Induction and Patterning in the Telencephalon

Abstract

Sitting at the front of the central nervous system (CNS), the telencephalon gives rise to many brain regions including the cerebral cortex – the seat of cognition, consciousness, and mind that uniquely defines us as human beings. For this reason, the telencephalon has long fascinated neuroscientists and the public alike. Given its special and complex functions, as well as some developmental oddities, the telencephalon was thought to play by its own developmental rules. On the other hand, organisms evolve by reusing and modifying preexisting developmental mechanisms rather than creating new ones from scratch. This led to a conundrum – does the developing telencephalon use shared rules, unique ones, or both? As it turns out, both shared and unique rules apply to the developing telencephalon. In this chapter, we go over these rules, with special attention paid to the developing cerebral cortex in an evolutionary context. Collectively, the rules of telencephalic development have allowed the cortex to grow in size and complexity, thereby expanding the cognitive capacity and abilities of this most remarkable of organs over evolutionary time.

Brief History of the Field

The foundations for understanding development began at the turn of the twentieth century with the concept of a morphogen – quite literally, a “form-generating substance.” Such substances were first envisioned for animal regeneration by Morgan (1901), then for normal development by Boveri (1901). Morphogen production sites called organizers were described by Spemann and Mangold (1924), after which the term “morphogen” was coined by Turing (1952). These seminal findings were combined by Wolpert into the famous “French Flag” model (1969), which formalized the idea that naïve tissues possess concentration gradients of morphogens produced by organizers (Fig. 1). Such gradients expose different parts of a tissue to different morphogen concentrations, which provides tissues their initial “pattern.” Garcia-Bellido (1975) then described how tissues translate the morphogens outside of cells into transcription factor codes inside them to specify cell types and adhesive properties. This culminates in cells of the same type aggregating within compartments or segments separated by sharp borders, just like the stripes of the French Flag.

Fig. 1

The French flag model. Schematic of early tissue induction and patterning by a morphogen. Morphogen (green) secreted from a localized source (organizer) forms a concentration gradient within a tissue. At high concentrations (above threshold 2), cells adopt “blue” fate. At intermediate concentrations (above threshold 1), responding cells adopt “white” fate. At very low levels of morphogen, cells maintain their default “red” fate. (From Grove and Monuki (2020))

The French Flag model applies to all developing organs, but controversial for the cerebral cortex. According to the model, segments (anatomical or molecular) should exist in the progenitor fields for every organ, and such segments exist elsewhere in the CNS, including other parts of the telencephalon as defined in the “prosomeric” model (Rubenstein et al. 1994). In contrast, cortical progenitor fields display no segments, only gradients. Only much later – when the cortex transforms into discrete functional areas with unique cellular architectures, as defined by Brodmann (1909) – does the cortex take on a segmented form. Given its evolutionary exceptionalism, could the cerebral cortex be an exception to the French Flag rule?

This same question arose a second time. The French Flag model implies that each tissue possesses its own information for patterning. In other words, patterning is “intrinsic” (internally contained) to the progenitor cells of every tissue. The concept of cortical areas being intrinsic to its progenitor cells was known as the “protomap” hypothesis (Rakic 1988). Around the same time, experimental evidence emerged for an opposing “protocortex” hypothesis (O’Leary 1989), which views the cortex as a “blank slate” that is extrinsically patterned by inputs from the thalamus, a brain region outside of the telencephalon entirely.

So does the telencephalon develop by shared rules, unique ones, or both? For both controversies just mentioned, the answer is both. For this reason, I divide this chapter into two parts. The first part covers the shared rules that apply just as well to the fly wing or your big toe as they do to the human telencephalon. In the second part, I cover the rules that are unique to the developing telencephalon.

The Telencephalon Follows Shared Rules of Development

Induction is the process of giving an organ, tissue, or cell its particular identity or “fate”; another term for induction is fate specification (e.g., specifying telencephalic fate). Broadly speaking, development can be viewed as a long series of inductions that give organs, tissues, and cells finer and finer identities over time. Patterning, on the other hand, is the process of subdividing an already-specified tissue into multiple parts. Thus, induction and patterning are two terms for the same process used in different contexts – individual fates are “induced,” while tissues are “patterned” into multiple fates. Induction and patterning are typically accompanied by changes in the sizes and shapes of tissues, which is referred to as morphogenesis.

For developmental biologists, the terms “morphogen” and “organizer” have strict definitions. For some events in telencephalic development, particularly earlier ones, these strict definitions have been met. For others, they have not. More generic terms such as “patterning signal” and “patterning center” could be used, but for clarity, conciseness, and readability, “morphogen” and “organizer” are used throughout this chapter. Be forewarned, however, when more precise word choices may be required, such as when discussing telencephalic development with a card-carrying developmental biologist!

The “French Flag” Model: Organizers, Morphogen Gradients, and Transcription Factors

The French Flag model underlies induction and patterning of all organs and tissues (Fig. 1). Organizers produce and secrete morphogens, most of which are proteins. The morphogens then diffuse away from their sources to create concentration gradients within adjacent tissues. (Although morphogens can get distributed in other ways, diffusion is hard to prevent and usually dominates.) Sometimes, organizers produce morphogen inhibitors rather than morphogens themselves, and morphogen degradation can also be regulated to sculpt the shapes of morphogen gradients. Regardless, within tissues, different amounts of morphogens accumulate outside of cells, which have receptors that translate the morphogen concentration information outside into signaling differences inside each cell. The signaling differences result in cells activating different cohorts of transcription factors, the master regulators of gene expression. In this way, cells end up with their unique gene expression profiles that determine the many facets of cell fate and identity.

Transcription factors are also central to the all-or-none (binary) nature of cell fate decisions. As depicted by the French Flag, cells are blue, white, or red, but not mixes of these colors. Although cells have different ways to make these binary decisions, it often comes down to the transcription factors competing with one another (mutual inhibition or cross-repression). Transcription factors also direct the formation of compartments, the other key feature of the French Flag model. As part of their job in specifying cell fate, transcription factors regulate cell adhesion molecules. Cells with the same transcription factors express the same cell adhesion molecules, which leads to cells sticking to one another (“homophilic” adhesion). This keeps cells together within compartments and limits their mixing or moving across compartment boundaries.

While transcription factors number in the 1000s (~10% of human genes encode for transcription factors), those regulated by morphogens are overrepresented in certain classes. These include the homeodomain, forkhead box (FOX), basic helix-loop-helix (bHLH), and zinc finger classes. In particular, homeodomain transcription factors have widespread roles in CNS patterning and can serve as selector genes (Garcia-Bellido 1975). Selector genes were so named, because they “select” a particular cell fate from a range of fate possibilities. This includes the selection of a cell’s adhesive properties. Hence, once morphogen information is translated into intracellular codes of selector genes and other transcription factors, tissues and cells have everything they need to create a French Flag.

Given the remarkable diversity of cell types in the animal kingdom, one might think that the number of different morphogens must be as impressive as that for transcription factors. In fact, the opposite is true – morphogen numbers are quite small. Rather, the cells responding to the morphogens account for the amazing diversity. As cells gain finer and finer identities, they increasingly lose “competence” for earlier identities, a process known as fate determination that is also regulated, in large part, by transcription factors. Ultimately, the vast ranges in cell competence across tissues and over time enable the same morphogens to be deployed over and over again in different tissues, or in the same tissue at different times, to diversify cell types in the telencephalon and elsewhere.

Anatomy of the Early CNS and Telencephalic Development

Before there is any “neural” identity, the embryo undergoes a process known as gastrulation to form its three basic layers: ectoderm, mesoderm, and endoderm (outer, middle, and inner layer, respectively). CNS development starts when part of the ectoderm is bestowed with neural identity and becomes the neural plate. The neural plate is a thin sheet of cells that runs from front-to-back (anterior-to-posterior) at the midline along the top (dorsal) edge of the embryo. The neural plate bends at its own midline to form the neural groove, which then fuses to form the neural tube (Fig. 2a). The open ends of this tube (neuropores) also fuse eventually. Initially, the neural plate and neural groove are covered by amniotic fluid, which surrounds the entire embryo. After neural tube closure, the CNS is covered by epidermal ectoderm (future skin), and its contact with amniotic fluid is greatly reduced. Choroid plexus inside the neural tube then takes over to produce cerebrospinal fluid (CSF), which fills the neural tube and bathes its outer surface.

Fig. 2

Anatomy of neural induction and the primary telencephalic organizers. (a) Formation of the neural groove and neural tube, whose roof plate and floor plate organizers are induced by non-neural organizers outside of the neural tube. BMPs from epidermal ectoderm induce the roof plate at its dorsal midline. Ventrally, SHH from the notochord induces the floor plate. Once induced, the roof plate and floor plate continue to produce BMPs and SHH, respectively. (b) Formation of the midline organizers of the telencephalon. In addition to the roof plate and floor plate, the ANR forms at the rostral midline and produces members of the FGF8 family. Abbreviations: ANR anterior neural ridge, BMP bone morphogenetic proteins, FGF fibroblast growth factor, SHH sonic hedgehog, RA retinoic acid. (Panels based on Medina (2009))

CNS induction, patterning, and morphogenesis begin in the neural plate and continue in the neural tube. From back to front, the neural tube first subdivides into four major subdivisions: the spinal cord, hindbrain (myelencephalon), midbrain (mesencephalon), and forebrain (prosencephalon) (Fig. 3a). The forebrain then subdivides into diencephalon and telencephalon, with telencephalon being the most anterior. As the anterior-posterior subdivisions form, they also become patterned in the top-bottom (dorsal-ventral) dimension (Fig. 3b). Dorsal-ventral patterning divides the telencephalon into left and right hemispheres and patterns each hemisphere into dorsal and ventral domains. The ventral and dorsal telencephalon give rise to several structures in the adult brain. Among these, the ventral telencephalon gives rise to the basal ganglia and amygdala, while the dorsal telencephalon forms the neocortex (“new” cortex), archicortex (“old” cortex, which is also known as hippocampus), paleocortex (“ancient” cortex, or piriform cortex), and olfactory bulb (Fig. 3c).

Fig. 3

Anatomy of the developing CNS and telencephalon. (a) Side views of the neural plate (top) through the emergence of telencephalon (bottom) as the neural tube is patterned into segments in its anterior-to-posterior dimension. (b) Side view of the telencephalon after segments have formed in its ventral domain, which is also known as the subpallium. (c) Cross sections of anterior and middle telencephalon. In the dorsal telencephalon (which is also known as the pallium), segments form for the hippocampus and piriform cortex, while the neocortex remains a single segment. Abbreviations: aep anterior entopeduncular area, db diagonal band, DI diencephalon, FB forebrain, HB hindbrain, LGE lateral ganglionic eminence, LP lateral pallium, MB midbrain, MGE medial ganglionic eminence, MP medial pallium, NP neural plate, poa preoptic area, SC spinal cord, TEL telencephalon, VP ventral pallium. (Panels (a) and (b) modified from Puelles et al. (2013). Panel (c) based on Medina (2009))

Induction and Patterning of the Early Telencephalon Follow the French Flag Model

Induction by Organizers Outside the Telencephalon

The earliest stages of CNS induction and patterning are directed by “non-neural” organizers – that is, organizers that are actually outside the CNS. The non-neural organizers reside next to or very close to the neural tissues they induce, and many reside along the embryonic midline. Neural plate induction is organized by the node, which produces FGFs (Fibroblast Growth Factors) and inhibitors (Chordin and Noggin) of morphogens known as the BMPs (Bone Morphogenetic Proteins).

Anterior-posterior patterning of the CNS is organized by mesoderm (axial and paraxial) and endoderm (anterior visceral endoderm). Axial and paraxial mesoderm produce FGFs, WNTs (Wingless-related integration site), and retinoic acid, which is unique among morphogens in being a vitamin derivative (vitamin A) rather than a protein. FGFs, WNTs, and retinoic acid specify spinal cord and hindbrain fates, the most posterior fates. Similar to the node, anterior visceral endoderm produces FGFs (specifically, members of its FGF8 subfamily) and inhibitors of BMPs and WNTs (Cerebrus, Dickkopf, and Tlc) to specify the forebrain and telencephalon, which are the most anterior fates. Dorsal-ventral patterning of the CNS is organized by mesoderm (notochord and prechordal mesoderm) and ectoderm (epidermal ectoderm) (Fig. 2a). Notochord and prechordal mesoderm pattern the ventral neural tube via the morphogens NODAL and SHH (Sonic hedgehog). Epidermal ectoderm patterns the dorsal neural tube via BMPs.

Patterning by Midline Centers Within the Telencephalon

In addition to specifying neural and telencephalic identities, the non-neural organizers induce new midline organizers within the telencephalon itself (Fig. 2b). The ventral and dorsal midline organizers of the telencephalon are known as the floor plate and roof plate, respectively. (Technically, the roof plate starts at the lateral edges of the neural plate before becoming the dorsal midline after neural tube closure.) An organizer known as the anterior neural ridge also forms at the rostral end of the telencephalon, creating a chain of organizers along the telencephalic midline. The ventral, rostral, and dorsal midline organizers produce SHH, FGF8, and BMPs, respectively. Interestingly, these telencephalic organizers produce the same morphogens that induced them. For example, the telencephalic floor plate produces SHH after being induced by SHH from prechordal mesoderm. This is an example of the “like inducing like” phenomenon known as homogenetic induction. Also known as the “community effect,” these homogenetic inductions of midline organizers represent the first of many ways that morphogens increase their ranges over space and time in the telencephalon.

Morphogen Gradients in the Early Telencephalon

After the midline organizers are induced, morphogen gradients are generated within telencephalic tissues. In the telencephalon, these gradients are best established for rostral FGF8 and dorsal BMP in the dorsal telencephalon, which nicely fit the French Flag model (Fig. 4). For example, the shapes of these gradients (declining exponentials) match those envisioned by Wolpert and others, and they occur when the telencephalon is relatively small. This is important, because morphogen gradients cannot pattern larger tissues on their own. As it turns out, the early gradient-containing telencephalon is roughly the same small size in mice and ferrets, even though their final telencephalic sizes are quite different. (The adult ferret has a much larger telencephalon than the mouse.) This fits the general developmental principle of “phylotypic” stages, which are developmental stages when species closely resemble one another. Typically, organs are just forming and still small at phylotypic stages, which is ideal for morphogenic patterning.

Fig. 4

Morphogen gradients in the early telencephalon. (a) The French Flag model applied to the BMP signaling gradient in the dorsal telencephalon. From high to low concentration, BMPs induce choroid plaque, choroid plexus, and cortical hem fates, then patterns the adjacent cortex. (b) The BMP signaling gradient as viewed by staining for phosphorylated Smad proteins, which transduce BMP signals. BMP signaling is highest at the dorsal midline, then declines more laterally in mouse dorsal telencephalon. This gradient is reduced and flattened in mutant mice with fewer BMP-producing roof plate cells. (c) Sagittal (side) view of the FGF8 protein gradient in the neocortical primordium (ncxp), shown in black-and-white (low-and-high signal, respectively). The ncxp is highlighted yellow. Scale bar, 0.1 mm. (d) The measured FGF8 gradient in ncxp, which is highest anteriorly, then declines more posteriorly. Abbreviations: mm millimeter, ncxp neocortical primordium, um microns. (Panels modified from Grove and Monuki (2020))

Maintaining Graded Information in Cortical Progenitor Fields

As mentioned near the beginning, a conundrum in the field had been the lack of anatomical or molecular segments in the cortical progenitor field. For transcription factors, this means that they themselves are expressed as continuous gradients rather than segments. This also implies that cortical progenitors can maintain graded morphogen information – that is, they can transfer the graded information in a linear or proportional manner without creating compartments and boundaries. Importantly, morphogen signaling pathways generally have linear designs that enable this proportional transfer of information. The BMP, WNT, and SHH signaling pathways do so by deploying transcription factors in an interesting way (Smad, Beta-catenin, and Gli transcription factors, respectively). Once the morphogens bind to their cell surface receptors, the receptors directly modify these transcription factors, which then move to the cell nucleus to regulate gene expression. In other words, the same transcription factors serve as both signal transducer and effector for these morphogen pathways, which avoids all sorts of interactions that amplify signals in nonlinear ways. The proportional transfer of morphogen information is even more direct for retinoic acid, which has no surface receptor. Rather, retinoic acid itself diffuses into cells and cell nuclei, where it directly binds to and regulates transcription factors (retinoic acid receptors). Although direct evidence for proportional information transfer is limited, this is established for the BMP-Smad system in dorsal telencephalic cells.

Converting Graded Information into Discrete Telencephalic Fates and Borders

While cortical progenitors can maintain graded information, the FGF8 and BMP gradients also induce multiple discrete fates at different concentration thresholds. Interestingly, when these fates are visible differs for the two gradients. For FGF8, this is not until functional areas emerge in the neocortex, which is discussed towards the end of this chapter. For BMPs, fate decisions are evident at the early progenitor stage like most other developing tissues.

Early on, the dorsal telencephalon can be subdivided into four parts. Starting from the midline, these are: (1) choroid plaque, (2) choroid plexus epithelium (CPE), (3) cortical hem, and (4) cortex (Fig. 4a), which itself can be subdivided into archicortex (hippocampus), neocortex, and paleocortex (Fig. 3c). At this early stage, high and intermediate BMP levels induce CPE and cortical hem fates, respectively. More laterally, lower BMP levels pattern the hippocampus and neocortex. Like all tissues, multiple mechanisms are used by dorsal telencephalic cells to convert graded BMP information into discrete CPE and hem fates. One mechanism is the rapid upregulation of certain homeodomain transcription factors (Msx1 and Msx2) by BMPs, which occurs within minutes. A second and much slower mechanism (hours to days) involves mutual inhibition between BMP and FGF signaling pathways, which converts the graded Msx1 and Msx2 changes into sharp “ultrasensitive” responses (Fig. 5a, b). The first mechanism is rapid, but primitive and crude; the second is slower, but more refined. Together, as revealed by systems biology (defined here as the combination of experiments with mathematical modeling and computer simulations), these mechanisms provide the system with “robustness,” or an insensitivity to noise and perturbations. These mechanisms also illustrate a few general truisms about development, which generally occurs in noisy, variable environments: (1) All developmental mechanisms have tradeoffs (the “no free lunch” principle), (2) all mechanisms work at different timescales, and (3) all robust developmental systems use multiple mechanisms to minimize the tradeoffs and work across timescales.

Fig. 5

Discrete telencephalic fates from interacting morphogens. (a) Schematic of a BMP target gene response (Msx1) with or without FGF. The Msx1 response without FGF is graded, but becomes sharp and ultrasensitive when FGF is present. Mutual BMP-FGF pathway inhibition accounts for Msx1 ultrasensitivity. (b) Schematic of dorsal telencephalic patterning by interacting BMP and FGF8 gradients. BMP-FGF mutual inhibition results in discrete domains with all-or-none cell fates. (c) Network of interactions between dorsal, rostral, and ventral morphogens of the telencephalon. (Panels (a) and (b) based on Srinivasan et al. (2014). Panel (c) modified from Monuki (2007))

In response to morphogens, telencephalic cells express transcription factors that determine their fates and identities. Early on, morphogens from ventral, rostral, and dorsal sources regulate Foxg1, which specifies telencephalic identity. Subsequently, transcription factors such as Gli3, Nkx2.1, and Gsx2 specify and subdivide the ventral telencephalon, while Msx1, Msx2, Lhx2, Emx2, Pax6, and others do the same in the dorsal telencephalon. Once telencephalic fates are specified, these same transcription factors help to refine and sharpen the borders between subdivisions. In the dorsal telencephalon, this includes the homeodomain transcription factors Lhx2 and Pax6, which act as classic selector genes. Lhx2 specifies cortical fate (hippocampus and neocortex together) and promotes sharpening of the cortex-hem boundary, while Pax6 sharpens a more lateral cortical boundary (pallial-subpallial boundary). Naturally, neither transcription factor works alone to sharpen these boundaries. For Pax6, Gsx2 (another homeodomain transcription factor) is a key partner for sharpening the pallial-subpallial boundary.

The Telencephalon Uses Additional Strategies Shared with Other Developing Tissues

Solving an “Ends” Problem: Morphogen Interactions

Being at the rostral end of the neural tube, the telencephalon has its own complex “ends” problem. This is solved, in part, by having its own rostral organizer (anterior neural ridge). In addition, for larger and more complex tissues, morphogens often work together to facilitate and coordinate tissue patterning, and such is the case for the telencephalon (Fig. 5c). Consistent with the French Flag model, morphogen interactions are strongest between adjacent organizers (ventral-rostral and rostral-dorsal pairs) where morphogen concentrations are higher. One important way that the morphogens interact is to regulate each other’s expression. For example, ventral SHH upregulates rostral FGF8 expression, and vice versa. Another way they interact is to co-regulate the same transcription factors. As mentioned, the midline morphogens co-regulate Foxg1 to specify telencephalic identity, then the BMP and FGF pathways interact to co-regulate Msx1 and Msx2 during CPE and cortical hem fate inductions.

Expanding Morphogenic Ranges: Expansion of Morphogen Families

Although morphogens are used over and over again in development, evolution has still diversified morphogen families to expand their specific roles and capabilities. The telencephalon is not unique in this regard, but does express large numbers of multiple morphogen families. This includes several dorsal BMPs and WNTs, rostral FGF8 family members, and several other FGFs and epidermal growth factors (EGFs). Often, members of the same family are expressed in a “nested” fashion – their expression epicenters coincide, often at the midline, then extend laterally to varying degrees. SHH is an interesting exception to this strategy. SHH does have family members (IHH and DHH), but they are not expressed in the telencephalon or in many other developing tissues.

Another family expansion worth mentioning is that for retinoic acid (RA). As mentioned earlier, RA is a vitamin derivative rather than secreted protein, so its availability as a morphogen is entirely different – it depends on vitamin A uptake from the diet, its distribution via the bloodstream, then local conversion into RA by cells that express enzymes known as retinaldehyde dehydrogenases (RALDHs). Thus, RA gradients are largely determined by the distribution of RALDHs, and the telencephalon expresses three RALDH genes in different locations and at different times. Conceptually, the unique features of RA should provide utility for patterning in unique telencephalic contexts, such as radial patterning and in larger tissues at later stages (e.g., during cortical arealization). We discuss some of these RA contributions later.

Expanding Morphogenic Ranges: Secondary Organizers

Since morphogen gradients can only pattern smaller tissues, an important objective for larger tissues is to establish secondary organizers. In the telencephalon, secondary organizers are induced bilaterally, which makes great sense for a symmetric organ that divides into two hemispheres (Fig. 6a). Ventrally, SHH from the midline floor plate induces bilateral SHH production in the preoptic areas, anterior entopeduncular areas, and medial ganglionic eminences. Rostrally, midline FGF8 induces the expression of FGF8 family members in bilateral septum and rostral poles of each hemisphere. Dorsally, BMPs from the midline roof plate induce the choroid plexus epithelium and cortical hem bilaterally, which produce BMPs and WNTs. Like the earlier midline inductions, the secondary organizer inductions are homogenetic – the secondary organizers produce the same morphogen that induced them. In the telencephalon, homogenetic induction can even extend past the secondary organizers into adjacent tissues. For example, BMP expression extends past the cortical hem into the adjacent hippocampal primordium. This expands morphogen ranges even further.

Fig. 6

Expanding morphogen ranges with secondary organizers and neuronal migrations. (a) Schematics of anterior and middle telencephalon indicating the bilateral positions of three secondary organizers – septum (yellow), antihem (red), and cortical hem (green). The secondary organizers continue to produce the morphogens that induced them as well as migrating neurons. (b) Schematic of the many neuronal migrations from ventral telencephalon. The LGE produces neurons destined for the olfactory bulb, while the MGE and CGE produce migratory cells that become cortical interneurons. Migratory neurons from dorsal and ventral subdivisions of LGE and CGE have somewhat different destinations. Abbreviations: CGE caudal ganglionic eminence, dCGE dorsal CGE, dLGE dorsal LGE, LGE lateral ganglionic eminence, MGE medial ganglionic eminence, vCGE ventral CGE, vLGE ventral LGE. (Panel (a) modified from Tole and Hebert (2020). Panel (b) modified from Wonders and Anderson (2006))

Notably, an additional organizer known as the antihem forms laterally at the dorsal-ventral boundary of the telencephalon (Fig. 6a). Among secondary organizers, the antihem is interesting for a few reasons. First, it produces a unique set of morphogens (EGF family members, FGF7, and the WNT inhibitor sFrp2). Second, being located laterally, it defies the rule of organizers being near the midline. Third, it seems to arise out of the blue – there is no nearby organizer that precedes it. It does, however, form at a tissue boundary, and looking across development, tissue boundaries are common places for secondary organizers to emerge. Apparently then, the lateral telencephalon has taken advantage of a new boundary to create the antihem, deploying yet another general strategy to expand morphogenic ranges in telencephalic space and time.

Expanding Morphogenic Ranges: Cell Migrations

Cell migration is another range-expanding mechanism used by many developing tissues. Using migration, cells can be locally induced by morphogens in bulk, then migrate to wherever they are needed. This provides another “economy of scale” in development and avoids the need to specify all cell types locally. Conceivably, organizer cells or the cells they induce could migrate. Both can occur, although the migration of induced cells is more common. Such is the case in the telencephalon, where the magnitude of neuronal migration is stunning.

Migratory neurons in the telencephalon come from two major sources: (1) secondary organizers and (2) the ventral telencephalon (Fig. 6a, b). The secondary organizers (cortical hem, antihem, septum, and another called the thalamic eminence) produce some of the first migratory neurons in the telencephalon, which are known as Cajal-Retzius cells. These cells migrate and occupy the very top layer (marginal zone) of the telencephalic wall. Cajal-Retzius cells migrating from secondary organizers end up covering the telencephalon, where they are thought to regulate telencephalic development in its radial dimension (see below).

The ventral telencephalon, on the other hand, produces migratory cells that become interneurons. Interneurons come from ventral telencephalic subdivisions that bulge into the ventricle and are known as the medial, lateral, and caudal ganglionic eminences (MGE, LGE, and CGE, respectively). All three eminences produce neurons that migrate in different streams to distribute widely throughout the ventral and dorsal telencephalon, but their destinations vary. LGE-derived neurons migrate mainly to the olfactory bulb, while neurons from CGE and especially MGE mainly end up in neocortex (Fig. 6b). Once settled at their final destinations, these interneurons make local inhibitory connections with other neurons using GABA (gamma aminobutyric acid).

Within neocortex, the MGE- and CGE-derived interneurons differ substantially from those generated by cortical progenitors. Rather than interneurons, cortical progenitors produce projection neurons that make long-distance excitatory connections to other CNS regions that can be quite distant. While interneurons make inhibitory GABA, projection neurons make excitatory glutamate. As the neocortex has expanded during evolution, its percentage of interneurons has also increased, requiring more and more migration from the ventral telencephalon.

The Telencephalon Has Also Adopted Its Own Unique Rules

Evolutionary Pressure for More Neocortex Without an Oversized Head

The evolutionary drive for more and more cognitive ability has led to more and more neocortex. This has come about in two ways – by increasing surface area (“tangential” expansion of the neocortical sheet) and by increasing cell complexity. As we have already discussed, the shared rules of development have helped this expansion effort, but have not been enough. Thus, to enable even more expansion, the neocortex has adopted its own unique rules. Applying the “no free lunch” principle, creating more neocortex must have tradeoffs, and one would be an oversized head. At some point, the head would become too big to balance on the neck or deliver via the birth canal. Accordingly, the telencephalon has adopted other mechanisms, particularly cortical folding, to become more compact and to keep head size manageable.

Adjacent Tissues Co-evolved to Support a Larger Neocortex

Support Inside and Out: The Superarachnoid Reticulum and Choroid Plexus “Bloom”

To make more neocortex, early neocortical progenitor numbers have dramatically increased in humans. Based on how the cortical sheet is built (see below under radial development), this progenitor expansion leads to substantial “ballooning” of the dorsal telencephalon at early stages. To accommodate this ballooning, a compliant tissue layer around the dorsal telencephalon known as the superarachnoid reticulum has conspicuously expanded during early human development (Fig. 7a). In adults, the superarachnoid reticulum becomes the subarachnoid space, which is the CSF space that surrounds the telencephalon and entire CNS.

Fig. 7

Tissue co-evolutions for an expanded neocortex. (a) The superarachnoid reticulum (yellow), the spongy and compliant tissue immediately surrounding the neocortex, is prominent during early embryonic development in humans. (b and c) Towards the end of embryogenesis (about 10 weeks gestation in humans), the choroid plexus “blooms” (green) within the telencephalic ventricles (blue) of humans and monkeys to support the thin and rapidly-expanding neocortex. (Panels (a) and (b) modified from images (courtesy of Shirley Bayer-Altman) or photographs (courtesy of Francisco Chanes) from Bayer and Altman (2008). Panel (c) modified from Smart et al. (2002))

Ballooning of the dorsal telencephalon has the tradeoff of potential collapse. Even in animals with smaller brains (rodents), the dorsal telencephalon collapses and cannot recover when its internal pressure is reduced (e.g., by draining ventricular CSF during early development). The dorsal telencephalic balloon in humans would be even more susceptible to such collapse. Perhaps unsurprisingly then, the choroid plexus inside the dorsal telencephalon has co-evolved to support the balloon (Fig. 7b). Presumably, this choroid plexus “bloom,” together with the CSF it produces, provides the internal pressure needed to prevent the human dorsal telencephalon from collapsing.

Patterning in the Radial Dimension Diversifies Neocortical Development

We now turn to radial patterning, or patterning across the cortical tissue (Fig. 8). Radial patterning is orthogonal to (90° from) the tangential patterning we have discussed thus far. At its earliest stages, the cortical sheet is composed of only one cell type, the neuroepithelial cell (NEC). NECs have slender bipolar processes that attach to the inner (ventricular) and outer (pial) surfaces of the cortical sheet. Thus, individual NECs themselves have a radial design. NECs also behave like “stem cells” – they divide and divide to generate more and more NECs (so-called “symmetric” cell divisions). Because of their radial design, NEC divisions lead to tangential expansion of the thin and single-layered cortical sheet. (This explains why the early dorsal telencephalon balloons in humans, as just mentioned.) NECs then differentiate into radial glia, which divide “asymmetrically” – in addition to making more radial glia, they produce other types of progenitors and neurons. The new neurons migrate outward, using the processes of radial glia as migratory scaffolds. As the young neurons reach the top of the cortical sheet, they form a distinct layer known as the cortical plate. Successive waves of migrating neurons then pass older ones to build more and more layers at the top of the cortical plate (Fig. 8a). In this way, the cortical plate grows in an “inside-out” fashion and eventually matures into the multilayered neocortex (six layers in humans) (Fig. 8c). As mentioned earlier under cell migration, these radially migrating neurons become the projection neurons of neocortex.

Fig. 8

Radial development of the neocortex. (a) Schematics of the developing neocortical wall at early and later stages. Radial glia and intermediate progenitors produce neurons that migrate outward along radial glial fibers to form the cortical plate. Younger cortical plate neurons pass older ones to occupy the very top layer of the cortical plate, thus building the neocortex in an inside-out pattern. (b) Schematics of the layered neocortical wall in monkeys and mice. (c) Illustrations of the six-layered neocortical plate in adult humans, showing the layering of neurons and other cell bodies (cytoarchitecture) and myelinated axons (myeloarchitecture) side-by-side. (Panel (a) modified from Striedter et al. (2015). Panel (b) modified from Smart et al. (2002). Panel (c) modified from Zilles et al. (2015))

Does the French Flag model apply to patterning in this radial dimension? Although preliminary, the answer appears to be “yes.” Logical morphogen sources for radial patterning would be CSF (in the ventricle and subarachnoid space) and cells (ependyma, meninges, and Cajal-Retzius cells) at the inner and outer surfaces of the cortical sheet, respectively. Morphogens secreted by choroid plexus into embryonic CSF (BMPs) are known to act upon cortical progenitors in rodents, as do morphogens emanating from the meninges (retinoic acid). Notably, choroid plexus-mediated patterning via the CSF means that the secondary organizers immediately adjacent to one another in the dorsal telencephalon – choroid plexus epithelium and cortical hem – pattern the cortex in orthogonal dimensions (radial and tangential, respectively). This would represent yet another evolutionary novelty in cortical patterning.

Intermediate Progenitors and Outer Radial Glia

As radial glia divide, they give rise to other progenitor subtypes known as intermediate progenitors (IPs) and “outer” radial glia (oRG) (Fig. 8a). Like radial glia, both of these progenitor subtypes generate radially migrating or radially-migrating neurons for the neocortex. IPs have short multipolar processes, while oRG have a single long process attached to the outer pial surface. Following the overall “inside-out” construction plan, the cell bodies of IPs and oRG occupy layers (subventricular zone and outer subventricular zone, respectively) above the radial glial layer at the bottom of the cortical sheet (ventricular zone) (Fig. 8b). Thus, new progenitor layers in the radial dimension arose during evolution. Presumably, the loss of one or both surface attachments for IPs and oRG facilitated the formation of these new layers. IPs are more abundant than oRG in smaller mammalian cortices, while oRG outnumber IPs in larger mammalian cortices, particularly those with more extensive cortical folding (see below). From an evolutionary perspective, the emergence of new progenitor subtypes in new radial layers makes sense. After all, as the choroid plexus bloom in humans suggests, there are limits to the number of cortical progenitors that can be accommodated by tangential expansion alone.

Cortical Folding Enables Even More Neocortex

Cortical folding (Fig. 9a) correlates strongly with brain size and cortical surface area across a large number of mammals. It is not hard to imagine why this is the case. As mentioned earlier, cortical folding solves the head size problem. In addition, cortical folding leads to more compact brains with shorter neuronal connections, which reduces energy costs and increases processing speeds. That being said, how cortical folding occurs has been more challenging to answer. Nonetheless, multiple mechanical and cellular factors have been implicated and can be understood as an interacting system of forces, processes, and biasing factors (Fig. 9b). Extrinsic mechanical factors include the skull and superarachnoid reticulum, while intrinsic forces and biases result from local differences in cell numbers, densities, proliferation, migration, and tensions exerted by neuronal processes. For layered organs such as the cortex, differences in stiffness and expansion rates between layers influence folding. The “buckling shell” model describes how structures with a distinct and rapidly growing shell, such as the cortical plate, will naturally buckle. This model is compelling for the neocortex, particularly given the shell expansion that occurs as waves of migrating neurons intercalate into the cortical plate (Fig. 9c). Ultimately, the different forces and biasing factors working together can explain both stereotypical and individualized folding of the neocortex.

Fig. 9

Development of cortical folding. (a) Progression of cortical folds in ferrets after birth and in humans before birth. (b) Model of interacting factors in cortical folding. External factors constrain intrinsic forces such as differential proliferation, differential tangential expansion, and axonal tension. (c) Model for intercalating young neurons as a proximate force in tangential expansion during outward and inward buckling of the cortical shell. Abbreviations: P postnatal day, wk weeks of gestation. (Panel (a) from Garcia et al. (2018). Panels (b) and (c) based on Striedter et al. (2015))

The Neocortical Areal Map Is Patterned by Both Intrinsic and Extrinsic Programming

While evolution acts to expand the neocortex for more cognitive power, development acts to create the functional areas of neocortex that provide this power. These functional areas are usually represented as maps of the cortical surface (Fig. 10a, b), and the process of creating this map is known as arealization. Uniquely, in addition to intrinsic patterning via the French Flag model, the cortical areal map is also patterned by extrinsic connections from the thalamus.

Fig. 10

Intrinsic and extrinsic influences in neocortical areal patterning. (a) Illustration of human neocortical areas designated by Brodmann. (b) Human and mouse cortical areas, viewed from above. Overall areal patterning is similar in the anterior-posterior dimension. (c) Comparison of the protomap and protocortex models of areal patterning. In the protomap model, ventricular zone (VZ) progenitors of the cortical primordium have the necessary patterning information for cortical areas. In the protocortex model, cortical VZ progenitors do not have this information, which is instead provided by connections that enter the cortex from different subdivisions of the thalamus. (d) Evidence for the protomap hypothesis. The normal FGF8 gradient is highest rostrally and acts on cortical VZ progenitors. When a posterior source of FGF8 is introduced in embryos, the cortical areal map in adults duplicates in a mirror image corresponding to the posterior FGF8 source. (e) Evidence for the protocortex hypothesis. After experimental removal of the superior and inferior colliculi, eye connections abnormally reroute to the MGN (part of the thalamus), thus relaying visual information to primary auditory cortex (A1). This leads to orientation maps (color-coded) in rewired A1 that resemble those in normal visual cortex (V1), including the presence of “pinwheels” (dotted circles). Abbreviations: A1 primary auditory cortex, CP cortical plate, Fr frontal cortex, IZ intermediate zone, LGN lateral geniculate nucleus, M1 primary motor cortex, MGN medial geniculate nucleus, S1 primary somatosensory cortex, V1 primary visual cortex, VZ ventricular zone. (Panel (a) from Zilles and Amunts (2010). Panel (b) modified from Chen et al. (2011). Panels (c), (d), and (e) modified from Grove (2020))

Intrinsic Patterning of Cortical Areas: Support for the “Protomap” Hypothesis

Under the “protomap” framework, overlapping transcription factor gradients in cortical progenitor fields collectively provide the intrinsic information for cortical areal patterning (Fig. 10c). Accordingly, manipulations of key morphogens or transcription factors in cortical progenitors alter the areal map, with FGF8 having a particularly profound effect (Fig. 10d). Normally, rostral FGF8 regulates the cortical progenitor field to specify rostral cortical areas (prefrontal and motor). When FGF8 is also produced posteriorly under experimental conditions in mice, the cortical areal map actually duplicates. The “new” rostral areas and their transcription factor codes emerge near the new posterior FGF8 source, as predicted by the French Flag model. Another morphogen implicated recently in the intrinsic specification of prefrontal and motor areas in humans is retinoic acid (RA). Synthesis of cortical RA has expanded in primates compared to mice, as has the upregulation of enzymes that breakdown RA in motor areas, which may limit RA activity to frontal regions.

Extrinsic Patterning of Cortical Areas: Support for the “Protocortex” Hypothesis

While an intrinsic program for arealization clearly exists, inputs from thalamic neurons also pattern cortical areas (Fig. 10c). In fact, some features of cortical areas depend entirely on this input. For example, a distinctive feature of primary somatosensory cortex in mice known as the “whisker barrels” do not form without this input. Dramatically, when visual inputs from thalamus are rerouted to auditory cortex, the re-wired auditory cortex adopts features of visual cortex, including visual representations and cellular architecture (Fig. 10e). That being said, the re-wired auditory cortex retains its normal auditory outputs and other aspects of its original auditory identity.

Collectively, these observations suggest an “extrinsic input, intrinsic output” model for cortical areal patterning. The job of neocortex is to receive and compute inputs, then send outputs. Neocortical output is provided by its projection neurons, which are intrinsically programmed via the French Flag model. On the other hand, thalamic neurons and their inputs have a completely different (and nontelencephalic) program. This leads to a model that combines the intrinsic programming of cortical outputs with extrinsic programming by inputs. In this sense then, the “protomap” and “protocortex” hypotheses are both correct, but not in isolation. Rather, they work together to create an intrinsic-extrinsic patterning model that is unique to neocortex.

The Neocortex Does Not Regenerate

Sometimes, what something cannot do is just as informative and interesting, if not more, than what it can. For neocortex, one interesting thing it cannot do is regenerate. Regeneration of body parts, including the CNS, occurs in invertebrates and amphibians such as the axolotl. Thus, the ability to regenerate is the more ancient condition, and some “older” parts of the telencephalon (hippocampus and olfactory bulb) partially retain this ability. For some reason, neocortex lost this ability during evolution. Could regeneration be a disadvantage or somehow nonadaptive for neocortex? It stands to reason that information stored by neurons and their circuits would be lost by neuronal turnover and regeneration. Thus, it may be that regeneration has been selected against in favor of long-term information storage and retrieval (i.e., memory).

Outlook

While much is known about the shared French Flag model of telencephalic induction and patterning, much remains to be learned. This includes gaps in our understanding of the mechanisms for expanding morphogen ranges, for interpreting morphogens, and for maintaining or converting graded morphogen information into discrete fates. In particular, the mechanisms underlying cortical arealization, which involves the conversion of graded information in progenitors into discrete areal fates in neurons, remains unknown. Holistic understanding of telencephalic development will benefit from systems biology approaches that combine experimentation with mathematical models and computer simulations, particularly for understanding time (which is harder to study than space) and system properties such as fragility, robustness, resilience, and evolvability.

Even more remains to be learned about the unique rules of telencephalic and neocortical development. Some only occur in humans or mammals with a large telencephalon, which are harder to study. Fortunately, some of the unique rules can be studied in smaller animal models, and more and more tools are now available to study the human telencephalon directly. These include cultures of embryonic brain slices, stem cell and organoid models, “big data” on postmortem and resected brain tissues, and rapid advances in neuroimaging. Ultimately, as they say, “we still don’t know what we don’t know.” If past predicts future, there is still a lot we don’t know. For this reason, unbiased observation and discovery studies of the telencephalon should continue to be highly informative. Increasingly, these discoveries will provide connections between normal telencephalic development and human malformations, including common disorders of telencephalic induction and patterning such as anterior neural tube defects, holoprosencephaly (which involves morphogens directly), and hydrocephalus (which impacts radial development directly).