs sharper boundaries or locally concentrated peaks of signaling (for instance, a narrow dorsal-most stripe of maximal Dpp signaling in the fly that specifies the amnioserosa). Tolloid is a zinc metalloprotease that specifically cleaves Chordin (and Sog) when it is in complex with BMP, releasing active, receptor-competent BMP at defined locations. This adds a second, spatially regulated control layer on top of diffusion and binding, letting the embryo sculpt sharper thresholds than passive diffusion alone would allow, and it also helps actively transport BMP toward regions of peak signaling (a "shuttling" mechanism).
How does this system relate to the broader concept of a morphogen gradient?
A morphogen gradient requires (1) a localized source, (2) a diffusible signal, and (3) concentration-dependent cell responses. BMP alone could satisfy this if its transcription were graded, but instead the embryo achieves a sharp, robust gradient by keeping BMP transcription broad and instead grading a diffusible antagonist from a point source (the organizer). This is a general strategy in developmental biology — using an antagonist gradient to convert a "blocky" or uniform ligand field into a smooth positional signal — and the same logic recurs in other systems, including limb bud patterning and the antagonist gradients that define Secondary Signaling Centers elsewhere in the embryo.
Does this mechanism operate identically in humans?
The core logic is conserved: human embryos use BMP4/BMP2 and Chordin/Noggin homologs to pattern the early D-V axis and neural plate, and Noggin is used routinely in human stem cell protocols specifically because it blocks BMP signaling in human cells the same way it does in frog cells. Direct experimental manipulation of live human embryos analogous to the amphibian organizer-graft experiments is not performed for ethical reasons, so human evidence relies on human genetic disease (Noggin mutations causing joint fusion disorders), cross-species conservation, and stem-cell/organoid models — all of which are consistent with, and did not overturn, the amphibian and fly mechanistic picture.
What's the difference between Chordin acting on ectoderm to make neural tissue and Chordin acting on mesoderm to make notochord/somite?
Both effects use the identical molecular mechanism — local suppression of BMP/Smad1/5/8 signaling — but the outcome differs because the two tissues have different "default" states and different target gene batteries available to respond. In ectoderm, low BMP unlocks a neural default program (proneural transcription factors). In dorsal mesoderm, low BMP allows expression of notochord and somite specification genes instead of the ventral mesoderm program (blood, mesenchyme) that high BMP would otherwise impose. The antagonist gradient is a shared upstream input; the tissue-specific response depends on the responding cells' pre-existing competence and transcriptional context, not on a different molecule.
Was Nieuwkoop's induction experiment superseded by the Chordin/Noggin discovery, or does it still stand?
It still stands and is complementary. Nieuwkoop's recombination experiments established that neuralization requires a signal from vegetal/dorsal tissue reaching the ectoderm — the phenomenon. Sasai, Harland, De Robertis, and colleagues later identified Chordin and Noggin as leading molecular candidates for that signal, and showed mechanistically that the "signal" works by removing an inhibitor (BMP) rather than delivering a positive instruction. So Nieuwkoop defined the logical requirement for an inducing interaction; the Chordin/Noggin work supplied the molecules and reversed the field's assumption about whether the interaction is instructive or permissive.