Neural Development and the Neural Tube

Neurulation is the morphogenetic process by which the flat neural plate, a thickened region of dorsal ectoderm induced by underlying mesoderm, transforms into a closed, fluid-filled cylinder — the neural tube. In amniotes this happens by primary neurulation (folding and fusion) along most of the body axis, and by secondary neurulation (cavitation of a solid cellular cord) in the posterior-most tail region. The neural tube is patterned along three axes, but the dorsoventral axis is where the clearest, most quantitatively understood morphogen logic operates.

Dorsoventral patterning refers to the process by which cells at different depths of the tube (dorsal = back, ventral = belly) acquire distinct identities: motor neurons ventrally, several classes of interneurons in the middle, and sensory-relay association neurons dorsally. This is achieved by two morphogens diffusing from opposite poles. Sonic hedgehog (Shh), a secreted signaling protein, is produced first by the notochord (an axial mesodermal rod beneath the tube) and then by the floor plate (a specialized ventral midline glial structure the notochord induces in the tube itself). BMPs (chiefly BMP4/BMP7) and Wnt proteins are produced by the roof plate at the dorsal midline, itself induced by overlying non-neural ectoderm. A cell's fate is set not by the absolute presence or absence of morphogen, but by the concentration and duration of exposure — the same logic Lewis Wolpert formalized as the "French flag model" of positional information. The key conceptual move is that a continuous, analog gradient is converted into discrete, sharply bordered domains of gene expression. This conversion is done by transcription factors that (1) are induced or repressed at different threshold concentrations of Shh, and (2) cross-repress each other's expression once turned on, so that any cell caught between two fates is forced into one or the other — there is no long-term intermediate state. This "gradient-to-switch" logic, worked out chiefly in the chick and mouse spinal cord by Thomas Jessell, Marysia Placzek, James Briscoe, and colleagues, is one of the clearest mechanistic demonstrations of positional information in any organ system.

Step 1 — Induction of the neural plate. During gastrulation, the dorsal mesoderm (organized by Spemann-Mangold Organizer-equivalent signals) secretes BMP antagonists — Noggin, Chordin, Follistatin — that block BMP signaling in the overlying ectoderm. Because default ectodermal fate under high BMP is epidermis, and low BMP permits neural fate, this local BMP inhibition is what specifies neuroectoderm as opposed to skin. The result is a spoon-shaped neural plate at the dorsal midline.

Step 2 — Neurulation: shaping the tube. The neural plate's lateral edges (neural folds) elevate, driven by a combination of intrinsic cell-shape changes (apical constriction, driven by actomyosin contraction at the apical surface of neuroepithelial cells, causing wedge-shaped cells) and extrinsic forces from adjacent non-neural ectoderm and underlying mesoderm. The folds converge dorsally and fuse in the midline. In mammals this fusion is not a single continuous zipper but initiates at multiple discrete points along the axis (in mouse, three principal closure sites) and proceeds bidirectionally from each, leaving transient unfused gaps (neuropores) at the very anterior and posterior ends that close last. Failure of posterior neuropore closure causes spina bifida; failure of anterior neuropore closure causes anencephaly. Once closed, the tube pinches free of the overlying surface ectoderm, and neural crest cells delaminate from the dorsal-most tip of the closing folds.

Step 3 — Establishing the two organizing centers. The notochord, an axial rod of mesoderm running the length of the embryo directly beneath the neural tube, contacts the ventral midline of the tube and induces the cells there to differentiate into the floor plate, a specialized wedge of non-neuronal glial-like cells. Critically, the floor plate itself then becomes a secondary, self-sustaining source of Shh, amplifying and relaying the ventralizing signal along the entire rostrocaudal length of the tube. Simultaneously, the dorsal-most tube cells, under influence of adjacent surface ectoderm, form the roof plate, which secretes BMP4, BMP7, GDF7, and Wnt1/Wnt3a.

Step 4 — Establishing the gradients. Shh protein is dually lipid-modified (cholesterol at the C-terminus, palmitate at the N-terminus), which restricts its diffusion and requires carrier mechanisms (lipoprotein particles, or transport via cytonemes/exosome-like particles) to move through tissue — this is why it forms a graded rather than uniform field, decaying in concentration from roughly nanomolar levels at the floor plate to nearly undetectable dorsally over a distance of some 200–300 μm in the chick spinal cord. BMPs diffuse similarly from the dorsal source, modulated by extracellular antagonists (Noggin) that sharpen the effective gradient. The two gradients are graded in opposite directions and overlap through the intermediate zone.

Step 5 — Cellular reception of Shh. Shh binds its receptor Patched1 (Ptch1), a twelve-pass transmembrane protein. In the absence of Shh, Ptch1 tonically inhibits a second transmembrane protein, Smoothened (Smo). When Shh binds Ptch1, this inhibition is relieved, Smo is stabilized and accumulates in the primary cilium (Hedgehog signal transduction in vertebrate cells occurs specifically within this microtubule-based organelle), and downstream signaling activates the Gli transcription factors (Gli1, Gli2, Gli3). At high Shh concentration, full-length Gli proteins act as transcriptional activators (GliA); at low or absent Shh, Gli3 is instead proteolytically processed into a truncated repressor form (Gli3R). The ratio of GliA:GliR across the dorsoventral axis is the actual biochemical readout cells use to sense their position — not Shh concentration per se, but the balance of activator and repressor Gli it produces.

Step 6 — The combinatorial transcription factor code. Progenitor cells express a battery of homeodomain and bHLH transcription factors falling into two functional classes:

• Class I proteins (e.g., Pax7, Dbx1, Irx3, Pax6) are repressed by Shh/GliA signaling; they are expressed by default and are switched off only where Shh is high enough.
• Class II proteins (e.g., Nkx6.1, Nkx2.2, Olig2) are induced by Shh/GliA; they require a threshold level of signaling to turn on.

Each factor has a specific concentration threshold at which it is induced or repressed. Critically, cognate pairs of Class I and Class II proteins that share a ventral boundary cross-repress one another at the protein/transcriptional level (e.g., Pax6 and Nkx2.2 repress each other's transcription; Irx3 and Olig2 repress each other). This mutual repression is a bistable switch: a cell cannot stably co-express both factors, so it commits sharply to one side of the boundary even though the underlying morphogen concentration changes continuously. This is the molecular explanation for how five or six sharply bordered progenitor domains emerge from one smooth gradient.

Step 7 — From progenitor domain to neuron. Along the ventral half of the tube, from floor plate to roof plate, progenitor domains are named p3, pMN, p2, p1, p0 (ventral to dorsal), each defined by a unique transcription factor combination, and each generating a distinct neuronal class:

Motor neurons (from pMN, marked by Olig2) are a special case: they arise at intermediate-high Shh and additionally require the bHLH factor Olig2, which — unusually — is induced by an intermediate level of Shh signaling but is flanked by both a higher-Shh domain (p3, Nkx2.2+) and lower-Shh domains (p2), so pMN itself is a "sandwiched" concentration-specific stripe, a clean illustration that it is a specific window of morphogen concentration, not simply "more is different," that specifies certain fates. Dorsally, BMP/Wnt signaling from the roof plate similarly establishes six or more dorsal interneuron classes (dI1–dI6) via a parallel combinatorial code (Math1/Atoh1, Ngn1, Mash1/Ascl1, and downstream factors like Lbx1), including the dI1–dI3 populations that become part of ascending sensory relay circuits.

Nicole Le Douarin & tissue grafting (1980s) — the notochord induces the floor plate. Using quail-chick chimeras (a technique in which donor and host cells are distinguishable by a nucleolar marker), researchers grafted an extra notochord adjacent to the lateral neural tube. An ectopic floor plate formed at the graft site, and motor neurons appeared nearby — showing that the notochord instructs adjacent neuroepithelium to become floor plate, and that floor plate in turn organizes ventral neuron identity.

Yamada, Placzek, Jessell et al. (late 1980s–1990s) — the floor plate itself is inductive and diffusible. Explant experiments showed that floor plate tissue, cultured next to intermediate neural plate explants without direct contact (separated by a filter), still induced motor neuron differentiation, demonstrating a diffusible ventralizing signal rather than a contact-dependent one. This was the first strong evidence for a soluble morphogen rather than juxtacrine signaling.

Echelard, Roelink, Chiang et al. (1993–1996) — identification of Shh as the diffusible signal. Cloning of the vertebrate hedgehog gene family identified Sonic hedgehog as expressed exactly where predicted: notochord, then floor plate. Roelink and colleagues (1995) showed that COS cells transfected with Shh, when grafted next to naive neural plate, were sufficient to induce floor plate and motor neurons — recreating the notochord's inductive effect with a single recombinant protein. This was the classic sufficiency experiment.

Dose-response bead implantation (Ericson, Briscoe, Jessell, and colleagues, mid-late 1990s). This is the single most decisive experiment for establishing Shh as a graded morphogen rather than a simple on/off switch. Beads soaked in different concentrations of recombinant Shh protein were implanted at different dorsoventral positions in neural plate explants. The result: low concentrations of Shh induced only the most dorsal ventral cell type (e.g., V0/V1 interneuron markers); intermediate concentrations induced motor neuron markers (Olig2); and the highest concentrations induced the most ventral fates (Nkx2.2, V3 interneurons, even floor plate-like cells) — reproducing the entire dorsoventral sequence of fates from a single signal simply by titrating its concentration. Critically, the same cells could be pushed to different fates depending purely on dose, proving concentration itself (not just presence/absence) carries positional information — the direct experimental proof of the French flag model for this system.

Loss-of-function genetics — Shh knockout mice (Chiang et al., 1996). Mice lacking a functional Shh gene fail to form a floor plate or motor neurons; the ventral neural tube instead adopts dorsal-like character throughout, and the mice display severe forebrain defects, including cyclopia (fusion of the eye fields, a phenotype long known from a distinct source: naturally occurring cyclopamine-induced cyclopia in livestock).

Cyclopamine — the pharmacological proof, discovered via an agricultural mystery. In the 1950s–60s, sheep grazing on corn lilies (Veratrum californicum) in Idaho gave birth to lambs with a single central eye (cyclopia) and other severe midline brain defects. The teratogenic compound was isolated and named cyclopamine. Decades later it was shown that cyclopamine directly binds and inhibits Smoothened, blocking all Hedgehog pathway signaling. Applying cyclopamine to developing embryos phenocopies Shh loss-of-function — ventral neural fates are lost, and eye fields fail to separate — closing the causal loop from gene to protein to receptor to phenotype, and incidentally explaining a real veterinary teratogenic disaster.

Gain-of-function — ectopic Shh expression. Conversely, retroviral misexpression of Shh in the dorsal neural tube, or genetic mutation causing constitutive Hedgehog pathway activation (e.g., loss of Ptch1, which normally represses Smo), causes ectopic ventral cell types (motor neurons, floor-plate-like cells) to appear dorsally — the mirror-image gain-of-function proof required alongside loss-of-function to establish both necessity and sufficiency.

Cross-repression circuitry (Briscoe et al., 2000; Sander et al., 2000, and later work). Misexpressing individual transcription factors (e.g., forcing Nkx2.2 expression outside its normal domain) was shown to repress Pax6 in a cell-autonomous manner and vice versa, directly demonstrating the mutual repression circuit that sharpens gradient-derived boundaries into discrete, bistable domains — turning a continuous signal into discrete cell types, exactly as predicted by mathematical models of the system. Together, these experiments satisfy all the classical criteria for establishing a morphogen: correct spatial expression, sufficiency (ectopic source recreates pattern), necessity (loss abolishes pattern), dose-dependence (different concentrations give different fates), and a mechanistic transduction pathway linking receptor occupancy to nuclear transcriptional output.

Example 1 — Predicting the phenotype of a partial Shh pathway mutation. Suppose a mouse embryo carries a hypomorphic (partial loss-of-function) allele of Smoothened that reduces maximal pathway activation to roughly 30% of wild-type levels, without eliminating signaling entirely. Reasoning: recall the dose-response bead data — floor plate and V3 interneurons (Nkx2.2+) require the highest Shh/Smo signaling levels; motor neurons (Olig2+) require an intermediate level; V0–V2 interneurons require progressively lower levels. If maximal signaling is capped at 30% of normal, the very highest-threshold fates (floor plate, V3) will never be reached anywhere in the tube — they require signaling levels the mutant cannot generate even immediately next to the notochord. Predicted phenotype: loss of floor plate and V3 interneurons (these domains are replaced by pMN- or p2-like character, since without high enough signal, cells cannot activate Nkx2.2, but many will still cross the lower threshold to activate Olig2); a ventrally-shifted but not eliminated domain of motor neurons, now positioned immediately adjacent to the notochord instead of separated from it by a V3/floor plate zone; dorsal domains largely unaffected, since they depend little on Shh. This exact "graded reduction of ventral-most fates first" logic has been confirmed in mouse and zebrafish hypomorphic Hedgehog pathway alleles and is the standard way to reason about partial pathway lesions in exam and research contexts.

Example 2 — A neural tube defect epidemiology calculation. Neural tube defects (NTDs, e.g., spina bifida, anencephaly) arise from failure of neural tube closure and are known to be reduced by maternal folic acid supplementation. Suppose in a population the baseline NTD incidence is 1.5 per 1,000 live births, and a randomized trial (modeled loosely on the real 1991 UK Medical Research Council Vitamin Study) shows folic acid supplementation before and during early pregnancy reduces recurrence risk by approximately 70%. Calculation: in a cohort of 200,000 births with no supplementation, expected NTD cases = 200,000 × (1.5/1000) = 300 cases. With supplementation reducing risk by 70%, new incidence = 1.5 × (1 − 0.70) = 0.45 per 1,000. Expected cases with supplementation = 200,000 × (0.45/1000) = 90 cases. Cases prevented = 300 − 90 = 210 cases per 200,000 births. Mechanistic link: folate is required for one-carbon metabolism, methylation reactions, and nucleotide synthesis, all of which support the very high proliferation rate of neuroepithelial cells during the third and fourth weeks of human gestation, precisely when neural fold elevation and fusion occur. This is why national folate fortification programs (mandatory in the US since 1998) target the periconceptional period, and why the effect is a population-level, quantifiable reduction in a specific structural birth defect rather than a general health claim — a clean example of translating basic neurulation mechanism directly into public health policy.

Much of this field's decisive work came from Thomas M. Jessell, working with colleagues including Marysia Placzek, Jane Dodd, Johan Ericson, and James Briscoe at Columbia University, alongside parallel contributions from Andrew McMahon, Philip Beachy, Cliff Tabin, and Matthew Scott, who cloned and characterized the vertebrate hedgehog genes. Jessell's laboratory combined chick neural tube explant culture, quail-chick chimera fate mapping, recombinant protein bead assays, and mouse genetics into a single coherent research program spanning nearly two decades, moving the field from "the notochord induces something" to a quantitative, receptor-to-transcription-factor mechanistic model with predictive power — a rare case in developmental biology where a full molecular pathway, from morphogen gradient to individual neuron subtype, was worked out end to end.

Regenerative medicine and stem-cell-derived neurons. Protocols to generate specific neuron types from human induced pluripotent stem cells (iPSCs) directly apply this pathway: treating stem cell cultures with defined concentrations and durations of Shh agonists (e.g., purmorphamine, smoothened agonist SAG) alongside retinoic acid or BMP inhibitors reproducibly generates motor neurons or specific interneuron subtypes in vitro, used to model amyotrophic lateral sclerosis (ALS) and spinal muscular atrophy, and to screen candidate drugs on patient-derived neurons.

Cancer therapeutics. Because uncontrolled Hedgehog pathway activation (often from loss-of-function mutations in PTCH1 or activating mutations in SMO) drives basal cell carcinoma and a subset of medulloblastoma, drugs derived directly from the cyclopamine story — vismodegib and sonidegib, both Smoothened inhibitors — are FDA-approved cancer therapies, a direct translational line from a sheep-teratogen curiosity to oncology clinics.

Public health and folate fortification. Understanding that neural tube closure depends on rapid neuroepithelial proliferation supported by one-carbon metabolism directly motivated mandatory folic acid fortification of grain products in many countries, measurably reducing spina bifida and anencephaly incidence at the population level.

Teratology and environmental/veterinary toxicology. The cyclopamine story remains a template for how agricultural/environmental teratogens are investigated: a population-level birth defect cluster, traced to a specific plant compound, traced to a specific molecular target (Smoothened), informs both livestock management and our understanding of a core developmental pathway.

Foreshadowing. The same dorsal neural tube domain that generates roof-plate BMP/Wnt signals also gives rise to neural crest cells, migratory cells built by an overlapping but distinct gene regulatory network — the subject of The Neural Crest and Its Derivatives. And a local evagination of the anterior neural tube, patterned by an analogous but distinct combination of signaling centers, becomes the optic vesicle, explored in The Vertebrate Eye as a Developmental Model.

Ask the tutor: How does the same Shh/BMP gradient logic get reused to pattern the vertebrate limb bud, and what changes between the two contexts?