The Nervous System · Part One · How It Is Built and Fueled
Lesson 19 / 61
Cortical Organization: Layers, Columns, and Migration Disorders
Where a cortical neuron is born decides which layer it will live in.
The cerebral cortex is assembled in a fixed order, and the order it was built in sets what it can do afterward. Neurons leave the wall of the embryonic ventricle, climb radial glial fibers, and stop just under the surface where Reelin tells them to stop. When that sequence fails the surface comes out smooth and the layers come out misbanded, and seizures and developmental delay follow. The Unified Model of Tone reads the finished arrangement as the constraint that decides which replies a patch of cortex can make.
Radial migration
Four phases in rat, one runs backward
Layer identity
TBR1 picks layer VI over V in mouse
Human amplifier
Outer SVZ radial glia, ARHGAP11B
Lissencephaly
Microtubules fail, the climb stops
Cortical organization
Six laminae stacked through the thickness of the cortical gray matter, crossed by radial columns that run from the white matter to the surface. Layer I is the molecular layer, settled first by Cajal-Retzius cells. Layers II and III carry traffic between cortical regions. Layer IV takes thalamic afferents. Layer V projects to the brainstem and the spinal cord, and layer VI addresses the thalamus.
The cortical column and tone
A vertical strip of cells that answers as one unit, from the white matter to the pial surface. In the Unified Model of Tone that strip is a constraint structure. Its laminar wiring sets the range of replies available to it, and a column that loses the arrangement loses the range, whatever its neuron count.
01The ventricular progenitor pool
The ventricular zone runs as a clock, and each round of division stamps a neuron with an identity
Every neuron in the cerebral cortex traces back to the ependymal layer of the embryonic ventricle. Rakic proposed in 1988 that this layer is built of proliferative units forming a proto-map of the prospective cytoarchitectonic areas Rakic 1988. Glial guides translate the output of each unit outward as an ontogenetic column. The number of columns an area ends up with stays open to revision, because afferent input can modify it.
The cell doing both jobs is the radial glial cell. Retroviral labeling in rat cortex produced clones made of mitotic radial glia and postmitotic neurons, and the neurons climbed radial glia they were clonally related to Noctor 2001. Time-lapse images caught proliferative radial glia generating neurons directly. The scaffold and the parent are the same cell, so the radial axis of the cortex carries a lineage as well as a direction.
Division changes character as the schedule advances. Time-lapse imaging of clonal cells in rat cortex split the source in two. Neurons arise directly from radial glia in the ventricular zone and indirectly from intermediate progenitors in the subventricular zone Noctor 2004. Symmetric divisions expand the progenitor pool early, and asymmetric divisions then yield one self-renewing stem cell and one basal progenitor that goes on to make neurons.
Migration runs in four phases, not one climb
Migrating neurons pause, and one phase carries them backward toward the ventricle before the final ascent. Time-lapse imaging in rat cortex resolved four distinct phases on the way to the cortical plate, that retrograde movement among them Noctor 2004. Each phase is a separate point at which the sequence can fail, which is what makes the migration disorders a family of conditions sharing one address.
The transcription factors expressed at the moment of birth stamp the daughter neuron with an identity it carries for life. TBR1 promotes the layer VI corticothalamic identity and suppresses the layer V subcerebral one by reducing Fezf2 and CTIP2 McKenna 2011. In Tbr1 null mice, neurons birthdated at embryonic day 11.5 expressed subcerebral markers and extended axons into subcerebral targets. One transcription factor decides whether a neuron will address the thalamus or the spinal cord.
The pool is finite, and its potential falls as the schedule runs, so the window for generating any given layer closes. Counted at the end, human cortex is an ordinary primate cortex by neuron number. The expansion was spent on surface area and connection instead, and the census behind those counts sits with the glia and the cell populations of the brain. In this model the ventricular schedule is where tone acquires an anatomy, since birth date is the input and laminar address is the record it leaves.
02Findings
What the research shows
03Inside-out lamination
Birth date decides laminar address, and Reelin released from layer I is what stops the climb
A neuron born early in the schedule sits in layer VI, and one born days later climbs past it into layer II. Cortical neurons settle by birth date, deepest first, with later-born cohorts migrating radially past earlier-born neurons into more superficial layers Rymar 2007. Because the timing of birth maps onto the final laminar address, a disturbance of birth order scrambles the wiring logic of a whole column.
Birth date is one input and cell phenotype is another. In rodent cortex, parvalbumin interneurons follow the inside-out gradient and are born isochronously with the projection neurons of the same layer, while calretinin interneurons populate the cortex on the opposite outside-in gradient Rymar 2007. Inhibitory cells are made in the proliferative zones of the ventral telencephalon, and their final depth depends on which subclass they belong to.
Layer I is settled first, by Cajal-Retzius cells, and those cells tell everyone else where to stop. Reelin is an extracellular matrix protein they synthesize in the marginal zone Frotscher 2017. It orients the leading process of the migrating pyramidal neuron toward the cortical surface, and it works through phosphorylation of cofilin to stabilize the actin cytoskeleton. Late-generated neurons in the reeler mouse never reach the upper layers.
Reelin acts at two separable steps Sekine 2014. One is the polarity switch when a neuron leaves multipolar migration for locomotion. The other is the terminal translocation of the cell body just beneath the marginal zone. The adhesion machinery is named: integrin alpha5beta1, N-cadherin and the nectin and afadin system. Migration is then terminated by degradation of Dab1 through SOCS7, Cullin5 and Rbx2.
Removing the cells proves what they do. Ablating Cajal-Retzius cells in organotypic hippocampal slice cultures prevented the ingrowth of entorhinal afferents while commissural afferents arrived normally Del Rio 1997. Inhibiting Reelin and examining reeler mutant mice reproduced the abnormality. The ablation took out one afferent system and left the other untouched, so layer I residents hand out addressing instructions that tell particular incoming fibers where to terminate.
Six layers is a mammalian variation, and it is not uniform
Cortical thickness climbs across sensory processing hierarchies and falls the other way through motor and frontal cortex, driven mainly by layers III, V and VI. BigBrain, a 3D histological atlas of a single human brain at 20 micrometer isotropic resolution, was segmented into cortical layers in both hemispheres by a convolutional neural network to measure that gradient Wagstyl 2020. The six-layered neocortex is itself a mammalian variation on a shared pallial program, produced by changes in the germinative zones and in radial and tangential migration Garcia-Moreno 2020.
Depth is an address. An afferent arriving in layer IV meets a different circuit from one arriving in layer I. In this model a thalamic volley and a modulatory volley count as two different inputs because they land at two different depths, and lamination is what keeps that difference legible.
04The cortical column
One whisker addresses one cortical column through the full thickness of the cortex
The cortex is organized in vertical columns as well as horizontal layers, and each column answers as one unit. High-resolution 2-deoxyglucose mapping in freely behaving mice showed that stimulating the single C3 whisker activated approximately 30 to 40 percent of the neurons in the C3 barrel column McCasland 1988. That is about twice the number labeled in the same column when all whiskers were stimulated together, and about ten times the number when the large whiskers were clipped.
The label ran the full thickness of the mouse cortex. Patches appeared above and below the layer IV barrels and matched cytoarchitectonic barrel boundaries more exactly than earlier work had shown McCasland 1988. Inside a single barrel, smaller densely labeled patches appeared that were unique to each cortex, so a column contains finer vertical units of its own.
The radial axis carries developmental history and the tangential axis carries geography. Radial glia deliver each clone to one strip of cortex, and those strips tile side by side into a somatotopic map. That is why a barrel field reads as an orderly image of the whisker pad. Damage that costs an area its columns costs the body part they served its territory on that map.
The outer subventricular zone is the human amplifier
Outer subventricular zone radial glia release the ventricular surface, keep a long basal process, and divide again far from the wall. In developing human neocortex the subventricular zone carries a massively expanded outer region built from those cells Hansen 2010. Real-time imaging and clonal analysis showed them dividing to self-renew and to generate neuronal progenitors that proliferate further. Amplification happens away from the ventricle.
The amplifier is sufficient on its own. Expressing the human-specific gene ARHGAP11B under its own human promoter in the fetal neocortex of the common marmoset increased basal radial glia in the outer subventricular zone Heide 2020. Upper-layer neurons increased, the neocortex enlarged, and it folded. One human gene placed into a primate produced a larger and gyrified cortex.
The column is where tone takes a shape you can measure. A column that commits a large fraction of its cells to one whisker holds a fixed range of possible replies. This model reads that range as set by how many columns were built and by how each one is wired from layer I to layer VI.
05Migration disorders
Each stage of cortical construction has its own failure, and the disorders name the stage
Cortical malformation begins after the tube is closed. The closure failure that precedes any cortex, its timing and the folate evidence that prevents most of it, is covered on Embryology and development. What this section covers starts later, when the neurons are made on schedule and cannot reach the layer they belong in.
Lissencephaly is a migration failure with named genes. Three forms with known genes cover most lissencephaly and are separable on imaging Guerrini 2002. LIS1 on chromosome 17 produces a posteriorly predominant form. XLIS and DCX produce an anteriorly predominant form in hemizygous males and subcortical band heterotopia in heterozygous females. Mutations in the reelin gene produce autosomal recessive lissencephaly with cerebellar hypoplasia, which is the human proof that the Cajal-Retzius signal is load-bearing.
The failure is cytoskeletal. LIS1 and DCX between them explain about 85 percent of classic lissencephaly, while TUBA1A accounts for about 1 percent of classic cases and about 30 percent of lissencephaly with cerebellar hypoplasia Kumar 2010. TUBA1A encodes a structural subunit of microtubules, and the disease mutations disrupt binding sites for microtubule-associated proteins. The neurons were made on schedule. The machinery that carries them upward is what failed.
Guerrini classified these as epileptogenic brain malformations, and that is the clinical shape of lost architecture. A cortex whose layers cannot constrain its own activity keeps collapsing into one state, and that state is the seizure. Developmental delay follows from the same loss, because a column that cannot sort its inputs by depth has a shorter list of replies to develop.
Some build faults stay quiet for decades. Tethered cord syndrome is abnormal tension on the spinal cord, and its findings run from a low-lying conus medullaris to a fatty filum terminale and split cord malformation. A fault laid down in the first weeks of gestation can surface in adult life as pain, weakness or bladder change. That is why scoliosis treats it as a red flag to rule out first.
06The adult cortical map
Holding the adult map costs continuous afferent traffic, and the cortex gives up tissue when the traffic stops
An adult cortical map is not a finished object. Rats had the A, B and C row whiskers trimmed from postnatal day 7 to 15. Layer 2/3 neurons in the spared D2 column then sent fewer supragranular axons into the deprived C row columns, while their infragranular projections were unchanged Broser 2008. Deprivation withdrew one projection at one depth and left the rest intact.
Adult reorganization has an anatomical ceiling, and enough deafferentation breaks it. In adult macaques, limited sensory deafferentation shifted cortical maps 1 to 2 millimeters mediolaterally, the dimension along which body parts are represented Pons 1991. That distance matches the mediolateral spread of the projection zones of individual thalamocortical axons. Extensive long-term deafferentation exceeded the ceiling by an order of magnitude.
A spinal cord injury registers in the human cortex within months. Thirteen patients were imaged over 12 months after acute spinal cord injury and compared with 18 controls Freund 2013. Spinal cord cross-sectional area declined at 0.46 square millimeters per month while the area in controls held steady. White matter volume fell faster in the cranial corticospinal tracts at the internal capsule and in the right cerebral peduncle. Gray matter fell in the left primary motor cortex at a Z score of 4.23 and p equal to 0.041. Better recovery on the SCIM and ISNCSCI scores tracked with less atrophy.
Rerouted input builds the modules the new input needs
Retinal projections rerouted into the auditory pathway of the ferret drive neurons in primary auditory cortex, and those neurons organize into orientation modules Sharma 2000. Orientation tuning inside those modules is comparable to tuning in primary visual cortex. The rewired map is less orderly than the visual one, and so are its patchy horizontal connections. In this model that residual disorder is the arrangement the auditory cortex was already holding when the new input arrived.
Input writes hardest early, and the window has an edge. Three months of monocular closure from birth in the kitten drops the share of cortical cells drivable by the deprived eye from 85 percent to about 7 percent Hubel 1970. Susceptibility begins near the start of the fourth week and disappears around the end of the third month. Three to four days of closure in the fourth and fifth weeks moves the share sharply, while a year of the same deprivation in an adult cat does nothing detectable.
The Unified Model of Tone reads those results as one claim. Segmental input holds the cortical map in place, and the map is edited at the level of named layers. The model predicts that a sustained change in segmental afferent traffic will register in cortical structure across months. It predicts that supragranular projections move before infragranular ones, and that the change follows the segment whose input changed. Freund's patients had complete acute cord lesions, so the human evidence sits at the severe end of that prediction and the ordinary end is still open.
The traffic in question has an address. A cortical column is fed by one peripheral territory through the thalamus, and that territory reports through the joints, muscles and skin of one spinal segment. Change what a segment sends and the column above it is the structure that pays. Losing that traffic costs the cortex measurable tissue inside a year.
07Build order sets function
Cortical function inherits the order in which the cortex was assembled
Neurons reach their layers in a fixed sequence, so the finished six-layer plan is the construction schedule made visible. Layer V carries cortical command to the brainstem and the spinal cord while layer VI addresses the thalamus. Which of the two a neuron becomes was settled by when it was born and by the factors expressed at the time McKenna 2011. Cajal-Retzius-seeded layer I coordinates the surface, layer IV receives, and layers II and III associate across regions.
A reeler cortex carries every neuron it should have. The order of arrival is what failed. Late-born cells that never reach the upper layers leave the laminar plan misread Frotscher 2017, and the variable that changed is arrangement while cell count held. Tone is that arrangement, which makes the reeler cortex the cleanest anatomical case the nervous system offers for it.
The laminated sheet turns raw thalamic input into perception and planning, and the same sheet sustains the adult capacity to reorganize. One column integrates afferent drive, internal state and descending control into a single output. The central integrative state of one neuron becomes, at the scale of the column, the state the whole vertical unit is holding.
The model stakes something checkable on that identification. It holds that the territory a body part keeps on the cortical map, the recruitment a column can muster for one input, and the autonomic output of the segment feeding that column move together. Deafferent a limb and the model expects all three to contract; restore the input and it expects all three to widen again, each on the schedule its own reserve permits. The tone research page carries the general commitment and the instruments that measure it.
The cost does not stop when the build does. A cortex loses measurable tissue inside a year of losing its input, and a column with less territory has a shorter list of replies to draw on. A failed climb in utero and a segment that stops reporting decades later read the same way here, as a loss of arrangement rather than of census.
A reeler cortex carries every neuron it should have. The order of arrival is what failed.
08Tone
How this system expresses tone
A column answers with the layers it has. Stimulating one C3 whisker recruits 30 to 40 percent of the neurons in that barrel column in the freely behaving mouse, through the full thickness of the cortex.
Constraint
Limited deafferentation moves an adult macaque map 1 to 2 millimeters, the spread of one thalamocortical axon. Extensive deafferentation clears that ceiling tenfold.
Input quality
Afferent traffic keeps paying to hold the map. Motor cortex gray matter fell at a Z score of 4.23 within 12 months of a human cord lesion.
Set point
Each cortical area defends a thickness. BigBrain, sampled at 20 micrometers across one human brain, shows thickness climbing through layers III, V and VI across sensory hierarchies.
Coupling: a single column ties a layer VI thalamic address to a layer V brainstem address inside one vertical unit. Prediction: the factors present at a neuron's birth commit it to layer VI or layer V, so Tbr1 null mouse neurons born at embryonic day 11.5 take the subcerebral fate. Gain: one progenitor's asymmetric divisions multiply into a whole column, and the outer subventricular zone raises that multiplier in the human fetus. Time course: the deprived eye's share of drivable cortical cells falls to about 7 percent in the kitten, and a year of the same closure in an adult cat changes nothing. Oscillation: spontaneous rhythmic activity crosses the cortical plate long before the eyes open, and lamination proceeds under that carrier. Load: the climb itself is the expense, since a neuron that pauses, reverses and then locomotes to the surface spends days of cytoskeletal work to buy one laminar address.
09Across the library
How this page relates to the rest of the library
The tube itself. How the neural plate folds and closes by the end of week three, the failure that produces spina bifida and anencephaly, and the folate evidence that prevents most of it.
What radial glia become once migration ends, and the cell census that settles how many neurons and how many non-neuronal cells a human brain actually holds.
The afferent stream that holds the cortical map, read at the receptor: spindle, Golgi tendon organ and joint afferent, and what each of them reports.
What an adult cortex can still change once lamination is finished, and how a rehabilitation program works against a map that reorganizes over 1 to 2 millimeters.
What the instruments read from a cortex already built. What an EEG electrode sums, what an fMRI voxel infers, and how the brain's energy budget divides between signaling and housekeeping.
What changes in the maternal nervous system across gestation, and why the exposures that matter most to a developing cortex land in the earliest weeks.
10Frequently asked
Questions about this topic
How is the cerebral cortex built from a single progenitor pool?
Every cortical neuron descends from proliferative units in the ependymal layer of the embryonic ventricle, and glial guides carry the output of each unit outward as an ontogenetic column. The radial glial cell is both the parent and the scaffold: retrovirally labeled clones in rat cortex contained mitotic radial glia together with postmitotic neurons climbing fibers they were related to. Symmetric divisions expand the pool early. Asymmetric divisions then yield one self-renewing stem cell and one basal progenitor that goes on to make neurons in the subventricular zone.
What does inside-out lamination mean?
Cortical neurons settle by birth date. The earliest-born cells occupy the deepest layers, and each later cohort migrates past its predecessors into more superficial layers, so layer VI exists before layer II. The gradient is not universal. In rodent cortex, parvalbumin interneurons follow the inside-out order and are born isochronously with the projection neurons beside them, while calretinin interneurons populate the cortex on the opposite outside-in gradient. Birth date sets laminar address for most cells, and phenotype is the second determinant of final depth.
What do Cajal-Retzius cells do?
Cajal-Retzius cells settle layer I first and secrete Reelin, an extracellular matrix protein that orients the leading process of each migrating pyramidal neuron toward the cortical surface. Reelin works through phosphorylation of cofilin to stabilize the actin cytoskeleton, and late-generated neurons in reeler mice never reach the upper layers. Ablating these cells in organotypic hippocampal slice culture stopped entorhinal afferents from growing in while commissural afferents arrived normally, so layer I hands out addressing instructions to particular incoming fiber systems.
What is a cortical column?
A cortical column is a radial array of neurons spanning all six layers that answers as one functional unit. In freely behaving mice, stimulating a single C3 whisker activated roughly 30 to 40 percent of the neurons in the C3 barrel column, about twice the number labeled when every whisker was stimulated together. The labeled patches ran above and below the layer IV barrels through the full cortical thickness, and smaller patches inside each barrel mark the minicolumns nested within the column.
Why is the human cortex so large?
Human cortical expansion comes from an extra progenitor zone. Developing human neocortex carries a massively expanded outer subventricular zone whose radial glia release the ventricular surface, keep a long basal process, and divide again away from the wall. That amplifier is sufficient on its own. Expressing the human-specific gene ARHGAP11B in the fetal neocortex of the common marmoset increased basal radial glia. Upper-layer neurons increased, the neocortex enlarged, and it folded. Surface area rather than neuron number is what the expansion bought.
What causes lissencephaly, or smooth brain?
Lissencephaly is a failure of neuronal migration. The neurons are made on schedule and cannot climb to the layers they belong in, so the cortical surface stays smooth. LIS1 on chromosome 17 gives a posteriorly predominant pattern. XLIS and DCX give an anteriorly predominant one in hemizygous males, and subcortical band heterotopia in heterozygous females. Reelin mutations give autosomal recessive lissencephaly with cerebellar hypoplasia. LIS1 and DCX between them explain about 85 percent of classic lissencephaly, and TUBA1A, a microtubule subunit, explains roughly 1 percent.
Does an adult cortex still need afferent input to hold its map?
The adult map needs traffic to keep its shape. In adult macaques, limited sensory deafferentation moved somatosensory maps 1 to 2 millimeters mediolaterally, matching the spread of individual thalamocortical axon projection zones, and extensive long-term deafferentation moved them ten times further. In humans, spinal cord cross-sectional area fell 0.46 square millimeters per month across the year after an acute cord injury, and gray matter shrank in the left primary motor cortex over the same window. Better neurological recovery tracked with less atrophy.
Can a brain have the right number of neurons and still not work?
A cortex can carry every neuron it should have and still fail, because function follows arrangement rather than census. Late-born cells in the reeler mouse never reach the upper layers, so the laminar plan is misread while the cell count holds. Depth is an address, and a thalamic volley arriving in layer IV meets a different circuit from one arriving in layer I. Afferent traffic then keeps paying to hold that arrangement, since motor cortex gray matter shrinks within twelve months of a human cord lesion.
11The sources
References
Sources: primary literature, linked inline.